HomeMy Public PortalAbout20110913 - Agenda Packet - Board of Directors (BOD) - 11-23 Midpeninsula Regional
' Open Space District
Meeting 11-23
SPECIAL MEETING
BOARD OF DIRECTORS
MIDPENINSULA REGIONAL OPEN SPACE DISTRICT
330 Distel Circle
Los Altos, California
Tuesday, September 13, 2011
SPECIAL MEETING OF THE MIDPENINSULA REGIONAL OPEN SPACE DISTRICT BEGINS AT 5:00 P.M.*
AGENDA
SPECIAL MEETING
5:00 ROLL CALL.
SPECIAL ORDERS OF THE DAY
SPECIAL MEETING OF THE BOARD OF DIRECTORS OF THE MIDPENINSULA REGIONAL OPEN SPACE DISTRICT—
STUDV SESSION
l. Acceptance of the District's Classification and Compensation Study Report—A. Spiegel and K. Kaneko
ADJOURN
* Times are estimated and items may appear earlier or later than listed. Agenda is subject to change of order.
TO ADDRESS THE BOARD: The President will invite public comment on agenda items at the time each item is
considered by the Board of Directors. You may address the Board concerning other matters during Oral
Communications. Each speaker will ordinarily be limited to three minutes.Alternately,you may comment to the
Board by a written communication, which the Board appreciates.
In compliance with the Americans with Disabilities Act,if you need assistance to participate in this meeting,please contact the District Clerk at(650)691-1200.
Notification 48 hours prior to the meeting will enable the District to make reasonable arrangements to ensure accessibility to this meeting.
Written materials relating to an item on this Agenda that are considered to be a public record and are distributed to Board members less than 72 hours prior to the
meeting,will be available for public inspection at the District's Administrative Office located at 330 Distel Circle,Los Altos,California 94022.
CERTIFICATION OF POSTING OF AGENDA
1,Michelle Radcliffe,District Clerk for the Midpeninsula Regional Open Space District(MROSD),declare that the foregoing amended agenda for the Special
Meeting of the MROSD Board of Directors was posted and available for review on September 9,2011,at the Administrative Offices of MROSD,330 Distel Circle,
Los Altos California,94022. 'The agenda is also available on the District's web site at http://www.openspace.org.
Signed this 9`'day of September,at Los Altos,California.
tµ: acy 4%.9.a: e.
District Clerk September 9,2011
i
I
Midpeninsula Regional
0 ' Open Space District
r
R-11-95
Meeting 11-24
September 13, 2011
STUDY SESSION AGENDA ITEM 1
AGENDA ITEM
Acceptance of the District's Classification and Compensation Study Report
GENERAL MANAGER'S RECOMMENDATIONS
The General Manager recommends that the Board:
1. Accept the Classification and Compensation Study Report prepared by Koff& Associates.
2. Direct the General Manager to:
a. Review and analyze Koff& Associates' Classification and Compensation Study Report
and identify recommended adjustments to the recommendations;
b. Review and analyze feedback submitted by employees since the July 13, 2011 Study
Session;
c. Return to the Board by December 2011 for approval and implementation of the
Classification and Compensation Study, including any recommended changes to the
report prepared by Koff& Associates.
SUMMARY
At the July 13`' Board Study Session, the Board reviewed the District's proposed Classification
and Compensation Study prepared by Koff& Associates and received comments from District
employees regarding the Study. This Report details the salary impacts on employees as well as
the total initial cost of implementing the compensation adjustments.
DISCUSSION
Study Process and Recommendations
After a year-long process, the District Board of Directors (Board) reviewed the Classification
and Compensation Study Report (Study) and recommendations prepared by Koff& Associates
(Consultant) at a Board Study Session on July 13, 2011 (Report R-11-67). At this Study
Session, Katie Kaneko, President of Koff& Associates, presented the Study's methodologies,
findings, and recommendations to the Board. Additionally, a large number of District employees
were in attendance and many provided oral feedback and comments regarding the Study. A
R-I 1-95 Page 2
chronological summary of the milestones in the process, including requests for information and
opportunities for staff input, is presented in Attachment A.
Below is a summary of the key recommendations included in the report prepared by the
Consultant (see Report R-1 1-67 for more detail):
Classification Changes
1. Title changes are recommended for 15 classifications to accurately reflect the actual job
responsibilities and duties performed by class incumbents and to be consistent with industry
terminology;
2. Six positions are recommended for reclassification due to level and scope of work and/or job
functions that have been added to the positions over time; and
3. Five positions are recommended to change from non-exempt (hourly) to exempt (salaried)to
comply with Fair Labor Standards Act (FLSA) rules.
Compensation Adjustments
1. The recommended salary model has 65 salary ranges, with each range comprised of ten steps
that are 2.5% apart from each other(Attachment B);
2. The top salary for each range is recommended to be at the 50th percentile of the market data
for the selected comparator agencies;
3. The top step salary is proposed to increase for 24 job classifications and to decrease for 18
classifications (Attachment C);
Employee Compensation Impacts
When the compensation plan is implemented, the Consultant recommends employees be placed
in their new salary ranges at their current salary or the next higher step, to ensure that no
employee's salary is reduced. Thus, of the 93 filled positions at the District (excluding seasonal
employees and Board Appointees), no employee will experience a salary reduction as a result of
implementation of the proposed compensation plan (Attachment D). Specifically:
• Seventy-four employees (80%) will receive a salary increase ranging from $3 to $470 per
month, with a median increase of 0.7% per employee.
• Currently, 56 employees (60%) are at top step; in the recommended new salary range
placements, 42 employees (45%) will be at top step. This means 14 employees who are
currently at top step will now be eligible for additional step increases under the
recommended compensation plan.
• Nineteen employees (20%) will be above the top step of their proposed new salary ranges.
These employees will remain at their current salaries and be Y-rated, which means that they
will remain at their current salaries until the top step of their range catches up to their current
salary through cost-of-living adjustments (COLAs). These employees would not receive
COLAs until the top step of their new salary range equals or exceeds their current salary.
Employee Feedback
At the July 13, 2011 Study Session, after hearing comments from 14 employees, the Board
invited staff to provide additional written feedback on the Study process, methodology, and
recommendations by August 19, 2011. As of August I 9th, eight employees had submitted
written feedback to the Board.
R-I 1-95 Page 3
On August 10, 2011, the Board received a letter from the Field Employees' Association's (FEA)
legal representative, Goyette & Associates, Inc., requesting that the Board postpone any
discussion of the employee feedback for at least a month so that the FEA could hire an
independent consultant to evaluate the Classification and Compensation Study prepared by the
Consultant. District General Counsel responded with a letter to Goyette stating that the Board is
not scheduled to discuss the Classification and Compensation Study until its Study Session on
September 13, 2011 at 5:00 p.m. and that the FEA may submit any additional information at that
time.
FISCAL IMPACT
The fiscal impact of implementing the Study's compensation recommendations in FY2012-13 is
$68,784 in salaries and $15,001 in CalPERS retirement costs, for a total increase of$83,785
PUBLIC NOTICE
Public notice was provided as required by the Brown Act. No additional notice is required.
CEQA COMPLIANCE
No compliance is required as this action is not a project under CEQA.
NEXT STEPS
It is recommended that the Board direct the General Manager to:
1. Review the employee feedback on the District Classification and Compensation Plan
proposed by the Consultant.
2. Return to the Board in December, 2011 with recommended adjustments to the Consultant's
proposed Classification and Compensation Plan.
Attachments:
1. Attachment A: Summary of Activities
2. Attachment B: Proposed Salary Schedule
3. Attachment C: Current vs. Proposed Salary Range Top Step
4. Attachment D: Employee Salary Impact
Prepared by:
Kate Drayson, Administrative Services Manager
Annetta Spiegel, Acting Human Resources Supervisor
Contact person(s):
Same as above
ATTACHMENT A: CLASSIFICATION AND COMPENSATION STUDY TIMELINE
Consultant Selection Process
Nov 13, 2009 Classification and Compensation Study RFP posted on District website
Jan 8, 2010 Final Filing date for RFP responses
Jan — Mar, 2010 HR staff reviewed and analyzed 15 RFP bids
April 27 & 29, 2010 Project Team interviewed CPS Human Resource Services, Bryce Consulting, and Koff& Associates
April 29, 2010 Benefits study request received from Anthony Correia on behalf of the FEA per MOA
May 7, 2010 Reference checks completed for all three consultants
May 11, 2010 Verbal quote of$6,000 received from Koff&Associates for addition of benefits study
May 7, 2010 Project Team Selection of Koff& Associates announced
June 9, 2010 Board of Directors authorized Koff&Associates to conduct the Class & Comp Study
June 21, 2010 Professional Services Agreement with Koff&Associates signed by General Manager
July 9, 2010 Cost proposal for Total Compensation Study received from Koff & Associates
Aug 11, 2010 Board of Directors authorized addition of benefits study to Koff&Associates contract
Classification Study
May 26, 2010 Met with FEA to discuss Class & Comp Study
June 29, 2010 AO Employee Orientation - distributed the Position Description Questionnaire (PDQ) and discussed process
June 29, 2010 Project Team discussed proposed benchmark classifications and proposed comparator agencies
July 1, 2010 FLSA Exemption Status explanation memo sent to all employees from Katie Kaneko
July 1, 2010 Pre-meeting for FFO and SFO Orientation with Anthony Correia, FEA President, Katie Kaneko of Koff&
Associates, and Acting HR Supervisor Annetta Spiegel,
July 7, 2010 FFO & SFO Employee Orientation - distributed the Position Description Questionnaire (PDQ) and discussed
process
July 23, 2010 Deadline for employees to submit completed PDQ to their supervisor or manager
July 30, 2010 Deadline for supervisors/managers to have reviewed/signed off on employee PDQ's
Aug 2, 2010 HR review of PDQ's and supervisor/manager comments completed
Aug 2, 2010 Employee PDQ's submitted to Koff& Associates
Aug 12, 2010 Koff& Associates interviews with employees completed
Aug — Oct, 2010 Job classifications Developed per PDQ's, employee interviews, and supervisor comments
Nov 10, 2010 Sent out Draft Classifications and cover letters to employees
Dec 6, 2010 Completed HR review of Final Classifications
Dec 8, 2010 Sent Final Classifications with changes and explanation cover letters to all employees
May 25, 2011 Received Final Classification Report: Volume 1 from Koff&Associates — sent to all staff and the Board via
email
1
Compensation Study
Sept 7, 2010 Received "proposed benefit data information to be collected" memo from FEA
Oct 8, 2010 Held Meet & Confer meeting with FEA to review Koff& Associates recommended comparator benchmark
agencies
Jan 7, 2011 Sent Top Monthly Salary, Benefits, and Total Compensation spreadsheets and cover letter to employees for
review
Jan 14, 2011 Deadline for employees' comments to salary and benefits benchmark data
Jan 24, 2011 Received response from Koff& Associates to employees' benchmark data comments
Feb 2, 2011 Received Koff &Associates Results Summary, proposed Range Placement Recommendations, and Salary
Schedule
Mar 17, 2011 Received proposed classification and compensation implementation plan
June 6, 2011 Received Final Range Placement Recommendations from Koff&Associates
June 13, 2011 Received Final Compensation Report: Volume 2 from Koff&Associates— sent to all staff via email
Board Communications
Jan 12, 2011 Katie Kaneko presented update on Classification and Compensation Study to the Board
July 13, 2011 Katie Kaneko presented Final Classification and Compensation Study Findings and Recommendations to the
Board
August 19, 2011 Employee deadline to submit their classification & compensation comments and concerns to the Board
President
* Project Team — Director Mary Davey, General Manager Steve Abbors, Acting HR Supervisor Annetta Spiegel, FEA President
Anthony Correia, and Acting Human Resources Analyst Sonya Siebe
i
2
ATTACHMENT B:
Proposed Salary Schedule
Range# Monthly Salary Range
Step 1 I Step 2 Step 3 1 Step 4 1 Step 5 1 Step 6 1 Step 7 Ste 8 1 Ste 9 Ste 10
1 $2,581 $2,646...... ......$2,712.............$2,780........... $2,849.............$2,921....... ......$2,994.............$3,068............$3,145.............$3,224......
Z........ ......$2,646.............
$2,712...... ......$2,780.............$2:849............$2:921....... ......$2;994...... ......$3,068.............$3:145............$3:224 $3,304..................... .....................
...........................
3 $2,710 $2,778..... ......$2,848.............$2,91.9............$2,992.............$3.067...... ......$3,143.............$3,222............$3.302.............$3,385......
4 .... ......$2,778.............$2,848 $2,919 ......$2,992.............$3,067.............$3,143...... ......$3,222.............$3,302............$3,385.............$3,470......
..................... ........................................I.............. ..........................
5 $2,846 $2.917 $2,990 ......$3,065....._.,..._$3,141..............$3.220.,.,.. ......$3,300.............$3,383............$3:467...... _....$......................... ....................................................... .............................9
6 $2,917 $2,990 $3,065 $3,141 $3,220 $3,300 $3,383 $3,467 $3:554 $3.643.._._.
..................... ........................... ........................... ................ .
7 $2,990 $3,065 $3,141 $3,220 $3,300.............$3,383,,..., ......$3,467...,,,,......$3,554 $3,643 $3,734
..................... ....................................................... ................. .
8 $3,063 $3,139..... ......$3,218......
.......$3.298......
......$3.381...............
....$3,465............
......$3,552.............
$3:641...
............$3,732......
....... 3,825......
9 ..$3,139 $3,218 $3,298 $3,381 $3,465 $3,552 $3,641 $3,732 $3,825 $3,921
..................... ....................................................... ................................................ . .......... ........................... .................................................................................... ...........................
10 $3,216 $3,296...... ......$3,379.............$3:463.............$3.550...... ......$3,639...... ......$3,730.............$3;823............$3:91.8.............$4,016......
11... ....$3,296.....,....$3,379...... ...,..$3463.............$3,550 $3,639 $3,730 $3823 $3918 $4,016 $4,117
.................................................................................................
12 $3,377 $3 461 $3,548 $3,637 $3,727 $3,821 $3,916 $4,014 $4,114 $4,217
.............. .................................................................................... . ...............................................................................................
13....... .....$3,461....... ....... 3,548...... ......$... ..............$3.727............$3,821..............$3.916...... ......$4,014.............$4.11..............$4,21.7.......... $4.323
14 $3,546 $3 634 $3 725 $3,818 $3,914 $4,012 $4 112 $4 215 $4,320 $4,428
............................................................................................... ....... ..................................................................
15 $3634............$3,725 $3,818............$3,914.............$4.01.2.............$4,112...... ......$4,215..............$4,320............$4,428.............$4,539......
..................... ....................................................... ................ .
16 $3,723 $3,816...... ......$3,911.............. ......$4.009............$4:109...........................$4212.............. ............$4318...........................$4:425...........
..........$4,536.............�4,649......
17....... ......$3,816.............$3,911 $4,009 .......$4,109............$4.21.�.............$4,318...... ......$4,425...... ......$4;536............$4.649.............$4.766......
..................... ........................... ........................... ...........................
18 $3,909 $4,007 $4,107.............$4,210.....,......$4,31.5,,,,,,,,,,...$4,423 $4,533....,...,,.,.$4,647 $4,763 $4,882
19 $4,007 $4,107 $4,210 $4,315 $4,423 $4,533 $4,647 $4,763 $4,882 $5,004
.........6....... ........................... ........................... ........4..,.,.............................6.............,.............. .......... ........................... ....................................................................................
20 $4,105 $4,207 $4,312 $4,420 $4,531..............$4,644 $4,760 $4,879 $5,001 $5,126
21 $4,207 $4,312 $4,420 $4,531 $4,644 $4,760 $4,879 $5,001 $5,126 $5,254
....... ................................................ .......................................................................... ... . . ........ ....................................................................................
....... .....$4:31. ...... ......$4,418...... ......$4,528...... ......$4,641.............$4,757.............$4,876...... ......$4,998.............$5:123............$5,251..............$5,382......
23 $4 418 $4 528 $4 641 $4,757 $4,876 $4,998 $5 123 $5 251 $5,382 $5,517
. ................................... ........................... ...............................................................................................................
24 $4,525 $4,638 $4 754 $4 873 $4,995 $5,120 $5,248 $5,379 $5,514 $5,651
....................................................................... .......................... .................................................................................... ...........................
...25....... ......$4,638 $4.754 $4,873 $4,995 $5,120 $5,248 $5,379 $5,514 $5,651 $5,793
................................................................................... ........................... .................................................................................... ...........................
26....... .....$4,752.............$4.870...... ............4,992.....................$5.117............$5,245......................$5.376...... ......$5,510.............$,5;648............$5,789...........I.$5.934......
27 $4 870 $4 992 5 117.............$5,245,.,,,.,,,...$5,376 $5,510 $5 648 $5 789 $5,934 $6,082
......................................................................................I...........
28...... $4,989 $5,114 $5,242 $5,373 $5,507 $5,645 $5,786 $5,931 $6,079 $6,231,
.......................................
..............................
.......
..........
...
......
. .......................................
... ......................
........29....... ...... .............$5.242...... ......$5,373..........................5,507.....................$5,645.............�5.786...... ......$5,93................$6.079............$6.......................$6.387..........
30 $5,239 $5,370 $5,504 5,641 $5,782 $5,927 $6,075 $6,227 $6,383 $6,542
..................... ....................................................... ...................
31 $5,370 $5.504 $5,641 $5,782 $5,92.. $6,075 $6,227 $6,383 $6,542 $6,706
32,,.,... $5,501 $5,638 $5,779 $5,923 $6,072 $6,223 $6,379 $6,538 $6,702 $6,869
..................... ............................................ ................................................................................................................
33 $5,638 $5,779 $5,923 $6,072 1 $6,223 $6,379 $6,538 1 $6,702 $6,869 $7,041
Page 1 of 2 ATTACH B_Proposed Salary Sched
ATTACHMENT B:
Proposed Salary Schedule
Monthly Sala Range
1.Range# Step 1 Step 2 Ste 3 Ste 4 Ste 5 Step 6 1 Step 7 Step 8 Ste 9 Ste 10
34....... $5,776 $5,920 $6,068 $6,220 $6,375 $6,535 $6,698 $6,865 $7,037 $7,213
35 $5,920...... ......$6,068..... ......$6,220.............$6,375.............$6,535.............$6,698..... ......$6,865.............$7,037.............$7,213.............$7,393.....
3s....... ,064.............$6,216..... ......$6,371..............$6,531..............$6,694.............$6,861...... ......$7,033.............$7,209.............$7,389.............$7,574...........
37....... ......$6,216.............$6,371...... ......$6,531..............$6,694.............$6,861.............$7,033..... ......$7,209.............$7,389.............$7,574...... ......$7,763.....
.........6....... ....................................................... ................................................................................................................ ................................................................................................................
38 $6,368 $6,527 $6,690.............$6,857,...........$71029 $7,204 $7,384.....,,,,...,$7,569...........$7,758 $7,952
........................ 6...... .................
39....... .............$6,690...... ......$6,857.............$7,029............$7.204.............$7,384...... ......$7,569.............$7,758.............$7.952.............$...151.......
.......,4.6....... .....$6,686 $6,853 $7024 ........,.$7 .00.... ......$7,380 $7,565 $7754 $7,947 $8,146 $8,350
..................................................................................................
41....... .....$6;853.............$7,024...... ......$7,200.............$7.380............$7,565.............$7,754...... ......$7,947.............$8,146............$8,350...... ......$8,559......
42 $7,020 $7 196 $7 376 $7 560 .......,..$7,749 $7,943 $8 141 $8 345 $8,553 $8,767
....... . ............. .. ...................................................................
43....... .....$7,196.............$7,376...... ......$7,560.............$7'749............$7,943.............$8,141....... ......$8,345.............$8,553............$8,767.............$8,986......
44....... $7,371 $7,556 $7 744 $7 938 $8.136 $8,340 $8 548 $8 762 $8,981 $9,206
.. .. .. . ...................................................................
45....... .....$7,556.............$7,744.................. ......$7,938...............$8,136............$8.340.............$8,548...... ......$8,762.............$8,981.............$9,206.............$9,436......
46....... .....$7,740.............$7,933...... ......$8 132.............$8 335 .........$8,543.............$8,757...... ......$8 976.............$9 200 $9,430 $9,666
...................................................................
47....... .....$7.933.............$8,132...... ...... 8,335.............$8.543............$8.757...... ......$8,976...... ...... 9,200.............$9:430............$9.666.
............�9,908......
48.,..... ,,.,,$8.127 $8,330 $8 538 $8 752.............$8,970 $9,195 $9 425 $9 660 $9,902 $10,149
.......................-...........................................
49 $8,330 $8,538 $8,752 $8,970 $9,195 $9,425 $9,660 $9,902 $10,149 $10,403
..................... ........................................................ ................................................................................................-.............. ........................... ....................................................................................
50....... .....$8,533.............$8,746...... ......$8,965.............$9.189............$9:41.9.............$9,654...... ......$9,896............$10,143..........$10,397..........$10,657.....
51 $8,746 $8,965 $9,189 $9,419 $9,654 $9,896 $10,143 $10,397 $10,657 $10,923
..................... ...........................................-......... ................................................................................................................ .................................................................................... ...........................
52....... .....$8,960.............$9,184...... ......$9,413.............$9,649.............$9,890...........$10,137..... .....$.10,391...........$10,650..........$10,917..........$1.�.,�.g0.....
53.,...., ,.,.,$9.184 $9,413 $9 649 $9 890 $10,137 $10 391 $10 650 $10 917 $11 190 $11,469,
54 $9,408 $9,643 $9,884 $10,131 $10,384 $10,644 $10,910 $11,183 $11,462 $11,749
...........................
55....... $9,643.............$9,884...... ....$10;131 $10,384 $10,644 $10:910.... .....$11,183..,,,....$11,462 $11,749 $12,043
56....... .....$9,878 $10,125 $10 378 $10 638 .$10,904 $11,176 $11,456 $11 742 $12 036 $12,336
57 $10,125 $10,378 $10,638 $10,904 $11,176 $11,456 $11,742 $12,036 $12,336 $12,645
58 $10,372 $10,631 $10.897 $11,170 $11.449 $11,735 $12,028 $12,329 „$12,637 $12,953
59 $10,631 $10,897 $11,170 $11,449 $11,735 $12,028 $12,329 $12,637 $12,953 $13,277
..................... ........................... ........................... ........................... ........................................................ ........................... .................................................................................... ...........................
60 $10,891 $11,163 $11,442 $11,728 $12,021 $12,322 $12,630 $12,946 $13,269 $13,601
..................... ....................................................... ................................................................................................................ ........................... ................................-.................................................
61 $11,163 $11,442 $11,728 $12,021 $12,322 $12,630 $12,946 $13,269 $13,601 $13,941
..................... ....................................................... ................................................................................................................ ................................................................................................................
62 $11,435 $11,721 $12,014 $12,314 $12,622 $12,938 $13,261 $13,593 $13,933 $14,281
................................. .................................................................................... ........................... ...................................................................-.............. ...........................
