ArticlePDF Available

Assessing Bird and Bat Mortality at the Forward Energy Center

Authors:

Steven M. Grodsky

University of California, Davis

David Drake

University of Wisconsin–Madison

. Methodology Summary for Post-Construction Carcass monitoring

…

Example of study plot with searched areas in grey and the turbine in the center of the plot.

…

Distance to turbine histogram for birds.

…

. Proportion of bat and mouse carcasses not scavenged by season.

…

. Summary of bird and bat fatalities found at the Forward Energy Center during monitoring studies July 15, 2008 through May 31, 2010.

…

Figures - uploaded by Steven M. Grodsky

Content may be subject to copyright.

Content uploaded by Steven M. Grodsky

Content may be subject to copyright.

Assessing Bird and Bat Mortality at the Forward

Energy Center

Final Report

Prepared by:

Steven M. Grodsky

David Drake, Ph.D.

Department of Forest and Wildlife Ecology

University of Wisconsin-Madison

1630 Linden Dr.

Madison, WI 53706

for

Forward Energy LLC

August 2011

PSC REF#:152052

Public Service Commission of Wisconsin

RECEIVED: 08/19/11, 3:41:18 PM

Forward Energy Bird and Bat Monitoring Final Report

STUDY PARTICIPANTS

David Drake Project Supervisor

Steve Grodsky Project Manager

Chris Jennelle Statistician

Aaron Meilicke Field Technician

Amanda Lipinski Field Technician

Craig Tomaszewski Field Technician

Emily Loew Field Technician

Tom Prestby Field Technician

Anna Wilson Searcher

Ben Calderon Searcher

Brian Hillers Searcher

Brooke Bogdanske Searcher

Carl Schroeder Searcher

Dan Rambo Searcher

Jeff Jarocki Searcher

Lana Raffensperger Searcher

Linda Lyon Searcher

Michael Giedd Searcher

Ryan Rysewyk Searcher

Shaleen Guell Searcher

Tom Underwood Searcher

REPORT REFERENCE

Grodsky, S.M. and D. Drake. 2011. Bird and Bat Mortality at the Forward Energy Center

in Southeastern Wisconsin. Final report prepared for Forward Energy LLC, Chicago, IL.

ACKNOWLEDGEMENTS

We would like to thank Laura Miner of Forward Energy LLC for her assistance with project

management. Forward Energy LLC staff Ken Drake, Amy Pearson, and the Forward Energy

Center wind technicians provided exceptional on-site logistical support. The United States Fish

and Wildlife Service Horicon Refuge staff was extremely helpful during the project; they

provided assistance with searches and maintained a cleared control site on refuge property. We

thank the University of Wisconsin’s Department of Forest and Wildlife Ecology graduate

students and faculty for helpful support and consultation. The Wisconsin Department of Natural

Resources, USFWS, and the Public Service Commission of Wisconsin provided input during

protocol development and useful commentary while reviewing our reports. This study was

funded by Forward Energy LLC as part of the permitting requirements determined by the Public

Service Commission of Wisconsin. Wisconsin Focus on Energy provided additional funding.

Forward Energy Bird and Bat Monitoring Final Report

TABLE OF CONTENTS

EXECUTIVE SUMMARY ...................................................................................................1

1.0 INTRODUCTION .....................................................................................................4

1.1 STUDY OBJECTIVES..............................................................................................4

2.0 METHODOLOGY ....................................................................................................4

2.1 STUDY DESIGN.......................................................................................................5

2.2 STUDY PLOTS.........................................................................................................5

2.3 CARCASS SEARCHES............................................................................................6

2.4 SEARCH INTERVALS.............................................................................................7

2.5 SEARCHER EFFICIENCY.......................................................................................7

2.6 SCAVENGER REMOVAL RATE............................................................................8

2.7 FATALITY ESTIMATION.......................................................................................9

2.8 COVARIATE ORIGINS AND STATISTICAL ANALYSIS ...................................12

3.0 RESULTS..................................................................................................................13

3.1 CARCASS SEARCHES............................................................................................13

3.2 SEARCHER EFFICIENCY.......................................................................................14

3.3 SCAVENGER REMOVAL RATE............................................................................14

3.4 BIRD MORTALITY..................................................................................................14

3.4.1 Characteristics of Bird Mortality..........................................................................14

3.4.2 Bird Mortality Estimates.......................................................................................15

3.4.3 Distribution of Bird Mortality...............................................................................16

3.5 BAT MORTALITY ..................................................................................................16

3.5.1 Characteristics of Bat Mortality...........................................................................16

3.5.2 Bat Mortality Estimates.........................................................................................17

3.5.3 Distribution of Bat Mortality................................................................................18

3.6 COVARIATE ANALYSIS.......................................................................................18

3.6.1 Bat Mortality as a function of Weather and Bat Activity......................................18

4.0 DISCUSSION............................................................................................................19

4.1 BIRD MORTALITY..................................................................................................19

4.2 BAT MORTALITY...................................................................................................20

4.3 CONCLUSIONS........................................................................................................21

5.0 LITERATURE CITED..............................................................................................22

Forward Energy Bird and Bat Monitoring Final Report

6.0 TABLES AND FIGURES.........................................................................................25

TABLE 1. METHODOLOGY SUMMARY FOR POST-CONSTRUCTION

CARCASS MONITORING .......................................................................................................25

