Chiroptera (English)‎ > ‎Methodology‎ > ‎Wind farms‎ > ‎

Critical review of the wildlife monitoring protocols in wind farms: state of the art and proposals for improvement

Abstract:

One of the biggest environmental impacts caused by wind farms is the direct mortality of flying fauna (birds and bats) inflicted by wind turbines. For this reason, the Administration usually requires a mortality monitoring during at least the first years of operation of each wind farm, as well as a previous study of the bird and bat populations within the framework of the environmental impact study in order to properly assess the likely impacts of the wind farm. However, the validity and effectiveness of the established protocols to perform this monitoring is rarely critically evaluated, so the obtained results can be distant enough to be representative of the reality. This article aims to critically review and highlight the necessary improvements to optimize the design of environmental monitoring plans of wind farms.

Keywords:

Wildlife monitoring, wildlife mortality, environmental impact, methodological review, wind farms, wind turbines, birds, bats.

Introduction

Wind farms can cause significant impacts on wildlife, primarily:

  • Mortality of birds and bats by wind turbines.
  • Mortality of terrestrial fauna during construction.
  • Habitat degradation or loss.

A general review of bird mortality at wind farms in Europe [11] shows the following figures: in Spain (6 wind farms), 21.2 ± 22.5 (4-64) individuals per turbine per year; in Belgium (7 wind farms), 16.0 ± 15.1 (1-43); in the Netherlands (6 wind farms), 16.6 ± 13.1 (2-34); in the UK (1 wind farm), 1.3, while a review for bats in southern Europe [17] the figures were: Portugal (28 wind farms), 2.7 ± 5.9 (0-26.3) individuals per turbine per year; in France (3 wind farms), 39.3 ± 29.0 (6.8-62.7); in Greece (9 wind farms), 14.9 ± 19.6 (1.6 to 13.9). On the other hand, SEO [13] estimated that an average between 300 and 1000 individuals (considering birds and bats jointly) die per turbine per year (one death per turbine every 9-29 hours) in Spain.

Mortality numbers may differ significantly between wind farms, so it is not advisable to extrapolate values.. It is also necessary to note that what really matters is not so much the overall mortality numbers but the impact on species population, since a small number of deaths may seriously affect rare, threatened or sensitive species.

Study of wildlife populations

It enables the detection and monitoring of endangered and protected species that may be affected, the estimation of the likely impacts of the wind farm on wildlife as part of the environmental impact study of the project, and the evaluation of the impacts caused during the operation phase.

Sampling frequency

Although the weekly sampling frequency is not exceptional, it is often carried out biweekly or monthly. In order to estimate its impact on species detection, we chose four wind farms (located in the Iberian peninsula) whose bird monitoring was conducted by TAXUS MEDIO AMBIENTE with a weekly frequency, and the data was resampled with biweekly and monthly frequency in order to study the consequences of reducing the sampling frequency on rare and endangered species which precisely can be the more seriously affected by the installation of the wind farm. (Figure 1)

The reduction of the frequency to biweekly involves the loss of the records of between 7 and 12 species, of which between 1 and 3 are endangered species. In turn, the reduction to a monthly frequency implies the loss of between 8 and 16 additional species (between 31% and 48% with respect to a weekly sampling). Taking into consideration that the general recommendation to establish the optimal level of sampling intensity is by a cumulative curve of the number of species in relation to the sampling effort (in which the optimal level is that from which the growth rate in the number of species asymptomatically approaches zero), is possible to conclude that the minimum required level of sampling effort is at least weekly.

Data quality

With the indicated frequency, the entire wind farm must be prospected. Prospecting should be done by transects on foot if the extension allows it; in the case of observation stations, all wind turbines must be less than 1 km from one station.

In order to obtain quality data, it is imperative that the field work be done by people with knowledge and experience identifying visually and by sound each group of wildlife, as well as through their tracks and signs of activity.

Additionally, it is important to have quality equipment (mainly optics in the case of birds and ultrasonic detectors in the case of bats).

Control zone

Due to the interest of making a pre-post analysis of the impact of wind farms (in order to assess the real impact of its construction and operation on wildlife populations by comparing the data collected during previous studies with those obtained during the operation phase), it is highly recommended to establish control zones (nearby locations with similar characteristics but in which there is no scheduled human activity that alters their conditions).

