Post-construction avian mortality monitoring at Project West Wind

Abstract

Post-construction avifauna investigations were undertaken at Project West Wind, Meridian Energy Limited's 62-turbine wind farm on the Wellington south coast. These investigations were required in accordance with the resource consent conditions to quantify the level of avian mortalities occurring at the wind farm, particularly in regard to New Zealand falcon (Falco novaeseelandiae), kākā (Nestor meridionalis) and kererū (Hemiphaga novaeseelandiae). This is the first comprehensive study at a New Zealand operating wind farm. The methods included three field components necessary to calculate annual estimates of mortalities across the wind farm site: routine turbine searches; carcass detection trials; and carcass removal trials. Results from years 1 and 2 of a three-year programme are presented. To date, mortalities have been recorded for 17 taxa at 18 of the 24 study turbines. There have been no recorded mortalities of falcon, kākā or kererū. Australasian harrier (Circus approximans) has been the species for which the most mortalities have been recorded. Overall estimated annual mortality rates for years 1 and 2 were calculated to be approximately six and five birds per turbine respectively.

Keywords:

• Project West Wind
• wind farm
• post-construction monitoring
• mortality estimates
• avifauna
• carcass detection rates
• carcass removal rates
• turbine searches

Introduction

The development of the wind industry in New Zealand is relatively new compared to that overseas, where considerable information has been collected for the purpose of both ecological assessments and post-construction monitoring. Potential impacts on avifauna associated with wind farms are habitat loss, disturbance and displacement, and collision fatalities (Drewitt & Langston 2006). Internationally, the avifauna groups that have been found to be most often impacted by wind farm developments have been swans, geese, ducks, waders, gulls, terns, large soaring raptors, owls and nocturnally migrating passerines (Percival 2000; Langston & Pullan 2003; Barrios & Rodríguez 2004; Stewart et al. 2004; Kingsley & Whittham 2005; Madders & Whitfield 2006; de Lucas et al. 2008; Garvin et al. 2010).

Most post-construction investigations have focused on the potential mortalities associated with collision with wind farm structures (e.g. turbines, overhead transmission lines, infrastructure). While some studies have reported high numbers of mortalities, not all wind farms are equally susceptible to bird strike. There are many contributing factors operating both temporally and spatially, including site topography, seasonal weather conditions, turbine design, number and spatial patterns of turbine distribution, wind farm orientation to valleys or passes, species present, movement patterns, and factors that affect their breeding, feeding and roosting behaviour (Drewitt & Langston 2006). Thus, it is now generally accepted that the potential impacts on avifauna can be reduced by the design and careful siting of a wind farm development.

To date, there have been no published studies of mortality at an operating wind farm in New Zealand (Boffa Miskell & Golder Associates 2009). While information obtained from international experiences may be used to guide New Zealand studies, the results of those studies are not directly comparable due to a number of differing variables, including the species present, their potentially unique behaviours, and site variability in terms of differences in terrain and habitat. As such, post-construction monitoring at New Zealand wind farm sites is required.

The most widespread method of monitoring bird collisions internationally involves the regular search for carcasses and remains under operating wind turbines. Perhaps the most significant difference between methods relates to search effort as different methodologies use different searcher schedules ranging from daily searches for set periods (e.g. weeks, months, seasons), to fortnightly, monthly or annual searches. A range of techniques are also used to account for both searcher efficiency (the probability of finding a carcass that is present) and carcass removal (the probability that the carcass will be removed by scavengers between search periods).

Project West Wind marked the first constructed New Zealand wind farm for which post-construction avifauna monitoring was a requirement of the consent conditions. The Project West Wind site lies on high ridgelines on the southwest peninsula of Wellington (Fig. 1), a rugged and isolated landscape approximately 5580 ha (55.8 km²) in area. The site is contained entirely on two private farm properties: Terawhiti Station in the south and Makara Farm in the north (Fig. 1). All of the land has been intensively farmed in the past; however, much of the southern area has been allowed to revert to shrublands and scrub. The majority of the site is located on steep greywacke hill country and broad rolling ridgelines. The northern (Makara) part is approximately 150–300 m above sea level, separated from the coast by high escarpments with the nearest turbine being 400 m from the coast; the southern (Terawhiti) part of the site is located along high ridgelines 200–400 m above sea level, with the closest turbine being approximately 700 m from the coast (Fig. 1).

Figure 1  Project West Wind study site.