TABLE 2. SUMMARY OF BIRD AND BAT FATALITIES FOUND DURING SCHEDULED CARCASS

SEARCHES AT THE FORWARD ENERGY CENTER, JULY 15, 2008 TO MAY 31, 2010..............25

TABLE 3. RESULTS OF SEARCHER EFFICIENCY TRIALS FOR BIRDS AND BATS OVERALL AND

BY SEASON AT THE FORWARD ENERGY CENTER.................................................................26

TABLE 4. PROPORTION OF BAT AND MOUSE CARCASSES NOT SCAVENGED BY SEASON.......26

TABLE 5. PROPORTION OF BIRD CARCASSES NOT SCAVENGED BY SEASON. ........................26

TABLE 6. DESCRIPTIONS OF BIRD EXPERIMENTAL BIAS TRIALS USED FOR (CARCASS

REMOVAL [CRT] AND SEARCHER EFFICIENCY [SEEF]), INCLUDING SAMPLE SIZE, TYPE OF

CARCASSES USED, AVERAGE REMOVAL TIME AND SEARCHER EFFICIENCY ESTIMATES. NA

REFERS TO DATA THAT IS NOT AVAILABLE ..........................................................................27

TABLE 7. SUMMARY OF BIRD AND BAT FATALITIES FOUND AT THE FORWARD ENERGY

CENTER DURING MONITORING STUDIES JULY 15, 2008 THROUGH MAY 31, 2010...............28

TABLE 8. COUNTS AND PERCENTAGES FOR CARCASS CONDITIONS FOR BIRDS AND BATS AT

FORWARD ENERGY CENTER. ..............................................................................................29

TABLE 9. PERCENTAGES OF SEX AND AGE COMPOSITION OF A SUBSET (N = 48) OF BAT

CARCASSES FOUND DURING CARCASS SEARCHES (SEX AND AGE VERIFIED BY

GROSS NECROPSY)...............................................................................................................29

TABLE 10. BIRD FATALITY ESTIMATES (FATALITIES/TURBINE/SPRING AND FALL) FOR

FORWARD ENERGY CENTER USING THE JAIN ESTIMATOR...................................................30

TABLE 11. BIRD FATALITY ESTIMATES (FATALITIES/TURBINE/SPRING AND FALL) FOR

FORWARD ENERGY CENTER USING THE HUSO ESTIMATOR WITH CARCASS REMOVAL

TIMES FROM BLUE SKY GREEN FIELD.................................................................................30

TABLE 12. BIRD FATALITY ESTIMATES (FATALITIES/TURBINE/SPRING AND FALL) FOR

FORWARD ENERGY CENTER USING THE MODIFIED HUSO ESTIMATOR.................................31

TABLE 13. BIRD AND BAT MORTALITY BY SECTION AND TURBINE (NOT INCLUDING

INCIDENTALS). ....................................................................................................................31

TABLE 14. GENERAL COMPASS DIRECTION IN RELATION TO STUDY PLOT ORIENTATION

OF BIRD AND BAT CARCASSES FOUND AT THE FORWARD ENERGY CENTER, WISCONSIN,

2008-2010. EXCLUDES INCIDENTALS.................................................................................33

TABLE 15. BAT FATALITY ESTIMATES (FATALITIES/TURBINE/SPRING AND FALL) FOR

FORWARD ENERGY CENTER USING THE JAIN ESTIMATOR...................................................33

TABLE 16. BAT FATALITY ESTIMATES (FATALITIES/TURBINE/SPRING AND FALL) FOR

FORWARD ENERGY CENTER USING THE HUSO ESTIMATOR WITH CARCASS REMOVAL

TIMES FROM BLUE SKY GREEN FIELD.................................................................................34

TABLE 17. BAT FATALITY ESTIMATES (FATALITIES/TURBINE/SPRING AND FALL) FOR

FORWARD ENERGY CENTER USING THE MODIFIED HUSO ESTIMATOR.................................34

TABLE 18. DESCRIPTIONS OF PREDICTOR VARIABLES USED IN THE ANALYSES FOR

ASSOCIATIONS BETWEEN WEATHER CHARACTERISTICS AND BAT MORTALITY.....................35

TABLE 19. LINEAR CORRELATIONS BETWEEN BAT MORTALITY LEVELS AT FORWARD

ENERGY CENTER AND WEATHER VARIABLES DURING FALL 2008 AND FALL 2009. .............36

TABLE 20. DISTRIBUTION OF BAT FATALITIES BY SEASON FROM SEVEN DIFFERENT WIND

POWER PROJECTS.................................................................................................................36

Forward Energy Bird and Bat Monitoring Final Report

FIGURE 1. MAP OF FORWARD ENERGY CENTER AND ITS PROXIMITY TO HORICON MARSH

AND NEDA MINE..................................................................................................................37

FIGURE 2. EXAMPLE OF STUDY PLOT WITH SEARCHED AREAS IN GREY AND THE

TURBINE IN THE CENTER OF THE PLOT.................................................................................38

FIGURE 3. DISTANCE TO TURBINE HISTOGRAM FOR BIRDS.................................................39

FIGURE 4. DISTANCE TO TURBINE HISTOGRAM FOR BATS...................................................40

FIGURE 5. TEMPORAL VARIATION IN BIRD AND BAT FATALITIES DURING THE

FALL 2008-2009 STUDY PERIODS. EXCLUDES INCIDENTALS................................................41

FIGURE 6. TEMPORAL VARIATION IN BIRD AND BAT FATALITIES DURING THE

SPRING 2010-2011 STUDY PERIODS. EXCLUDES INCIDENTALS. ...........................................42

FIGURE 7. COMPARISON OF BAT FATALITY RATES TO OTHER PROJECTS IN THE

MIDWEST REGION. ..............................................................................................................43

Forward Energy Bird and Bat Monitoring Final Report

EXECUTIVE SUMMARY

The Forward Energy Center (Center) consists of 86 General Electric 1.5MW turbines in

southeastern Wisconsin. Each turbine has an 80-meter hub height and 120-meter rotor-tip height.

The turbines are located in agricultural land (corn/soybean rotations being the predominant crop

types) in southern Fond du Lac County and northern Dodge County. The Center is

approximately 5 km east of Horicon National Wildlife Refuge, 63 km west of Lake Michigan, 22

km north of Neda Mine State Natural Area, and 21 km south of Lake Winnebago.

Steve Grodsky and Dr. David Drake conducted a two year bird and bat mortality study. The

primary objectives of our study were to:

1) Assess bird and bat mortality at the Center,

2) Provide corrected mortality estimations for birds and bats using the most recent

statistical estimator, and

3) Correlate observed mortality rates with select weather variables, proximity to

Horicon Marsh and Neda Mine, turbine operating status, and bird and bat activity

at the Center.