63 $11:721 $12,014 $12,314 $12,622 $12,938 $13,261 $13,593 $13,933 $14,281 $14,638
. .. ......... ....... . ......................... ......................... . ............. ........................... ................................................................................................................
64 $12,007 $12,307 $12,615 $12,930 $13,253 $13,585 $13,924 $14,273 $14,629 $14,995
......... ....... . ............... ........................................................ ...........................
65 $12,307 $12,615 $12,930 $13,253 $13,585 $13,924 $14,273 $14,629 $14,995 $15,370
Page 2 of 2 ATTACH B_Proposed Salary Sched
i
ATTACHMENT C:
Current vs. Proposed Salary Range Top Step
Current Proposed Monthly Salary
I
Current Class Title(Proposed Title) Maximum Maximum Percent
Monthly Salary Monthly Salary Change
I
Assistant General Manager $12,178 $14,281 17.3%
Administrative Services Manager $10,177 $11,190 9.9%
Operations Manager $10,177 $11,190 9.9%
Planning Manager $10,177 $11,190 9.9%
Real Property Manager $10,177 $10,149 -0.3%
Public Affairs Manager $10,177 $10,149 -0.3%
Assistant General Counsel(Assistant General Counsel ll) $9,491 $10,657 12.3%
Assistant General Counsel I Proposed $9,666 N/A
District Clerk $8,850 $8,559 -3.3%
Senior Management Analyst $8,850 $8,151 -7.9%
Area Superintendent $8,590 $9,206 7.2%
Human Resources Supervisor Proposed $9,206 N/A
Senior Resource Management Specialist $8,172 $8,986 10.0%
Senior Planner $8,172 $8,986 10.0%
j Management Analyst-HR(HR Analyst) $7,697 $6,706 -12.9%
Management Analyst(Management Analyst-Operations) $7,697 $7,041 -8.5%
Resource Management Specialist(Resource Mgmt Specialist II) $7,396 $7,041 -4.8%
Support Services Supervisor $7,396 $7,041 -4.8%
GIS Coordinator(GIS Administrator) $7,396 $8,350 12.9%
Open Space Planner 11 (Planner ll) $7,396 $8,151 10.2%
Real Property Specialist $7,396 $7,393 0.0%
Maintenance&Resource Supervisor(Maint., Construction&Resource Supv.) $6,694 $7,393 10.4%
Supervising Ranger $6,694 $7,393 10.4%
Open Space Planner I (Planner I) $6,368 $6,706 5.3%
Website Content Coordinator(Website Administrator) $6,368 $7,574 18.9%
Communications Specialist(Media Communications Specialist) $6,368 $7,041 10.6%
Public Affairs Specialist $6,368 $7,041 10.6%
Resource Specialist I (Resource Mgmt. Specialist l) $6,368 $6,387 0.3%
Senior Accounting Specialist(Accountant) $6,242 $6,706 7.4%
1
ATTACHMENT C:
Current vs. Proposed Salary Range Top Step
Current Proposed Monthly Salary
Current Class Title(Proposed Title) Maximum Maximum Percent
Monthly Salary Monthly Salary Change
Equipment Mechanic/Operator $6,242 $5,517 -11.6%
Network Specialist(lT Specialist) $6,058 $7,574 25.0%
Docent Coordinator(Docent Program Manager) $6,058 $6,706 10.7%
Volunteer Coordinator(Volunteer Program Manager) $6,058 $6,231 2.9%
Training and Safety Coordinator(Training and Safety Specialist) $5,880 $7,574 28.8%
Deputy District Clerk/Office Manager(Senior Administrative Assistant) $5,880 $5,651 -3.9%
Human Resources Technician $5,880 $5,382 .8.5%
Ranger $5,764 $5,793 0.5%
Planning Technician $5,764 $5,517 -4.3%
Lead Open Space Technician $5,706 $5,517 -3.3%
Assistant Program Coordinator(Public Affairs Program Coordinator) $5,538 $5,793 4.6%
Accounting Clerk(Accounting Technician) $5,268 $4,766 -9.5%
Administrative Assistant $5,268 $5,126 -2.7%
Farm Maintenance Worker $5,164 $5,004 -3.1%
Open Space Technician $5,164 $5,004 -3.1%
Seasonal Open Space Technician $3,643 $3,385 -7.1%
2
ATTACHMENT D:
EMPLOYEE WAGE IMPACT
Current Current Proposed Proposed Proposed Monthly Monthly
Monthly Salary Salary Salary Monthly Wage Percent
Salary Step Range Step Salary Increase Increase
1 $5,268 6 17 10 $4,766 $0 Y-rated
2 $4,913 5 19 10 $5,004 $91 1.85%
3 $5,164 6 19 10 $5,004 $0 Y-rated
4 $5,164 6 19 10 $5,004 $0 Y-rated
5 $5,164 6 19 10 $5,004 $0 Y-rated
6 $5,164 6 19 10 $5,004 $0 Y-rated
7 $4,913 5 19 10 $5,004 $91 1.85%
8 $4,674 4 19 8 $4,763 $89 1.90%
9 $5,164 6 19 10 $5,004 $0 Y-rated
10 $4,913 5 19 10 $5,004 $91 1.85%
11 $4,674 4 19 8 $4,763 $89 1.90%
12 $4,913 5 19 10 $5,004 $91 1.85%
13 $4,913 5 19 10 $5,004 $91 1.85%
14 $5,268 6 20 10 $5,126 $0 Y-rated
15 $4,536 3 20 6 $4,644 $108 2.38%
16 $5,012 5 20 10 $5,126 $114 2.27%
17 $5,268 6 20 10 $5,126 $0 Y-rated
18 $5,268 6 20 10 $5,126 $0 Y-rated
19 $4,768 4 20 8 $4,879 $111 2.33%
20 $5,012 5 20 10 $5,126 $114, 2.27%
21 $4,316 5 20 4 $4,420 $104 2.41%
22 $4,316 5 20 4 $4,420 $104 2.41%
23 $5,063 3 22 6 $5,123 $60 1.19%
24 $6,242 6 23 10 $5,517 $0 Y-rated
25 $6,242 6 23 10 $5,517 $0 Y-rated
26 $6,242 6 23 10 $5,517 $0 Y-rated
27 $5,706 6 23 10 $5,517 $0 Y-rated
28 $5,706 6 23 10 $5,517 $0'Y-rated
29 $5,594 6 23 10 $5,517 $0'Y-rated
30 $5,706 6 23 10 $5,517 $0 Y-rated
31 $5,594 5 24 10 $5,651 $57 1.02%
32 $5,764 6 25 10 $5,793 $29 0.50%
33 $5,764 6 25 10 $5,793 $29, 0.50%
34 $4,963 3 25 4 $4,995 $32 0.64%
35 $4,963 3 25 4 $4,995 $32 0.64%
36 $4,963 3 25 4 $4,995 $32' 0.64%
37 $5,764 6 25 10 $5,793 $29 0.50%
38 $4,721 2 25 2 $4,754 $33 0.70%
39 $5,764 6 25 10 $5,793 $29 0.50%
40 $5,764 6 25 10 $5,793 $29 0.50%
41 $5,764 6 25 10 $5,793 $29 0.50%
42 $5,764 6 25 10 $5,793 $29 0.50%
43 $5,764 6 25 10 $5,793 $29 0.50%
44 $5,764 6 25 10 $5,793 $29' 0.50%
45 $5,764 6 25 10 $5,793 $29 0.50%
46 $5,764 6 25 10 $5,793 $29 0.50%
47 $5,764 6 25 10 $5,793 $29 0.50%
48 $4,963 3 25 1 4 $4,995 $32' 0.64%
Page 1 of 2
ATTACHMENT D:
EMPLOYEE WAGE IMPACT
Current Current Proposed Proposed Proposed Monthly Monthly
Monthly Salary Salary Salary Monthly Wage Percent
Salary Step Range Step Salary Increase Increase
49 $5,764 6 25 10 $5,793 $29 0.50%
50 $5,216 4 25 6 $5,248 $32 0.61%
51 $4,721 2 25 2 $4,754 $33 0.70%
52 $5,538 6 25 9 $5,651 $113 2.04%
o
53 $6,058 6 28 9 $6,079 $21 0.35/o
54 $5,216 2 29 2 $5,242 $26 0.50%
55 $5,764 4 29 6 $5,786 $22 0.38%
56 $5,483 3 31 2 $5,504 $21 0.38%
57 $5,429 2.5 31 2 $5,504 $75 1.38%
58 $6,058 6 31 6 $6,075 $17 0.28%
59 $7,697 6 33 10 $7,041 $0 Y-rated
60 $7,396 6 33 10 $7,041 $0'Y-rated
61 $6,368 6 33 6 $6,379 $11 0.17%
62 $5,483 3 33 1 $5,638 $155 2.83%
63 $6,694 6 35 6 $6,698 $4 0.06%
64 $6,694 6 35 6 $6,698 $4 0.06%
65 $6,694 6 35 6 $6,698 $4 0.06%
66 $6,694 6 35 6 $6,698 $4 0.06%
67 $6,694 6 35 6 $6,698 $4' 0.06%
68 $6,694 6 35 6 $6,698 $4 0.06%
69 $6,694 6 35 6 $6,698 $4 0.06%
70 $6,694 6 35 6 $6,698 $4 0.06%
71 $7,322 5.5 35 10 $7,393 $71 0.97%
72 $6,368 6 36 3 $6,371 $3 0.05%
73 $6,058 6 36 1 $6,064 $6 0.10%
74 $5,594 3 36 1 $6,064 $470 8.40%
75 $7,250 2 39 6 $7,384 $134 1.85%
76 $7,396 6 39 7 $7,569 $173 2.34%
77 $7,396 6 39 7 $7,569 $173 2.34%
78 $7,036 5 39 5 $7,204 $168 2.39%
79 $7,036 5 40 4 $7,200 $164 2.33°l0
80 $7,250 2 41 4 $7,380 $130 1.79%
81 $8,172 6 41 9 $8,350 $178 2.18%
82 $8,172 6 43 7 $8,345 $173 2.12%
83 $8,172 6 43 7 $8,345 $173, 2.12%
84 $8,050 5.75 43 6 $8,141 $91 1.13%
85 $8,504 6 44 7 $8,548 $44 0.52%
86 $8,590 6 44 8 $8,762 $172 2.00%
87 $8,172 5 44 6 $8,340 $168 2.06%
88 $10,177 6 48 10 $10,149 $0 Y-rated
89 $10,177 6 48 10 $10,149 $0 Y-rated
90 $9,491 6 50 6 $9,654 $163 1.72%
91 $9,211 4 52 3 $9,413 $202 2.19%
92 $9,976 5.6 52 6 $10,137 $161 1.61%
93 $10,177 6 52 1 7 $10,391 $214 2.10%
Page 2 of 2
Ii
a
I
I
Comments Received by the District Clerk
from District Staff
l
�t
Micen Space D Regional Memorandum
" Open Spare District
I
i
i
To: Board of Directors
From: Casey Cleve, GIS Coordinator, and Ana Ruiz, Planning Manager
Date: 2011 August 19
Re: Revised Planning Technician Job Description
Definition of the Problem:
In December 2010, Koff&Associates released a draft job description for the Planning Technician
position showing proposed updates and revisions to bring the job description, which was last revised in
2000, up to date.The December 2010 draft job description accurately detailed the current responsibilities
and job duties that are expected from and routinely delivered by Planning Technicians. Both the Planning
Manager and department staff agreed with the December 2010 draft job description.
In June 2011,Koff&Associates released a second revised draft job description for the Planning
Technician position that removed essential job functions and education requirements that were included
in the December 2010 version,and many of which were also included in the original 2000 job description
by which current and fonner Planning Technicians have been hired,working,and evaluated against for
more than ten years. The latest June 2011 version significantly lowered the essential core duties as
compared to both the December 2010 and 2000 versions and lowered the education requirement as
compared to the December 2010 version. If the June 2011 draft job description is adopted,the District
will create an immediately negative effect on the Planning Department's ability to complete Key Action
Plan projects by reducing current staff project management capacity(i.e.the District would no longer be
able to allow Planning Technicians to manage small scale planning projects)and by reducing the level of
GIS services that are provided(i.e.the District would no longer be able to delegate GIS analysis to
Planning Technicians).
Recommendation:
1) The Planning Department recommends adopting the Koff and Associates December 2010 draft
Planning Technician job description and eliminating the June 2011 version.
2) Change the class level in the December 2010 job description back to entry level.
Fiscal Impact:
Implementing this recommendation has no fiscal impact to the District. The December 2010 job
description and its associated compensation plan did not propose an increase to the current Planning
Technician's salary.
Reasons why the June 2011 draft job description is not in the District's best interest:
1) Reduces the Number of Key Action Plan Projects Completed by the Planning Department
For over ten years, Planning Technicians have been hired(per the 2000 job description)with the
intent and expectation of taking on project management responsibility for small-scale planning
projects for up to 50%of their time. This work includes project scoping, preparing draft Board
i
I
I
reports,conducting site visits,communicating with field staff,preparing site plans and
construction drawings; monitoring contractors and vendors and monitoring budget expenditures.
These essential responsibilities have been removed in the proposed June 2011 job description.
The impact of removing these job responsibilities from the Planning Technician positions are as
follows:
The Planning Department would no longer have the staff capacity to carry the current
load of Key Action Plan Projects. For this fiscal year,a number of Key Projects that are
now managed by Planning Technicians would need to be removed from the Action Plan.
These include:
• Beatty Property Demolition(asbestos testing,CEQA analysis and
permitting)
• Hydrological investigations for the South Area Field Office Site.
• Schilling Lake Dam Tree Removal Project
To help explain the magnitude of this potential change,below is a list of Key Projects that
have been completed by Planning Technicians that would otherwise have not been possible
without impacting other Key Projects had this position not been allowed to manage small scale
projects:
• Beatty Property and Bear Creek Stables Archaeological Surveys and
Historical Assessments
• Fletcher Demolition
• Regnart Road Culvert Replacement
• Red Barn Remodel
2) Downgrades GIS Services Offered will be Reduced at the District
The draft June 2011 job description proposes to eliminate the Planning Technicians' current role
to assist with the development and implementation of systems and equipment related to the
District's Geographic information System (GIS)and conduct GIS Analysis. Over the last 10
years,the District's GIS services have greatly increased since the current 2000 job description
was developed. Prior to 2000,GIS at the District was largely completed by a GIS consulting
company; now all GIS work is completed internally. In addition,the GIS program now provides a
sophisticated level of GIS products including GIS analysis, 3-D rendering and flyovers, and GIS
web work. Continued Planning Technician participation in this level of work is essential to
meeting District needs.
3) Removes Planning Technicians from the Professional Planning Series Track.
The June 2011 draft job description also changed the jab level for the Planning Technician by no
longer making it part of the professional planning series,and instead separating this position from
the planning series as a para-professional position. This change eliminates the District's ability to
train its own staff to eventually be equipped to manage more complex planning projects and
promote into other positions with greater responsibilities. In fact, historically many of the
District's Planning Technicians were successfully trained and provided the project management
experience necessary to promote to Open Space Planners,and in the last 10 years,two Planning
Managers actually began at the District as Planning Technicians.
of 3
4) Lowers Required Education Level that is Needed to Meet District Demands
The June 2011 draft job description reduced the education requirement from that proposed in the
December 2010 version to require an Associate's degree versus a Bachelor's degree.Currently
Planning Technicians are only required to have an Associate's Degree. Over the last 10 years,
however,given the advances in GIS,the increasing demand for complex GIS analysis and
graphic media,and the increasing complexity of small scale projects(i.e.increased regulatory
scrutiny, increased permitting constraints, increased public involvement,etc),the Planning
Department has recognized that the problem-solving,communication,and technical skills gained
by a four-year degree are now needed for the Planning Technician position. In fact,all Planning
Technicians in the last 10 years have started with a minimum of a Bachelor's degree.Moreover,
the GIS Intern position,which is one level lower than Planning Technician,requires a four-year
degree.
As part of the Class and Compensation Study, a number of positions at the District received this
recommendation of an increased education requirement.The December 2010 job description
including this recommendation is consistent with the results of other positions and the overall
goal of the Class and Compensation Study.
5) Weakens the Quality of Future Applicant Pools
The June 2011 draft job description is likely to reduce the quality of future applicant pools during
recruitment since the responsibilities and requirements are significantly downgraded.
Recommended Amendment to the December 2010
level. Since 2000,Planning Technicians have been hired as an entry level class position. These
staff did not/have not received nor been required to attend training sessions when hired,as is
warranted for a trainee level class. There are no other trainee level positions at the District.Even
the Seasonal Open Space Technicians are entry level.
In addition, of the 19 positions that the Planning Technicians were compared to by Koff&
Associates, 16 of them were entry-level or higher.Two of the comparables(both Open Space
Planner 1)were entry/trainee level,and only one comparable(Open Space Planner I)was trainee
level.
The entry level class rather than the trainee level class is more consistent with the level of
experience and knowledge that is expected from a newly hired Planning Technician.
Thank you for your time and consideration.
4*
Casey Clevg,'GIS Coordinator Ana Ruiz,Planning Manager
Page 3 of 3
nsu
egio
O i epn Space iDistriict nal Memorandum
• ' p p
� a
To: MROSD Board of Directors
From: Planning Department
Copy: Ana Ruiz, Planning Manager
Date: 2011 August 19
Re: Planning Series Titles for Classification and Compensation Study
Purpose:
To align District Planning series titles to be consistent with industry standards.
Recommendation:
The Planning Department recommends that the Board of Directors re-title the existing
Planningseries titles to match the titles of the comparable positions chosen by Koff Et
p
Associates in the Classification and Compensation Study in order to establish consistency with
equivalent job duties, positions, and titles found at outside agencies.
District management:
Benefits to g
1) Establish more accurate industry standard title descriptions for planning positions
2) Attract more appropriate labor market applicant pool for District job vacancies
3) Establish more streamlined collaboration with outside agencies due to more clearly
defined peer-to-peer working relationships
4) Increase public clarity regarding salaries for position levels
5) Improve employee morale at very limited cost to District
Fiscal Impact:
Limited staff time to revise titles because this is a request for re-titling, not re-classification.
This change would not affect the proposed adjustments regarding duties and salaries as
recommended by Koff Et Associates.
Discussion:
For the purposes of this letter, the Planning Technician job description refers to Koff Et
Associates' December 2010 draft job description (for specific concerns related to the Planning
Technician job description, see comment letter titled "Revised Planning Technician Job
Description" dated 2011 August 19 and submitted to the Board separately). Board Report R-
11-67 states that Koff Et Associates is recommending title changes for other positions in the
District in order to match industry standards. The recommendations below carry the same
intent of these other title changes.
Based on duties and responsibilities, Koff Et Associates determined that the Planning positions
and salaries should be compared to positions that have titles and salaries one level higher at
other agencies (see Attachment A for a list of comparable positions and Attachment B for
industry standard titles). For example, Planner Its were primarily compared to Planner Ills
and Senior Planners, and Planning Technicians were primarily compared to Planner Is/GIS
Technicians. Although the Planning series' salaries were adjusted to these higher levels
based on comparable duties, the Planning series' titles were not. Planning staff expressed
their concerns in writing that their title series is misaligned with the title series in the
industry based on job responsibilities and duties. Adjusting the titles to the higher level would
align with industry standards (see Attachment B for a link to industry standard job
descriptions established by the American Planning Association).
The response to these concerns about re-titling was that the Planning Technician position was
benchmarked to a technician position and thus, the entire Planning series was anchored on
that position and could not be adjusted one level upward.
The Planning Technician position was bracketed primarily between two types of entry level
job descriptions, Planner I and GIS Technician. Note that Planning Technicians, considered
trainee level, if they exist at other agencies, were not used for comparables because they did
not have equivalent job responsibilities and duties. The District's Planning Technician
position requires knowledge of multiple fields and is unique because of the combination of
two distinct skill sets of what would be two different career tracks at another agency. There
are no equivalent comparable positions like the District's Planning Technician position which
combine planning and GIS skill sets. Therefore, in order for Koff Et Associates to accurately
benchmark the job responsibilities, the two different types of jobs were analyzed and
combined together for 8 out of 11 comparable government agencies. In total, 19 jobs were
used to benchmark the Planning Technicians' salary range.
Instead of holding back the entire Planning series because of the GIS Technician comparable,
it is the department's opinion that the Planning Technician position would be more correctly
titled if it were re-titled Planner I/GIS Technician. This title more accurately reflects the
comparable duties and responsibilities of the whole position, namely that of a Planner I and
that of a GIS Technician.
This memo proposes the following title changes for the Planning series. These
recommendations are based on the titles of comparable positions selected by Koff &
Associates. This issue has been discussed with the Planning Manager who agrees the Planning
series should reflect industry standards. If the District deems that re-titling is not
appropriate, Planning staff respectfully requests clarification for the reasons behind this
decision.
Current District Planning Comparable Position Title Position Title Proposed by position Title Proposed by
Position Title from Outside Agencies Koff&Associates Planning Dept
Chosen by Koff&Associates
Planning Technician Planner I/Various GIS Titles Planning Technician Planner I/GIS Technician
Open Space Planner I N/A Planner I Planner II
Open Space Planner II Planner III Planner II Planner III
Senior Planner N/A Senior Planner Principal Planner
Senior Resource Planner
Miegio
Opendp eninsula Space DistrictR nal Memo ran d u m
ri
Thank you for your time and consideration.
Kffk Lenin & 26nior'Resource Planner Meredith Manning, Senior Planner�
Matt Baldzikowski, OS Planner 11 Lisa—Bankosh, OS Planner 11
AinaCoony, OS Planner 1[-"' Tina Hugg, OS436nner 11
Casey Cleve, GIS Coordinator Aie Andersen, source Planner I
Grefchen Laustsen, OS Planner Zachary Alexander, Planning Technician
Gatti Basson, Planning Technician
I
Attachment A: Comparable Position Titles for District Planning Series
The following is a list of all comparable positions, based on job duties and responsibilities,
which the consultant chose as equivalent to the District Planning positions which were
benchmarked in the Compensation Study.