Pre-construction baseline point counts at Project West Wind recorded 32 species, with 81% (n=3388) of recorded observations comprising introduced birds (Boffa Miskell 2007). The counts were dominated by common starlings (Sturnus vulgaris), finches and house sparrows (Passer domesticus). Only a few native birds of open country and farmland were recorded, including Australasian harrier (Circus approximans; hereafter ‘harrier’), New Zealand pipit (Anthus novaeseelandiae), paradise shelduck (Tadorna variegata), spur-winged plover (Vanellus miles) and welcome swallow (Hirundo tahitica). Native species recorded in areas of regenerating bush included morepork (Ninox novaeseelandiae), fantail (Rhipidura fuliginosa), and grey warbler (Gerygone igata).

Both the pre-consenting ecological assessment and the results of the pre-construction baseline point counts concluded that, with regard to the avifauna at the West Wind site, some bird strike of common introduced and native resident species such as southern black-backed gull (Larus dominicanus) and harrier was likely to occur (Boffa Miskell 2005, 2007). Furthermore, it was predicted that occasionally coastal birds such as shearwaters or mollymawks may cross the site, particularly as a result of storms or low fog, and that fatal encounters with turbines might occur on rare occasions (Boffa Miskell 2005, 2007).

Construction of the 62-turbine wind farm commenced in 2007. The turbine models used are 111 m tall Siemens 2.3 megawatt wind turbines, consisting of a 67 m high tower and 40 m long blades. The three-year post-construction avian monitoring programme began in September 2009 when all 62 turbines began operating. This paper presents the findings of years 1 (September 2009 to August 2010) and 2 (September 2010 to August 2011) of the post-construction mortality investigations at Project West Wind—the first such comprehensive study at an operating wind farm in New Zealand. The key avian species of concern identified in the consent conditions for Project West Wind were falcon, kākā and kererū. As such, the aim of this study was to quantify the level of avian mortalities occurring at the wind farm, particularly of these species.

Methods

For calculating wind farm avian mortality rates, a number of variables require consideration. These include mortality searches to collect carcasses from around turbines, carcass detection rates to estimate the percentage of carcasses found by searchers, and carcass removal rates to estimate the length of time carcasses remain in the search area between search sessions.

Mortality/carcass searches

This study was carried out at the northern end (Makara Farm, see Fig. 2) of the site after the pre-construction baseline study showed an absence of native species at the southern end (Terawhiti Station). This approach was agreed with the Department of Conservation.

Figure 2  Turbine locations and carcass removal trial sites within the Project West Wind study area.

Fortnightly carcass searches were undertaken at the 24 turbine sites on Makara Farm over the course of each year of monitoring. Year 1 searches were undertaken from September 2009 to August 2010 and year 2 searches from September 2010 to August 2011. Each turbine was visited once a fortnight, the full search of all 24 turbines taking three days working on site between 0800 h and 1700 h. For the purpose of this study, the definition of a ‘carcass’ includes whole birds or body parts (e.g. head, wings, tarsi), as well as remains of tail or flight feathers. While not reported on here, post-construction point counts were also undertaken at a number of the study turbines prior to the carcass searches. The point counts had a seasonal and temporal component, and as such the order of the turbine searches was determined by the timing requirement of the point counts.

During all search sessions, the search focused on a circular area centred on each turbine with a minimum search radius of 80 m (equivalent to two blade lengths). Observers searched each site by walking transects that radiated out (wedge-shaped) from the base of the turbine, searching the area on either side of each transect. This method ensures that the area closest to the base of the turbine, where most carcasses are likely to be found, is searched most thoroughly, but potentially reduces the search efficiency at the end of each transect. This radial search method was used due to the terrain of the site.

All carcasses were photographed as found and their location within the search area mapped before being removed and stored in a freezer. Any sign (e.g. featherspot, partial and full carcasses) of a possible mortality was assumed to be a result of a turbine collision. This may result in an over-estimate of turbine collisions. The following details were recorded for each carcass: date and time located; identification number; species; age; sex; carcass condition (intact, scavenged, dismembered, featherspot); probable cause of death; turbine number; distance to turbine; bearing to turbine; distance from bush; and weather conditions (e.g. wind direction, speed, temperature, precipitation and cloud cover).

A Chi square test was used to investigate any statistical association between mortality and bird size (very small 5–100 g; small 100–500 g; medium 500–1000 g; large > 1 kg). Individual species’ body weights were derived from Heather and Robertson (2000). Due to the small sample sizes, the following methods were used to investigate the data with regard to the distance mortality was detected from a turbine and bird size: 1) a one-way ANOVA to test if mean distance from a turbine differed between bird size groups; 2) a non-parametric Kruskal-Wallis test to see if spatial distributions were different between bird size groups; and 3) a Chi square test by grouped (10 m) distances to test for an association between bird size and distance from the turbine. Following these, a Fisher's test of the equivalence of two variances was undertaken. Because of the small sample sizes on which these variance estimates were based, a permutation test was performed which generated 1000 samples of the data randomly ‘redistributing’ the observed distances between the groups.