Of the 86 wind turbines at the Center, 29 wind turbines (34%) were searched for dead birds and

bats during the study periods July 15 – November 15, 2008, July 15 – October 15, 2009, and

April 15 – May 31, 2009 and 2010.

A total of 122 bat fatalities was recorded during mortality searches. Of these bats, a majority of

the mortality was comprised of migratory tree-roosting bats including the eastern red bat

(Lasiurus borealis), hoary bat (Lasiurus cinereus), and silver-haired bat (Lasionycteris

noctivagens). Bat mortality was positively correlated with humidity, dew point, and bat activity,

and negatively correlated with power output, which is a proxy for wind speed. Searcher

efficiency and scavenger removal trials were conducted to quantify bias in our mortality

estimates, along with search interval and proportion of area searched.

A mortality estimate for bats was calculated using a modified version of the Huso estimator

(2010). Modifications were necessary because carcass removal data were limited to 5 days at the

Forward project. This modified estimator produced estimates of 26.2 bats/turbine/spring and fall

combined during the first year (90% ci: 20.55 to 31.85), 20.68 bats/turbine/spring and fall

combined during the second year (90% ci: 13.78 to 27.58), and a two-year average of 23.44

bats/turbine/spring and fall combined (90% ci: 17.16 to 29.72). The corresponding values in

bats/MW/spring and fall combined are 17.41 for the first year (90% ci: 13.02 to 21.87),13.85 for

the second year (90% ci: 9.3 to 18.5), and a two-year average of 15.63 (90% ci: 11.16 to 20.19).

Bat mortality consistently peaked during late August and early September.

The environmental consulting firm WEST, Inc. was hired by Forward Energy to calculate 2

additional mortality estimators. The first alternative estimator for bats was the Jain estimator.

Adjusted estimates for bat mortality using the Jain estimator were 33.47 bats/turbine/spring and

fall combined for the first year of study (90% ci: 24.82 to 43.51), and 21.06 bats/turbine/spring

and fall combined for the second year of study (90% ci: 14.98 to 28.75). This is equivalent to

Forward Energy Bird and Bat Monitoring Final Report

22.31 (90% ci: 16.55 to 29.01) and 14.04 (90% ci: 9.99 to 19.17) bats/MW/spring and fall

combined, respectively. Although the estimates between the two years appear to be quite

different, the confidence intervals overlap, suggesting that they are not statistically significantly

different. The two-year average was 27.26 bats/turbine/spring and fall combined (90% ci: 22.37

to 33.83) or 18.17 (90% ci: 14.91 to 22.55) bats/MW/spring and fall combined. The second

alternative estimator used the Huso (2010) estimator with carcass removal data from Blue Sky

Green Field( located in close proximity to Forward) and bat carcass and searcher efficieny data

from Forward. Average removal time for bats at BSGF was 3.49 days. Using the Huso

estimator produced estimates of 27.40, 14.97 and 21.18 bats/turbine/spring and fall combined for

the first year, second year, and two-year average, respectively. In the corresponding metric, this

is 18.27 bats/MW/spring and fall combined for the first year, 9.98 bats/MW/spring and fall

combined for the second year, and a two-year average of 14.12 bats/MW/spring and fall

combined. These results are in agreement with estimates calculated using the Jain estimator and

the modified Huso estimator. Confidence limits were not available for these estimates since only

a fixed average removal time was available.

A total of 20 bird fatalities was recorded during mortality searches. Because very little bird

mortality was observed, correlation analyses with weather and turbine variables were deemed

unreliable and were not performed. Bird mortality estimates were quantified in a similar manner

as described above for bat mortality.

Using the modified Huso estimator we calculated 5.6 birds/turbine/spring and fall combined

during the first year (90% ci: 2.34 to 9.82), 0.93 birds/turbine/spring and fall combined during

the second year (90% ci: -0.62 to 2.25), and a two-year average of 3.27 birds/turbine/spring and

fall combined (90% ci: 0.86 to 6.04). These values are equivalent to 3.73 birds/MW/spring and

fall combined for the first year (90% ci: 2.34 to 6.08), 0.63 bird/MW/spring and fall combined

for the second year (90% ci: -0.67 to 1.93), and a two-year average of 2.18 birds/MW/spring and

fall combined (90% ci: 0.84 to 4.01).

Using the Jain estimator, WEST, Inc. calculated 5.14 birds/turbine/spring and fall combined for

the first year of study (90% ci: 2.75 to 8.37), and 1.00 birds/turbine/spring and fall combined for

the second year of study (90% ci: 0 to 2.32). This is equivalent to 3.43 (90% ci: 1.83, 5.58) and

0.67 (90% ci: 0, 1.55) birds/MW/spring and fall combined, respectively. The confidence

intervals between the two years do not overlap for birds, suggesting that the estimates are

statistically significantly different. For unknown reasons, 12 of the 20 total bird fatalities were

found during spring of 2009. Searcher efficiency values were significantly lower during spring

2010 than they were during any other season. These confounding factors may serve to explain

the differences between the estimates. The two-year average was 3.07 birds/turbine/spring and

fall combined (90% ci: 1.77 to 4.84) or 2.05 (90% ci: 1.18, 3.23) birds/MW/spring and fall

combined. The Huso estimator was also considered for birds using carcass removal data from

Blue Sky Green Field and bird carcass and searcher efficiency data from Forward. Large birds

had an average removal time of 11.59 days at Blue Sky Green Field, while small birds had an

average removal time of 10.62 days. Using these values in the Huso estimator in conjunction

with Forward searcher efficiency data produced estimates of 2.71, 0.79 and 1.75

birds/turbine/spring and fall combined for the first year, second year, and two-year average,

respectively. These values are equivalent to 1.81, 0.53, and 1.17 birds/MW/spring and fall

Forward Energy Bird and Bat Monitoring Final Report

combined. These results are somewhat lower than estimates calculated using the Jain estimator

and the modified Huso estimator; however, bird mortality estimates are in agreement regardless

of which estimator was used.