Current District Comparable Position Titles Chosen by Position Title Proposed
Planning Consultant by Planning Dept
Position Title
Planning Information Technology Analyst / Planner I Planner I
Technician Assistant Planner / Engineering Technician I
Systems Coordinator - Technician / Assistant
Planner
OSD Conservation Analyst / OSD Assistant
Planner
Resource Specialist - GIS / Assistant Open
Space Planner
GIS Technician I / Planner I*
GIS Technician I / Planner I*
Resource Monitoring Technician
Geographic Information Systems Technician
GIS Specialist II / Environmental Planner I
GIS Project Manager I
Open Space N/A Planner II
Planner I
Environmental Planner Planner III
Open Space Plan e
P P
Planner II P
Senior Planner / Landscape Architect/Park
Planner
Senior Environmental Planner
Stewardship Coordinator
Senior Planner*
Senior Planner*
Planner III
Senior Planner N/A Principal Planner
Senior Resource
Planner
*Duplicate titles indicate that the same title was used by more than one agency.
i
Attachment B: American Planning Association Standard Position Title Series
Source: American Planning Association Website
(URL: http://www.plannin$.org/ontheiob/descriptions/ - Accessed June 15, 2011)
Planning Intern
Planning Technician
Planner I / Assistant Planner
Planner II / Associate Planner
Planner III / Senior Planner
Planner IV / Principal Planner / Planning Manager
Planning Director
�
Michelle Radcliffe
From: Renee Fitzsimons
Sent: Friday, August 1S. 2O11 5:49PyW
To: Michelle Radcliffe
Subject: Class and Compensation Board Feedback-R. Fitzsimons
Subject: Classification and Compensation Study Feedback and Next Steps
To: District Board of Directors
From: Ren6e Fitzsimons, Docent Programs Coordinator
Thank you for your interest in feedback from District staff about the Classification and Compensation Study undertaken
by the District in 20110. | respectfully submit the following for your review and consideration.
� I have provided input to the District's consultant at each of the opportunities that review and feedback was solicited. I
�
thinkthataUoftheinforrnationprovidedbyrn�'asweU input fronnother xtaffthat has been provided during the
�
� process, should be made available to the Board if they wish to review it to get a sense of the overall nature of the input
� provided by staff. I do not choose at this juncture to re-provide all my prior comments and input.
�
My comments here reflect what I do and don't think have been adequately addressed about the position I hold, and also
a few more general comments about the process overall.
Position Specific:
° The District has chosen to have two exempt positions,the Docent Programs Coordinator and the Volunteer
� Programs Coordinator(current dt|ex/dassifications\'with complex responsibilities, be the lowest paid salaried
� '
� positions that the District has.The salaries have not kept pace with new hires in the Public Affairs Department �
� (or agency wide)—positions with less responsibilities and less professionally experienced individuals. This is
� devaluing personally and professionally tome.
�
These above two positions do not benchmark accurately with any other comparators because of the complexity
� and variety of knowledge, skills,and abilities required by each position. I repeatedly asked the consultant to
�
understand and evaluate the Docent Programs Coordinator position as hybrid position for the purposes of
� finding a comparable position—to analyze two positions,e.g. Interpretive Program Manager and Docent
� Program Manager.
The Docent Programs Coordinator oversees over 100 docents/non-paid employee-like volunteers and is not
considered a supervisory position;the Public Affairs Specialist position is non-exempt(has been for years) and
occasionally supervises interns, and is considered a supervisory position.
� • When the Community Program Supervisor position was eliminated numerous duties and responsibilities came
under my position (Docent Programs Coordinator) including: Management of the Daniels Nature Center
�
� (facilities oversight and coordination of maintenance needs; refurbishment of exhibitry and displays; overseeing
� work of contractors, etc.); Oversight and management of consultant contracts for interpretive projects,and
� overall project management. | would be happy to provide a highlighted full comparison of the position
� descriptions to fully detail all of transferred, but not acknowledged or compensated, duties and responsibilities.
� * A re-classification of the Volunteer Programs Coordinator position was requested several years ago by the
person in the position following the elimination of Community Program Supervisor;the analysis was denied and
� was not considered for consultant evaluation. Had that re-classification been considered, I would have
� requested the same analysis for the Docent Programs Coordinator position. Both positions took on significant
� other roles without out of class pay ur additional salary increases.
�
� w The Community Programs Assistant position holder asked to be reclassified and was considered,and evaluated
� by an outside consultant.This is a position that in prior year consultant analysis was recommended to report to
� the Docent Programs Coordinator.The reclassification was granted.This position is now considered a
1 �
�
�
"coordinator" position.That said the Docent Programs Coordinator and the Volunteer Programs Coordinator are
devalued in my opinion—these two positions are professional positions not administrative
General Comments:
• Salaries should keep pace with the general market and employees should not top out of their salaries after
3-5years. COLAs are not a replacement. Merit opportunities aside,the District's salaries are not keeping
pace nor are current employees made whole when new hires are brought in.
• 1 have heard many comments through the process of this study that the consultant would respond and reply
to staff input and comments and this was done inconsistently; I know many fellow employees do not feel
heard.
• 1 appreciate that changes in title and salary are being recommended for the Docent Programs Coordinator
position. I do request that the professional nature and status of the position be brought into line with other
professional classifications at the District. It is not just a "volunteer coordinator"job.
Thank you for your time and consideration,
Renee
Renee Fitzsimons
Docent Programs Coordinator
rfitzsimons(?Dopenspace.org
Midpeninsula Regional Open Space District
330 Distel Circle, Los Altos, CA 94022
P: (650) 691-1200 - F: (650) 691-0485
www.Openspace.Orq twitter: Camrosd
2
Michelle Radcliffe
From: Jaime Villarreal
Sent: Sunday, July 17. 2O11 10:06AM
To: Michelle Radcliffe
Subject: class/comp comments
I am happy that my position's roles and responsibilities were updated but not too happy nor surprised that there was no
comparable index for the Farm Maintenance position. l feel it falls into its own classification because of its unique
| requirements: maintenance 0'rnentoring youth gnoups+educationpnograno+visitorservices /higheatmfa|| pnas�nm�).
� ` '
Also, as a board member pointed out, two important agencies were not included in the comparison study and so these
two should be added asan addendum to the report.
Jaime Villarreal(Radio 9MI4)
Farm Maintenance Worker
Deer Hollow Farm Office:650-903-6430
Foothills Field Office: 6f0-691-2165 �
Cell. 409-460-5518 �
�
�
�
�
�
1 |
IT SPECIALIST RECLASSIFICATION [RENNY HSIEH.. NETWORK SPECIALIST]
Reclassification from Non-Exempt to Exempt
Most IT related jobs in the market are generally classified as an exempt position.
Reclassification from Network Specialist to IT Specialist
The job description appears capture the duties and responsibilities in all with an exception of the
assigned title.
My main concern is that the job has evolved to a point where the Network Specialist is responsible for
the entire District IT infrastructure with minimal supervision and direction.This not only includes the
main District office but three separate offices:Foothills Field Office,Skyline Field Office,and the newly
provisioned South Area Outpost. Other private organizations and many local government agencies I've
found to be similarly sized to ours (e.g. City of El Cerrito)generally assign these tasks to those of an IT
Manager or equivalent under a similarly sized IT Division.
General Comments
From what I understand,Koff&Associates mentioned that these new titles are for internal reference
and the District,at its own option,can use a more appropriate title when recruiting. Believe this will
eventually cause some confusion between all parties involved.
Date: August 19,2011
To: Board of Directors and General Manager,Midpeninsula Regional Open Space District
From: Lynn Tottori,Sr.Management Analyst
RE: Response to the Koff&Associates Classification&Compensation Study,June 2011
i
1. Background: The Senior Management Analyst position's maximum salary has been
recommended to be decreased 7.9%by Koff&Associates,as stated in their Classification&
Compensation Study completed June 2011. This was after only(1)year since the District's
Human Resources Dept.originally benchmarked the position. This was a newly created position
and the salary range determination was very recent,only 1-yr.old,as opposed to other older,
existing positions throughout the District. This salary range was used as basis of District
recruitment and advertisement; it was also the basis for acceptance of the position.
a. Original salary range: $82,763-$106,205/year
b. Recommended maximum salary by Koff&Assoc.: $97,812/year
2. Key factor: agencies used for comparison by the District differed from those used by Koff&
Assoc.
3. Key Factor: Intent of position's roles&responsibilities by District
a. The intent of the District was to create a senior position and be part of a leadership
team providing District fiscal leadership. This position is at the top of the Management
Analyst Series,with a focus on District-wide Budget and Financial Analysis.
My financial role at the District is a very visible position to the Board of Directors,
Committees and Staff. I regularly present financial information and recommendations
that impact the District's administrative, budgetary and financial programs and policies.
Additionally,with a small finance staff and a part-time Controller,this position requires
a high level of independence and initiative.
Very significant,is that this is one of the most challenging economic times as the current
recession continues well into its 3.d year. The District is fiscally challenged to continue
its mission with high quality. Tough decisions will require more analysis,substantiating
this position at its original range.
4. 1 was interviewed for the Classification and Compensation study only(3)months after hiring
into the District. Since then, my position has evolved with an expectation to grow as the District
faces new challenges and launches new projects that address these challenges:
� .
Key Factor: My Position has expanded beyond the original job description with the following
Key projects and intent of the District:
o. Nominated for and accepted into the 2011 Santa Clara County Leadership Academy
b. Long-Term Strategic Plan member and on implementation team on Key projects
�
District objectives.
|
c. !AFD(Integrated Accounting&Finance Software Project): �
I. Core Team member—|mplementadon&Deployment toStaff.
H. Subject Matter Expert—Budgeting,Accounting&Financial,Reporting.
�
� iii Requires: Business Process Re-engineering Qarchitectureof the District's .
reporting General ledger and Reporting needs for District-wide and Depts.
d. Future Funding needs: Funding Measure project,refinance of bonds,Guideline change
needs,etc.
'that will require financial analysis.
e. Growth into Revenue forecasting and assistance tu the District Controller.
5. Organization and setting the annual budgetary calendar—Determining deadlines and
deliverables,while defining scope of requirements. Coordination and communications with all
departments for budget&project activities.
a. Provide change management leadership needed at the District—by initiating and
executing key projects: Change required in adopting the new|ASF(Integrated
Accounting&Finance Software system)—that will require changing the business
� process ranging from: Approving Purchase Orders,entering hours worked byeach
� employee,ded�nnewreportinQ,etc |
� ' . �
b. Initiates&executes Districts new policies&Guidelines
8. Key Factor: resulting in
an incomplete study of this position. Range stated @15%above Management Analyst'This
� 15%percent appears discretonan/arbitrary. (Other nun-benchmarked positions,Koff&Asswc.
� applied 96 ranging between 1O~4O%above/below adjacent positions.
� 7. Key Factor: �
� prog[essions. The District has a skeleton staff compared with other agencies,resulting in staff
� taking nn more roles Q responsibilities asneeded. Koff&Assoc.salary ranges for the following
pos}don»also appear lower than market:
�
�
� |
'
�
a. For example: the District Controller was originally rated as a Financial Analyst following
their initial interview with the Controller,and then later adjusted. Koff&Assoc.
recommended salary range for the District Controller(the most senior finance
professional on staff,a Board Appointee, reporting directly to the General Manager) is
/ 17%below the Department Manager's salary range.
�
! i District Controller(Koff&Assoc. recommended Top 5tap): $I11,63G'Top step
i
b17%below Dept. Manager's Top Step
ii Department Manager(Koff&^Assoc. recommended Top Step): $134,508 �
b. |e: The Accounting Technician top step was also recommended tm6e decreased by
9.53%by Koff&Associates; however,is being recommended by management to retain
the current,higher salary range.
8 Key�octor Ko�&6�ssoc necocnmendationis |imitedtoxa|aryrmn�eandnottmt�(bene�t �
. . �
package. Koff&Assoc.did state that the District's benefit package was lower than competitors. �
A. Request to review Koff&Assoc.study and re-evaluate their recommendation on this Senior �
Management Analyst position. There also exists the ability to Grandfather-In the current salary �
range for this position,as originally posted.
�
�
�
�
�
�
Classification and Compensation Study Board Feedback
August 18,2011
Dear MROSD Board:
I have thought heavily about the outcome of the Classification and Compensation(C&C)Study, its
objectives,and methodology prior to writing this letter to the Board.After much consideration, I have
decided to present to the Board my unique situation. My objective is to provide the Board with a different
perspective and additional information so the Board can have a well-informed study session on September
13,201 1.
Summary
The C&C Study recommendation of Senior Administrative Assistant classification was a snapshot in
time. That classification no longer reflects my job responsibilities. My responsibilities have permanently
changed. It is recommended that the Board authorize Administration to assess my job responsibilities,
evaluate whether a re-classification is necessary,and make a proposal.
Situation
I am one of the employees who received a suggested re-classification that, in my opinion,does not truly
reflect the work that I have done,am currently doing,and will be doing. Due to changes in the
Administration department the past three years, I have done assignments that span multiple areas beyond
my current classification of Deputy District Clerk/Office Manager.The C&C Study captured my job
functions at a snapshot in time, when Administration was in transition with the recent hires of Michelle
Radcliffe as a full-time District Clerk, Lynn Tottori as the new Senior Management Analyst,and the
absence of a department head. At that point in time, I had not fully launched into special projects assigned
to me and was in an administrative support role. Since then, I have done duties and assignments that not
only fall outside the responsibilities of my current classification as Deputy District Clerk/Office Manager,
but also beyond the classification of Senior Administrative Assistant recommended in the report by Koff
and Associates.
Work History
During the last two and a half years of my employment, I have been supervised by a total of six different
supervisors. During the transition between department heads,resources were thin and I was chosen to
assist where needed.As a result, my responsibilities changed significantly. My work consisted of a
variety of responsibilities in multiple areas that include but were not limited to the following from
February 2009 to April 2010:
• Annual budget and action plan co-ordination for the Administration department and the District;
• Research and analysis for complex and specialized projects,e.g. Request for Proposal(RFP)for
the Integrated Accounting and Financial System(IAFS)software vendor and consulting services,
evaluation and selection of the IAFS,and Capital Improvement Project(GIP)cost accounting best
practices;
• Formulation and implementation of departmental and Board processes,e.g. Administrative
policies,Public Record Act requests,and public communications to the Board;
• Cost analysis for Brown Act agenda reimbursement and Board election costs;
• Budget oversight,control and variance analyses for Administration Services and Supplies,where
I presented FY09-10 Year-end review of budget and action plan to the full Board and
Administration FYI 0-11 Midyear Action Plan to the Administration and Budget Committee;
• Business process study and workflow recommendation for Board Packet Agenda and Standing
Committee process;
• Acting District Clerk duties for one full year(approximately mid-2009 to mid-2010);
1 of 3
i
Classification and Compensation Study Board Feedback
• Prepared and presented several board reports,e.g. appointment of auditor for CY2009 and
CY2010,election of Board officers for CY2010;appointment of Standing and Ad Hoc Board
Committees and Financing Authority;and
• Procurement of office equipment and contractual services for AO,e.g. lease and maintenance
services for multi-functional printers for AO and Field Offices.
At its meeting on November 18,2009(see report R-09-121),the Board approved a new position,a full-
time District Clerk, in the Administration Department.Creation of this dedicated position was to ensure
continuity in addressing all Board-related matters and relieve the Administration/HR Manager, later re-
titled to Administrative Services Manager,of functions not typically included in the duties of that
classification.On April 14,2010,the General Manager appointed Michelle Radcliffe as the District Clerk.
4 2010 and
Annetta Spiegel was the Actin Administrative/HR Manager at that time. Between April 14,
g S P
November 10,2010(the date the C&C Study concluded), I was in an administrative support role.
On November 29,2010, Kate Drayson was hired as the Administrative Services Manager,and
Administration underwent yet another transition where my role changed again.At the same time,the
next phase of evaluation. From November 2010 until today,m
IAFS project progressed to the Y Y
P J P g P
assignments were as follows:
• IAFS Project:
o Bidding and selection process for IAFS Consultant Services;
o Educational workshop for analysts and managers re: IAFS project scope and terminology
o Educational webinar demo for analysts on New World Systems' (NWS)business
analytics software;
o Coordination and oversight of Phase I and II of consultant work: needs assessment and
r cards software
' n scripts and score workshops,development of demonstration s ,
functional
> P
P P
demonstration score analysis, in-house payroll integration research, initial data migration
and chart of accounts assessment;>
o Coordination,analysis,and oversight of Phase 11 of IAFS software vendor evaluation
process: evaluation of 3 short-listed vendors, software demonstration facilitation,site j
visit at the City of El Cerrito,NWS reference checks,contract and cost negotiation,work
plan creation and review,resource estimation,cash flow projection;and
o Overall risk and performance management.
• Other Special Projects:
o Cost analysis for AO copying/printing;
o Wells Fargo banking cost reduction analysis;
o Hawthorns Trust investment vehicle research;
o Contract negotiation for AO multi-functional printer lease and maintenance services;and
o Member of T3 Committee and Strategic Plan Working Group.
In February 2011, I was re-assigned to be the IAFS Project Manager,and I've been performing in that
role since,carrying the project forward to its current state.
C&C Study
Between June and November 2010,the C&C Study was conducted and concluded,and on November 10,
2010,Koff and Associates delivered their proposed position classification to all District employees. I
received a proposed Senior Administrative Assistant classification,to which I responded back
immediately to Katie Kaneko, President of Koff and Associates. I presented to Ms. Kaneko my unique
situation and concerns with supporting documentation.Those documents were:
• a detailed list of duties and responsibilities with my qualifications;
• several sample job descriptions of the Administrative/Management Analysts series,where
similar duties and responsibilities were noted as essential job functions of that classification;and
I
2 of 3
Classification and Compensation Study Board Feedback
• my request for Koff and Associates to re-evaluate its recommendation given the information.
Shortly after, Ms. Kaneko replied back acknowledging my unique situation and the difficulty she had
when assessing my classification. However, Ms. Kaneko's response still fixated on the Senior
Administrative Assistant classification offering no alternative recommendation. Koff and Associates then
proceeded to conduct a compensation study for me using that classification.
Current Situation
1 currently have a working title of Project Manager, which was approved by the General Manager,
Administrative Services Manager and Human Resources Supervisor, in recognition that my current
classification of Deputy District Clerk/Office Manager no longer applies to what I do. I am receiving
actin a to appropriately com ensate me for the work I am performing. Going forward,my assignments
acting P
will be as follows:
• IAFS Project:
o Project management for the implementation and training through January 2013;and
o Evaluate District's business process and recommend improvements,as necessary.
• Research banking services to optimize cost-effectiveness for the District;
es g
• Provide support for Board-related matters and act as District Clerk in her absence;
• Oversight of spending and control for all Administration Services and Supplies accounts,a
$435,000 budget;
• In support of the Administrative Services Manager,co-ordinate fiscal year budget and action plan
development and review for Administration and the District;and
• Ad hoc research and analysis on an as-needed basis.
I hope this new information gives the Board a better picture for the September 13`"study session. Please
p g P P
let me know if the Board has any questions and/or needs additional information. District employees,
including myself, look forward to the Board's fair and equitable assessment.
Respectfully Yours,
Anna Duong
IAFS Project Manager
3 of 3
I
;fie,
Midpeninsula Regional
Open Space District
i
To: Board of Directors
From: Stephen E. Abbors
Date: September 9, 2011
Re: FYIs
Midp eninsula Regional
Openp
n Space District Memorandum
r ,
DATE: September 14, 2011
I
MEMO TO: MROSD Board of Directors
THROUGH: Stephen E. Abbors,General Manager
FROM: Tina Hugg �YJf
SUBJECT: Cooley Landing—Project update, schedule,and Preserve closure for construction
Since the July 27th Board meeting when the Board approved Amendments to the Comprehensive Use&
Management Plan for Ravenswood Open Space Preserve(Preserve)and the Partnership Agreement, the
City of East Palo Alto(City)has made significant progress towards preparing the Cooley Landing project
for construction.
I
I
In addition to completing the construction drawing set for Phase 1 of the project,the City has obtained the
1
401 Water Quality Certification from the Regional Water Quality Control Board, approval from the Cities
of Menlo Park and East Palo Alto for the Joint Permitting Agreement, and U.S. Fish and Wildlife
Service's Biological Opinion.
inion. The City anticipates receiving the Bay Conservation Development
Commission's permit at the commission's meeting on September 15th with the Army Corps of
Engineers' permit following two weeks later. Final permits are thus expected by late September, and the
City expects to release a Request for Bids in early September. Award of contract is contingent upon
receiving these permits.
Bidding, award of contract,and mobilization at the site are anticipated to be complete by mid-October.
As indicated at the July 27th Board meeting, because the City anticipates significant construction traffic
due to the import of soil for the soil cap, Ravenswood Open Space Preserve will be closed for the duration
I
of Phase I construction(approximately eleven months). The Preserve is expected to reopen along with
the new Cooley Landing area in August 2012.
District staff will notify neighbors, including the City of Palo Alto, via mail of the anticipated closure and
upcoming construction. We will also notify our docents and volunteers via email and a notice will be
placed on our website for the general public. We will also notify the Association of Bay Area
Governments,who oversees the regional Bay Trail, of this closure.
Mideninsua
Opepn Space lDist get naI Memorandum
DATE: September 14,2011
MEMO TO: Board of Directors
THROUGH: Steve Abhors, General Manager
FROM: Gretchen Laustsen
SUBJECT: Parking issues along Purisima Creek Road
r r
lot a popular staging area for local residents to access the 3 360 acre
The lower Purisima Creekparking
D P P g €�
Purisima Creek Redwoods Open Space Preserve, provides seven parking spaces. The limited parking can
fill up quickly on high use days(weekends, holidays), resulting in visitors parking along the edges of
Purisima Creek Road,just outside the Preserve boundary. Preserve neighbors and other members of the
public have raised concerns about the roadside parking and requested that the District provide additional
parking in the Preserve. In response to these comments and concerns, District staff explored the potential
for expanding the existing parking lot that is located adjacent to Purisima Creek.
In October 2009, District staff met onsite with San Mateo County planners to evaluate the site and discuss
the planning and permitting issues and constraints that affect the area, which lies within the Coastal Zone.
Development within the Coastal Zone is closely guided by the local government's(in this case, San
Mateo County)Local Coastal Program(LCP), in partnership with the California Coastal Commission.
The San Mateo County LCP allows for only very specific, resource-based uses within or near riparian
habitat. Riparian"corridors"are defined by the limit of riparian vegetation. Corridors must contain at
least 50%cover of riparian vegetation (i.e. red alder,jaumea, pickleweed, big leaf maple, narrow-leaf
cattail,arroyo willow, broadleaf cattail, horsetail,creek dogwood, black cottonwood, and box elder).
Within riparian corridors,trails and scenic overlooks are the greatest level of permissible development,
except for bridges, pipelines, and emergency infrastructure. Within the riparian"buffer," which is
defined as a zone extending 50 feet from the edge of riparian habitat, permissible development is slightly
broader,and can include parking in certain zones if no other feasible alternative exists. As such County
planners expressed concerns regarding the feasibility of obtaining permitting clearance to expand the
parking lot given its location with the coastal zone, its close proximity to the creek,and the high
likelihood that part if not all of the site could be defined as part of the riparian corridor,and therefore lie
within the most protective zone and outside of the riparian buffer.
Recognizing the potential permitting constraints, District staff concluded that a discussion with the
Coastal Commission was necessary to understand whether or not the area is considered riparian habitat or
whether it lies within the riparian buffer. To begin these discussions, District staff presented a
preliminary conceptual design for an expanded parking lot to the Coastal Commission in the fall of 2010 j
that minimized impacts to native vegetation to the maximum extent feasible. At this time, Coastal
Commission staff stressed the need for a riparian survey to determine the precise limit of riparian
vegetation and riparian buffer in relation to the conceptual parking lot design. District staff then
! proceeded to contract with Coast Range Biological, who has extensive experience working in the Coastal
!