Carcass detection trials

Detection rates of carcasses were measured through the placement of wild (both native and introduced) bird carcasses within the study search areas. Possible variation in detection rates between habitat types and bird size was also investigated. Six trials were conducted between June and October 2010.

In order to randomise carcass placement, each turbine search area was divided into four sub-areas (A–D), resulting in a total of 96 sub-areas (i.e. four sub-areas for each of the 24 study turbines) across the site. For each of the 96 sub-areas, the predominant vegetation type (i.e. pasture or scrub/shrub) was then determined and random numbers allocated to each sub-area.

Each carcass detection trial consisted of placing a single bird of each size class within four pre-determined sub-areas (see Table 1), thus making a total of 24 ‘detection’ carcasses. In most cases one bird of each size class was put out, except on two occasions. The first was when a lack of medium-sized birds resulted in the use of a soft-plumaged petrel (Pterodroma mollis, 300 g) as this was the largest ‘small’ bird that was available at the time of the trial. On the second occasion, two medium-sized birds were put out during the trial on 17 October 2010 to compensate for two small-sized birds having been put out on the previous trial (2 October 2010).

Table 1  Location of carcass placement for carcass detection trials undertaken at study turbines at Project West Wind during years 1 and 2 post-construction investigations.

Date of trial set-up	Turbine sub-area	Vegetation	Size class	Carcass	Scientific name
29/6/10	B05-C	Shrub	VS	Juvenile tūī	Prosthemadera novaeseelandiae
	D08-C	Pasture	S	Adult tūī	Prosthemadera novaeseelandiae
	G01-D	Shrub	M	Juvenile blue penguin	Eudyptula minor
	D03-A	Shrub	L	Adult blue penguin	Eudyptula minor
03/8/10	G02-B	Pasture	VS	Yellowhammer	Emberiza citrinella
	G03-B	Shrub	S	Adult tūī	Prosthemadera novaeseelandiae
	D02-A	Pasture	S	Soft-plumaged petrel	Pterodroma mollis
	D09-D	Shrub	L	Adult blue penguin	Eudyptula minor
19/9/10	D07-A	Pasture	VS	Blackbird	Turdus merula
	D06-D	Pasture	S	Adult tūī	Prosthemadera novaeseelandiae
	D06-A	Pasture	M	Sooty shearwater	Puffinus griseus
	B02-D	Shrub	L	Adult blue penguin	Eudyptula minor
02/10/10	F01-A	Shrub	VS	Juvenile tūī	Prosthemadera novaeseelandiae
	D09-C	Shrub	S	Adult tūī	Prosthemadera novaeseelandiae
	G09-A	Pasture	S	Morepork	Ninox novaeseelandiae
	G01-B	Pasture	L	Adult blue penguin	Eudyptula minor
17/10/10	G07-D	Shrub	VS	Juvenile tūī	Prosthemadera novaeseelandiae
	D02-D	Shrub	M	Harrier	Circus approximans
	F01-D	Pasture	M	Juvenile blue penguin	Eudyptula minor
	D04-D	Pasture	L	Adult blue penguin	Eudyptula minor
31/10/10	B02-A	Pasture	VS	Silvereye	Zosterops lateralis
	G05-D	Shrub	S	Tūī	Prosthemadera novaeseelandiae
	G03-A	Pasture	M	Kererū	Hemiphaga novaeseelandiae
	E04-A	Pasture	L	Adult blue penguin	Eudyptula minor

In most cases, the carcasses were placed out the day before a scheduled search period. Searchers were not told of the number or location of carcasses, or when the trial was being conducted. Each carcass was tied and pinned to the ground so that it could be identified as a ‘detection’ carcass and to prevent scavengers from removing it. Because not all turbines were checked the day following a carcass placement, it was assumed that all birds not detected by searchers were in fact present when the turbine was searched (i.e. the carcass had not been removed prior to the search). Checks for the detection carcasses not recorded by searchers were not undertaken and, once again, it was assumed the carcass was still present during subsequent search sessions.

Chi square tests were used to test for differences in detection rates with respect to the bird size classes and habitat type. Agresti's method of adjustment (Agresti & Coull 1998) to the normal theory confidence interval was used to calculate confidence intervals for detection rates by bird size.

Carcass removal rates

Rates of scavenging, decomposition or removal of carcasses were measured through the placement of wild (both native and introduced) bird carcasses. The important measure in these trials was rate of removal of a carcass during the interval in between the search sessions (i.e. 14 days). In order to incorporate effects of varying weather, vegetation conditions and scavenger densities, the carcass removal trials were undertaken in autumn (16–30 April 2010) and spring (13–27 September 2010). Each trial monitored the fate of 40 birds, consisting of 10 carcasses from each of the four size classes.