When comparing corrected mortality estimates, bird and bat mortality rates recorded at the

Forward Energy Center were similar to those of neighboring wind farms and other studies

throughout the Midwest. In order to understand corrected mortality estimates in the proper

context comparisons with other wind project mortality studies are presented with the caveat that

direct comparisons are difficult and potentially misleading due to the crucial differences between

study methodologies, especially those used in searches and mortality estimators.

Forward Energy Bird and Bat Monitoring Final Report

1.0 INTRODUCTION

The Forward Energy Center (Center) consists of 86 General Electric 1.5MW turbines in

southeastern Wisconsin. Each turbine has an 80-meter hub height and 120-meter rotor-tip

height. The turbines are located in agricultural land (corn/soybean rotations being the

predominant crop types) in southern Fond du Lac County and northern Dodge County. The

Center is approximately 5 km east of Horicon National Wildlife Refuge, 63 km west of Lake

Michigan, 22 km north of Neda Mine State Natural Area, and 21 km south of Lake Winnebago.

The following report presents the final results of a 2-year post-construction bird and bat mortality

study at the Center. The primary analyses were conducted to estimate bird and bat mortality at

the Center, and identify which meteorological, temporal, spatial, species-specific, and turbine-

specific covariates were associated with bird and bat mortality. Data were collected for two fall

study periods (July 15 – November 15, 2008 and July 15 – October 15, 2009) and two spring

study periods (April 15-May 31, 2009 and 2010) to assess mortality during bird and bat

migratory periods. The results of this study will be useful to future planning and wildlife

management efforts of wind farms in Wisconsin and the Upper Midwest.

1.1 STUDY OBJECTIVES

The primary objectives of our study were to:

1) Assess bird and bat mortality at the Center,

2) Provide corrected mortality estimations for birds and bats using the most recent

statistical estimator, and

3) Correlate observed fatality rates with select weather variables, proximity to Horicon

Marsh and Neda Mine, turbine operating status, and bird and bat activity at the Center.

2.0 METHODOLOGY

The design used in this study was adapted from the bird and bat mortality studies conducted at

the Top of Iowa Wind Farm, IA (Koford et al. 2004), Crescent Ridge Wind Farm, IL (Kerlinger

et al. 2007), and the Maple Ridge Wind Farm, NY (Jain et al. 2007), which share similar regional

and land use attributes with the Center. These methods were consistent with protocols from

California, US and Canada that were released in 2007 (Canadian Wildlife Service, 2007), as well

as recommended standards for post-construction monitoring at terrestrial wind facilities in

Wisconsin as described by the USFWS, DNR, and Commission (Table 1). Additionally, the

methods used in this study were comparable to those used at concurrent bird and bat mortality

studies at neighboring wind farms Blue Sky Green Field, WI (Gruver et al. 2009) and Cedar

Ridge, WI (BHE Environmental, Inc. 2010).

Forward Energy Bird and Bat Monitoring Final Report

The study periods (fall and spring) were established under the assumption that the majority of

mortality occurs during peak migration periods for birds and bats, as has been observed in past

studies (Howe et al. 2002, Johnson 2005, Kunz et al. 2007, Arnett et al. 2008). Additionally,

searches for carcasses of Sandhill Cranes (Grus canadensis) were conducted from October 15 -

November 15, 2008 because the USFWS was concerned about impacts from the Center on late

migrants such as cranes.

2.1 STUDY DESIGN

Twenty-nine of the 86 (34%) wind turbines at the site were randomly selected and searched in a

stratified sample. This level of sampling was chosen to provide adequate spatial coverage using

a representative sample of turbines within the project area. The Center was divided into three,

north-south oriented sections, each of which was approximately 3.5 kilometers wide, which

allowed for comparison of mortality rates as distance increased eastward from Horicon Marsh

and northward from Neda Mine (Figure 1). The number of selected turbines in each section was

proportional to the total number of turbines in each section. Because the western, central, and

eastern sections contained 48%, 38%, and 14% of the total number of turbines at the wind farm,

respectively, 14, 11, and 4 turbines were selected to be searched within each respective section.

2.2 STUDY PLOTS

The total search area was defined identically for all 29 study plots, with each plot consisting of a

160 m by 160 m square (6.3 acres) centered on the wind turbine (Figure 2). In order to minimize

impacts on crops and landowners, 26 of the 29 searched turbine plots had 19% (1.2 acres) of the

total searchable area searched using five parallel 160 m by 5 m transects. Transects were

randomly selected from the total searchable area. The five parallel transects were perpendicular

to the turbine access road. The access road itself plus an extension and the pad of the turbine

served as a 6th search transect.

At the remaining three turbine plots, the entire 160m by 160m plot was searched, which allowed

for determination of the number of carcasses potentially missed within plots where only a portion

of the plot was searched. Additionally, three, 1.2-acre control sites were searched to measure

background mortality. One control site and one fully cleared study plot were present within each

of the three, north-south oriented sections of the study area, but each control site was located

outside of the wind farm boundaries.

All of the turbines monitored during this study were located in active agricultural fields, with the

study plots mostly located within corn and soybean crops. Other crop types present included

alfalfa, wheat, timothy grass, and hay, in addition to Conservation Reserve Program habitat.

Transects were marked with posts and flagging during and in between study periods to ensure

consistency of location throughout the duration of our 2-year study. Search transects were

cleared of vegetation by mowing with a tractor. Mowing effort was generally minimal during

the spring study periods because the majority of the crops were still in early stages of growth (a

notable exception was alfalfa, which required more frequent mowing). The fall study periods

Forward Energy Bird and Bat Monitoring Final Report

necessitated one large-scale mowing effort to remove developed crops (e.g. corn, soy), but

afterwards most plots did not require many additional cuttings. Crop types such as corn and

soybeans were cut down to a height of 4 – 8 inches after mowing, while alfalfa was generally cut

lower as it grew back faster. At all plots, the mowing effort ensured adequate visibility (i.e. low,

uniform vegetation levels or bare soil), and variation in visibility between crop types was

captured in the searcher efficiency trials. All transects were searched in their entirety despite

difficult conditions such as mud or flooding.