Estimate of expected impacts

The purpose of the environmental impact assessment procedure is to identify and assess the likely environmental impacts of the future wind farm. In the case of birds and bats it seeks (among other things) to estimate the expected direct mortality caused by wind turbines, and as the mechanism for doing this is by using of collision risk models (because there is no simple relationship between abundance and mortality), the field work should be focused on collecting the data needed for its calculation. The most used model for mortality estimates is the Scottish Natural Heritage model developed primarily by William Band. Like any model of this nature, the final results are very sensitive to the quality of the input data, and in this specific case the two most significant variables [2] are: the numerical values ​​of time in the risk zone for each species obtained from field sampling, and the "evasion rates" used to adjust the raw results of the model to reality. Regarding this last factor, there is only data for a few species, a fact which forces to assume that it can be extrapolated to the rest, and besides its calculation is strongly affected by the first. In order to standardize the collection of field data and avoid bias arising from the use of different criteria, the Scottish Natural Heritage itself published a guide [24] which provides a methodology based on observation points with a minimum sampling effort of 36 hours at each point at each period in which it divides the year (breeding, migration, no breeding) and counting flights in a range of 200 m on either side of the line of wind turbines. Regarding the methodology for bats [1], Eurobats [15] indicates a detailed work schedule using hand-held bat detectors in addition to the use of fixed recording stations.

When the estimates of mortality are used as a criterion for assessing the environmental impact of a wind farm, it must be assessed at the level of impact on the populations of each species (with special attention to endangered or rare ones) by using demographic models in order to estimate the impact of this new cause of mortality (population viability analysis).

In addition to the use of collisions forecast models, it is also recommended to apply spatial vulnerability indices [12, 20].

Monitoring of direct mortality

When studying the direct mortality of the flying vertebrate fauna (birds and bats) in a wind farm, it is necessary to note that the mortality detected by searching for carcasses on the field is only a fraction of the real mortality. On the one hand, some of the accidents do not cause death in situ, but the injuries allow the animal to fly away and die at a distance from the wind farm some minutes or hours later (ex situmortality). On the other hand, from the moment that the body falls to the ground, scavengers, decomposers and weather agents begin to act causing its disappearance. Finally, the technicians’ efficiency to detect carcasses is not perfect, and often within the search area there are different vegetal covers with different carcass detectability. Based on these aspects, the following relationship can be deduced:

R ~ E, C, P, D

Where:

  • R = Real mortality.
  • E = Ex situ mortality.
  • C = Found carcasses.
  • P = Temporal rate of disappearance of carcasses.
  • D = Efficiency of carcasses detection by the technicians.

D (technicians’ efficiency to detect carcasses) and P (temporal rate of disappearance of carcasses) variables vary depending on factors specific to the carcass (such as size), the environment (characteristics of the vegetation cover) and dynamic agents (variations throughout the year in the activity of scavengers, decomposers and weather agents), and in the first case, on the technicians.

A critical factor to adjust the estimates of real mortality to reality is the sampling frequency of carcass searching in the environmental monitoring program of the wind farm, which (due to the shape of the temporal pattern of carcass disappearance and the associated accumulation of uncertainty) is recommended to be at least equal to the period of time in which 50% of carcasses disappear according to the results of disappearance experiments. According to the bibliography consulted, this period is 2.89 ±2.35 (0.8-8; n=7) in birds and 2.17 ±0.45 (1.7-2.8; n=4) in bats. (Figures 2-3)

The sampling surface around each monitored turbine is generally defined as a square whose apothem is equal to the maximum height of the wind turbine (tower height plus rotor radius). The square rather than circular design facilitates the realization of linear transects for carcass searching.

A general model to calculate the estimates of real mortality is:

Ck,l,m,n = Rk,l,m · ( 1 – Ek ) · P(t)k,l,m · Dk,l,m,n

Where:

  • P(t) = Average proportion of carcasses which persists at time t.
  • k = Referring to the k species, or otherwise the k group of species defined according to their similarity in size.
  • l = Referring to each different vegetal cover existing on the sampling surface.
  • m = Referring to each differentiable period of the year depending on variations in the temporal rate of carcass disappearance and on the technicians’ efficiency to detect carcasses.
  • n = Referring to the technicians’ different efficiency to detect carcasses.