Carcasses were set out at 10 sites located outside of the turbine search areas (Fig. 2). These 10 sites were located systematically along the access road away from turbine search areas in locations that were safe for the field team to access. The 10 sites comprised five within pasture (sites 2, 5, 6, 7 and 9), two adjacent to pines (sites 1 and 3), two adjacent to shrubland (sites 8 and 10), and one adjacent to bush/scrub margin (site 4).

At each of the 10 sites, four birds (one from each size class) were placed at least 10 m apart. Carcasses were left in situ for 14 days to allow the observation of how quickly they were removed, scavenged or decomposed. Each carcass was photographed and monitored daily for 14 consecutive days; however, day 14 and day 4 observations are missing from the autumn and spring trials respectively due to closure of the wind farm site as a result of extremely high winds. Carcasses were then assessed again on day 21. On each occasion, the daily condition of each carcass was recorded using the following condition categories:

1.	Whole carcass: a carcass that is completely intact, is not badly decomposed, and shows little or no sign of being fed upon by a predator or scavenger.
2.	Scavenged: an entire carcass showing signs of being fed upon by a predator or scavenger, or a dismembered carcass in one location (e.g. wings, skeletal remains, legs, pieces of skin).
3.	Featherspot: 10 or more feathers.
4.	Decomposing: carcass showing signs of decomposition.
5.	Nothing: carcass completely removed, no sign remaining.

Agresti's method of adjustment (Agresti & Coull 1998) to the normal theory confidence intervals for a binomial proportion was used to calculate confidence intervals around our estimates of carcass removal rates (i.e. time after which no carcass sign is present), which were incorporated into our models for mortality.

Estimated wind farm mortality

We researched a variety of international reports which calculated wind farm related mortality rates. In general, an estimate of the mortality associated with a wind farm is obtained by adjusting observed carcass rates for removal bias (affected by scavenging) and carcass detection bias (affected by searcher detection). Several statistical models have been developed to combine these measures to produce a single figure for turbine mortality rates; however, there is little consensus over the most appropriate or accurate method. A number of models were found to differ only in the arrangement of the parameters (e.g. Jacques Whitford 2008; Anon 2009; Boffa Miskell & Golder Associates 2009).

We applied the Stony Creek wind farm model (Anon 2009) to the West Wind data, whereby the estimated mortality (M) for a subgroup of birds is estimated by:

where uM =observed mortality; SC=proportion of birds removed (from carcass removal trials); E=proportion of carcasses missed by searchers (carcass detection trials); and P=proportion of turbines searched. In addition, 95% confidence intervals for SC and E were determined, from which we calculated the 95% confidence interval for mortality. As can be seen from the formula, if SC or E are high (close to 1), the estimate of mortality can be quite unstable, as it involves dividing by a number close to zero.

The mortality estimates were based on the results obtained from the post-construction studies which were located in Makara Farm, the area known to contain the main concentration of native species. The estimated mortality rates obtained for that area were then extrapolated across the entire wind farm, including areas where native species were rarely recorded during the baseline study. Therefore, the potential mortalities occurring outside of the Makara Farm area are likely to comprise a larger proportion of introduced species.

Results

Mortality searches

Fifty-three turbine-related fatalities involving 18 taxa were recorded at the study turbines over the two years of monitoring, seven of which—harrier, southern black-backed gull, paradise shelduck, chaffinch (Fringilla coelebs), redpoll (Carduelis flammea), yellowhammer (Emberiza citrinella) and finch sp.—have been recorded in both years (Table 2). No falcon, kākā or kererū mortalities were recorded. The most frequent fatality was harrier, with 12 mortalities at eight turbines. Fairy prion (Pachyptila turtur) was the only At Risk species detected at the study turbines; however, sooty shearwater (Puffinus griseus), another At Risk species, was incidentally found at a non-study turbine site.

Table 2  Summary of recorded fatalities in year 1 and year 2 from standardised search areas around 24 study turbines at Project West Wind.