2.3 CARCASS SEARCHES

A trained crew consisting of searchers, field technicians, and S. Grodsky performed carcass

searches for birds and bats. At the 26 turbines where 1.2 acres were searched, each transect was

divided into two approximate halves longitudinally (each 2.5 m wide), and searched up the first

half while scanning ahead and towards the transect center, and then searched back down the

second half of the transect while scanning ahead and towards the center of the transect. All five

transects plus the road and pad at each of the 26 turbines were searched in this manner, which

allowed for complete transect coverage while maintaining a slow and constant search pace. The

6.3-acre, fully cleared plots were searched by walking parallel transects 5 m apart in a snaking

pattern from one side of the square plot to the other. Meanwhile, the searcher was scanning

approximately 2.5 m to each side of the search line. Prior to searches, weather conditions and

vegetation height were recorded, in addition to other data. All plots were cleared of any

carcasses by performing a clearing search prior to beginning the first search of each season.

When a carcass was found, the animal was assigned a unique carcass identification number that

included the turbine number and the date, placed in a re-sealable bag, and the level of

decomposition was estimated based on the following scale:

freshly killed - unaltered by scavenging animals, no signs of fly larva infestation

(approximately 1 – 2 days on the landscape)

scavenged - signs of insect infestation, partially degraded and/or consumed

(approximately 3 – 5 days on the landscape)

decomposed - severely decayed and/or scavenged (> 5 days on the landscape)

The distance of the carcass from the base of the turbine was estimated by searcher measurement.

To do this, searchers were provided hard copies of maps for each study plot. Once a carcass was

found, the searcher paced off the distance from where the carcass lay to the base of the turbine.

Each grid cell was 4.6 m in length, and searchers applied their individual pace to determine

distance on the map to the nearest 4.6 m grid cell. This method of mapping was chosen because

the number of searchers and budget limitations precluded equipping searchers with GPS units.

Additional information pertaining to the appearance and location of the carcass were recorded,

and photographs of the disposition of the carcass were taken before moving it. When possible,

carcasses were identified to species in the field by S. Grodsky prior to storage at the Center’s

main office facilities in a designated freezer.

Forward Energy Bird and Bat Monitoring Final Report

All birds and bats were frozen and saved for future use in searcher efficiency and scavenger

removal trials, with the exception of bats found during the fall 2009 study period. These bats

were refrigerated and transported to the Wisconsin Veterinary Diagnostic Laboratory (WVDL)

for further analysis. A subset of the carcasses were sexed and aged either at the University of

Wisconsin Zoology Museum (Museum) or the WVDL. Bat carcasses that were too decomposed

to be identified were sent to the Museum to be accurately identified using skull morphology by

academic curator Paula Holohan. Bird and bat carcasses found outside of designated searchable

areas or discovered at turbines outside of the study area (e.g. reported by Forward Energy

technicians) were considered incidental finds. While incidental carcasses not found on turbines

scheduled for search were excluded from mortality estimates, they were included in the general

results as supplementary information. Required collection and salvage permits for the transport

and possession of deceased wildlife were obtained from the DNR and the USFWS. Carcasses

not used in searcher efficiency or scavenger removal trials by the end of the study were stored in

a designated freezer for potential use in future studies in the area. Carcasses that were too

decomposed to be considered useful were discarded.

2.4 SEARCH INTERVALS

To minimize the potential for carcass removal by diurnal scavenging animals, searches began

approximately 30 minutes before sunrise and generally concluded prior to noon. Searched

turbines were randomly selected for one of three search schedules for the duration of the study:

11 (38%) were searched everyday, 9 (31%) were searched every three days, and 9 (31%) were

searched every five days. Search intervals were chosen so that all 29 turbines could be searched

in a cost-effective manner. Some daily searches were necessary to correlate weather with

mortality. The search intervals were randomly distributed throughout the three study sections.

One each of the three fully cleared sites was searched every day, every 3 days, and every 5 days.

The order in which the turbines were searched during the day was randomized to prevent bias

from time of day effects.

2.5 SEARCHER EFFICIENCY

Searcher efficiency trials were conducted to estimate the proportion of available carcasses

discovered by searchers. A searcher’s efficiency was calculated as the proportion of trial bird or

bat carcasses found and recorded by the searcher relative to the total number used for the trials.

Trials coincided with actual mortality searches and were comprised of one to four bird and bat

carcasses per trial and placed by a field technician at randomly selected locations within the

searchable area of the study plot prior to each day’s searches. The trial carcasses were not

physically marked, but were explicitly mapped on a grid along with notation of identifying

attributes, such as species, appearance, and condition of carcass, to differentiate the trial carcass

from turbine-related fatalities. The field technician returned after searches were completed for

that day to determine whether searchers had successfully located trial carcasses. In the event that

a trial carcass was determined to have been removed by a scavenger prior to the standard carcass

search, that carcass was removed from the trials and not counted in the final results. That trial

Forward Energy Bird and Bat Monitoring Final Report

was repeated on a subsequent day. Trials were implemented at randomly selected turbines.

Approximately 100 trials were conducted throughout each study period, with the timing and

carcass placement unknown to searchers. Bats collected during mortality searches and from the

Wisconsin State Laboratory of Hygiene (WSLH; post-rabies testing) were used as trial carcasses.

There were far fewer bird mortalities compared to bats, and consequently, there was a shortage

of bird carcasses available to use as trial carcasses. Thus, Brown-headed Cowbirds (Molothrus

ater) provided by the United States Department of Agriculture – Wildlife Services (Wisconsin

office) were used for trial bird carcasses. The condition of trial carcasses varied from fresh to

partially decomposed, and generally simulated the conditions of carcasses found as mortalities.