Thus, for the wind turbine xi in the interval of time between samplings   Δti = ti – ti–1  the estimated total real mortality is equal to the summation of the estimated partial real mortalities for the different combinations of k, l, m, and n

R(xi,Δti) = Σ(xi,Δti) Rk,l,m,n

The efficiency of carcass detection (D) and the temporal rate of carcass disappearance (P) need to be properly estimated experimentally for the different combinations of the following factors: species or group of species (k), vegetation cover type (l), period of the year (m) and the teachnician in charge of the search (n). To this end, field experiments consistent in the use of dead wild birds and bats of various sizes (from deaths in wind farms and roads; otherwise animals bred in captivity such as mice, quails and other cage and pen birds, all fresh and brown) randomly scattered on the sampling surface of wind turbines must be conducted. These experiments must be conducted with adequate sample size, randomness and sampling frequency (at least once in each season of the year). In order to prevent the possible "feeling under examination" effect in the technicians who participate in the experiments (which may artificially inflate detection efficiency), it is possible to extend the experiments throughout the routine sampling of carcass searching rather than making one-off experiments. Because of the significant differences found occasionally between close locations, the values ​​obtained in nearby wind farms must not be used. If any vegetation cover cannot be sampled or it has a zero carcass detection efficiency (D), the real mortality is estimated for the remaining portion of the sampling surface and the result is extrapolated to the total. Since extrapolations should be avoided in order to achieve the most accurate estimates, it is recommended that an area around each wind turbine with a radius equal to the maximum height of the wind turbine (tower height plus rotor radius) be maintained without vegetation or with low vegetation (<10 cm high).

Due to insufficient data, we suggest using a value of ex situ mortality (E) equal to 10% (0.1).

In order to estimate the real mortality according to the specified model, as value of P(t) must be used the average proportion of carcasses which, according to the data from the disappearance experiments, persists for a time equal to half of the interval between each samplings ( t = Δti / 2 ). Several authors have proposed different alternative models (see for example those collected in [5], [8], [14]); an adjustment to the model described above which has demonstrated an outstanding ability to estimate (even better than more complex models) when the sampling frequency of carcass searching is not optimal is the proposal of Gareth Jones et al. [6] with the additional indications of Julien Cornut & Stéphane Vincent [14]:

P(t) = exp ( – 0.5 · min(Δti, Δtefect.) / ) · min(1, (Δtefect./Δti))

Where:

  • = Average time of cascass persistence (according to the disappearance experiments).
  • Δtefect. = – log (0.01) ·

In order to experimentally test the utility and robustness of the proposed model in real situations, we take two very detailed and complete real mortality data sets (with high and medium average rates of carcass disappearance) and generate 42 scenarios by resampling these data with different sampling frequency of carcasses searching (daily, 2, 3, 4, 8, 16 and 32 days) and reducing the real mortality by random selection (25, 50 and 100 dead individuals per year). We apply the different models cited in the literature and assess the results according to the level of adjustment to the real numbers of dead individuals. The results show that the above proposed model is the one that provides in a greater number of scenarios (N=36; 86% of total) an estimated mortality value within the range of ±30% of the real mortality value and an estimated mortality value that is >60% the real mortality value in a smaller number of scenarios (in just 1 scenario which is the most unfavourable scenario of all). The Manuela Huso’s model [16] provides significant results too (values within the range of ±30% in 79% of scenarios, and >60% in 2 scenarios). (Tables 1-4)

Bibliographic references:

[1] Alexis Puente Montiel. Recomendaciones para el seguimiento de murciélagos en la evaluación de impacto ambiental de parques eólicos. Online publication in http://www.miscelaneanatural.org/estudios-naturales/recomendaciones-para-el-seguimiento-de-murcielagos-y-aves-en-parques-eolicos/Recomendaciones_para_el_seguimiento_de_murci%C3%A9lagos_en_parques_e%C3%B3licos_-_1.0.2.pdf?attredirects=0 (2010).