Species	Scientific name	Threat status^1	Size class	Year 1	Year 2	Total
Australasian harrier	Circus approximans	Native – Not Threatened	M	5	7	12
Chaffinch	Fringilla coelebs	Introduced	VS	4	2	6
Paradise shelduck	Tadorna variegata	Native – Not Threatened	L	1	4	5
Southern black-backed gull	Larus dominicanus	Native – Not Threatened	L	2	2	4
Mallard	Anas platyrhynchos	Introduced	L		4	4
Redpoll	Carduelis flammea	Introduced	VS	3	1	4
Finch sp.		Introduced	VS	1	2	3
Yellowhammer	Emberiza citrinella	Introduced	VS	2	1	3
Skylark	Alauda arvensis	Introduced	VS		2	2
Tūī	Prosthemadera novaeseelandiae	Native – Not Threatened	S		2	2
Dunnock	Prunella modularis	Introduced	VS		1	1
Spur-winged plover	Vanellus miles	Native – Not Threatened	S		1	1
Blackbird	Turdus merula	Introduced	VS	1		1
Fairy prion	Pachyptila turtur	Native – At Risk	S	1		1
Goldfinch	Carduelis carduelis	Introduced	VS	1		1
Greenfinch	Carduelis chloris	Introduced	VS	1		1
Eastern rosella	Platycercus eximius	Introduced	VS	1		1
Song thrush	Turdus philomenos	Introduced	VS	1		1
Total				24	29	53
Note: ^1Threat status according to Miskelly et al. 2008.

While there was no statistically significant association between mortality sign (i.e. whole carcass, partial carcass, featherspot) and bird size in the individual years (year 1: χ²=10.99, d.f.=6, P =0.09; year 2: χ²=10.57, d.f.=6, P =0.10), a significant difference was detected in the combined year 1 and year 2 data (χ²=13.55, d.f.=6, P =0.04): medium-sized birds persisted as whole carcasses significantly longer than other bird size categories (Table 3).

Table 3  Carcass state and bird sizes of year 1 and year 2 mortalities recorded within the search area of 24 study turbines at Project West Wind.

		Size class
Year	Carcass state	VS	S	M	L	Total
1	Whole carcass	7	0	5	3	15
	Partial carcass	3	0	0	0	3
	Featherspot	4	2	0	0	6
2	Whole carcass	4	1	6	1	12
	Partial carcass	3	1	1	4	9
	Featherspot	2	1	0	5	8
Total		23	5	12	13	53

The distribution of carcass detection according to month is shown in Fig. 3; a peak was recorded in November in both years. In terms of seasonal variability, while the highest number of carcasses was recorded over the spring months in both years 1 and 2 (Fig. 4), there was no significant difference between seasons (χ²=3.72, d.f.=3, P =0.30).

Figure 3  Monthly detections of avian carcasses at Project West Wind during years 1 and 2 post-construction mortality searches.

Figure 4  Seasonal detections of avian carcasses at Project West Wind during years 1 and 2 post-construction mortality searches.

Carcasses were detected at 12 of the 24 turbines in year 1, and 18 of the 24 study turbines in year 2. While D08 and G03 were the two turbines at which the most carcases were recorded in year 1 (five and four respectively), the highest number of carcasses were detected at E08 (five) in year 2.

With regard to the distance from the turbine base at which carcasses were detected, birds of different size classes were spread across the entire range of the search area, with an apparent peak at 41–50 m (Fig. 5). However, using combined year 1 and 2 data, no significant difference was found in terms of the distance from the turbine tower at which different sized birds are detected (ANOVA: F(3, 49) = 0.60, P =0.62; Kruskall Wallis: χ²=1.68, d.f.=3, P =0.64).

Figure 5  Distance from turbine tower at which avian carcasses were detected at Project West Wind during years 1 and 2 post-construction mortality searches.

It is likely that some carcasses fell outside of the 80 m search area and, as such, there is the potential to introduce bias into the results. The limitations of these search areas and the resulting implications on the estimated mortality rates are discussed later.

Carcass detection trials

Ten of the 24 detection carcasses were not found by searchers (Table 4). With regards to the different size classes, 50% of large, 0% of medium, 43% of small and 67% of very small birds were not detected (Table 5). Despite this variability, there was no significant difference in detection rates of different bird sizes (χ²=5.29, d.f.=3, asymptotic P =0.15, Monte Carlo P =0.17). In terms of habitat type, 62% of the carcasses placed in pasture were detected, compared to 55% in shrub/scrub; there was no significant difference between detection rates and habitat type (χ²=0.12, d.f.=1, asymptotic P =0.73, Monte Carlo P =1.00).

Table 4  Results of carcass detection trials undertaken at study turbines at Project West Wind during years 1 and 2 post-construction investigations.