� Zone,to perform a riparian survey using the guidelines and definitions found in the [[P. The results of
this survey are provided as Exhibit A. According k»the survey, the potential parking expansion would
impact approximately 02 ocncs(0,43O oq.ft.)ofred alder forest which,although it is physically separated
from PuriyiroaC»cek, meets the LCP definition of riparian habitat. An additional O.2acres, including the
entire access driveway(which would likely require at least partial widening), is within the 50-foot
riparian buffer.
The riparian survey results were submitted to the Coastal Commission in Pcbruury20l | ao part ofu pre-
application project revievv. Coastal Commission staff indicated that the permanent conversion ofred
alder riparian forest is not permissible under the LCP, and that this would likely preclude approval ofu
permit application for the project.
� Given the Coastal Commission's concern over the unavoidable impacts to red alder forest, District staff �
� recommends seeking un alternative solution to the existing parking uituation. Because there iacuorndyu �
�
lack ofany feasible expanded parking on District property near the existing lower Purisima Creek �
tnui|hcud, District staff recommends to continue seeking purchasing opportunities in the area that can
provide expanded parking. In the interim, field staff recently improved the available roadside parking.
Working with un adjacent neighbor, District field staff moved the existing fence approximately 5 feet
back toward the creek to allow more room in the road right mf way for parked cars. Also, four additional
no parking signs were installed to |inoh parking on the narrowest section of roadway. These
improvements,while not increasing the amount of available parking, Arcu1|y improve pedestrian and
vehicular safety along PurininmoCreek Road.
�
�
�
|
�
GENERAL MANAGER
Stephen E.Abbors
Regional
OpenSpaee I Midpeninsula Regional Open Space District I BOARD OF DIRECTORS
Pete Siemens
Yoriko Kishimoto
Jed Cyr
Curt Riffle
Nanette Hanko
Larry Hassett
Cechy Harris
I
September 6, 2011
The Honorable Edward G. Brown, Jr.
Governor, State of California
State Capitol Building
Sacramento, CA 95814
RE: AB 703—Request for Signature
Dear Governor Brown:
I am writing to request your signature on AB 703 which will delete the January 1, 2013
sunset date of the property tax welfare exemption for natural resources and open space lands
and leave the exemption in place for another ten years.
The legislature has granted an exemption from this tax for property owned by nonprofit
organizations that is open to the general public and used for the preservation of native plants
or animals, or used solely for recreation and for the enjoyment of scenic beauty.
The most recent extension, which is the fifth extension since the original one expired in 1982,
set the repeal date to January 1, 2013 and AB 703 will extend this date by ten years setting it
to January 1, 2023.
We urge you to sign this important bill because some of the property currently owned by
nonprofit organizations would most likely be transferred to the state if the exemption were
not renewed, putting these important lands at risk because of the increased land management
and administrative costs that would have to be carried during a time when public resources
are tight. In addition, state-owned property is not subject to the property tax, as you know,
which is why the net effect of such an action would be the same as keeping the exemption in
place.
g
�a
3W3y
iS
33o Distel Circle Los Altos,CA 94022 1 11650,691 1200 1 F 6so.69i 048S www openspace.org
Nonprofit organizations hold thousands of acres of open space lands and it is important that
the exemption remains in place so that they can focus their strapped resources on managing
these lands for the public.
Thank you for your consideration of this legislation. Should you have any questions or need
additional information, please do not hesitate to contact me.
Sincerely,
Stephen E. Abbors
General Manager
cc: Ralph Heim, Public Policy Advocates
MROSD Board of Directors
i
i
Vol.14:179-191,2011 ENDANGERED SPECIES RESEA���Pubtished
doi:10.3354/esr00348 Endang Species Resine August 31
Addressing biased occurrence data in predicting
potential Sierra Nevada red fox habitat for survey
prioritization
maw"Ma" John Perrine2'*, Barbara Holzman', Ellen Hines'
'Department of Geography and Human Environmental Studies,San Francisco State University, 1600 Holloway Avenue,
HSS Room 279,San Francisco,California 94132,USA
'Biological Sciences Department,California Polytechnic State University, f Grand Avenue,San Luis Obispo,
California 93407-0401,USA
ABSTRACT:The Sierra Nevada red fox Vulpes vulpes neca tor is listed as a threatened species under
the California Endangered Species Act. It originally occurred throughout California's Cascade and
Sierra Nevada mountain regions. Its current distribution is unknown but should be determined in
order to guide management actions.We used occurrence data from the only known population,in the
Lassen Peak region of northern California,combined with climatic and remotely sensed variables,to
predict the species'potential distribution throughout its historic range. These model predictions can
guide future surveys to locate additional fox populations.Moreover,they allow us to compare the rel-
ative performances of presence-absence (logistic regression) and presence-only(maximum entropy,
or Maxent)modeling approaches using occurrence data with potential false absences and geograph-
ical biases. We also evaluated the recently revised Maxent algorithm that reduces the effect of geo-
graphically biased occurrence data by subsetting background pixels to match biases in the occur-
rence data. Within the Lassen Peak region, all models had good fit to the test data, with high values
for the true skill statistic (76-83%),percent correctly classified (86-92%), and area under the curve
(0.94-0.96),with Maxent models yielding slightly higher values. Outside the Lassen Peak region,the
logistic regression model yielded the highest predictive performance,providing the closest match to
the fox's historic range and also predictinga site where red foxes were subsequently
ubse uen tl detected in I
autumn 2010, Subsettin background q Y
g g and pixels in Maxent reduced but
t did not eliminate the effect that
biased occurrence data geographically had on prediction results relative to the Maxent model using
full background pixels.
KEY WORDS: California • GIS • Logistic regression • Maxent • Species distribution model
Vulpes vulpes necator
Resale or republication not permitted without written consent of the publisher
INTRODUCTION 2006, Rachlow & Svancara 2006, Ferreira de Siqueira
et al. 2009, Menon et al. 2010, Rebelo & Jones 2010).
Determining the composition and distribution of Because a rare species will likely be absent from
suitable habitat is crucial for the successful manage- most sample locations in a simple random sample,
ment of rare or endangered species (Guinan & Zim- targeting survey locations to areas with high proba-
mermann 2000). Species distribution models have bility of species occurrence represents a more effi-
become an important tool to identify optimal survey cient use of limited conservation resources (Guisan et
areas for these species and to increase the probabil- al. 2006).
ity of locating previously unknown populations The Sierra Nevada red fox Vulpes vulpesnecatoris a
(Engler et al. 2004, Edwards et al. 2005, Guisan et al, medium-sized canid that historically occurred at low
'Corresponding author.Email:jperrine@calpoly.edu 0 Inter-Research 2011 •www.int-res.com
180 Endang Species Res 14: 179-191,2011
densities throughout the high elevations of the Sierra Barbosa et al. 2009); presence-only models such as
Nevada and southern Cascade mountain ranges of Maxent have made improvements in their transfer-
California and Oregon, USA (Grinnell et al. 1937, Hall ability (Phillips 2008), with their performance being
1981, Sacks et al. 2010).Within this broad range,Grin- positively correlated with the similarity between the
nell et al. (1937) reported 3 population centers: the occurrence data region and the projection area (Bul-
Mount Shasta/Lassen Peak region in northern Califor- luck et al. 2006).
nia, the central Sierra near Mono Lake and Yosemite Addressing these biases provided an opportunity to
National Park, and the southern Sierra near Mount explicitly compare the performance of 2 different mod-
Whitney(now mostly in the Sequoia and Kings Canyon eling approaches: logistic regression based on P
res-
National Parks). In 1980, due to a noticeable decline in ence-absence data versus maximum entropy based on
numbers, the Sierra Nevada red fox was listed as a presence-only data. Logistic regression has been
state 'threatened' species; the factors causing its widely used in species distribution modeling (Mlade-
decline are unknown (CDFG 1996, 2004). Despite its noff et al. 1999,Nielsen et al.2002,Johnson et al.2004,
former extent, the verified detections since 1991 have Posillfco et al. 2004, Olivier & Wotherspoon 2006).
all been in the Lassen Peak region of northern Califor- Logistic regression uses presence-absence data to
nia (Perrine et al. 2007).A recent conservation assess- model the probability of species occurrence as a func-
ment for this species (Perrine et al. 2010) recom- lion of its predictor variables, which can be continuous
mended that targeted surveys for the Sierra Nevada or categorical (MacKenzie et al. 2006). Its output is
red fox be conducted throughout its historic range to confined to values between 0.0 and 1.0.
determine whether any additional populations exist. Maximum entropy methods,although relatively new
Prior attempts to detect this species as part of multi- and not as widely used as logistic regression, can out-
species carnivore inventories have been unsuccessful, perform logistic regression (Elith et al.2006) and have
even in the Lassen Peak region where the species has successfully identified locations of previously undis-
persisted (Zielinski et al. 2005). Surveys targeting the covered populations (Rebelo & Jones 2010). The most
Sierra Nevada red fox and focusing on areas with high widely used maximum entropy approach, in the pro-
probabilities of the species' occurrence may increase gram Maxent (Phillips et al. 2006), estimates the spe-
detection probability and survey efficiency. cies' probability distribution that is most dispersed
Our goal was to predict the extent and distribution within the constraints of the target population informa-
of suitable habitat for the Sierra Nevada red fox lion.Like logistic regression,Maxent can use both cat-
throughout its historic range, based on the character- egorical and continuous predictor variables, and the
istics of occuP ied habitat in its current known range. output can provide information on the relative contri-
The available data were collected during a compre- button of each predictor variable (Phillips et al. 2006).
hensive ecological study of the Lassen Peak popula- Maxent can utilize presence-only data, sidestepping
lion (Perrine 2005). Unfortunately, these detection the problem of false absences, but geographical and
data contained 2 important biases. First, the survey environmental biases in the occurrence data can intro-
data contained potential 'false absences;' surveys duce considerable error in presence-only models
(Phillips . Presence-only models such as Maxent s 2008
have faile
d to detect this species in areas where popu- ( p )
P
Absence data for draw background pixel values from the entire study
ations were known to be resent. Ab g
1 P
rare species may result from the rarity of the species region, while presence data values are drawn only
rather than its true absence (MacKenzie et al. 2006). from a small portion of the study area (Phillips et al.
False absences are problematic in species distribution 2006). The resulting predictions may therefore under-
modeling because they do not indicate unsuitable represent habitat suitability outside of the occurrence
habitat or confirm that the species does not occur at a data area(Peterson et al.2007, Phillips 2008).Substan-
given site (Guisan & Zimmermann 2000, Engler et al. tiai improvements can be made in Maxent models
2004, MacKenzie et al. 2006). Second, the only avail- derived from biased occurrence data by selecting
able occurrence data for the Sierra Nevada red fox background data with similar biases as the occurrence
were from the Lassen Peak region, which represents data (Phillips 2008). This approach has recently been
only a small portion of the species' historic range. implemented in Maxent (Phillips 2008), but its effects
Researchers have cautioned against using geographi- have had little empirical validation.
cally biased occurrence data or transferring models to Here we compared the performance and output of a
broad unsampied regions (Peterson et al. 2007, Bar- presence-absence logistic regression model versus
bosa et al. 2009);however, such projections have suc- presence-only Maxent models with and without the
cessfully increased detection rates (Guisan et al. transferability improvements. We then combined the
2006). General linear models (GLMs) may transfer output from the logistic regression and updated Max-
quite well to unsampi (
Randin et al. 2006, ent model to generate an ensemble prediction, lever-
ed areas
Cleve et al.:Predicting potential Sierra Nevada red fox habitat 181
aging the strengths of each model while minimizing Peak region (6455 km') includes Lassen Volcanic Na-
their respective weaknesses (Aratijo & New 2007, tional Park (LVNP), the surrounding Lassen National
Stohlgren et al. 2010). In addition, we explored the Forest (LNF), and the immediately adjacent lands of
use of unclassified spectral data as a predictor vari- various ownerships(Fig. 1).This montane area is domi-
able in place of predetermined classification schemes nated by conifers such as Jeffrey pine Pin us jeffreyi and
(e.g. vegetation or canopy cover categories).Although Ponderosa pine P. ponderosa, red fir Abies magnifica
classified maps are commonly used predictor vari- and white fir A. concolor,and mountain hemlock Tsuga
ables, wildlife may respond to continuous environ- mertensiana,along with•wet alpine meadows and talus
i mental gradients that are not captured in the class- slopes. This area has a Mediterranean climate with
ification schemes (Laurent et al. 2005). By using warm dry summers and cold wet winters. Most of the
unclassified spectral data, species' occurrence can be annual precipitation occurs as snow from November
predicted by spectrally detectable components of through April, with snowpacks at the higher elevations
their habitat, rather than predetermined classification often exceeding 3 m in depth and persisting into the
schemes that may inaccurately delineate boundaries summer months.
between cover types and under-represent habitat het- Fox survey data.Sierra Nevada red fox locations were
erogeneity (St-Louis et al. 2006). Using unclassified determined using 4 detection methods: radio teleme-
spectral reflectance in the distribution model may try,scats(feces),and camera surveys using opportunis-
minimize errors in the resulting predictive maps (Lau- tic and stratified random sampling designs. Each
rent et al. 2005). detection method contained biases.
MATERIALS AND METHODS
Study area. Our model prediction area
covers the area within and immediately Study area
surrounding the historic range of the r' ® Area outside of
Sierra Nevada red fox (Fig. 1). This in- study region
cludes the Sierra Nevada and the south- Lassen Peak
region
ernmost extent of the Cascade Range in Approximate
California. The prediction area spans 2 historic range
Major Land Resource Area (MLRA) .+
ecoregions: the Sierra Nevada ecore-
gion and the California portion of the
Eastern Cascade Slopes and Foothills
ecoregion (USDA-NRCS 2006). The
Sierra Nevada ecoregion extends from
just south of Lassen Peak to the
Tehachapi Pass near Bakersfield. The
majority of this ecoregion is comprised
of elevations ranging from 450 to
2750 m, with the highest peak being
Mount Whitney(4419 m).The California Pacific
portion of the Eastern Cascade Slopes Ocean
and Foothills ecoregion represents the
southernmost extent of the Cascade N
Mountain Range, extending from the
Central Cascade Mountains to the Sierra
Nevada. The majority of this ecoregion
is comprised of elevations ranging from
450 to 2500 m, with the highest peak 0 25 50 100 miles
being Mount Shasta 4318 m).
Sierra Nevada red fox survey data
were collected from the area surrounding Fig.1. Vulpes vulpes necator.Model prediction area for the Sierra Nevada red
Lassen Peak (3187 m), the southernmost fox relative to its historic range (Grinnell et al. 1937) and the Lassen Peak
peak in the Cascade Range. The Lassen region.Note that scale bar is in miles(1 mile=ca. 1.6 km)
182 Endang Species Res 14: 179-191,2011
Five Sierra Nevada red foxes were captured and the same location can introduce conflicting information
fitted with VHF collars and tracked by aerial and in the model, which may lower its predictive power. To
ground-based telemetry from 1998 through 2002 (Per- remove pseudoreplicates and correct for conflicting in-
rine 2005). All capture and handling activities were in formation,we used ArcGIS(ESRI)to label a location as a
accordance with California Department of Fish and presence if a fox was detected there at any time during
Game and University of California Berkeley protocols. the 10 yr period. Similarly, we deleted duplicate de
The field team collected 586 independent ground tections at the same location.
telemetry locations using a Trimble GeoExplorer II After pooling these data, the sampling intensity var-
GPS and LOCATE II telemetry software (Nams 2001), ied between habitat types. Since survey data were
and aerial telemetry provided 123 additional locations. plentiful(-2000 records), we randomly subsampled the
In addition, a total of 227 Sierra Nevada red fox scats data from each environmental zone to balance the sam-
were collected opportunistically from June 1998 pling intensity (Guisan &Zimmermann 2000, Araujo&
through December 2002, primarily in association with Guisan 2006).The environmental zones were based on
ground telemetry. The telemetry and scat locations a combination of 10 elevation zones and 19 California
were clustered in the western half of LVNP and the Wildlife Habitat Relationship (CWHR; Mayer & Lau-
adjacent LNF lands (Perrine 2005). denslayer 1988) types. This subsam- - reduced the
on- dataset from 2000 to 1200 points (600 per presence and
The opportunistic camera station survey was c
III PP
ducted between 1992 and 2002 and consisted of 968 absence, respectively) and reduced but did not elimi-
baited TrailMaster (Goodson and Associates) camera nate sample clustering. To reduce clustering bias, we
stations throughout LVNP and the LNF (Perrine 2005). used Thiessen polygons (Rhynsburger 1973) to down-
This and For- weight points that occurred close together.
primarily b Pa
This survey was conducted pn y y g
est Service biologists following the standard protocol Environmental predictor variables.We based our envi-
j for surveying forest carnivores (Zielinski & Kucera ronmental predictor variables on Sierra Nevada red fox
i
1995). Although the cameras were widely distributed ecology and the availability of digital data. The Sierra
throughout the region, sampling biases arose due to Nevada red fox is associated with high-elevation
the opportunistic nature of this survey. For example, conifer forests, subalpine woodlands, talus slopes, and
the southwest portion of LVNP and the LNF east of the barren areas above treeline (Grinnell et al. 1937,
Caribou Wilderness were heavily sampled, whereas Schempf & White 1977, Perrine 2005). To represent
the northern portion of the region had the least sam- these environmental conditions, we used a variety of
pling effort. Samples were also biased toward roads. GIS layers containing vegetation, climate, hydrology,
This survey yielded 50 Sierra Nevada red fox detection and forest structure data.Specifically,to represent veg-
locations,with multiple detections at some locations. etation and forest stand structure,we used CalFire Fire
The stratified random camera survey was conducted Resource and Assessment Program's Multi-Source
in the summers of 2001 and 2002.This survey consisted Land Cover Data(MSLC),which contained CWHR veg-
of 24 sites stratified by elevation and randomly placed etation type, total tree canopy closure, tree size class,
throughout the Lassen Peak region. Each site con- and tree density class attributes (www.frap.cdf.ca.gov).
tained 2 baited TrailMaster cameras approximately We also derived Tasseled-cap greenness and wetness
1.6 km apart,following the standard protocol for forest from Landsat 5 imagery as an additional variable
carnivores (Zielinski & Kucera 1995). This survey (software: ERDAS 2008 Leica geosystems geospatial
yielded 3 red fox detection locations (Perrine 2005). imaging, Atlanta, GA). Tasseled-cap transformation
We combined and subsampled these occurrence data variables represent a continuous environmental gradi-
to use as the species response variable.Combining these ent that is highly correlated with stand age and struc-
data reduced the effect of the sampling biases inherent tural complexity (Hansen et al. 2001). Pixels containing
in each method,but new errors arose as a result of com- high greenness and wetness values are associated with
bining 4 different collection methods that spanned dense vegetation having high leaf area index, while
multiple years. Some locations contained 2 presence lower values indicate sparsely vegetated areas such as
points,because a fox was detected in the same area by 2 barren areas or regions of snow and ice (White et al.
different methods. These P seudoreplicates violate the 1997,Waring&Running 1998).We used Spatial Analyst st
assumption that training data points are independent, (ESRI)to calculate the Euclidean(straight-line)distance
which in turn can bias model results(Gursa n&Zimmer- from the center of each raster cell to the nearest water
from the
National H dro rah dataset
P Y
mane 2000 .Additionally,false absences likely occurred featureY 9
}
at some camera locations due to the elusiveness of this (http://nhd.usgs.gov) and to derive slope from a 30 m
species. Within the 10 yr sampling period, several digital elevation model (USGS 2000). The Sierra
camera locations detected foxes in one year but not an- Nevada red fox's elevational limits,relationship to snow
other. Having detection and non-detection records at pack, and phylogeography (Aubry et al. 2009) suggest
Cleve et al.:Predicting potential Sierra Nevada red fox habitat 183
an affinity for specific climatic conditions.We used grid- entropy method(Maxent) with full region background
ded climate data derived from the Parameter-elevation pixels(hereafter,MFB),a Maxent model using a subset
Regressions on Independent Slopes Model (PRISM) to of background pixels with similar biases as the occur-
predict the species'response to different climatic condi- rence data(MSB), and a spatially-weighted presence-
tions; specifically, we used mean monthly precipitation absence logistic regression model(LRW). We used the
and monthly average daily minimum and maximum default parameters for both Maxent models and gener-
temperatures from 1971 through 2000(Daly et al. 1994). ated outputs in the logistic regression format.
After selecting the initial environmental variables For each approach, we developed the model with a
based on red fox ecology, we used the R statistical random subset of 70% of the data and withheld the
package (R Development Core Team 2005) to deter- remaining 30%for model evaluation.To determine the
mine correlations between continuous variables and to classification accuracy of each model, we used the
identify interaction terms. If 2 variables were corre- evaluation data to identify the optimum cutoff value
lated (Pearson's correlation coefficient > 0.3), only the that corresponded with high red fox habitat suitability.
variable with the lower p value was retained. We then Optimum cutoff values were determined by caiculat-
identified pairwise interaction terms using a classifica- ing the true skill statistic (TSS) across the entire range
tion and regression tree(Miller&Franklin 2002).Clas- of potential cutoff values,and the cutoff value that cor-
sification and regression trees determine a set of if- responded with the highest TSS was selected as the
then statements that define class membership,and can optimum cutoff(Allouche et al.2006,Jones et al.2010).
express complex non-linear and non-additive relation- We then calculated the AUC to assess predictive
ships among the predictor variables (Miller&Franklin performance (Buckland et al. 1997).
2002). We included 9 environmental variables in the Predictive accuracy could only be tested in the
classification and regression tree model: CWHR type, Lassen Peak region because of the limited geographic
total tree canopy closure, tree size class, tree density extent of the available data.To compare model perfor-
class, slope, February precipitation, minimum Decem- mance outside the Lassen Peak region, we compared
ber temperature, Tassled-cap greenness, and distance the distribution and abundance of each model's suit-
to water. able habitat to the other models and to the historic
After removing correlated predictor variables and range map for the Sierra Nevada red fox(Grinnell et al.
determining interactions, we selected significant pre- 1937).We used the optimum cutoff value to determine
dictor variables using iterative manual stepwise logis- the appropriate suitable habitat threshold. To create
tic regression;at each run,the least significant variable the ensemble prediction,the values of each model that
was removed until only significant variables remained fell below the optimum cutoff value were given a value
(Hastie & Pregibon 1992, Hosmer& Lemeshow 2000). of 0.0. The mean probability value from both models
In addition, at each step the Akaike information crite- was then assigned to each cell of the study area.
non (AIC) was used to select the best fitting model.