No large bird carcasses were used in the trials. Most of the bird carcasses found during searches

were of similar size to Brown-headed Cowbirds. However, since large birds tend to have higher

searcher efficiency rates than small birds, the lack of large bird carcasses in the searcher

efficiency trials may have lead to a bias of our estimated bird mortality. Searcher efficiency

trials were conducted through the duration of each field season to account for any temporal

variation in searcher efficiency. Searcher efficiency rates were averaged for all searchers, and

these values were used in the corrected mortality estimator.

2.6 SCAVENGER REMOVAL RATE

Similar to searcher efficiency trials, scavenger removal rate trials were designed to account for

the bias associated with the removal of carcasses by scavenging animals before searchers

encountered them. Although the scavenger removal trials were initially designed to account for

scavenging by animals such as coyote, fox, skunk, raccoon, feral cat, the trials also accounted for

other types of removal, such as tilling, plowing, mowing, and weather conditions such as the

flooding of study transects. Brown-headed cowbirds were used for all bird trial carcasses.

Because bat carcasses were difficult to acquire, and the utilization in trials of retrieved bat

mortality carcasses precluded data acquisition for determining their cause of death, we used

black and grey weanling mice (20 to 25 days old; Rodentpro®.com) as surrogates for bat

carcasses. Mice alone were used during the spring 2009 and fall 2009 study periods, and in

combination with bat carcasses in the spring 2010 study period. The fall 2008 study period used

bats only. Paired studies at other wind resource areas have validated the use of mice as

surrogates for bats in bias trials (Jain et al. 2008). Between 1 and 3 birds or bats/mice were used

on each trial date, with approximately 100 trials performed during each study period. Similar to

the searcher efficiency trials, the scavenger removal trials were evenly distributed throughout the

entire study period to account for temporal variation in scavenger abundance and composition.

The turbine, date, and placement of trial carcasses were all selected and mapped using the same

methods as described for the searcher efficiency trials. The scavenger removal trial carcasses

were generally placed prior to noon. Unlike with searcher efficiency trials, each searcher and

field technician had a copy of the mapped locations for each scavenger removal carcass at each

study plot. This allowed field technicians to record the results of the trial, and prevented

searchers from treating trial carcasses as turbine fatalities. If a searcher mistakenly removed a

scavenger removal trial carcass, the trial was repeated at the same turbine at a later date. The

duration of the trial corresponded to the search interval of the respective turbine. For instance,

turbines searched every day had scavenger removal trials lasting 24 hours. If the trial carcasses

persisted throughout the 24 hour period, a field technician removed the carcass. The status

Forward Energy Bird and Bat Monitoring Final Report

(presence or scavenged) of trial carcasses was checked every 24 hours. Thus, turbines searched

for mortality every three days were checked up to three times, 24 hours apart, while turbines

searched for mortality every five days had up to five checks, 24 hours apart. Trials were

concluded once all the trial carcasses were removed or the trial period was completed, whichever

happened first. The scavenger removal rate was averaged across turbines per season and study

period and incorporated into the mortality estimate.

2.7 FATALITY ESTIMATION

The following variables are used in the equations for fatality estimation below:

the estimated probability that a carcass is both available to be found during a search

and is found, as determined by the removal trials and the searcher efficiency trials

the average time (in days) a carcass remains in the study area before it is removed, as

determined by the removal trials

p = the estimated proportion of detectable carcasses found by searchers, as determined by

the searcher efficiency trials

I = the average interval between standardized carcass searches, in days

N = number of carcasses placed for carcass removal trials

Nc = number of carcasses with right censoring during carcass removal trials, ie. carcasses

that persist to the end of the trial

di = total number of days that carcass i persists during removal trials

Sc = proportion of carcasses not scavenged halfway through the time interval between

searches

average observed number of fatalities per turbine

A = proportion of the search area of a turbine actually searched

M = adjusted fatality estimate in fatalities/turbine

First, Huso’s estimator for the probability of availability and detection is:

where (Huso, 2010).

In this estimator, carcass removal times are calculated using the standard survival analysis

method for averages with censoring:

Forward Energy Bird and Bat Monitoring Final Report

When the protocol was developed in early 2008, the study was designed with similar

methodology to other studies being conducted at the time. It was intended to use Huso’s

estimator, which was expected to be published in 2008. However, Huso’s final version was not

published and publicly available until 2010 (Huso, 2010). Because of this delay, it was unknown

until the end of the Forward study that carcass removal data would need to be collected over an

extended period (on the order of 40 days) for use with the Huso estimator. Our carcass removal

trials lasted a maximum of 5 days. Therefore, we modified Huso’s (2010) estimator and did so

by not using the exponential component in the equation. Instead of the exponential component,

we substituted the 2-year average for the 1-, 3-, and 5-day scavenger removal rates for birds (1-

day removal rate = 0.42, 3-day = 0.77, and 5-day = 0.81) and bats (1-day = 0.18, 3-day = 0.5,

and 5-day = 0.65).

Huso (2010) defined the effective search interval as "the length of time beyond which the

probability of a carcass persisting is less than or equal to 1%". A carcass can persist until either

retrieved by a searcher or a scavenger. Our average scavenger removal rates for the Forward

study indicated that 81% of all birds and 65% of all bats were scavenged by the end of day 5.

Extrapolating those rates, by day 10, at the latest, the probability of both birds and bats

persisting would most likely have been less than or equal to 1%, considerably sooner than the 40

or so days suggested by Huso (2010). Thus, we are confident that modifying the Huso estimator

as we did is a suitable way to handle our carcass removal data.

Our estimates along with 90% confidence intervals were based on 3000 bootstrap samples with

replacement from the original data and calculated using SAS software (Version 9.2, SAS

Institute, North Carolina, US). Bootstrapping is a computer simulation technique that is useful

for calculating point estimates, variances, and confidence intervals for complicated test statistics.

The lower 5th and upper 95th percentiles of the bootstrap estimates are estimates of the lower

limit and upper limit of 90% confidence intervals. Confidence intervals for estimated mortality

were obtained by using the Delta method approximation for variance estimates (Powell 2007).