Detected	Size class	Species	Scientific name	Turbine sub-area	Veg-etation cover	Set-up date	Days between set-up and first search	No. searches until detected
Yes	L	Adult blue penguin	Eudyptula minor	D09-D	Shrub	03/08/10	36	1
Yes	L	Adult blue penguin	Eudyptula minor	E04-A	Pasture	31/10/10	2	1
Yes	L	Adult blue penguin	Eudyptula minor	D04-D	Pasture	17/10/10	1	1
Yes	M	Juvenile blue penguin	Eudyptula minor	G01-D	Shrub	29/06/10	6	1
Yes	M	Australasian harrier	Circus approximans	D02-D	Shrub	17/10/10	3	1
Yes	M	Sooty shearwater	Puffinus griseus	D06-A	Pasture	19/09/10	3	1
Yes	M	Kererū	Hemiphaga novaeseelandiae	G03-A	Pasture	31/10/10	2	1
Yes	M	Juvenile blue penguin	Eudyptula minor	F01-D	Pasture	17/10/10	1	1
Yes	S	Morepork	Ninox novaeseelandiae	G09-A	Pasture	02/10/10	4	2
Yes	S	Adult tūī	Prosthemadera novaeseelandiae	G03-B	Shrub	03/08/10	3	1
Yes	S	Adult tūī	Prosthemadera novaeseelandiae	G05-D	Shrub	31/10/10	3	1
Yes	S	Adult tūī	Prosthemadera novaeseelandiae	D08-C	Pasture	29/06/10	1	1
Yes	VS	Blackbird	Prosthemadera novaeseelandiae	D07-A	Pasture	19/09/10	3	2
Yes	VS	Juvenile tūī	Prosthemadera novaeseelandiae	G07-D	Shrub	17/10/10	1	1
No	L	Adult blue penguin	Eudyptula minor	B02-D	Shrub	19/09/10	15	–
No	L	Adult blue penguin	Eudyptula minor	D03-A	Shrub	29/06/10	6	–
No	L	Adult blue penguin	Eudyptula minor	G01-B	Pasture	02/10/10	4	–
No	S	Soft-plumaged petrel	Pterodroma mollis	D02-A	Pasture	03/08/10	36	–
No	S	Adult tūī	Prosthemadera novaeseelandiae	D09-C	Shrub	02/10/10	4	–
No	S	Adult tūī	Prosthemadera novaeseelandiae	D06-D	Pasture	19/09/10	3	–
No	VS	Juvenile tūī	Prosthemadera novaeseelandiae	B05-C	Shrub	29/06/10	6	–
No	VS	Juvenile tūī	Prosthemadera novaeseelandiae	F01-A	Shrub	02/10/10	3	–
No	VS	Yellowhammer	Emberiza citrinella	G02-B	Pasture	03/08/10	3	–
No	VS	Silvereye	Zosterops lateralis	B02-A	Pasture	31/10/10	1	–

Table 5  Mean non-detection rates (and confidence intervals) for different bird size classes obtained from carcass detection trials at Project West Wind.

Size class	Non-detection rates	95% CI
VS	0.67	(0.21, 0.99)
S	0.43	(0.06, 0.85)
M	0.00	(0.00, 0.55)
L	0.50	(0.10, 0.90)
Combined	0.42	(0.03, 0.82)

Each turbine search occurred sometime over a three-day search period. This, combined with some sessions being abandoned due to extremely high winds, resulted in a period of 1–36 days between when the detection carcasses were placed in the field and when the turbines were searched (Table 4). All successful detections occurred either during the first or second search period subsequent to the carcass being placed in the field.

The small sample sizes used for this trial and the timing of turbine searches following detection carcass deployment are likely to be in part responsible for the wide confidence intervals that have been calculated. The limitations of these trials and the resulting implications on the estimated mortality rates are discussed later.

Carcass removal trials

Differences were observed in the proportion of birds of different size classes removed within the first 14 days (Table 6; Figs. 6, 7). In general, a higher number of smaller birds were removed within 14 days compared to the larger birds. Considerable seasonal variation was noted in the proportion of large birds removed by day 14 during the autumn (40%) and spring (10%) trials (Figs. 6, 7). Overall, most birds that were present on day 14 persisted to day 21 (Table 6).

Figure 6  Daily state of carcasses observed during autumn carcass removal trial. A, Very small. B, Small. C, Medium. D, Large birds.

Figure 7  Daily state of carcasses observed during spring carcass removal trial. A, Very small. B, Small. C, Medium. D, Large birds.

Table 6  Mean removal rate (and associated confidence intervals) of birds removed after 14 and 21 days of the carcass removal autumn and spring trials.

		Size class
Season	Days since trial began	VS	S	M	L	Combined
Autumn	14 days	50%	50%	10%	40%	38%
	21 days	60%	60%	10%	50%	45%
Spring	14 days	70%	40%	10%	10%	33%
	21 days	70%	40%	10%	10%	33%
Mean removal rate (14 days)		0.60	0.45	0.10	0.25	0.36
95% CI		(0.41, 0.84)	(0.28, 0.72)	(0.00, 0.33)	(0.13, 0.54)	(0.18, 0.61)

Again, the small sample sizes used for this trial are likely to be in part responsible for the wide confidence intervals that have been calculated (Table 6). The limitations of these trials and the resulting implications on the estimated mortality rates are discussed later.