AIC is a standardized score used to compare models
for best fit relative to the number of parameters in the RESULTS
model;lower AIC values indicate better fit(Burnham&
Anderson 2002). The 9 environmental variables listed The classification and regression tree identified 2
in the-previous paragraph, along with the 2 pairwise significant pairwise interaction terms: February pre-
interaction terms identified in the classification and cipitation with minimum December temperature, and
regression tree model (see 'Results'), were included in February precipitation with image greenness. The
the stepwise logistic regression weighted by Thiessen stepwise logistic regression weighted by Thiessen
polygon area.The resulting set of significant predictor polygon area reduced our environmental variable set
variables was used in both the Maxent and logistic to the following6 variables: February precipitation,
regression models.This allowed for direct comparison minimum December temperature, Tasseled-cap
between the 2 modeling approaches. This 2-step greenness, distance to water, the interaction between
method of using a GLM to select predictor variables February precipitation and minimum December tem-
followed by Maxent modeling has been shown to cre- perature, and the interaction between February pre-
ate predictions with very high area under the receiver cipitation and greenness. We used these 6 variables to
operating characteristic (ROC) curve (AUC) values generate the species distribution models via the 3
(Wollan et al. 2008). High AUC values indicate low approaches described in the above section 'Distribu-
error, while lower values indicate lower predictability tion models and model evaluation'.
(Pearce&Ferrier 2000). Within the Lassen Peak region, all models had good
Distribution models and model evaluation.We gener- fit to the test data, as indicated by high values for TSS,
ated 3 distribution models: a presence-only maximum percent correctly classified,and AUC(Table 1).For all 3
184 Endang Species Res 14: 179-191,2011
Table 1. Summary of model results for all 3 models of suitable habitat for Sierra Nevada red fox Vulpes vulpes necator. MFB:
spatially-weighted Maxent with full background pixels; MSB: Maxent with subset background pixels;LRW: logistic regression,
AUC:area under the curve;TSS:true skill statistic;LPR SH:suitable habitat(ha)for Lassen Peak Region;SA SH:suitable habitat
(ha)for entire study area;SA-LPR SH:suitable habitat(ha)outside the Lassen Peak Region.Suitable habitat is defined as the area
that contains a probability of red fox occurrence equal to or greater than the optimum cutoff value.Optimum cutoff values were
determined by calculating the TSS across the entire range of potential cutoff values,and the cutoff value that corresponded with
the highest TSS was selected as the optimum cutoff
Model Correctly classified(%) AUC TSS(%) Optimum cutoff LPR SH(ha) SA SH(ha) SA-LPR SH(ha)
MFB 91.5 0.9579 83.1 0.333 92187 660 479 568 292
MSB 90.8 0.9537 81.9 0.157 79 754 935 362 855 608
LRW 86.0 0.9438 75.8 0.184
115 530 1 546 021 1430 491
Table 1
t statistics, the MFB model yielded sifted outside of the Lassen Peak region ( ),
accuracy assessment s Yl
Y
the highest values, with the MSB model slightly lower, where the MSB and MFB models predicted approxi-
followed by the LRW model. All 3 models had low opti- mately 40 to 60% less suitable habitat than the LRW
mum cutoff values, ranging from 0.157 for the MSB model. All 3 models approximated the historic range
model to 0.333 for the MFB model(Table 1). boundary for the Sierra Nevada red fox,with the LRW
jDespite their similar accuracy,the 3 models varied in model yielding the closest fit and the MFB model
the location and extent of the suitable habitat area they having the sparsest fit (Fig. 2).
predicted (Table 1, Fig. 2). Within the Lassen Peak The Maxent models (MFB and MSB) selected
� region, the MFB and MSB models predicted approxi- regions with lower minimum December temperatures
mately 20 and 30% less suitable habitat, respectively, and higher February precipitation than the entire
than the LRW model(Table 1).This pattern was inten- study region (Table 2). The interaction between
MFB MSB LRW
Mount,', Mount Suitability
Mount' .• y
Shalt - Suitability Shalt .-,.�; Suitability Shlsta,,. -w
High �.. '�.. High , High
low i
Low Low
Lassen''. -T 'Lassen ` Lassen a California.,?,_ - Calflomia
,..Region ': I pundary R gion ,- tbpoxmate R goon .. °' APB xl t
Peak /, i ,. boundary
........
historic i......:historic i....•.• historic y
range range range
Ni
O Ixy t'. 0^
J'X �
m
k. Sequoia and Kings '
Sequoia and Kings Sequoia and Kingstir
Canyon
National Canyon National ;. Canyon National
y
Parks Region
Parks Re ion
g
Parks Region g N
N �• N I
A
0 50 100 200 0 50 100 200
0 50 100 200 m
i
. Maxe
nt background
' rra Nevada red fox based on 3 models' 9
for 5>e
es vuI es necator.Predicted suitable habitat
Fig.2. Vulp p
pixels (MFB), Maxent with subsetted background pixels (MSB), and spatially-weighted logistic regression (LRW).The dashed
line represents the historic range(Grinnell et al.1937).Figs.3 to 5 display in detail the 3 geographic regions identified in this map
Cleve et al.:Predicting potential Sierra Nevada red fox habitat 185
I
1 Table 2.Range,mean,and SD of environmental variables(minimum temperature in December,precipitation in February,greenness,
j and distance to nearest body of water)for the entire study area and for suitable habitat areas for Sierra Nevada red fox Vulpes vulpes
necator as predicted by each of the 3 models.MFB:Maxent with full background pixels;MSB:Maxent with subset background pixels;
LRW:spatially-weighted logistic regression
Environmental — Study area MFB MSB LRW
variables(units) Min/Max Mean t SD Min/Max Mean t SD Min/Max Mean t SD Min/Max Meant SD
MIN.TEMP. -15.3/5.2 -3.9 t 3.4 -10.4/-2.4 -5.3 t 1.0 -15.3/-3.6 -6.8 t 2.3 -15.3/5.3 -7.4 t 2.6
DEC(°C)
FEBTRECIP 26.8/481.6 183.3 t 80.5 137.7/471.3 240.7 t 64.0 147.7/471.3 239.3 t 52.8 34.3/471.3 223.9 t 57.6
GREENNESS -2266/1205 52 t 247 -1906/1030 39 t 217 -2134/881 -125 t 375 -2134/896 -168.6 t 356.3
(Derived)
DIST.TO WATER(m) 0/23350 3912 t 3402 0/10331 2051 t 1483 0/10657 2038 t 1441 0/13829.3 2715.6 t 2110.2
December minimum temperature and February pre- spectral data that are not captured in predetermined
cipitation accounted for 50% of the predictor variable classification schemes. In our study, image greenness
contribution to the Maxent models (Table 3).The LRW emerged as a predictor variable over 3 predetermined
model also included this interaction and predicted classification schemes:tree density,tree size class,and
areas with a lower minimum temperature than the CWHR category. This suggests that the vegetation
study region as a whole. But unlike the Maxent mod- associations of the Sierra Nevada red fox are better
els, the LRW model had a February winter precipita- characterized by the unclassified spectral data than by
tion range similar to that of the entire study area, these predetermined classification schemes.This find-
which included areas that received lower precipitation ing is significant because few studies leverage the ben-
(Table 2).The remaining environmental variable sum- efits of remotely-sensed data in their species distribu-
mary statistics were comparable across the 3 models: Ition modeling(Turner et al.2003, Gillespie et al.2008).
suitable habitat was within 2 to 3 km of a water feature, In addition, predetermined classification schemes are
and greenness values were lower than the study area often not available, are time consuming and expensive
mean`(Table 2). These variables represented 12 and to produce, and are often inconsistent between
2%,respectively,of the predictor variable contribution regions. Not having to rely on vegetation maps for
in the Maxent models(Table 3). habitat modeling can greatly reduce needed resources
because satellite imagery is often free and readily
available.Additionally, using satellite imagery instead
DISCUSSION of predetermined classification schemes creates a
more parsimonious model by decreasing the number of
We used presence-absence (logistic regression) and variables needed.
presence-only maximum entropy) methods to create
P Y ( eate
PY) Both the logistic regression and maximum entropy
the first spatially explicit habitat suitability model,based modeling approaches indicated that climate was a
on climatic and remotel
y sensed variables,
ambles for the Sierra
Y major component of habitat suitability for the Sierra
Nevada red fox,a threatened species under the Califor- Nevada red fox. Recent phylogenetic analyses have
nia Endangered S _
g Species Act. The resulting model pre indicated that the Sierra Nevada red fox and its con-
dictions can guide future surveys to locate additional specifics in the Cascade and Rocky Mountains (Vulpes
populations of this rare subspecies, and also illustrate vulpes cascadensis and V. v. macroura, respectively)
the relative performance of these modeling approaches. comprise a distinct genetic lineage separate from much
Our analysis is one of the first empirical tests of the of the rest of North America (Aubry et al. 2009). This
revised algorithm in Maxent to reduce the effect of geo- lineage was more widespread during the height of the
graphically biased occurrence data and improve its abil- Pleistocene glaciation but retracted to the high eleva-
ity to transfer to new study regions (Phillips 2008). tions of the western mountains when the glaciers
retreated.
These
historic range expansions and con-
tractions coincide with regional climate change, indi-
Predictor variables cating that the Sierra Nevada red fox may be physic-
logically and ecologically constrained to subalpine
Our findings support the conclusion by Laurent et al. climate zones (Aubry et al. 2009).
(2005) that wildlife may have a response to the contin- Both modeling approaches identified areas with low
uous environmental gradient present in unclassified winter minimum temperatures, but the models varied
186 Endang Species Res 14: 179-191,2011
Table 3.Variable coefficients for logistic regression(LR)model and percent con- the species' entire range and at a more
tributions for Maxent models.See Table 2 for definitions of variables and units. localized scale of the original popula-
na:not applicable lion centers of the Sierra Nevada red
fox as described by Grinnell et al.
Variable LR: Maxent: (1937). Although all 3 models pre-
coefficient %contribution dieted high habitat suitability through-
Constant 0.3067 na out the historic range of the Sierra
FEB.PRECIP -0.0001483 30.4 Nevada red fox, the MSB model pre-
MIN.TEMP.DEC 0.008398 5.3 dicted slightly more habitat than the
GREENNESS -0,00483 2.0 MFB model, and the LRW model pre-
DIST.TO WATER -0.0003856 11.9 dicted far more suitable habitat than
Interaction:FEB.PRECIP x MIN.TEMP.DEC -0.000000636 50.3 either Maxent model. For example,
Interaction:FEB.PRECIP x GREENNESS 0.000000144 0.1
north of Lassen Park, the LRW model
predicted all of Mount Shasta to be
in how they represented the effect of precipitation.The suitable habitat, whereas the MSB model selected
Maxent models predicted areas with winter precipita- Shasta's eastern slope and a small portion of its peak
tion well above the regional minimum, whereas the and western slope, and the MFB model selected only
LRW model predicted areas with lower winter precipi- its eastern slope (Fig. 4). The discrepancy became
tation (Table 2). This likely accounted for much of the more pronounced with increased distance from Lassen
spatial difference between the model predictions. Peak,the location of the occurrence data.In the south-
Within the Lassen Peak region, for example, the LRW ernmost historic population center, the Sequoia and
model predicted more suitable habitat than the Max- Kings Canyon National Parks region of the southern
ent models, and predicted more habitat in the eastern Sierra Nevada, the LRW identified most of the region
portion of the area (Fig. 3). Field surveys occasionally as suitable habitat (196855 ha; Fig. 5). In contrast, the
detected red fox east of Lassen Peak, but far less fre- MFB predicted only a small amount of suitable habitat
quently than in the western portion of the area region, (23 234 ha),and the MSB model predicted an interme-
where the projections of the Maxent models were con- diate amount (84 496 ha). The pattern of Maxent and
centrated(Perrine 2005). logistic regression models yielding similar AUC values
Outside of the Lassen Peak region,the differences in but predicting slightly different suitable habitat areas
prediction area were exacerbated, both on the scale of is consistent with prior findings(Gibson et al. 2007).
MFB MSB LRW
Suitability Test data Suitability Test data Suitability Test data
High ♦ Absence High • Absence High • Absence
O Presence c, O Presence c, O Presence
Low
"
Low ®
.�
Low • •� •
Lassen Peak
OLassen Peak O Lassen Peak O
Region boundary Region boundary Region boundary
9 ,
• • � � _ •"
-q.
" �" M• " 2 io •` -
IY
' 9 �q
d�. L
•. '• L. GP do • :•.. .p .� .
� •
r ^ a N
,1g ``�• 0 15 30 km .'`ram" 0 15 30 km ti'L' 0 15 30 km
Fig.3. Vulpes vulpes necator.Predicted suitable habitat for Sierra Nevada red fox in the Lassen Peak region,relative to presence-
absence test data. MFB: Maxent with full background pixels; MSB: Maxent with subset background pixels; LRW: spatially-
weighted logistic regression.See Fig.2 for the location of the Lassen Peak region in California
Cleve et al.:Predicting potential Sierra Nevada red fox habitat 187
MFB MSB LRW
A:
.F
Suitability Suitability S Suitability d
I ,A' High High High
Low Low
Fig.4. Vulpes vulpes necator.Predicted suitable habitat for Sierra Nevada red fox in the Mount Shasta region of northern Califor-
nia.MFB: Maxent with full background pixels; MSB:Maxent with subset background pixels; LRW:spatially-weighted logistic
regression.See Fig.2 for the location of Mount Shasta in California
MFB MSB LRW
Suitability h''t# Suitability r Suitability
High na d IH!ghx High
i S
Kt. , LOW
LOW - ��t�
a rAi' Low
h j 1
411�r� 'r.�,
I!: �Y S,�i. "� 4 1 $ ail?•
I•
NN N ;
0 10 20 0 10 20 k 0 10 20 m
i i i
Fig.5. Vulpes vulpes necator.Predicted suitable habitat for Sierra Nevada red fox in the Sequoia and Kings Canyon National Parks
region in the southern Sierra.MFB:Maxent with full background pixels;MSB:Maxent with subset background pixels;LRW:spa-
tially-weighted logistic regression.See Fig.2 for the location of the Sequoia and Kings Canyon National Parks region in California
Because evaluation data were available only within in the occurrence data, yielded a prediction area and
the Lassen Peak region, our assessment of model per- extent that was intermediate between the MFB model
formance beyond this region was limited to compar- and the LRW model (Fig. 2).Although the MFB model
isons between models and with the species' historic yielded high accuracies within the occurrence data re-
range (Grinnell et al, 1937). The MSB model, which gion, it predicted less suitable habitat in areas farther
used subsetted background pixels to match the biases away from the occurrence data than the MSB model.
188 Endang Species Res 14: 179-191,2011
These results are consistent with general expectations results indicate that while subsetting significantly
for geographically biased presence data (Phillips reduces the effects of geographically biased presence
2008). The MSB model, which used a subset of back- data in Maxent, it does not completely resolve the
ground pixels, greatly reduced the effect of geographic problem. Despite the risk of false absences, the LRW
bias.In contrast to the Maxent models,the LRW model model did quite well predicting both within and out-
predicted abundant suitable habitat outside of the side of the Lassen Peak region,leading us to conclude
Lassen region, confirming prior findings that logistic that in this case the potentially false absences were less
regression models can transfer well to similar study problematic than the transferability issues inherent to
areas(Randin et al.2006,Barbosa et al.2009). the Maxent models.
Our range predictions for the Sierra Nevada red fox,
like all species distribution models,are hypotheses that
should be tested by the collection of new data, espe- Management implications
cially in predicted areas that were previously unsam-
pled (Olivier & Wotherspoon 2006, Parra & Monahan As a carnivore closely associated with montane tree-
2008, Wollan et al. 2008, Costa et al. 2009). The line habitats,the Sierra Nevada red fox may be acutely
increased use of automatic camera stations for carni- affected by climate changes. Since 1920, California's
vore inventory and monitoring on public lands average annual temperature has warmed by 1.7°C,
throughout the Sierra Nevada holds great promise for with greater warming occurring with daily minimum
additional detections of Sierra Nevada red fox. Fortu- temperatures and at higher elevations(Kapnick&Hall
itously, in autumn 2010 Forest Service biologists 2009). In the central Sierra Nevada, December mini-
obtained photographs of several putative Sierra mum temperatures have increased by 0,23°C decade-'
Nevada red foxes in the Sonora Pass region of the in Yosemite Valley (1220 m elevation) and 0.32°C
northern Sierra Nevada, approximately 75 km south decade' at Lake Tahoe (1900 m; Thorne et al. 2006).
of Lake Tahoe (A. Rich and S. Lisius pers. comm.). Recent surveys in the Sierra Nevada have documented
Although final confirmation via molecular genetic range shifts by small mammals (Moritz et al. 2008),
methods is still underway, preliminary analyses and birds (Tingley et al. 2009), butterflies (Forister et al.
the fact that several foxes were detected in dose prox- 2010),and conifers(Thorne et al.2006),consistent with
imity indicate that these animals represent the first ver- responses to climate warming. Despite the recent
ified population of Sierra Nevada red fox detected out detection of at least 1 Sierra Nevada red fox in the
of the Lassen area in nearly 2 decades (B. Sacks pers. northern Sierra, the lack of recent documented detec-
comm.). These detections lend further support to the tions in the southern extent of its historic range (Per
9 -
lo istic regression model over the 2 Maxent models. rine et al. 2010) may indicate that its range has
The logistic regression identified these detection sites retracted northward in response to climate change.It is
as high quality habitat for Sierra Nevada red fox; the unclear whether climate has a direct or indirect impact,
subalpine woodland is virtually identical structurally to such as facilitating coyotes Canis la trans or other com-
occupied Sierra Nevada red fox habitat in the Lassen petitors or changing understory structure through
region, despite being composed of whitebark pine altered fire regimes.
Pinus al bicaulus as opposed to mountain hemlock in If such range retraction has already occurred, the
' rra Nev
ada red fox may have little future in Califor-
the Lassen region. The Maxent model with subsetted Sierra Y
background pixels identified the Sonora Pass site as nia, Climate warming is expected to continue if not
medium-quality habitat for Sierra Nevada red fox, accelerate in the coming century, although forecasts of
whereas the original Maxent algorithm did not identify the amount and rate of change depend greatly on the
it as potential habitat at all. specific climate model and emissions scenario used
The ability to subset background pixels is relatively (Hayhoe et al. 2004). Nevertheless, montane regions
new to Maxent (Phillips 2008), and few studies to date are likely to experience the greatest warming (Snyder
have used or evaluated these methods (Anderson & et al. 2002), with mean annual temperatures in the
Raza 2010). Our findings support the conclusion of Sierra Nevada and southern Cascades projected to rise
Anderson & Raza (2010) that calibrating the study 3.0 to 3.5°C by 2070 to 2099(Ackerly et al.2010).These
region or subsetting background pixels to the occur- changes will likely have a profound effect on the
rence data area yields a Maxent model with a larger ranges, elevations, and associations of California's
predicted area that is less concentrated around the biota (e.g. Loarie et al. 2008, Parra & Monahan 2008,
occurrence data region. Our study emphasizes the Wiens et al. 2009, Ackerly et al. 2010, Forister et al.
importance of subsetting background pixels in Maxent 2010) The Sierra Nevada red fox may follow the
when using geographically biased presence data, pattern of another alpine-associated carnivore, the
which are common with rare species. Moreover, our wolverine Gulo gulo,which disappeared from Califor-
Cleve et al.:Predicting potential Sierra Nevada red fox habitat 189
nia by the 1930s, with populations persisting only in include the Mount Shasta region (Fig. 4), which is the
higher latitude states such as Washington, Montana, largest predicted area north of the extant Lassen pop-
and Idaho (Aubry et al. 2007). Predicting the future ulation, and the Sequoia and Kings Canyon region,
range of Sierra Nevada red fox based upon its historic which is the southernmost of the historic population
range and the anticipated climate changes is an impor- centers (Grinnell et al. 1937) and an area of profound
tant next step toward its effective conservation. model discrepancy. Such surveys should incorporate
Locating additional Sierra Nevada red fox popula- the collection of specimens for genetic analysis
bons outside of the Lassen Peak area is critically because photographs alone cannot conclusively iden-
important (Perrin et al. 2010) to better document the tify Sierra Nevada red foxes (Perrine et al. 2010).
taxon's true spatial extent and to acquire additional Additionally, the Lassen population should be closely
specimens to refine its phylogenetic relationships with monitored for changes in size or extent, as it remains
other red fox populations (e.g., Perrine et al, 2007, the only known actively reproducing population of
Aubry et al. 2009, Sacks et al. 2010). One of our pri- this endangered taxon (Perrine et al. 2010).If no other
mary goals for these analyses was to provide guid- reproductive populations exist outside of the Lassen
ance for range-wide field surveys. To test our model region, then the Sierra Nevada red fox likely warrants
predictions, field surveys should target areas of model a higher level of state or federal protection and active
discrepancy, such the Sequoia and Kings Canyon area management than it currently receives.
of the southern Sierra Nevada (Fig. 5). However,
surveys attempting to locate additional populations
as efficiently as possible should instead target areas LITERATURE CITED
of model agreement, as indicated by our 2-model 1p Ackerly DD,Loarie SR,Cornwell WK,Weiss SB,Hamilton H,
ensemble (Fig. 6). The highest survey priority should Branciforte R,Kraft NJB (2010)The geography of climate
be to determine the extent of the newly discovered change:implications for conservation biology.Divers Dis-
Sonora Pass population.Additional high priority areas trib 16:476-487
>Allouche O, Tsoar A, Kadmon R (2006) Assessing the accu-
racy of species distribution models:prevalence,kappa and
the true skill statistic(TSS).J Appl Eco143:1223-1232
r. Suitability > Anderson RP, Raza A (2010) The effect of the extent of the
(Ensemble model) study region on GIS models of species geographic distrib-
utions and estimates of niche evolution: preliminary tests
High:0.969074 with montane rodents (genus Nephelomys)in Venezuela.
j J Biogeogr 37:1378-1393
Low:0 Arau'o MB,Guisan A 2006 Five or so challenges for species
> l ( ) ( ) g
distribution modelling.J Biogeogr 33:1677-1688
California ' Aradjo MB, New M (2007) Ensemble forecasting of species
+A boundary
distributions.Trends Ecol Evol 22:42-47
Approximate > Aubry KB, McKelvrey KS, Copeland JP (2007) Distribution
historic range and broadscale habitat relations of the wolverine in the
1
I
contiguous United States.J Wil dl ana g M 71:2147-2158
g
q� i
i.