[2] Dan Chamberlain, Steve Freeman, Mark Rehfisch, Tony Fox, Mark Desholm. Appraisal of Scottish Natural Heritage’s wind farm collision risk model and its application. British Trust for Ornithology, Research Report 401 (2005).

[3] David Mazuelas Benito. Programa de vigilancia ambiental parque eólico “Elgea-Erkilla” (Eraba-Álava). Control de las afecciones sobre la avifauna. Fase de funcionamiento - Informe final. Año 2010. Abies, Recursos Ambientales, S. L. (2011).

[4] Federico Faci Miguel. Parque eólico Acampo Armijo. Zaragoza. Estudio de la incidencia sobre la fauna del Parque Eólico Acampo Armijo. Fase II. Año 2004. Galerida Consultora S. L. (2005).

[5] Fränzi Korner-Nievergelt, Pius Korner-Nievergelt, Oliver Behr, Ivo Niermann, Robert Brinkmann, Barbara Hellriegel. A new method to determine bird and bat fatality at wind energy turbines from carcass searches. Wildlife Biology, 17 (4): 350-363 (2011).

[6] Gareth Jones, Rachael Cooper Bohannon, Kate Barlow, Katie Parsons. Determining the potential ecological impact of wind turbines on bat populations in Britain. Scoping and method development report. Final. University of Bristol & Bat Conservation Trust (2009).

[7] Jesús María Lekuona Sánchez. Uso del espacio por la avifauna y control de la mortalidad de aves y murciélagos en los parques eólicos de Navarra durante un ciclo anual. Dirección General de Medio Ambiente, Gobierno de Navarra (2001).

[8] Joana Bernardino, Regina Bispo, Hugo Costa, Miguel Mascarenhas. Estimating bird and bat fatality at wind farms: a practical overview of estimators, their assumptions and limitations. New Zealand Journal of Zoology, 40 (1): 63-74 (2013).

[9] Joana Bernardino, Regina Bispo, Paulo Torres, Rui Rebelo, Miguel Mascarenhas, Hugo Costa. Enhancing carcass removal trials at three wind energy facilities in Portugal. Wildlife Biology in Practice, 7(2): 1-14 (2011).

[10] Jon Domínguez, Francisco Cervantes, Juan Erans. Seguimiento de la mortalidad de quirópteros en un parque eólico de la provincia de Albacete: técnicas y sugerencias. I Congreso Ibérico sobre Energía Eólica y Conservación de la Fauna, Jerez de la Frontera (Cádiz, Spain) (2012).

[11] Joris Everaert, Eckhart Kuijken. Wind turbines and birds in Flanders (Belgium). Preliminary summary of the mortality research results. Research Institute for Nature and Forest (2007).

[12] José C. Noguera, Irene Pérez, Eduardo Mínguez. Impact of terrestrial wind farms on diurnal raptors: developing a spatial vulnerability index and potential vulnerability maps / Impacto de campos eólicos terrestres sobre rapaces diurnas: desarrollo de un índice de vulnerabilidad espacial y mapas de vulnerabilidad potencial. Ardeola, 57(1): 41-53 (2010).

[13] Juan Carlos Atienza, Isabel Martín Fierro, Octavio Infante, Julieta Valls, Jon Domínguez. Directrices para la evaluación del impacto de los parques eólicos en aves y murciélagos (versión 3.0). SEO/BirdLife, web version 3.1 (2012).

[14] Julien Cornut, Stéphane Vincent. Suivi de la mortalité des chiroptères sur deux parcs éoliens du sud de la région Rhône-Alpes. LPO Drôme (2010).

[15] Luísa Rodrigues, Lothar Bach, Marie-Jo Dubourg-Savage, Branko Karapandža, Dina Kovac, Thierry Kervyn, Jasja Dekker, Andrzej Kepel, Petra Bach, Jan Collins, Christine Harbusch, Kirsty Park, Branko Micevski, Jeroen Minderman. Guidelines for consideration of bats in wind farm projects. Revision 2014. EUROBATS, Publication Series 6 (2015).

[16] Manuela Huso. An estimator of wildlife fatality from observed carcasses. Environmetrics, 22: 318-329 (2010).