Estimation of mortality across the wind farm site

When allowing for different size classes, an annual mortality rate of 5.83 birds per turbine was calculated for year 1 (Table 7). Given that 62 turbines are operating at West Wind, this equates to an annual mortality rate of approximately 363 birds for the wind farm comprising 308 very small, 18 small, 15 medium and 22 large birds. In comparison, analysis of the year 2 data yielded a slightly lower estimated annual mortality rate of 4.64 birds per turbine, equating to an annual mortality rate of 289 birds for the wind farm (comprising 156 very small, 28 small, 28 medium and 77 large birds).

Table 7  Annual turbine and wind farm mortality estimates (with 95% confidence intervals) for Project West Wind.

		Estimated annual mortality
Year	Size class	Per turbine	Project West Wind
1	VS	4.96 (1.25, 444.49)	308 (77.19, 27558.50)
	S	0.29 (0.12, 2.00)	17.95 (7.56, 124.18)
	M	0.23 (0.19, 0.69)	14.25 (11.58, 42.99)
	L	0.35 (0.16, 2.70)	21.98 (9.79, 167.34)
2	VS	2.51 (0.80, 232.63)	155.93 (49.51, 14423.08)
	S	0.45 (0.18, 2.95)	27.97 (11.37, 183.15)
	M	0.45 (0.26, 0.96)	27.72 (16.17, 59.53)
	L	1.23 (0.53, 8.99)	76.54 (32.75, 557.41)

Discussion

Mortality searches

The first two years of the post-construction mortality monitoring at Project West Wind recorded mortalities of 18 taxa within the study area. Harrier was the species for which the most fatalities were detected during this period. Internationally, raptor mortalities have been recorded at a number of wind farms, most often when wind farms are located near high breeding densities or where high numbers of bird pass through on migration (Madders & Whitfield 2006; de Lucas et al. 2008; Garvin et al. 2010). While this is unlikely to be the issue for indigenous New Zealand raptors, raptors are also known to fixate on prey while hunting, a behavioural trait which makes them further vulnerable to collision with structures (Madders & Whitfield 2006; Seaton 2007). In New Zealand, Stirnemann and Kessels (2008) reported harriers at Te Āpiti wind farm flying within the rotor swept area. Thus, the high number of harrier mortalities reported at West Wind are consistent with previous observations of raptor behaviour at operating wind farm sites in New Zealand and overseas.

International studies have shown a relationship between bird size and distance a carcass is detected from the turbine base. A model constructed by Hull and Muir (2010) predicted that, for large turbines such as those at Project West Wind, a peak (approximately 16%) of small bird carcasses will occur 10–20 m from the turbine base, and for large birds a peak (approximately 8%) 70–80 m from the turbine base. However, two years of data collected at Project West Wind do not support this pattern for this site; though not statistically significant, an apparent peak (30%) of carcasses was detected 41–50 m from the turbine base, irrespective of bird size. Furthermore, the mean distances (43.1 m and 43.8 m) recorded for years 1 and 2 respectively were similar, being slightly more than the 40 m blade length of the West Wind turbines.

Internationally there is significant variability in the area and pattern of searching, with both factors having the potential to introduce bias into the results through birds not being detected or occurring outside of the search area. Typically, a grid search using parallel transects is used on flat to rolling landscapes. However, at West Wind, turbines are located on narrow ridgelines and spurs, and search areas extend onto slopes on either side. Overseas, more difficult wind farm sites have used radial transects extending from the base of the turbine out to the extent of the search area (Gauthreaux 1996; AUSWEA 2005), thus allowing for searchers to move up and down slopes rather than sidling across them. Consequently, the decision to use the radial search pattern at Project West Wind was based on the practicality and safety issues of the site, while acknowledging that this method has the potential for slightly poorer coverage towards the end of each radial transect and, therefore, a potential bias in the distance at which carcasses were detected.

Carcass detection trials

The overall detection rate for these trials was approximately 60%, with the highest detection rate being for medium sized birds and the lowest for very small birds (which also had the broadest confidence interval). This pattern is consistent with Smallwood's (2007) review of detection trials which reported improved carcass detection around wind turbines with increasing body size.