� Aubry KB, Statham MJ, Sacks BN Perrine JD Wisely SM> Y
.� Y I
O (2009)Phylogeography of the North American red fox:vic-
7�. ariance in Pleistocene forest refugia. Mol Ecol 18:
y 2668-2686
\� D• Barbosa AM, Raimundo R, Vargas JM (2009) Transferability
of environmental favourability models in geographic
space: the case of the Iberian desman (Galemys pyre-
naicus)in Portugal and Spain.Ecol Model 220:747-754
Pacific \ > Buckland ST,Burnham KP,Augustin NH(1997)Model selec-
Ocean 1' tion: an integral part of inference. Biometrics 53:603-618
> Bulluck L, Fleishman E, Betrus C, Blair R (2006) Spatial and
N temporal variations in species occurrence rate affect the
J
accuracy of occurrence models. Glob Ec
ol Biogeogr 4
Bio r 15:
27-38 S
Burnham DP,Anderson DR(2002)Model selection and multi-
0 50 100 km model inference: a practical information-theoretic
approach,2nd edn.Springer,New York,NY
L I CDFG (California Department of Fish and Game) (1996)The
status of rare, threatened and endangered animals and
Fig. 6. Vulpes vulpes necator. Ensemble habitat suitability plants of California. California Department of Fish and
model for Sierra Nevada red fox,based on spatially-weighted Game,Sacramento,CA
logistic regression (LRW) and Maxent with subset back- CDFG (2004) Special animals. California Department of Fish
ground(MSB)models and Game,Sacramento,CA
190 Endang Species Res 14: 179-191,2011
I. Costa GC, Nogueira C, Machado RB, Colli GR (2009) Sam- y Loarie SR Carter BE,Ha Yhoe K,McMahon S,Moe R,Knight
pling bias and the use of ecological niche modeling in con- CA,Ackerly DD (2008) Climate change and the future of
servation planning: a field evaluation in a biodiversity California's endemic flora.PLoS ONE 3:e2502
hotspot.Biodivers Consery 19:883-899 MacKenzie D,Nichols J,Royle JA,Pollock K,Bailey L,Hines
> Daly CS, Neilson RP, Phillips DL (1994) A statistical-topo- J (2006) Occupancy estimation and modeling. Academic
graphic model for mapping climatological precipitation Press,London
over mountainous terrain.J Appl Meteorol 33:140-158 Mayer KE, Laudenslayer WF Jr (eds) (1988) A guide to
Edwards TC,Cutler DR,Zimmermann NE,Geiser L,Alegria J wildlife habitats of California. California Department of
(2005) Model-based stratifications for enhancing the Fish and Game,Sacramento,CA
detection of rare ecological events.Ecology 86:1081-1090 Menon S, Choudhury B, Khan ML, Peterson AT (2010)
Elith J, Graham C, Anderson R, Dudik M and others (2006) Ecological niche modeling and local knowledge predict
Novel methods to improve prediction of species' distribu- new populations of Gymnocladus assamicus a critically
tions from occurrence data.Ecography 29:129-151 endangered tree species. Endang Species Res 11:
j p. Engler R, Guisan A, Rechsteiner L (2004) An improved 175-181
approach for predicting the distribution of rare and endan- ' Miller J, Franklin J (2002) Modeling the distribution of four
gered species from occurrence and pseudo-absence data. vegetation alliances using generalized linear models and
J Appl Ecol 41:263-274 classification trees with spatial dependence. Ecol Model
1. Ferreira de Siqueira M,Durigan G,de Marco Junior P,Peter- 157:227-247
son AT(2009) Something from nothing: using landscape )p. Mladenoff D, Sickley T, Wydeven A (1999) Predicting gray
similarity and ecological niche modeling to find rare plant wolf landscape recolonization: logistic regression models
species.J Nat Consery 17:25-32 vs.new field data.Ecol Appl 9:37-44
Forister L, McCall AC, Sanders NJ,Fordyce JA and others ),. Moritz C, Patton JL, Conroy CJ, Parra JL, White GC,
M
(2010)Compounded effects of climate change and habitat Beissinger SR(2008)Impact of a century of climate change
alteration shift patterns of butterfly diversity. Proc Natl on small-mammal communities in Yosemite National Park,
Acad Sci USA 107:2088-2092 USA,Science 322:261-264
Gibson L,Barrett B,Burbidge A(2007)Dealing with uncertain Nams VO(2001)Locate II:user's guide.Pacer,Truro,NS
absences in habitat modeling: a case study of a rare Nielsen SE, Boyce MS, Stenhouse GB, Robin H, Munro M
ground-dwelling parrot.Divers Distrib 13:704-713 (2002) Modeling grizzly bear habitats in the Yellowhead
1r. Gillespie TW, Foody GM, Rocchini D, Giorgi AP, Saatchi S Ecosystem of Alberta: taking autocorrelation seriously.
(2008) Measuring and modeling biodiversity from space. Ursus 13:45-56
Prog Phys Geogr 32:203-221 g► Olivier F,Wotherspoon SJ(2006)Distribution and abundance
Grinnell J,Dixon J,Linsdale J(1937)Fur-bearing mammals of of Wilson's storm petrels Oceanites oceanicus at two loca-
California.University of California Press,Berkeley,CA tions in East Antarctica: testing habitat selection models.
A
Guisan A,Zimmermann NE(2000)Predictive habitat distrib- Polar Bio129:878-892
ution models in ecology.Ecol Model 135:147-186 ' Farra JL,Monahan WB (2008)Variability in 20th century cli-
i
Guisan A, Broenniam g ann O Engler R Vust M, Yoccoz N, mate change reconstructions and its consequences for
�.
ann N 2006 Using niche-based predicting geographic responses of California mammals.
Lehmann A, Zimmermann ( I g
models to improve the sampling of rare species. Consery Glob Change Biol 14:2215-2231
Biol20:501-511 Pearce J, Ferrier S (2000) Evaluating the predictive perfor-
Hall ER (1981) The mammals of North America, 2nd edn. mance of habitat models developed using logistic regres-
John Wiley&Sons,New York,NY sion.Ecol Model 133:225-245
n M 2001 For- Perrine JD 2005) Ecology of red fox (Vulpes vulpes) in the
sma C Peterson ( g
Hansen MJ,Franklin SE Woud )
n
Columbia mou n- Lassen Peak region of California, USA. PhD dissertation,
est structure classification in th
e North Co 9
tains using Landsat TM tasseled cap wetness component. University of California,Berkeley,CA
Can J Rem Sens 27:20-32 q. Perrine JD, Pollinger J, Sacks B, Barrett R, Wayne R (2007)
Hastie TJ, Pregibon D (1992) Generalized linear models. In: Genetic evidence for the persistence of the critically
Chambers JM, Hastie TJ (eds) Statistical models in S. endangered Sierra Nevada red fox in California.Consery
Wadsworth& Brooks/Cole,Pacific Grove, CA,p 195-248 Genet 8:1083-1095
q. Hayhoe K,Cayan D,Field CB,Frumhoff PC and others(2004) Perrino JD, Campbell LA, Green GA (2010) Sierra Nevada
Emissions pathways, climate change,and impacts on Cal- red fox (Vulpes vulpes necator): a conservation assess-
ifornia.Proc Nall Acad Sci USA 101:12422-12427 ment.RS-FR-010.USDA Forest Service,Vallejo,CA
Hosmer DW, Lemeshow S (2000)Applied logistic regression. Peterson AT, Papes M, Eaton M (2007) Transferability and
John Wiley&Sons,New York,NY model evaluation in ecological niche modeling:a compar-
Johnson CJ, Seip DR, Boyce MS (2004) A quantitative ison of GARP and Maxent.Ecography 30:550-560
approach to conservation planning: using resource selec- y Phillips SJ (2008) Transferability, sample selection bias and
tion functions to map the distribution of mountain caribou background data in presence-only modelling: a response
at multiple spatial scales.J Appl Ecol 41:238-251 to Peterson et al. (2007).Ecography 31:272-278
;. Jones CC, Acker SA, Halpern CB (2010) Combining local- >. Phillips SJ, Anderson RP, Schapire RE (2006) Maximum
and large-scale models to predict the distributions of inva- entropy modeling of species geographic distributions.Ecol
sive plant species.Ecol Appl 20:311-325 Model 190:231-259
Kapnick S, Hall A (2009) Observed changes in the Sierra Posillico M, Meriggi A, Pagnin E, Lovari S, Russo L (2004)A
Nevada snowpack:potential causes and concerns.CEC-500- habitat model for brown bear conservation and land use
2009-016-F,California Energy Commission,Sacramento,CA planning in the central Apennines. Biol Consery 118:
1,, Laurent E, Shi H, Gatziolis D, LeBouton J, Walters M, Liu J 141-150
(2005) Using spatial and spectral precision of satellite R Development Core Team(2005)R:a language and environ-
imagery to predict wildlife occurrence patterns. Remote ment for statistical computing.R Foundation for Statistical
Sens Environ 97:249-262 Computing,Vienna.Available at www.R-project.org
Cleve et al.:Predicting potential Sierra Nevada red fox habitat 191
Rachlow JL, Svancara LK (2006) Prioritizing habitat for )..Tingley MW, Monahan WB, Beissinger SR, Moritz C (2009)
surveys of an uncommon mammal:a modeling approach Birds track their Grinnellian niche through a century of
applied to pygmy rabbits.J Mammal 87:827-833 climate change. Proc Nall Acad Sci USA 106 Suppl 2:
10., Randin CF, Dirnbo T, Dullinger S, Zimmermann NE, Zappa 19637-19643
M, Guisan A (2006) Are niche-based species distribution )p. Turner W, Spector S, Gardiner N, Fladeland M, Sterling E,
models transferable in space?J Biogeogr 33:1689-1703 Steininger M (2003) Remote sensing for biodiversity sci-
)► Rebelo H,Jones G(2010)Ground validation of presence-only ence and conservation.Trends Ecol Evol 18:306-314
modelling with rare species: a case study on barbastelles USGS (US Geological Survey) (2000) 10 m digital elevation
Barbastella barbastellus(Chiroptera: Vespertilionidae). J model.Menlo Park,CA
Appl Ecol 47:410-420 USDA-NRCS (US Department of Agriculture-Natural Re-
)m,. Rhynsburger D(1973)Analytic delineation of Thiessen poly- sources Conservation Service) (2006) Major Land Re-
gons.Geogr Anal 5:133-144 source Area (MLRA). Land resource regions and major
y. Sacks BN, Statham MJ, Perrine JD, Wisely SM, Aubry KB land resource areas of the United States, the Caribbean,
(2010) North American montane red foxes: expansion, and the Pacific Basin. US Department of Agriculture
fragmentation, and the origin of the Sacramento Valley Handbook 296
red fox.Consery Genet 11:1523-1539 Waring RH,Running SW(1998)Forest ecosystems:analysis at
Schempf PF, White M (1977) Status of six furbearer popula- multiple scales,2nd edn.Academic Press, San Diego,CA
tions in the mountains of northern California. USDA For- )P.White J,Running S,Nemani R,Keane R,Ryan K(1997)Mea-
est Service,California Region,San Francisco,CA surement and remote sensing of LAI in Rocky Mountain
y. Snyder MA, Bell JL, Sloan LC, Duffy PB, Govindasamy B montane ecosystems.Can J Res 27:1714-1723
(2002)Climate responses to a doubling of atmospheric car- p,. Wiens JA,Stralberg D,Johgsomjit D,Howell CA,Snyder MA
bon dioxide for a climatically vulnerable region.Geophys (2009) Niches,models,and climate change:assessing the
Res Lett 29:1514 doi:10.1029/2001GL014431 assumptions and uncertainties. Proc Natl Acad Sci USA
v. St-Louis V, Pidgeon AM, Radeloff VC, Hawbaker TJ, Clay- 106(Suppl 2):19729-19736
ton MK(2006) High-resolution image texture as a predic- y Wollan AK, Bakkestuen V, Kauserud H, Gulden G,
for of bird species richness. Remote Sens Environ 105: Halvorsen R (2008) Modeling and predicting fungal dis-
299-312 tribution patterns using herbarium data. J Biogeogr 35:
)p. Stohlgren TJ, Ma P, Kumar S, Rocca M, Morisette JT, 2298-2310
Jarnevich CS,Benson N(2010)Ensemble habitat mapping Zielinski WJ,Kucera TE(1995)American marten,fisher,lynx,
of invasive plant species.Risk Anal 30:224-235 and wolverine: survey methods for their detection. PSW-
Thorne JH,Kelsey R,Honig J,Morgan B(2006)The develop- GTR-157. USDA Forest Service, Pacific Southwest
ment of 70-year-old Weislander vegetation type maps and Research Station,Albany,CA
an assessment of landscape change in the central Sierra >. Zielinski WJ, Truex R, Schlexer F, Campbell L, Carroll C
Nevada. CEC 500-2006-107, California Energy Commis- (2005) Historical and contemporary distributions of carni-
sion,PIER Energy-related Environmental Program,Sacra- vores in forests of the Sierra Nevada, California, USA. J
mento,CA Biogeogr 32:1385-1407
Editorial responsibility:Luigi Boitani, Submitted:November 8,2010,Accepted:March 24,2011
Rome,Italy Proofs received from au :July y 23,
thor s J 2011
i
GENERAL MANAGER
Stephen E.Abbors
Regional
OpenSpaee 1 Midpeninsula Regional Open Space District BOARD OF DIRECTORS
Pete Siemens
Yoriko Kishimoto
Jed Cyr
Curt Riffle
Nonette Hanko
Larry Hassell
Cecily Harris
September 6, 2011
The Honorable Edward G. Brown, Jr.
Governor, State of California
State Capitol Building
Sacramento, CA 95814
RE: AB 703—Request for Signature
Dear Governor Brown:
I am writing to request your signature on AB 703 which will delete the January 1, 2013
sunset date of the property tax welfare exemption for natural resources and open space lands
and leave the exemption in place for another ten years.
The legislature has granted an exemption from this tax for property owned by nonprofit
organizations that is open to the general public and used for the preservation of native plants
or animals, or used solely for recreation and for the enjoyment of scenic beauty.
The most recent extension, which is the fifth extension since the original one expired in 1982,
set the repeal date to January 1, 2013 and AB 703 will extend this date by ten years setting it
to January 1, 2023.
We urge you to sign this important bill because some of the property currently owned by
nonprofit organizations would most likely be transferred to the state if the exemption were
not renewed, putting these important lands at risk because of the increased land management
and administrative costs that would have to be carried during a time when public resources
are tight. In addition, state-owned property is not subject to the property tax, as you know,
which is why the net effect of such an action would be the same as keeping the exemption in
place.
t�
33o Distel Circle Los Altos.CA 94022 6$0_691.1200 � r 65a.691 043S www_openspace_org
Nonprofit organizations hold thousands of acres of open space lands and it is important that
the exemption remains in place so that they can focus their strapped resources on managing
these lands for the public.
Thank you for your consideration of this legislation. Should you have any questions or need
additional information, please do not hesitate to contact me.
Sincerely,
Stephen E. Abbors
General Manager
cc: Ralph Heim, Public Policy Advocates
MROSD Board of Directors
Vol.14:179-191,2011 ENDANGERED SPECIES RESEARCH
doi:10.3354/esr00348 Endang Species Res Published online August 31
Addressing biased occurrence data in predicting
potential Sierra Nevada red fox habitat for survey
prioritization
John Perrine2'', Barbara Holzman', Ellen Hines'
'Department of Geography and Human Environmental Studies,San Francisco State University, 1600 Holloway Avenue,
HSS Room 279,San Francisco,California 94132,USA
2Biological Sciences Department,California Polytechnic State University,1 Grand Avenue,San Luis Obispo,
California 93407-0401,USA
ABSTRACT:The Sierra Nevada red fox Vulpes vu
lpes ulpes necatoris listed as a threatened species under
the California Endangered Species Act. It originally occurred throughout California's Cascade and
Sierra Nevada mountain regions. Its current distribution is unknown but should be determined in
order to guide management actions.We used occurrence data from the only known population,in the
Lassen Peak region of northern California,combined with climatic and remotely sensed variables,to
predict the species'potential distribution throughout its historic range. These model predictions can
guide future surveys to locate additional fox populations.Moreover,they allow us to compare the rel-
ative performances of presence-absence (logistic regression) and presence-only(maximum entropy,
or Maxent)modeling approaches using occurrence data with potential false absences and geograph-
ical biases. We also evaluated the recently revised Maxent algorithm that reduces the effect of geo-
graphically biased occurrence data by subsetting background pixels to match biases in the occur-
rence data.Within the Lassen Peak region, all models had good fit to the test data, with high values
for the true skill statistic (76-83%),percent correctly classified (86-92%), and area under the curve
(0.94-0.96),with Maxent models yielding slightly higher values.Outside the Lassen Peak region,the
logistic regression model yielded the highest predictive performance,providing the closest match to
the fox's historic range and also predicting a site where red foxes were subsequently detected in
autumn 2010. Subsetting background pixels in Maxent reduced but did not eliminate the effect that
geographically biased occurrence data had on prediction results relative to the Maxent model using
full background
gr pixels.els.
KEY WORDS: California • GIS • Logistic regression • Maxent • Species distribution model
Vulpes vulpes necator
Resale or republication not permitted without written consent of the publisher
INTRODUCTION 2006, Rachlow & Svancara 2006, Ferreira de Siqueira
et al. 2009, Menon et al. 2010, Rebelo & Jones 2010).
Determining the composition and distribution of Because a rare species will likely be absent from
suitable habitat is crucial for the successful manage- most sample locations in a simple random sample,
ment of rare or endangered species (Guisan & Zim- targeting survey locations to areas with high proba-
mermann 2000). Species distribution models have bility of species occurrence represents a more effi-
become an important tool to identify optimal survey cient use of limited conservation resources (Guisan et
areas for these species and to increase the probabil- al. 2006).
ity of locating previously unknown populations The Sierra Nevada red fox Vulpes vulpes necatoris a
(Engler et al. 2004, Edwards et al. 2005, Guisan et al. medium-sized canid that historically occurred at low
'Corresponding author.Email:jperrine@calpoly.edu 0 Inter-Research 2011 •www.int-res.com
180 Endang Species Res 14: 179-191,2011
densities throughout the high elevations of the Sierra Barbosa et al. 2009); presence-only models such as
Nevada and southern Cascade mountain ranges of Maxent have made improvements in their transfer-
California and Oregon,USA (Grinnell et al. 1937, Hall ability (Phillips 2008), with their performance being
1981,Sacks et al.2010).Within this broad range,Grin- positively correlated with the similarity between the
nell et al. (1937) reported 3 population centers: the occurrence data region and the projection area (Bul-
Mount Shasta/Lassen Peak region in northern Califor- luck et al. 2006).
nia, the central Sierra near Mono Lake and Yosemite Addressing these biases provided an opportunity to
National Park, and the southern Sierra near Mount explicitly compare the performance of 2 different mod-
Whitney(now mostly in the Sequoia and Kings Canyon eling approaches: logistic regression based on pres-
National Parks).In 1980,due to a noticeable decline in ence-absence data versus maximum entropy based on
numbers, the Sierra Nevada red fox was listed as a presence-only data. Logistic regression has been
state 'threatened' species; the factors causing its widely used in species distribution modeling (Mlade-
. 1999 Nielsen eta 1.2002 Johnson et al,2 004
i noff et al ,
r unknown CDFG 1996 2004 . Despite is ,
decline are ( ) p
former extent, the verified detections since 1991 have Posillico et al. 2004, Olivier & Wotherspoon 2006).
all been in the Lassen Peak region of northern Califor- Logistic regression uses presence-absence data to
nia (Perrine et al. 2007).A recent conservation assess- model the probability of species occurrence as a func-
ment for this species (Perrine et al. 2010) recom- tion of its predictor variables,which can be continuous
mended that targeted surveys for the Sierra Nevada or categorical (MacKenzie et al. 2006). Its output is
red fox be conducted throughout its historic range to confined to values between 0.0 and 1.0.
determine whether any additional populations exist. Maximum entropy methods,although relatively new
Prior attempts to detect this species as part of multi- and not as widely used as logistic regression, can out-
species carnivore inventories have been unsuccessful, perform logistic regression(Elith et al. 2006) and have
even in the Lassen Peak region where the species has successfully identified locations of previously undis-
persisted (Zielinski et al. 2005). Surveys targeting the covered populations (Rebelo &Jones 2010). The most
Sierra Nevada red fox and focusing on areas with high widely used maximum entropy approach, in the pro-
probabilities of the species' occurrence may increase gram Maxent (Phillips et al. 2006), estimates the spe-
detection probability and survey efficiency. cies' probability distribution that is most dispersed
Our goal was to predict the extent and distribution within the constraints of the target population informa-
of suitable habitat for the Sierra Nevada red fox lion.Like logistic regression,Maxent can use both cat-
throughout its historic range, based on the character- egorical and continuous predictor variables, and the
istics of occupied habitat in its current known range. output can provide information on the relative contri-
The available data were collected during a compre- button of each predictor variable (Phillips et al. 2006).
hensive ecological study of the Lassen Peak popula- Maxent can utilize presence-only data, sidestepping
lion (Perrine 2005). Unfortunately, these detection the problem of false absences, but geographical and
data contained 2 important biases. First, the survey environmental biases in the occurrence data can intro-
data contained potential 'false absences;' surveys duce considerable error in presence-only models
have failed to detect this species in areas where popu- (Phillips 2008). Presence-only models such as Maxent
lations were known to be present. Absence data for draw background pixel values from the entire study
rare species may result from the rarity of the species region, while presence data values are drawn only
rather than its true absence (MacKenzie et al. 2006). from a small portion of the study area (Phillips et al.
False absences are problematic in species distribution 2006). The resulting predictions may therefore under-
modeling because they do not indicate unsuitable represent habitat suitability outside of the occurrence
habitat or confirm that the species does not occur at a data area(Peterson et al.2007,Phillips 2008).Substan-
given site (Guisan & Zimmermann 2000, Engler et al. tial improvements can be made in Maxent models
2004, MacKenzie et al. 2006). Second, the only avail- derived from biased occurrence data by selecting
able occurrence data for the Sierra Nevada red fox background data with similar biases as the occurrence
were from the Lassen Peak region, which represents data (Phillips 2008). This approach has recently been
only a small portion of the species' historic range. implemented in Maxent (Phillips 2008), but its effects
Researchers have cautioned against using geographi- have had little empirical validation.
cally biased occurrence data or transferring models to Here we compared the performance and output of a
broad unsampled regions (Peterson et al. 2007, Bar- presence-absence logistic regression model versus
bosa et al. 2009);however, such projections have suc- presence-only Maxent models with and without the
cessfully increased detection rates (Guisan et al. transferability improvements. We then combined the
2006). General linear models (GLMs) may transfer output from the logistic regression and updated Max-
quite well to unsampled areas (Randin et al. 2006, ent model to generate an ensemble prediction, lever-
Cleve et al.:Predicting potential Sierra Nevada red fox habitat 181
aging the strengths of each model while minimizing Peak region (6455 km') includes Lassen Volcanic Na-
their respective weaknesses (Araujo & New 2007, tional Park (LVNP), the surrounding Lassen National
Stohlgren et al. 2010). In addition, we explored the Forest (LNF), and the immediately adjacent lands of
use of unclassified spectral data as a predictor vari- various ownerships(Fig. 1).This montane area is domi-
able in place of predetermined classification schemes nated by conifers such as Jeffrey pine Pin us jeffreyi and
e. vegetation or canopy cover categories).Although Ponderosa( g• 9 PY 9 )• g pine P. ponderosa, red fir Abies magnifica
classified maps are commonly used predictor vari- and white fir A. concolor,and mountain hemlock Tsuga
ables, wildlife may respond to continuous environ- mertensiana,along with wet alpine meadows and talus
mental gradients that are not captured in the class- slopes. This area has a Mediterranean climate with
ification schemes (Laurent et al. 2005). By using warm dry summers and cold wet winters. Most of the
unclassified spectral data, species' occurrence can be annual precipitation occurs as snow from November
predicted by spectrally detectable components of through April, with snowpacks at the higher elevations
their habitat, rather than predetermined classification often exceeding 3 m in depth and persisting into the
schemes that may inaccurately delineate boundaries summer months.
between cover types and under-represent habitat het- Fox survey data. Sierra Nevada red fox locations were
erogeneity (St-Louis et al. 2006). Using unclassified determined using 4 detection methods: radio teleme-
spectral reflectance in the distribution model may try,scats(feces),and camera surveys using opportunis-
minimize errors in the resulting predictive maps (Lau- tic and stratified random sampling designs. Each
rent et al. 2005). detection method contained biases.