[17] Marie-Jo Dubourg-Savage, Luísa Rodrigues, Helena Santos, Panagiotis Georgiakakis, Elena Papadatou, Lothar Bach, Jens Rydell. Pattern of bat fatalities at wind turbines in Europe comparing north and south. Conference on Wind Energy and Wildlife Impacts, Trondheim (Norway) (2011).

[18] Peter Shoenfeld. Suggestions regarding avian mortality extrapolation. West Virginia Highlands Conservancy (2004).

[19] Robert Brinkmann. Untersuchungen zu möglichen betriebsbedingten Auswirkungen von Windkraftanlagen auf Fledermäuse im Regierungsbezirk Freiburg / Survey of possible operational impacts on bats by wind facilities in Southern Germany. Regierungspräsidium Freiburg - Referat 56 Naturschutz und Landschaftspflege (2006).

[20] Roberto Toffoli, Paola Culasso, Paolo Oberto. Wind farms and preventive evaluation of impacts on bats: a case study. II Convegno Internazionale Fauna Problematica: Conservazione e Getione, Genazzano (Italy) (2011).

[21] Anónimo. Estudio de la incidencia sobre la avifauna del parque eólico de Badaia. Programa de vigilancia ambiental. Control de las afecciones sobre la fauna durante la fase de funcionamiento. Año 2008 - Informe final. Consultora de Recursos Naturales, S. L. (2009).

[22] Anónimo. Estudio de la incidencia sobre la avifauna del parque eólico de Elgea-Urkilla. Programa de vigilancia ambiental. Control de las afecciones sobre la fauna durante la fase de funcionamiento. Año 2008 - Informe final. Consultora de Recursos Naturales, S. L. (2009).

[23] Anónimo. Estudio de la incidencia sobre la avifauna del parque eólico de Oiz. Programa de vigilancia ambiental. Control de las afecciones sobre la fauna durante la fase de funcionamiento. Año 2008 - Informe final. Consultora de Recursos Naturales, S. L. (2009).

[24] Anónimo. Survey methods for use in assessing the impacts of onshore windfarms on bird communities. Scottish Natural Heritage (2010).

Figures and tables

 

Figure 1. Result of the reduction of the sampling frequency on the number of detected species (are indicated the species no longer detected and their IUCN threatened categories in Spain).

 

Figure 2. Percentage of disappearance/persistence of carcasses in function of time.
Sources: [3], [7], [9], [19]

 

Figure 3. Average rate of disappearance of carcasses in different studies.
Sources: [3], [4], [7], [9], [10], [14], [19], [21], [22], [23]

 


Sampling frequency of carcasses searching


Daily

2 days

3 days

4 days

8 days

16 days

32 days

Global


>60%

≤30%

≤15%

>60%

≤30%

≤15%

>60%

≤30%

≤15%

>60%

≤30%

≤15%

>60%

≤30%

≤15%

>60%

≤30%

≤15%

>60%

≤30%

≤15%

>60%

30%

15%

Simple exponential

0%

100%

100%

0%

100%

100%

0%

33%

0%

0%

33%

17%

100%

0%

0%

100%

0%

0%

100%

0%

0%

43%

38%

33%

Peter Shoenfeld (2004)

0%

100%

100%

0%

100%

50%

0%

100%

33%

17%

50%

33%

50%

33%

33%

83%

0%

0%

100%

0%

0%

36%

55%

36%

Manuela Huso (2010)

0%

100%

100%

0%

100%

50%

0%

100%

67%

0%

83%

67%

0%

100%

83%

0%

50%

0%

33%

17%

0%

5%

79%

52%

Fränzi Korner-Nievergelt et al. (2011)

0%

100%

33%

0%

100%

83%

0%

100%

67%

0%

67%

33%

0%

50%

17%

17%

33%

33%

67%

17%

17%

12%

67%

43%

Proposed in this paper

0%

100%

100%

0%

100%

67%

0%

83%

33%

0%

100%

33%

0%

100%

50%

0%

83%

33%

17%

33%

17%

2%

86%

48%

Table 1. Summary of the model results. For each model is indicated the percentages of the results are within the range of ≤15%, ≤30% and >60% of the real mortality values.