Because we had to modify Huso’s (2010) estimator to fit our data, Forward Energy hired the

environmental consulting firm WEST, Inc. to calculate bird and bat mortality using 2 different

estimators to verify the mortality we estimated using Huso’s (2010) modified equation. The first

additional estimator used was Huso’s estimator as published in 2010, without any modifications.

First, Huso’s estimator for the probability of availability and detection is:

where (Huso, 2010).

Forward Energy Bird and Bat Monitoring Final Report

In this estimator, carcass removal times are calculated using the standard survival analysis

method for averages with censoring:

In order to calculate this value, it is suggested that carcasses be left out for an extended period of

time (generally 30 to 40 days). Because carcasses were left on the ground according to the

agreed upon protocol for a maximum period of 5 days rather than a longer period such as 40

days, estimated average removal times could not be calculated for use with the Huso estimator.

Therefore, carcass removal rates from the published Blue Sky Green Field study were used in

place of the Forward carcass removal data in the Huso estimator. Because the Blue Sky Green

Field project area is also located in Wisconsin, and is in close proximity to Forward Energy

Center, average removal times could be expected to be similar to what would be found at

Forward. The number of retrieved bird and bat carcasses and searcher efficiency data from

Forward Energy Center were used with the Blue Sky Green Field carcass removal data. It was

not possible to produce confidence intervals for these estimates because only the point estimate

was available for average removal time.

In accordance with the method of data collection, a second estimate was calculated by WEST,

Inc. using the same estimator as was used in the studies at Top of Iowa, Maple Ridge and

Crescent Ridge, after which this protocol was designed (Koford et al. 2004, Kerlinger et al. 2007,

and Jain et al. 2007). For this estimator the probability of availability and detection of carcasses

is calculated by:

For both estimators, the adjusted fatality estimate is calculated by:

For the estimators calculated by WEST, Inc., confidence intervals for estimated mortality were

obtained by bootstrap sampling with 1000 repetitions.

None of the incidental mortalities or carcasses found on non-search plots were included in

fatality estimates. For each of birds and bats, per turbine mortality was estimated by year and a

two-year average. Mortality was estimated using data collected from four study periods: Fall

2008, spring 2009, fall 2009, and spring 2010. Because carcass searches did not occur

throughout the calendar year, the results are conditioned on sampling periods, which primarily

coincide with migratory periods of bats and birds. Therefore, the final estimate is not an estimate

of annual mortality, but an estimate of mortality during spring and fall combined. However, in

general, very few bat fatalities are expected during summer and winter, and few bird fatalities are

expected during winter. Therefore, the estimates for spring and fall combined are most likely

similar to what would be obtained by year-round sampling.

Forward Energy Bird and Bat Monitoring Final Report

Sub-lethal effects or crippling bias (a bat or bird is injured by a turbine but dies out of the search

area or at a later time from injuries suffered as a result of a turbine, see Grodsky et al. 2011) was

assumed to be low, and was not accounted for in our mortality estimations. Levels of

background mortality, such as death by natural causes, are also assumed to be low, and no bird or

bat carcasses were found in any of our control plots during the 2-year study period. Because it is

impossible to accurately assess the undetected mortality rate and very difficult to determine the

background mortality rate, it is general practice to disregard these potentially biasing factors. If

levels of background mortality are high, then estimates may be biased high, because not all

carcasses found will be from deaths attributable to the wind farm. If undetected mortality rates

are high, then the estimates will be biased low.

2.8 COVARIATE ORIGINS AND STATISTICAL ANALYSIS

Post-construction weather data including hourly records of visibility, temperature, ceiling height,

relative humidity, dew point, barometric pressure, precipitation, and wind speed were collected

by S. Grodsky from the University of Wisconsin and obtained from the National Oceanic and

Atmospheric Administration service station at the Fond du Lac, Wisconsin airport, located 17

km from the project area (Table 16). These data were collected from the airport as on-site

meteorological tower data were not available. Turbine operating status, including hourly power

output (MW) and revolutions per minute of the rotor (rpm), were obtained from Mike Liska of

Invenergy, LLC via a GE data-logger (Table 16). All of the hourly data were averaged over a 12-

hour period from 1900 to 0700 each night, when bats are active and neotropical migrant birds are

typically migrating. Bat activity data in the form of bat passes/detector-night, determined from

sonograms recorded by Anabat detectors (Titley, Inc. Sydney, Australia, Drake et al. 2010) were

obtained from Mike Watt. Distances of each study plot from Horicon Marsh and Neda Mine

were measured using spatial analyst tools in ArcMap software (Version 9.2, ESRI ArcGIS 9,

Redlands, CA). Because we did not have turbine-specific weather and bat activity data, all

covariate values were extrapolated across all study plots for any analysis involving these specific

data and mortality at each turbine was pooled on a daily basis.

All statistical analyses were performed by WEST, Inc. using R software (Version 2.7.2, R

Foundation for Statistical Computing, Vienna, Austria). Covariates were evaluated for co-

linearity using Pearson’s correlation coefficients (r), and all correlated variables were excluded

from our modeling. We then evaluated generalized linear models (glm) with Poisson and

negative binomial distribution to determine individual covariate relationships with bird, bat,

migratory bat, and non-migratory bat mortality. The fitted negative binomial and Poisson

models all had log link and were of the form:

which related the behavior of the natural logarithm of the mean number of bat fatalities per

search, to a linear function of the set of predictor variables 1,..., p

x. The '

are the parameters

that specify the nature of the relationship. The program R was used to fit several alternative

models. In particular, step-wise model selection methods based on AIC were used to determine

the best fitting Poisson and negative binomial models. The correlation matrix was obtained for

Forward Energy Bird and Bat Monitoring Final Report

all continuous main effects listed in Table XX. Variables with pairwise correlations ≥ 0.6 were

not allowed to be present in the models at the same time. The Poisson and negative binomial

models were built using a forward and backward stepwise approach in which main effects

entered or left the model based on the AIC value. The first step began with the full model

containing all parameters. In the next step, covariates were added or subtracted from the model

one at a time. If the model AIC decreased, the change in covariates was retained. If AIC

increased, that change was discarded and the next covariate was tested. This procedure was

repeated until none of the covariate changes produced a lower AIC. Poisson and negative

binomial models with the same parameter sets were compared using the Vuong test to determine

whether they were distinguishable from one another (Vuong 1989). We interpreted p-values of

less than 0.05 as statistically significant.