MATERIALS AND METHODS
4
Study area. Our model prediction area
covers the area within and immediately ..:` Stud Y area
surrounding the historic range of the Area outside of
J� study region
Sierra Nevada red fox (Fig. 1). This in-
cludes the Sierra Nevada and the south- R p Lassen Peak
ernmost extent of the Cascade Range in - region
�. Approximate
California. The prediction area spans 2 historic range
Major Land Resource Area (MLRA) "
ecore 'ons: the Sierra Nevada ecore-
gion
e-
gion and the California portion of the
Eastern Cascade Slopes and Foothills s
ecoregion (USDA-NRCS 2006). The
Sierra Nevada ecoregion extends from
just south of Lassen Peak to the
Tehachapi Pass near Bakersfield. The
majority of this ecoregion is comprised
of elevations ranging •r:.
g from 450 to
2750 m, with the highest peak being
Mount Whitney419 m .The California
Pacific
Y( )
a a
r ,
portion of the Eastern Cascade Slopes Ocean
and Foothills ecoregion represents the
southernmost extent of the Cascade
Mountain Range, extending from the
A
Central Cascade Mountains to the Sierra
Nevada. The majority of this ecoregion
is comprised of elevations ranging from
450 to 2500 m, with the highest peak 0 25 50 100 miles "
being Mount Shasta (4318 m). ,
Sierra Nevada red fox survey data
were collected from the area surrounding Fig. 1. Vulpes vulpes necator.Model prediction area for the Sierra Nevada red
Lassen Peak (3187 m), the southernmost fox relative to its historic range (Grinnell et al. 1937) and the Lassen Peak
peak in the Cascade Range. The Lassen region.Note that scale bar is in miles(1 mile=ca.1.6 km)
182 Endang Species Res 14: 179-191,2011
Five Sierra Nevada red foxes were captured and the same location can introduce conflicting information
fitted with VHF collars and tracked by aerial and in the model, which may lower its predictive power. To
ground-based telemetry from 1998 through 2002 (Per- remove pseudoreplicates and correct for conflicting in-
rine 2005).All capture and handling activities were in formation,we used ArcGIS(ESRI)to label a location as a
accordance with California Department of Fish and presence if a fox was detected there at any time during
Game and University of California Berkeley protocols. the 10 yr period. Similarly, we deleted duplicate de-
The field team collected 586 independent ground tections at the same location.
telemetry locations using a Trimble GeoExplorer II After pooling these data, the sampling intensity var-
GPS and LOCATE Il telemetry software (Nams 2001), ied between habitat types. Since survey data were
and aerial telemetry provided 123 additional locations. plentiful(-2000 records),we randomly subsampled the
In addition, a total of 227 Sierra Nevada red fox scats data from each environmental zone to balance the sam-
were collected opportunistically from June 1998 pling intensity (Guisan&Zimmermann 2000,Araujo&
through December 2002, primarily in association with Guisan 2006).The environmental zones were based on
ground telemetry. The telemetry and scat locations a combination of 10 elevation zones and 19 California
were clustered in the western half of LVNP and the Wildlife Habitat Relationship (CWHR; Mayer & Lau-
adjacent LNF lands(Perrine 2005). denslayer 1988) types. This subsampling reduced the
The opportunistic camera station survey was con- dataset from 2000 to 1200 points (600 per presence and
ducted between 1992 and 2002 and consisted of 968 absence, respectively) and reduced but did not elimi-
baited TrailMaster (Goodson and Associates) camera nate sample clustering. To reduce clustering bias, we
stations throughout LVNP and the LNF(Perrine 2005). used Thiessen polygons (Rhynsburger 1973) to down-
This survey was conducted primarily by Park and For- weight points that occurred close together.
est Service biologists following the standard protocol Environmental predictor variables.We based our envi-
for surveying forest carnivores (Zielinski & Kucera ronmental predictor variables on Sierra Nevada red fox
1995). Although the cameras were widely distributed ecology and the availability of digital data. The Sierra
throughout the region, sampling biases arose due to Nevada red fox is associated with high-elevation
the opportunistic nature of this survey. For example, conifer forests, subalpine woodlands, talus slopes, and
the southwest portion of LVNP and the LNF east of the barren areas above treeline (Grinnell et al. 1937,
represent
r 'n 2005 . To e
White 1977 Perrine Caribou Wilderness were heavily sampled, whereas Schempf & ) p
the northern portion of the region had the least sam- these environmental conditions, we used a variety of
pling effort. Samples were also biased toward roads. GIS layers containing vegetation, climate, hydrology,
This survey yielded 50 Sierra Nevada red fox detection and forest structure data.Specifically,to represent veg-
locations,with multiple detections at some locations. etation and forest stand structure,we used CalFire Fire
The stratified random camera survey was conducted Resource and Assessment Program's Multi-Source
and 2002.This survey consisted Land Cover Data MSLC),which contained CWHR veg-
in the summers of 2001 a (
Y
of 24 sites stratified by elevation and randomly placed etation type, total tree canopy closure, tree size class,
site con- and tree density class attributes www.fra .cdf.ca. ov).
throughout the Lassen Peak region. Each s ( P 9
9 g Y
tained 2 baited TrailMaster cameras approximately We also derived Tasseled-cap greenness and wetness
1.6 km apart,following the standard protocol for forest from Landsat 5 imagery as an additional variable
P 9
carnivores (Zielinski & Kucera 1995). This survey (software: ERDAS 2008 Leica geosystems geospatial
yielded 3 red fox detection locations (Perrine 2005). imaging, Atlanta, GA). Tasseled-cap transformation
We combined and subsampled these occurrence data variables represent a continuous environmental gradi-
to use as the species response variable.Combining these ent that is highly correlated with stand age and struc-
I data reduced the effect of the sampling biases inherent tural complexity(Hansen et al. 2001).Pixels containing
in each method,but new errors arose as a result of com- high greenness and wetness values are associated with
bining 4 different collection methods that spanned dense vegetation having high leaf area index, while
multiple years. Some locations contained 2 presence lower values indicate sparsely vegetated areas such as
points,because a fox was detected in the same area by 2 barren areas or regions of snow and ice (White et al.
different methods. These pseudoreplicates violate the 1997,Waring&Running 1998).We used Spatial Analyst
assumption that training data points are independent, (ESRI)to calculate the Euclidean(straight-line)distance
which in turn can bias model results(Guisan&Zimmer- from the center of each raster cell to the nearest water
mann 2000).Additionally,false absences likely occurred feature from the National Hydrography dataset
at some camera locations due to the elusiveness of this (http://nhd.usgs.gov) and to derive slope from a 30 m
species. Within the 10 yr sampling period, several digital elevation model (USGS 2000). The Sierra
camera locations detected foxes in one year but not an- Nevada red fox's elevational limits,relationship to snow
other. Having detection and non-detection records at pack, and phylogeography (Aubry et al. 2009) suggest
I
Cleve et al.:Predicting potential Sierra Nevada red fox habitat 183
an affinity for specific climatic conditions.We used grid- entropy method(Maxent) with full region background
ded climate data derived from the Parameter-elevation pixels(hereafter,MFB),a Maxent model using a subset
Regressions on Independent Slopes Model (PRISM) to of background pixels with similar biases as the occur-
predict the species'response to different climatic condi- rence data(MSB), and a spatially-weighted presence-
tions; specifically,we used mean monthly precipitation absence logistic regression model (LRW).We used the
and monthly average daily minimum and maximum default parameters for both Maxent models and gener-
temperatures from 1971 through 2000(Daly et al. 1994). ated outputs in the logistic regression format.
After selecting the initial environmental variables For each approach, we developed the model with a
based on red fox ecology, we used the R statistical random subset of 70% of the data and withheld the
package (R Development Core Team 2005) to deter- remaining 30%,for model evaluation.To determine the
mine correlations between continuous variables and to classification accuracy of each model, we used the
identifyinteraction terms.
s. If 2 variables were come- evaluation data to identify the optimum cutoff value
lated (Pearson's correlation coefficient > 0.3), only the that corresponded with high red fox habitat suitability.
variable with the lower p value was retained. We then Optimum cutoff values were determined by calculat-
identified pairwise interaction terms using a classifica- ing the true skill statistic (TSS) across the entire range
tion and regression tree(Miller&Franklin 2002).Clas- of potential cutoff values,and the cutoff value that cor-
sification and regression trees determine a set of if- responded with the highest TSS was selected as the
then statements that define class membership,and can optimum cutoff(Allouche et al.2006,Jones et al.2010).
express complex non-linear and non-additive relation- We then calculated the AUC to assess predictive
ships among the predictor variables(Miller&Franklin performance(Buckland et al. 1997).
2002). We included 9 environmental variables in the Predictive accuracy could only be tested in the
classification and regression tree model: CWHR type, Lassen Peak region because of the limited geographic
total tree canopy closure, tree size class, tree density extent of the available data.To compare model perfor-
class, slope, February precipitation, minimum Decem- mance outside the Lassen Peak region, we compared
i ber temperature, Tassled-ca
p greenness, and dis
tance the distribution and
abundance ofe each model'ssu suit
-
to to water. able habitat to the other models and to the historic
After removing correlated predictor variables and range map for the Sierra Nevada red fox(Grinnell et al.
determining interactions, we selected significant pre- 1937).We used the optimum cutoff value to determine
dictor,variables using iterative manual stepwise logis- the appropriate suitable habitat threshold. To create
tic regression;at each run,the least significant variable the ensemble prediction,the values of each model that
was removed until only significant variables remained fell below the optimum cutoff value were given a value
(Hastie &Pregibon 1992, Hosmer & Lemeshow 2000). of 0.0. The mean probability value from both models
In addition, at each step the Akaike information crite- was then assigned to each cell of the study area.
rion (AIC) was used to select the best fitting model.
AIC is a standardized score used to compare models
for best fit relative to the number of parameters in the RESULTS
model;lower AIC values indicate better fit(Burnham&
Anderson 2002). The 9 environmental variables listed The classification and regression tree identified 2
in the previous paragraph, along with the 2 pairwise significant pairwise interaction terms: February pre-
interaction terms identified in the classification and cipitation with minimum December temperature, and
regression tree model (see 'Results'),were included in February precipitation with image greenness. The
the stepwise logistic regression weighted by Thiessen stepwise logistic regression weighted by Thiessen
polygon area.The resulting set of significant predictor polygon area reduced our environmental variable set
variables was used in both the Maxent and logistic to the following 6 variables: February precipitation,
regression models.This allowed for direct comparison minimum December temperature, Tasseled-cap
between the 2 modeling approaches. This 2-step greenness, distance to water, the interaction between
method of using a GLM to select predictor variables February precipitation and minimum December tem-
followed by Maxent modeling has been shown to cre- perature, and the interaction between February pre-
ate predictions with very high area under the receiver cipitation and greenness. We used these 6 variables to
operating characteristic (ROC) curve (AUC) values generate the species distribution models via the 3
(Wollan et al. 2008). High AUC values indicate low approaches described in the above section 'Distribu-
error, while lower values indicate lower predictability tion models and model evaluation'.
(Pearce&Ferrier 2000). Within the Lassen Peak region, all models had good
Distribution models and model evaluation.We gener- fit to the test data, as indicated by high values for TSS,
ated 3 distribution models: a presence-only maximum percent correctly classified,and AUC(Table 1).For all 3
184 Endang Species Res 14: 179-191,2011
Table 1. Summary of model results for all 3 models of suitable habitat for Sierra Nevada red fox Vulpes vulpes necator.MFB:
Maxent with full background pixels;MSB:Maxent with subset background pixels;LRW:spatially-weighted logistic regression;
AUC:area under the curve;TSS:true skill statistic;LPR SH:suitable habitat(ha)for Lassen Peak Region;SA SH:suitable habitat
(ha)for entire study area;SA-LPR SH:suitable habitat(ha)outside the Lassen Peak Region.Suitable habitat is defined as the area
that contains a probability of red fox occurrence equal to or greater than the optimum cutoff value.Optimum cutoff values were
determined by 4 calculatin the TSS across the entire range of potential cutoff values,and the cutoff value that corresponded with
the highest TSS was selected as the optimum cutoff
it
Model Correctly classified(%) AUC TSS(%) Optimum cutoff LPR SH(ha) SA SH(ha) SA-LPR SH(ha)
MFB 91.5 0.9579 83.1 0.333 92187 660 479 568 292
MSB 90.8 0.9537 81.9 0.157 79 754 935 362 855 608
LRW 86.0 0.9438 75.8 0.184 115530 1546021 1430491
accuracy assessment statistics, the MFB model yielded sified outside of the Lassen Peak region (Table 1),
the highest values, with the MSB model slightly lower, where the MSB and MFB models predicted approxi-
followed by the LRW model.All 3 models had low opti- mately 40 to 60% less suitable habitat than the LRW
mum cutoff values, ranging from 0.157 for the MSB model. All 3 models approximated the historic range
model to 0.333 for the MFB model(Table 1). boundary for the Sierra Nevada red fox,with the LRW
Despite their similar accuracy,the 3 models varied in model yielding the closest fit and the MFB model
the location and extent of the suitable habitat area they having the sparsest fit(Fig.2).
predicted (Table 1, Fig. 2). Within the Lassen Peak The Maxent models (MFB and MSB) selected
region, the MFB and MSB models predicted approxi- regions with lower minimum December temperatures
mately 20 and 30% less suitable habitat, respectively, and higher February precipitation than the entire
than the LRW model(Table 1).This pattern was inten- study region (Table 2). The interaction between
MFB MSB _ LRW
Mount' Mount,"' Mount Suitability
Shast , Suitability Sfaasta� ''•.,_.,.�; Suitability Shjsta '......:
1 `��;•` High High .. * High
K
LOW LOW LOW
Lassen -- Lassen''. '. _ Lassen
Peak '� California Peak Q California Peak Z. ., California
>
Re ion boundary Re ion . boundary Region ,. .� boundary
Z
g _ � g
r1 Approximate Approximate �� 4 _!Ap IXocmate
• histonc
-."-"'historic '' .
range N; range { range
C, '1r C,
O Or 0-
1>
4> d
-lt_L,
Sequoia and Kings ✓� Sequoia and Kings Sequoia and Kings
Canyon National Canyon National\A Canyon National
N Parks Region N Parks Region :tea N Parks Region
A A A
0 50 100 200 0 50 100 200�rn 0 50 100 200 m
Fig.2. Vulpes vulpes necator.Predicted suitable habitat for Sierra Nevada red fox based on 3 models:Maxent full background
pixels (MFB), Maxent with subsetted background pixels(MSB),and spatially-weighted logistic regression (LRW).The dashed
line represents the historic range(Grinnell et al.1937).Figs.3 to 5 display in detail the 3 geographic regions identified in this map
Cleve et al.:Predicting potential Sierra Nevada red fox habitat 185
Table 2.Range,mean,and SD of environmental variables(minimum temperature in December,precipitation in February,greenness,
and distance to nearest body of water)for the entire study area and for suitable habitat areas for Sierra Nevada red fox Vulpes vulpes
necator as predicted by each of the 3 models.MFB:Maxent with full background pixels;MSB:Maxent with subset background pixels;
LRW:spatially-weighted logistic regression
Environmental — Study area MFB MSB LRW
variables(units) Min/Max Mean t SD Min/Max Mean t SD Min/Max Mean t SD Min/Max Mean t SD
MIN.TEMP. -15.3/5.2 -3.9 t 3.4 -10.4/-2.4 -5.3 t 1.0 -15.3/-3.6 -6.8 t 2.3 -15.3/5.3 -7.4 t 2.6
DEC(°C)
FEBTRECIP 26.8/481.6 183.3 t 80.5 137.7/471.3 240.7 t 64.0 147.7/471.3 239.3 t 52.8 34.3/471.3 223.9 t 57.6
(mm)
GREENNESS -2266/1205 52 t 247 -1906/1030 39 t 217 -2134/881 -125 t 375 -2134/896 -168.6 t 356.3
(Derived)
DIST.TO WATER(m) 0/23350 3912 t 3402 0/10331 2051 t 1483 0/10657 2038 t 1441 0/13829.3 2715.6 t 2110.2
December minimum temperature and February pre- spectral data that are not captured in predetermined
cipitation accounted for 50% of the predictor variable classification schemes. In our study, image greenness
contribution to the Maxent models(Table 3).The LRW emerged as a predictor variable over 3 predetermined
model also included this interaction and predicted classification schemes:tree density,tree size class,and
areas with a lower minimum temperature than the CWHR category. This suggests that the vegetation
study region as a whole. But unlike the Maxent mod- associations of the Sierra Nevada red fox are better
els, the LRW model had a February winter precipita- characterized by the unclassified spectral data than by
tion range similar to that of the entire study area, these predetermined classification schemes.This find-
which included areas that received lower precipitation ing is significant because few studies leverage the ben-
(Table 2).The remaining environmental variable sum- efits of remotely-sensed data in their species distribu-
mary statistics were comparable across the 3 models: lion modeling(Turner et al.2003, Gillespie et al.2008).
suitable habitat was within 2 to 3 km of a water feature, In addition, predetermined classification schemes are
and greenness values were lower than the study area often not available, are time consuming and expensive
mean'(Table 2). These variables represented 12 and to produce, and are often inconsistent between
2%,respectively,of the predictor variable contribution regions. Not having to rely on vegetation maps for
in the Maxent models(Table 3). habitat modeling can greatly reduce needed resources
because satellite imagery is often free and readily
available.Additionally, using satellite imagery instead
DISCUSSION of predetermined classification schemes creates a
more parsimonious model by decreasing the number of
We used presence-absence (logistic regression) and variables needed.
presence-only (maximum entropy) methods to create Both the logistic regression and maximum entropy
the first spatially explicit habitat suitability model,based modeling approaches indicated that climate was a
on climatic and remotely sensed variables,for the Sierra major component of habitat suitability for the Sierra
Nevada red fox,a threatened species under the Califor- Nevada red fox. Recent phylogenetic analyses have
nia Endangered Species Act. The resulting model pre- indicated that the Sierra Nevada red fox and its con-
dictions can guide future surveys to locate additional specifics in the Cascade and Rocky Mountains (Vulpes
populations of this rare subspecies, and also illustrate vulpes cascadensis and V. v. macroura, respectively)
the relative performance of these modeling approaches. comprise a distinct genetic lineage separate from much
Our analysis is one of the first empirical tests of the of the rest of North America (Aubry et al. 2009). This
revised algorithm in Maxent to reduce the effect of geo- lineage was more widespread during the height of the
graphically biased occurrence data and improve its abil- Pleistocene glaciation but retracted to the high eleva-
ity to transfer to new study regions(Phillips 2008). tions of the western mountains when the glaciers
retreated. These historic range expansions and con-
tractions coincide with regional climate change, indi-
Predictor variables cating that the Sierra Nevada red fox may be physio-
logically and ecologically constrained to subalpine
Our findings support the conclusion by Laurent et al. climate zones (Aubry et al. 2009).
(2005) that wildlife may have a response to the contin- Both modeling approaches identified areas with low
uous environmental gradient present in unclassified winter minimum temperatures, but the models varied
186 Endang Species Res 14: 179-191,2011
Table 3.Variable coefficients for logistic regression(LR)model and percent con- the species`entire range and at a more
tributions for Maxent models.See Table 2 for definitions of variables and units. localized scale of the original popula-
na:not applicable tion centers of the Sierra Nevada red
fox as described by Grinnell et al.
Variable LR: Maxent: (1937). Although all 3 models pre-
coefficient %contribution dicted high habitat suitability through-
Constant 0.3067 na out the historic range of the Sierra
FEB.PRECIP -0.0001483 30.4 Nevada red fox, the MSB model pre-
MIN.TEMP.DEC 0.008398 5.3 dicted slightly more habitat than the
GREENNESS -0.00483 2.0 MFB model, and the LRW model pre-
DIST.TO WATER -0.0003856 11.9 dicted far more suitable habitat than
Interaction:FEB.PRECIP x MIN.TEMP.DEC -0.000000636 50.3
Interaction:FEB.PRECIP x GREENNESS 0.000000144 0.1 either Maxent model. For example,
north of Lassen Park, the LRW model
predicted all of Mount Shasta to be
in how they represented the effect of precipitation.The suitable habitat, whereas the MSB model selected
Maxent models predicted areas with winter precipita- Shasta's eastern slope and a small portion of its peak
tion well above the regional minimum, whereas the and western slope, and the MFB model selected only
LRW model predicted areas with lower winter precipi- its eastern slope (Fig. 4). The discrepancy became
tation (Table 2). This likely accounted for much of the more pronounced with increased distance from Lassen
spatial difference between the model predictions. Peak,the location of the occurrence data.In the south-
Within the Lassen Peak region, for example, the LRW ernmost historic population center, the Sequoia and
model predicted more suitable habitat than the Max- Kings Canyon National Parks region of the southern
ent models, and predicted more habitat in the eastern Sierra Nevada, the LRW identified most of the region
portion of the area (Fig. 3). Field surveys occasionally as suitable habitat (196 855 ha; Fig. 5). In contrast, the
detected red fox east of Lassen Peak, but far less fre- MFB predicted only a small amount of suitable habitat
quently than in the western portion of the area region, (23 234 ha), and the MSB model predicted an interme-
where the projections of the Maxent models were con- diate amount (84496 ha). The pattern of Maxent and
centrated(Perrine 2005). logistic regression models yielding similar AUC values
Outside of the Lassen Peak region,the differences in but predicting slightly different suitable habitat areas
prediction area were exacerbated, both on the scale of is consistent with prior findings(Gibson et al. 2007).
MFB MSB LRW
Suitability Test data Suitability Test data Suitability Test data
High ♦ Absence High ♦ Absence High ♦ Absence
4. O Presence ,4 O Presence a. O Presence
Low ,Low I Low
• Lassen Peak Lassen Peak Q Lassen Peak
Region boundary Region boundary Region boundary
.sv �a :.', • 7
• 0.
A A-
t. .. , t• ti. �• ••
A.