There has been criticism internationally about the methodology adopted for a number of carcass detection trials (Smallwood 2007). With regard to the West Wind trials, we attempted to avoid introducing bias through randomisation of carcass placement, use of different sized carcasses, spreading the trial over a period of six months, and not providing the searchers with any information regarding the trials. However, there were a number of limitations to the methodology used at West Wind. The most significant was that not all turbines were searched the day immediately after the carcasses were set out (necessary to maintain the blind testing of searchers) and the small sample sizes that were used. These factors are likely to have resulted in a reduced carcass detection rate and contributed to the large confidence intervals calculated for both the detection rates and estimated annual mortalities. International studies have shown that the use of dogs results in significantly higher carcass detection rates compared to human searches (Homan et al. 2001; Arnett 2006; Paula et al. 2011). Given the increase in uncertainty of mortality estimates associated with the low human detection rates, we suggest that dogs be used in future avifauna and bat post-construction mortality investigations at other New Zealand wind farms.

Carcass removal

There is considerable variability in the interval between mortality searches at wind farm sites (Smallwood 2007). The duration of persistence of a carcass between mortality searches is necessary to determine the proportion of carcasses for which there is potentially no sign during the turbine searches. The results of these trials showed the general pattern of removal with regard to bird size: very small birds are typically removed from site, while larger birds were typically scavenged in situ, leaving remains present for long periods. More frequent searches reduce the variance; however, the observed removal of some birds (mostly very small) within the first 24 hours of the trial indicates that at West Wind even daily searches would not detect all carcasses.

While numerous carcass removal trials have been undertaken overseas, there has been recent criticism about how these trials have been conducted, and of factors that can introduce potential bias into the results (Morrison 2002; Smallwood 2007). The West Wind trial was designed to avoid some of these biases through the placement of carcasses over a long transect (minimum of 10 m spacings between four carcasses along a transect), by using carcasses of naturally killed wild birds as opposed to battery hens/chicks, and through the careful separation of bird sizes into categories which we feel represent the range of birds present on site.

Nevertheless, we acknowledge that use of the carcass removal rates obtained for this study is based on the results of a small sample size and, as such, are contributing to the large confidence intervals calculated for both the removal rates and estimated annual mortalities. We note that the very small bird size class had the highest removal rate, and the broadest confidence interval. Consequently, these estimates should be viewed with some caution.

West Wind mortality rate estimates

Estimated annual mortality rates for years 1 and 2 were calculated to be approximately six and five birds per turbine respectively. This equates to an annual mortality rate of 363 and 289 birds for the wind farm in years 1 and 2 respectively. In both years, the majority of these estimated mortalities are predicted to be very small birds.

While year 2 mortality estimates were relatively similar to those obtained for the year 1 data, the confidence intervals for these estimates remain large. This is particularly true for very small birds which, through the course of this study, were found to have the lowest detection rate and highest removal rate of the four bird size classes.

Despite the same carcass removal and carcass detection rates being used to model year 1 and year 2 data, and the fact that more carcasses were recorded in year 2 (29 vs. 24), the estimated mortality rate was lower in year 2. This difference resulted from the size distributions of the birds found. In year 1, 14 of the 24 birds were very small, which has the highest ‘inflation factor’ (i.e. low detection rate and high removal rate) of all the bird size groups, whereas in year 2 only nine of the 29 were very small. Conversely, three of the 24 carcasses recorded in year 1 were large birds (lowest ‘inflation factor’), compared to 10 out of 29 in year 2. An additional year's data collection will further assist with establishing the confidence intervals around these estimates.

Whilst the low sample sizes in the carcass removal and carcass detection trials gave the mortality estimates broad confidence intervals, making differences hard to detect, it should be noted that decisions about the impact of wind farms on avifauna are commonly made on the basis of similar sample sizes. The confidence intervals calculated here serve to emphasise that such decisions should be made with caution, and that perceived differences may not be real.

Conclusions

While we acknowledge a number of limitations in the results obtained from this study, the first comprehensive avian monitoring programme at an operating wind farm in New Zealand, we believe that they form a good basis on which to continue to refine appropriate methodologies. However, given the differences in sites and species composition at individual wind farm sites, we do not advocate the use of the estimates obtained in the West Wind study for other wind farm projects in New Zealand.

Acknowledgements

We would like to thank the following people for their contribution to the project: Ewen Roberston, Mike Ohs and Annette Richards (Meridian Energy Limited); members of the post-construction field team including Ben Hancock, Emma-Jane Goldsworthy, Thomas Barber, Jeannine Fischer, Joseph Azar, Nicky Fitzgibbon and Heather Constable; Barb Risi, Martin Pecher and Pen Moore (Boffa Miskell) for producing the maps and helping with the field work; Lynn Adams, Brent Tandy, Jeanine Bishop and Sandra Burles (DOC) for permit processing and assistance with obtaining carcasses for trials. A Department of Conservation (DOC) Low Impact, Research and Collection Permit was obtained, allowing for the collection and storage of native and exotic bird carcasses found on the site.