Based on the presence of adequate bat mortality, weather and bat activity data, 139 nights during

the study were used in the analysis. All nights were selected from the fall 2008 and fall 2009

study periods. These periods showed the most concentrated levels of bat mortality. Only fresh

bat fatalities on daily search turbines were used in the analysis, a total of 37 bats. This was

necessary to precisely correlate fatality and weather patterns. None of the models selected were

allowed to contain both temperature and dewpoint. Only bat activity from all frequencies was

considered since it was highly correlated with high frequency and low frequency bat activity, and

bat mortality rates were most highly correlated with all bat activity. These exceptions were

necessary due to perceived high correlations between the pairs of variables (Neter et al. 1996).

To determine whether distance to Horicon marsh and/or Neda mine had a significant correlation

with mortality, a data set was created containing turbine number, distance to marsh, distance to

mine, count of fatalities, and count of searches. An indicator variable was added which was “1”

if the plot was one of the 3 fully cleared plots and “0” otherwise. A Poisson regression was then

performed to determine what, if any, significance each variable had in determining mortality

rates.

3.0 RESULTS

3.1 CARCASS SEARCHES

A total of 3,763 carcass searches was completed throughout the duration of the 2-year study

(Table 2). By search interval, 2,562 carcass searches (mean = 233/ turbine) occurred for turbines

searched daily, 724 carcass searches (mean = 80.44/turbine) were conducted for turbines

searched every 3 days, and 477 carcass searches (mean = 53/turbine) took place at turbines

searched every 5 days. A total of 135, 43, and 29 carcass searches were completed at the control

sites searched daily, every 3 days, and every 5 days, respectively, for a grand total of 207 carcass

searches at all control sites.

Forward Energy Bird and Bat Monitoring Final Report

3.2 SEARCHER EFFICIENCY

A total of 399 trial bird carcasses and 396 trial bat carcasses were used to test searcher efficiency

from all study periods (Table 3). Searcher efficiency was lower for the spring of 2010 for both

birds and bats (Table 3). This difference was less pronounced for bats than for birds.

Researchers noted that mowing was less frequent during spring 2010 and vegetation growth was

greater than in the previous spring. A test of differences between seasons for searcher efficiency

showed a significant difference between seasons for birds (chi-squared = 17.74, df = 3, p-value =

0.0005) at the alpha = 0.05 level. Overall searcher detection rates were 61.9% for birds and

37.9% for bats.

3.3 SCAVENGER REMOVAL RATE

A total of 402 bat and mouse carcasses and 403 bird carcasses were used to estimate scavenger

removal rate for bats and birds (Tables 4 and 5). During spring 2010 both bats (N = 57) and

mice (N = 42) were used for scavenger removal bias trials. These data were used to test for

differences between bats and mice. Of the 57 bats used for removal trials, 96% remained after 1

day, 89% remained after 2 days, and 74% remained after 3 days. For mice, 88% remained after

1 day, 71% remained after 2 days, and 50% remained after 3 days. The counts for each day are

not statistically significantly different at the alpha = 0.05 level (day 1: chi-squared = 0.18, df =

1, p-value = 0.67; day 2: chi-squared = 0.34, df = 1, p-value = 0.56; day 3: chi-squared = 0.81,

df = 1, p-value = 0.37). This suggests that roughly the same proportions of bats and mice were

remaining at days 1, 2 and 3. These results agree with the results of previous studies and validate

the use of mice as surrogates for bats in carcass removal trials (Jain et al. 2008).

Testing scavenging proportions for bats and mice by season at the alpha = 0.05 level, the values

for days 1, 2, and 3 are not significantly different at the alpha = 0.05 level . For birds, days 1

and 2 are significantly different between seasons (chi-squared = 12.00, df = 3, p-value = 0.007;

chi-squared = 11.55, df = 3, p-value = 0.009). Day 3 for birds was not significantly different

between seasons.

Since only brown-headed cowbirds were used for carcass removal trials, bias trial data was not

split by bird size. Large birds generally have a longer removal time than small birds (Table 6).

Data was not available for determining how different proportion not scavenged might be

between bird sizes. Estimates may be biased due to this factor.

3.4 BIRD MORTALITY

3.4.1 Characteristics of Bird Mortality

A total of 20 birds was recorded during scheduled carcass searches (Table 7). One Tree Swallow

(Tachycineta bicolor) and 3 Red-tailed Hawks were also recorded as incidental finds.

Approximately half of the bird carcasses (55%) were found fresh, while 20% and 25% of the

carcasses were found scavenged and decomposed, respectively (excludes incidental finds; Table

Forward Energy Bird and Bat Monitoring Final Report

8). The majority of the birds found during carcass searches were adults (includes incidentals; N

= 20; 83%), with few juveniles (N = 4; 17%) (Table 9). Sex was determined for the Red-Tailed

Hawks by necropsy, and all were females with ovaries in the non-reproductive phase (N = 3).

Carcasses were found from 15 unique species (Table 7). Of these, Red-tailed Hawk, Tree

Swallow, Ruby-crowned Kinglet (Regulus calendula), Black-and-white Warbler (Mniotilta

varia), and Red-eyed Vireo (Vireo olivaceus) were the most common fatalities with 2 individuals

each; there was one fatality of each of the other species (Table 7). A majority of the bird species

found on scheduled searches occurred within the taxonomic orders Passeriformes (N = 15; 75%)

and Buteos (N = 2; 10%), with comparatively little representation from other orders. None of the

birds found during mortality searches were listed as Wisconsin state threatened or endangered

species. However, the Black-billed Cuckoo (Coccyzus erythropthalmus), Bobolink (Dolichonyx

oryzivorus), and Ruby-crowned Kinglet are considered state species of concern.

3.4.2 Bird Mortality Estimates