A Ali 4-
b� bpi b
N e N N>a
. + _, .
r
A ,.A•A.� A A
�"• `" 0 15 30 km " 0 15 30 km • �, 0 15 30 km
Fig.3. Vulpes vulpes necator.Predicted suitable habitat for Sierra Nevada red fox in the Lassen Peak region,relative to presence-
absence test data. MFB: Maxent with full background pixels; MSB: Maxent with subset background pixels; LRW: spatially-
weighted logistic regression.See Fig.2 for the location of the Lassen Peak region in California
Cleve et al.:Predicting potential Sierra Nevada red fox habitat 187
MFB MSB LRW
F
Suitability Suitability Suitability
High a ,High High
n
Low �i;.�fy,� � �r Low r Low
� •s
3r�
r,
Fig.4. Vulpes vulpes necator.Predicted suitable habitat for Sierra Nevada red fox in the Mount Shasta region of northern Califor-
nia. MFB:Maxent with full background pixels; MSB: Maxent with subset background pixels; LRW:spatially-weighted logistic
regression.See Fig.2 for the location of Mount Shasta in California
MFB MSB LRW
Suitability " Suitability Suitability
,High " ,High ,High
Low ?4 ##y,, Low T Low
IL f
r
f"
N N N ..
A A A
0 10 20 0 10 20 k 0 10 20 fi
Fig.5. Vulpes vulpes necator.Predicted suitable habitat for Sierra Nevada red fox in the Sequoia and Kings Canyon National Parks
region in the southern Sierra.MFB:Maxent with full background pixels;MSB:Maxent with subset background pixels;LRW:spa-
tially-weighted logistic regression.See Fig.2 for the location of the Sequoia and Kings Canyon National Parks region in California
Because evaluation data were available only within in the occurrence data, yielded a prediction area and
the Lassen Peak region, our assessment of model per- extent that was intermediate between the MFB model
formance beyond this region was limited to compar- and the LRW model (Fig. 2).Although the MFB model
isons between models and with the species' historic yielded high accuracies within the occurrence data re-
range (Grinnell et al. 1937). The MSB model, which gion, it predicted less suitable habitat in areas farther
used subsetted background pixels to match the biases away from the occurrence data than the MSB model.
188 Endang Species Res 14: 179-191,2011
I
These results are consistent with general expectations results indicate that while subletting significantly
for geographically biased presence data (Phillips reduces the effects of geographically biased presence
2008). The MSB model, which used a subset of back- data in Maxent, it does not completely resolve the
ground pixels,greatly reduced the effect of geographic problem. Despite the risk of false absences, the LRW
bias.In contrast to the Maxent models,the LRW model model did quite well predicting both within and out-
predicted abundant suitable habitat outside of the side of the Lassen Peak region,leading us to conclude
Lassen region, confirming prior findings that logistic that in this case the potentially false absences were less
regression models can transfer well to similar study problematic than the transferability issues inherent to
areas(Randin et al.2006,Barbosa et al.2009). the Maxent models.
Our range predictions for the Sierra Nevada red fox,
like all species distribution models,are hypotheses that
should be tested by the collection of new data, espe- Management implications
cially in predicted areas that were previously unsam-
pled (Olivier & Wotherspoon 2006, Parra & Monahan As a carnivore closely associated with montane tree-
2008, Wollan et al. 2008, Costa et al. 2009). The line habitats,the Sierra Nevada red fox may be acutely
increased use of automatic camera stations for carni- affected by climate changes. Since 1920, California's
vore inventory and monitoring on public lands average annual temperature has warmed by 1.7°C,
throughout the Sierra Nevada holds great promise for with greater warming occurring with daily minimum
additional detections of Sierra Nevada red fox. Fortu- temperatures and at higher elevations (Kapnick&Hall
itously, in autumn 2010 Forest Service biologists 2009). In the central Sierra Nevada, December mini-
obtained photographs of several putative Sierra mum temperatures have increased by 0.23°C decade-'
Nevada red foxes in the Sonora Pass region of the in Yosemite Valley (1220 m elevation) and 0.32°C
northern Sierra Nevada, approximately 75 km south decade' at Lake Tahoe (1900 m; Thorne et al. 2006).
v documented
Nevada have n surveysin the Sierra
. Rich and S. Lisius pers. com
m.). Recent of Lake Tahoe A
( P )
Although final confirmation via molecular genetic range shifts by small mammals (Moritz et al. 2008),
methods is still underway, preliminary analyses and birds (Tingley et al. 2009), butterflies (Forister et al.
the fact that several foxes were detected in close prox- 2010),and conifers(Thorne et al.2006),consistent with
imity indicate that these animals represent the first ver- responses to climate warming. Despite the recent
ified population of Sierra Nevada red fox detected out detection of at least 1 Sierra Nevada red fox in the
of the Lassen area in nearly 2 decades (B. Sacks pers. northern Sierra, the lack of recent documented detec-
comm.). These detections lend further support to the tions in the southern extent of its historic range (Per-
logistic regression model over the 2 Maxent models. rine et al. 2010) may indicate that its range has
The logistic regression identified these detection sites retracted northward in response to climate change.It is
as high quality habitat for Sierra Nevada red fox; the unclear whether climate has a direct or indirect impact,
subalpine woodland is virtually identical structurally to such as facilitating coyotes Canis latrans or other com-
occupied Sierra Nevada red fox habitat in the Lassen petitors or changing understory structure through
region, despite being composed of whitebark pine altered fire regimes.
Pinus albicaulus as opposed to mountain hemlock in If such range retraction has already occurred, the
the Lassen region. The Maxent model with subsetted Sierra Nevada red fox may have little future in Califor-
background pixels identified the Sonora Pass site as nia. Climate warming is expected to continue if not
medium-quality habitat for Sierra Nevada red fox, accelerate in the coming century,although forecasts of
whereas the original Maxent algorithm did not identify the amount and rate of change depend greatly on the
it as potential habitat at all. specific climate model and emissions scenario used
The ability to subset background pixels is relatively (Hayhoe et al. 2004). Nevertheless, montane regions
new to Maxent (Phillips 2008), and few studies to date are likely to experience the greatest warming (Snyder
have used or evaluated these methods (Anderson & et al. 2002), with mean annual temperatures in the
Raza 2010). Our findings support the conclusion of Sierra Nevada and southern Cascades projected to rise
Anderson & Raza (2010) that calibrating the study 3.0 to 3.5°C by 2070 to 2099(Ackerly et al.2010).These
region or subsetting background pixels to the occur- changes will likely have a profound effect on the
rence data area yields a Maxent model with a larger ranges, elevations, and associations of California's
predicted area that is less concentrated around the biota (e.g. Loarie et al. 2008, Parra & Monahan 2008,
occurrence data region. Our study emphasizes the Wiens et al. 2009, Ackerly et al. 2010, Forister et al.
importance of subsetting background pixels in Maxent 2010) The Sierra Nevada red fox may follow the
when using geographically biased presence data, pattern of another alpine-associated carnivore, the
which are common with rare species. Moreover, our wolverine Gulo gulo,which disappeared from Califor-
Cleve et al.:Predicting potential Sierra Nevada red fox habitat 189
nia by the 1930s, with populations persisting only in include the Mount Shasta region(Fig. 4), which is the
higher latitude states such as Washington, Montana, largest predicted area north of the extant Lassen pop-
and Idaho (Aubry et al. 2007). Predicting the future ulation, and the Sequoia and Kings Canyon region,
range of Sierra Nevada red fox based upon its historic which is the southernmost of the historic population
range and the anticipated climate changes is an impor- centers (Grinnell et al. 1937) and an area of profound
tant next step toward its effective conservation. model discrepancy. Such surveys should incorporate
Locating additional Sierra Nevada red fox popula- the collection of specimens for genetic analysis
tions outside of the Lassen Peak area is critically because photographs alone cannot conclusively iden-
important (Perrine et al. 2010) to better document the tify Sierra Nevada red foxes (Perrine et al. 2010).
taxon's true spatial extent and to acquire additional Additionally, the Lassen population should be closely
specimens to refine its phylogenetic relationships with monitored for changes in size or extent, as it remains
other red fox populations (e.g., Perrine et al. 2007, the only known actively reproducing population of
Aubry et al. 2009, Sacks et al. 2010). One of our pri- this endangered taxon (Perrine et al. 2010).If no other
mary goals for these analyses was to provide guid- reproductive populations exist outside of the Lassen
ance for range-wide field surveys. To test our model region, then the Sierra Nevada red fox likely warrants
predictions, field surveys should target areas of model a higher level of state or federal protection and active
discrepancy,such the Sequoia and Kings Canyon area management than it currently receives.
of the southern Sierra Nevada (Fig. 5). However,
surveys attempting to locate additional populations
as efficiently as possible should instead target areas LITERATURE CITED
of model agreement, as indicated by our 2-model s Ackerly DD,Loarie SR,Cornwell WK,Weiss SB,Hamilton H,
ensemble (Fig. 6). The highest survey priority should Branciforte R,Kraft NJB (2010)The geography of climate
be to determine the extent of the newly discovered change:implications for conservation biology.Divers Dis-
Sonora Pass population.Additional high priority areas trib 16:476-487
)j. Allouche O, Tsoar A, Kadmon R (2006) Assessing the accu-
racy of species distribution models:prevalence,kappa and
_-..__.._ the true skill statistic(TSS).J Appl Ecol 43:1223-1232
Suitability )► Anderson RP, Raza A (2010) The effect of the extent of the
(Ensemble model) study region on GIS models of species geographic distrib-
utions and estimates of niche evolution: preliminary tests
'l 4 High:0.969074 with montane rodents (genus Nephelomys)in Venezuela.
F ! J Biogeogr 37:1378-1393
Low:0 > Aratijo MB,Guisan A(2006)Five(or so)challenges for species
distribution modelling.J Biogeogr 33:1677-1688
Califom ). Araujo MB, New M (2007) Ensemble forecasting of species
boundarydistributions.Trends Ecol Evol 22:42-47
Approximate � Aubry KB, McKelvrey KS, Copeland JP (2007) Distribution
1 historic range and broadscale habitat relations of the wolverine in the
contiguous United States.J Wildl Manag 71:2147-2158
(` )o. Aubry KB, Statham MJ, Sacks BN, Perrine JD, Wisely SM
p (2009)Phylogeography of the North American red fox:vic-
ariance in Pleistocene forest ia. Mol Ecol 18:refu
g
"y 2668-2686
�� �► Barbosa AM, Raimundo R, Vargas JM (2009) Transferability
of environmental favourability models in geographic
space: the case of the Iberian desman (Galemys pyre-
Pacificnaicus)in Portugal and Spain.Ecol Model 220:747-754
)• Buckland ST,Burnham KP,Augustin NH(1997)Model selec-
tion: an integral part of inference. Biometrics 53:603-618
y Bulluck L, Fleishman E, Betrus C, Blair R (2006)Spatial and
N temporal variations in species occurrence rate affect the
accuracy of occurrence models. Glob Ecol Biogeogr 15:
27-38
1 Burnham DP,Anderson DR(2002)Model selection and multi-
0 50 100 km \ model inference: a practical information-theoretic
1 approach,2nd edn.Springer,New York,NY
i i ! CDFG (California Department of Fish and Game) (1996)The
status of rare, threatened and endangered animals and
Fig. 6. Vulpes vulpes necator. Ensemble habitat suitability plants of California. California Department of Fish and
model for Sierra Nevada red fox,based on spatially-weighted Game,Sacramento,CA
logistic regression (LRW) and Maxent with subset back- CDFG (2004) Special animals. California Department of Fish
ground(MSB)models and Game,Sacramento,CA
190 Endang Species Res 14: 179-191,2011
�. Costa GC, Nogueira C, Machado RB, Colli GR (2009) Sam- '► Loarie SR,Carter BE,Hayhoe K,McMahon S,Moe R,Knight
pling bias and the use of ecological niche modeling in con- CA, Ackerly DD (2008) Climate change and the future of
servation planning: a field evaluation in a biodiversity California's endemic flora.PLoS ONE 3:e2502
hotspot.Biodivers Consery 19:883-899 MacKenzie D,Nichols J,Royle JA,Pollock K,Bailey L,Hines
>. Daly CS, Neilson RP, Phillips DL (1994) A statistical-topo- J (2006) Occupancy estimation and modeling. Academic
graphic model for mapping climatological precipitation Press,London
over mountainous terrain.J Appl Meteorol 33:140-158 Mayer KE, Laudenslayer WF Jr (eds) (1988) A guide to
pr. Edwards TC,Cutler DR,Zimmermann NE,Geiser L,Alegria J wildlife habitats of California. California Department of
(2005) Model-based stratifications for enhancing the Fish and Game,Sacramento,CA
detection of rare ecological events.Ecology 86:1081-1090 D,. Menon S, Choudhury B, Khan ML, Peterson AT (2010)
)s. Elith J, Graham C, Anderson R, Dudik M and others (2006) Ecological niche modeling and local knowledge predict
Novel methods to improve prediction of species' distribu- new populations of Gymnocladus assamicus a critically
tions from occurrence data.Ecography 29:129-151 endangered tree species. Endang Species Res 11:
)t. Engler R, Guisan A, Rechsteiner L (2004) An improved 175-181
approach for predicting the distribution of rare and endan- y Miller J, Franklin J (2002) Modeling the distribution of four
gered species from occurrence and pseudo-absence data. vegetation alliances using generalized linear models and
J Appl Ecol 41:263-274 classification trees with spatial dependence. Ecol Model
�,. Ferreira de Siqueira M,Durigan G,de Marco Junior P,Peter- 157:227-247
son AT(2009) Something from nothing: using landscape y. Mladenoff D, Sickley T, Wydeven A (1999) Predicting gray
similarity and ecological niche modeling to find rare plant wolf landscape recolonization: logistic regression models
species.J Nat Consery 17:25-32 vs.new field data.Ecol Appl 9:37-44
Forister ML,McCall AC, Sanders NJ,Fordyce JA and others p. Moritz C, Patton JL, Conroy CJ, Parra JL, White GC,
(2010)Compounded effects of climate change and habitat Beissinger SR(2008)Impact of a century of climate change
alteration shift patterns of butterfly diversity. Proc Nall on small-mammal communities in Yosemite National Park,
Acad Sci USA 107:2088-2092 USA.Science 322:261-264
b. Gibson L,Barrett B,Burbidge A(2007)Dealing with uncertain Nams VO(2001)Locate II:user's guide.Pacer,Truro,NS
absences in habitat modeling: a case study of a rare Nielsen SE, Boyce MS, Stenhouse GB, Robin H, Munro M
ground-dwelling parrot.Divers Distrib 13:704-713 (2002) Modeling grizzly bear habitats in the Yellowhead
. Gillespie TW, Foody GM, Rocchini D, Giorgi AP, Saatchi S Ecosystem of Alberta: taking autocorrelation seriously.
(2008) Measuring and modeling biodiversity from space. Ursus 13:45-56
Prog Phys Geogr 32:203-221 fe. Olivier F,Wotherspoon SJ(2006)Distribution and abundance
Grinnell J,Dixon J,Linsdale J(1937)Fur-bearing mammals of of Wilson's storm petrels Oceanites oceanicus at two loca-
California.University of California Press,Berkeley,CA tions in East Antarctica: testing habitat selection models.
,y. Guisan A,Zimmermann NE(2000)Predictive habitat distrib- Polar Biol 29:878-892
ution models in ecology.Ecol Model 135:147-186 Parra JL,Monahan WB (2008)Variability in 20th century cli-
�. Guisan A, Broenniamann O, Engler R, Vust M, Yoccoz N, mate change reconstructions and its consequences for
Lehmann A, Zimmermann N (2006) Using niche-based predicting geographic responses of California mammals.
models to improve the sampling of rare species. Consery Glob Change Biol 14:2215-2231
Biol 20:501-5 11 Pearce J, Ferrier S (2000) Evaluating the predictive perfor-
Hall ER (1981) The mammals of North America, 2nd edn. mance of habitat models developed using logistic regres-
John Wiley&Sons,New York,NY sion.Ecol Model 133:225-245
Hansen MJ,Franklin SE,Woudsma C,Peterson M(2001)For- Perrine JD (2005) Ecology of red fox (Vulpes vuipes) in the
j est structure classification in the North Columbia moun- Lassen Peak region of California, USA. PhD dissertation,
tains using Landsat TM tasseled cap wetness component. University of California,Berkeley,CA
Can J Rem Sens 27:20-32 > Perrine JD, Pollinger J, Sacks B, Barrett R, Wayne R (2007)
Hastie TJ, Pregibon D (1992) Generalized linear models. In: Genetic evidence for the persistence of the critically
Chambers JM, Hastie TJ (eds) Statistical models in S. endangered Sierra Nevada red fox in California.Consery
Wadsworth&Brooks/Cole,Pacific Grove, CA,p 195-248 Genet 8:1083-1095
�. Hayhoe K,Cayan D,Field CB,Frumhoff PC and others(2004) Perrine JD, Campbell LA, Green GA (2010) Sierra Nevada
Emissions pathways,climate change,and impacts on Cal- red fox (Vulpes vuipes necator): a conservation assess-
ifornia.Proc Nall Acad Sci USA 101:12422-12427 ment.R5-FR-010.USDA Forest Service,Vallejo,CA
Hosmer DW,Lemeshow S(2000)Applied logistic regression. Peterson AT, Papes M, Eaton M (2007) Transferability and
John Wiley&Sons,New York,NY model evaluation in ecological niche modeling:a compar-
Johnson CJ, Seip DR, Boyce MS (2004) A quantitative ison of GARP and Maxent.Ecography 30:550-560
approach to conservation planning: using resource selec- Phillips SJ (2008) Transferability, sample selection bias and
tion functions to map the distribution of mountain caribou background data in presence-only modelling: a response
at multiple spatial scales.J Appl Ecol 41:238-251 to Peterson et al. (2007).Ecography 31:272-278
>. Jones CC, Acker SA, Halpern CB (2010) Combining local- ar. Phillips SJ, Anderson RP, Schapire RE (2006) Maximum
and large-scale models to predict the distributions of inva- entropy modeling of species geographic distributions.Ecol
sive plant species.Ecol App120:311-325 Model 190:231-259
Kapnick S. Hall A (2009) Observed changes in the Sierra Posillico M, Meriggi A, Pagnin E, Lovari S, Russo L(2004)A
Nevada snowpack:potential causes and concerns.CEC-500- habitat model for brown bear conservation and land use
2009-016-F,California Energy Commission,Sacramento,CA planning in the central Apennines. Biol Consery 118:
Laurent E, Shi H, Gatziolis D, LeBouton J, Walters M, Liu J 141-150
(2005) Using spatial and spectral precision of satellite R Development Core Team(2005)R:a language and environ-
imagery to predict wildlife occurrence patterns. Remote ment for statistical computing.R Foundation for Statistical
Sens Environ 97:249-262 Computing,Vienna.Available
at www.R-project.org
9
Cleve et al.:Predicting potential Sierra Nevada red fox habitat 191
y. Rachlow JL, Svancara LK (2006) Prioritizing habitat for y Tingley MW, Monahan WB, Beissinger SR, Moritz C (2009)
surveys of an uncommon mammal:a modeling approach Birds track their Grinnellian niche through a century of
applied to pygmy rabbits.J Mammal 87:827-833 climate change. Proc Natl Acad Sci USA 106 Suppl 2:
10. Randin CF, Dirnbo T, Dullinger S, Zimmermann NE, Zappa 19637-19643
M, Guisan A (2006) Are niche-based species distribution y. Turner W, Spector S, Gardiner N, Fladeland M, Sterling E,
models transferable in space?J Biogeogr 33:1689-1703 Steininger M (2003) Remote sensing for biodiversity sci-
W Rebelo H,Jones G(2010)Ground validation of presence-only ence and conservation.Trends Ecol Evol 18:306-314
modelling with rare species:a case study on barbastelles USGS (US Geological Survey) (2000) 10 m digital elevation
Barbastella barbastellus (Chiroptera: Vespertilionidae). J model.Menlo Park,CA
Appl Ecol 47:410-420 USDA-NRCS (US Department of Agriculture-Natural Re-
y Rhynsburger D (1973)Analytic delineation of Thiessen poly- sources Conservation Service) (2006) Major Land Re-
gons.Geogr Anal 5:133-144 source Area (MLRA). Land resource regions and major
D Sacks BN, Statham MJ, Perrine JD, Wisely SM, Aubry KB land resource areas of the United States, the Caribbean,
(2010) North American montane red foxes: expansion, and the Pacific Basin. US Department of Agriculture
fragmentation, and the origin of the Sacramento Valley Handbook 296
red fox.Consery Genet 11:1523-1539 Waring RH,Running SW(1998)Forest ecosystems:analysis at
Schempf PF, White M (1977) Status of six furbearer popula- multiple scales,2nd edn.Academic Press, San Diego, CA
tions in the mountains of northern California. USDA For- y.White J,Running S,Nemani R,Keane R,Ryan K(1997)Mea-
est Service,California Region,San Francisco,CA surement and remote sensing of LAI in Rocky Mountain
)• Snyder MA, Bell JL, Sloan LC, Duffy PB, Govindasamy B montane ecosystems.Can J Res 27:1714-1727
(2002)Climate responses to a doubling of atmospheric car- )p. Wiens JA,Stralberg D,Johgsomjit D,Howell CA,Snyder MA
bon dioxide for a climatically vulnerable region.Geophys (2009) Niches,models, and climate change:assessing the
Res Lett 29:1514 doi:10.1029/2001GL014431 assumptions and uncertainties. Proc Natl Acad Sci USA
y,. St-Louis V, Pidgeon AM, Radeloff VC, Hawbaker TJ, Clay- 106(Suppl 2):19729-19736
ton MK(2006)High-resolution image texture as a predic- )p. Wollan AK, Bakkestuen V, Kauserud H, Gulden G,
for of bird species richness. Remote Sens Environ 105: Halvors
en sen R 2008( ) Modeling and predicting fungal dis-
299-312 tribution patterns using herbarium data. J Biogeogr 35:
y Stohlgren TJ, Ma P, Kumar S, Rocca M, Morisette JT, 2298-2310
Jarnevich CS,Benson N(2010)Ensemble habitat mapping Zielinski WJ,Kucera TE(1995)American marten,fisher,lynx,
of invasive plant species.Risk Anal 30:224-235 and wolverine: survey
methods for their detection. PSW-
Thorne JH,Kelsey R,Honig J,Morgan B(2006)The develop- GTR-157. USDA Forest Service, Pacific Southwest
ment of 70- ear-old Weislander vegetation type maps
Y Research Station
9 YP P ,Albany,CA
an assessment of landscape change in the central Sierra y Zielinski WJ, Truex R, Schlexer F, Campbell L, Carroll C
Nevada. CEC 500-2006-107, California Energy Commis- (2005) Historical and contemporary distributions of carni-
sion,PIER Energy-related Environmental Program,Sacra- vores in forests of the Sierra Nevada, California, USA. J
mento,CA Biogeogr 32:1385-1407
Editorial responsibility:Luigi Boitani, Submitted:November 8,2010,Accepted:March 24,2011
Rome,Italy Proofs received from author(s):July 23,2011
i
i
i
I