 


Sampling frequency of carcasses searching

25 deaths / year

Daily

2 days

3 days

4 days

8 days

16 days

32 days


High rate

Medium rate

High rate

Medium rate

High rate

Medium rate

High rate

Medium rate

High rate

Medium rate

High rate

Medium rate

High rate

Medium rate

Simple exponential

25,60

24,07

28,17

24,08

32,78

32,54

39,34

36,07

129,00

66,38

1849,00

176,25

0,00

1941,47

Peter Shoenfeld (2004)

24,72

21,79

24,46

19,61

18,50

23,68

13,88

23,31

6,94

23,13

2,31

9,25

0,00

2,31

Manuela Huso (2010)

24,72

21,79

24,46

19,61

26,19

23,68

26,19

23,31

26,19

25,99

17,46

20,79

0,00

10,40

Fränzi Korner-Nievergelt et al. (2011)

29,63

26,39

26,46

24,22

24,27

29,78

22,34

29,78

20,59

33,53

13,63

25,02

0,00

6,12

Proposed in this paper

25,64

21,87

28,26

19,86

30,77

24,38

30,77

24,55

30,77

30,54

20,52

24,43

0,00

12,22

Table 2. Results of the models for the 25 deaths / year scenarios.

 


Sampling frequency of carcasses searching

50 deaths / year

Daily

2 days

3 days

4 days

8 days

16 days

32 days


High rate

Medium rate

High rate

Medium rate

High rate

Medium rate

High rate

Medium rate

High rate

Medium rate

High rate

Medium rate

High rate

Medium rate

Simple exponential

56,01

49,41

53,78

46,55

73,76

52,87

52,46

59,26

215,00

99,57

3698,00

308,43

3418799,42

7765,87

Peter Shoenfeld (2004)

54,07

44,73

46,70

37,90

41,63

38,48

18,50

38,29

11,57

34,70

4,63

16,19

2,31

9,25

Manuela Huso (2010)

54,07

44,73

46,70

37,90

58,92

38,48

34,91

38,29

43,64

38,98

34,91

36,38

34,91

41,58

Fränzi Korner-Nievergelt et al. (2011)

64,81

54,17

50,51

46,83

54,60

48,40

29,79

48,93

34,32

50,30

27,26

43,78

27,26

49,78

Proposed in this paper

56,09

44,88

53,94

38,41

69,24

39,62

41,03

40,34

51,29

45,81

41,03

42,76

41,03

48,87

Table 3. Results of the models for the 50 deaths / year scenarios.

 


Sampling frequency of carcasses searching

100 deaths / year

Daily

2 days

3 days

4 days

8 days

16 days

32 days


High rate

Medium rate

High rate

Medium rate

High rate

Medium rate

High rate

Medium rate

High rate

Medium rate

High rate

Medium rate

High rate

Medium rate

Simple exponential

110,42

105,16

112,67

104,33

135,23

117,95

157,38

139,13

473,00

252,24

7396,00

1013,43

3418799,42

25239,09

Peter Shoenfeld (2004)

106,59

95,20

97,85

84,96

76,33

85,83

55,51

89,90

25,44

87,90

9,25

53,20

2,31

30,07

Manuela Huso (2010)

106,59

95,20

97,85

84,96

108,01

85,83

104,74

89,90

96,01

98,76

69,83

119,55

34,91

135,14

Fränzi Korner-Nievergelt et al. (2011)

127,78

123,88

105,82

116,19

100,09

122,56

89,38

133,24

75,51

156,07

54,52

181,55

27,26

6,40

Proposed in this paper

110,59

95,52

113,02

86,08

126,94

88,39

123,09

94,71

112,83

116,06

82,06

140,50

41,03

158,82

Table 4. Results of the models for the 100 deaths / year scenarios.

 

 


 

 

ą
Alexis PM,
24 oct. 2017 10:35
ą
Alexis PM,
24 oct. 2017 10:35
ą
Alexis PM,
24 oct. 2017 10:35
ą
Alexis PM,
24 oct. 2017 10:36
ą
Alexis PM,
24 oct. 2017 10:36
ą
Alexis PM,
24 oct. 2017 10:36