workgroup SESSIONS
This section of the meeting consisted of four Workgroup sessions, held two at a time such that each meeting attendee could participate in two of the four Workgroups. The Workgroup topics were as follows:
- Site evaluation and pre-permit research and planning: What types of avian research ought to be conducted before deciding whether a site should be developed? What methodologies ought to be used?
- Operational monitoring: Once a site is developed, what types of research can help estimate and predict the number of birds killed by wind turbines? What methodologies ought to be used?
- Modeling and forecasting, including population dynamics: What research studies will help model or forecast where wind energy developments may conflict with priority species or with large numbers of species or individuals? Are population models helpful? What models ought to be developed and used?
- Avian behavior and mortality reduction: What research should be conducted to better understand why birds are killed and whether and what technology can mitigate this impact?
The results of Workgroups (1) and (2) follow immediately. Results from Workgroups (3) and (4) are summarized below.
The main recommendations from each Workgroup were discussed at a concluding plenary session, and a short list of recommended "Next Steps" was compiled by the meeting as a whole. That list is given in the "Next Steps" section, following the Workgroup summaries.
Workgroups 1 and
2.
Site Evaluation and Monitoring
The assigned tasks for Workgroups (1) and (2), as listed above, were to identify the types of avian research needed during pre- and post-construction studies, respectively. Most of the same people participated in both Workgroups. They noted that topics (1) and (2) are closely related. One reason for this is that pre- and post-construction data need to be collected in consistent ways in order to facilitate comparisons and evaluation of impact. Therefore, the members of these two Workgroups decided to compile a combined set of conclusions and recommendations covering both topics. The combined Workgroup consisted mainly of federal and state regulatory representatives, wind energy developers, and researchers. It was noted that environmental advocacy organizations were not represented in the group.
Regulators' Primary QuestionsThe Workgroups discussed the key questions about avian - wind power interactions that need to be answered in approving and siting a wind plant, monitoring its initial operations, and deciding whether future expansion would be acceptable. In addition, the regulators attending the meeting held a side-meeting to discuss these topics from their perspective (Appendix 3).
After considerable discussion, the following were identified as the regulators' primary questions:
 |
What is the net impact on individuals and populations? How many birds are predicted (pre-operational) or observed (during operations) to be killed, by species? |
|
Is project impact reasonably minimized? |
|
Is sufficient information available for decisionmaking? |
In discussing the general types of data that would be needed, several general principles were identified:Leading (0,.08) on
|
Consistency: Data collected before and during operation of the wind plant need to be consistent and directly comparable in order to permit evaluation of impact. |
|
Statistical Validity: The study design, both before and during operation, must be appropriate for detecting and quantifying changes in the parameters of interest—bird populations and bird mortality. |
|
BACI Design: A BACI (Before-After Control-Impact) design is likely to be most appropriate. This approach can detect and quantify any changes on the wind plant from the before- to the during-operation phase, and can provide data helpful in assessing whether changes are attributable to the wind plant or to some other factor. |
|
Risk: The data need to be suitable for analyses of the risk posed by the wind plant to the species of concern. Different types of data may be needed in different areas depending on the species of main concern.Leading off. |
Many workgroup participants mentioned the need for standard observation protocols. This point was made during discussion of both site-selection issues and operational monitoring issues. It had also been discussed repeatedly during the preceding plenary sessions (see above). In addition, workgroup members knew that the interview process preceding the meeting, involving an even broader range of stakeholders, had identified a widespread desire for standard observation protocols (see the paper by A. Arnold and C. Behr in these Proceedings). Participating regulators discussed how standard protocols would help them make informed judgements about the adequacy of proposed, ongoing, and completed studies. In addition, standard protocols would facilitate comparisons among datasets obtained at different times and places. This would be desirable both for long-term studies at one site and for regional comparisons involving different researchers working at different sites.
Several members mentioned the difficulties in developing statistically reliable sampling designs while keeping study costs to a practical level. A related point is the question, "Should all potentially relevant variables be measured in every study, to facilitate across-study comparisons, or should each study focus on the key variables at that particular site?" Some participants emphasized the value of collecting the maximum amount of standardized information to facilitate across-study comparisons and analyses that were not planned until after the study was initiated. Other participants cautioned that this approach can increase overall project cost and may reduce the effort that can be devoted to the most important issues at each specific site.
Pre-Construction IssuesThe group noted that the pre-construction phase is generally divided into an initial site-screening and comparison phase, when a number of candidate sites may be examined in a preliminary way; and a follow-on site-evaluation phase involving more detailed assessment, often of only the single "preferred" site. Workgroup participants discussed the extent to which bird information should be sought and considered during the site-screening phase, and the types of studies needed during the site-evaluation phase.
Participants noted that site-selection and site-evaluation follow (or could follow) a sequence involving a gradual narrowing of focus combined with a gradual increase in the level of detailed investigation of relevant factors, including birds:
|
identify wind resource areas in the region of interest; |
|
evaluate markets, accessibility, connectivity; |
|
review existing general data on wind, land use, and birds at candidate sites; |
|
identify key species; |
|
review habitats at potentially acceptable sites; |
|
identify a preferred site; |
|
implement specific studies to evaluate habitat quality, bird abundance and bird use (along with wind potential and other factors relevant to the wind developer); and |
|
focus on species of special concern; conduct specific studies as necessary. |
Site-Screening Phase.—The group noted that bird issues should be considered early during the site-screening phase. The objective should be to identify areas with an optimum combination of wind potential, accessibility to markets, and minimal potential impacts on birds. By considering birds during the site-screening phase, it may be possible to reduce impacts on birds, reduce the need for costly mitigation, and reduce subsequent regulatory delays. Existing data or local knowledge of bird populations, movements, refuge locations, etc., can be used as an initial screen.
Some participants believed that limited field surveys are desirable at the site-selection stage to compare bird populations and movements at candidate sites. It was noted that in Minnesota, a radar study is planned to obtain comparative bird use information for various windy sites. Other participants said that, in their areas, the candidate areas were too numerous and large to allow meaningful field surveys of birds utilizing all candidate sites. Also, they noted the complications when a number of prospective wind developers are involved, and they raised questions about who could or would fund wide-ranging site comparisons.
Some attendees commented that, with or without any preliminary bird surveys, it can be useful to consider the habitats on the candidate sites in relation to general knowledge or specific models of habitat preferences for species of concern. Habitats are often less costly to survey than are bird populations. Remote sensing methods are often useful.
The types of information mentioned above as being potentially obtainable during the site-selection phase may provide an indication of relative risks to birds from alternative wind plant proposals. Even this limited information could be valuable in the site-selection process, and cost-effective to all concerned. However, general information of these types cannot be used for quantitative estimates of risk and impact.
Site-Evaluation Phase.—Participants agreed that more detailed and systematic bird studies are needed at this phase. There was general consensus that these studies should be designed to provide both
|
the data needed for permit decisions and |
|
the baseline data that would be needed for comparison with post-operational data if the wind plant is constructed. |
As noted earlier, pre- vs. post-operational comparisons are best done with a BACI study design. Given the requirements of this design, it is important to obtain pre-construction data both from the preferred site and from one or more nearby sites that can serve as control or reference sites.
Several participants emphasized that, for a meaningful prediction of risk and to provide meaningful baseline data, an intensive avian study of at least one year's duration is required. Topics that require study or evaluation include
|
the species and numbers of birds present, |
|
the expected changes if a wind plant is developed (including the expected incremental mortality), |
|
how those changes and additional deaths will affect the populations (especially of species listed as endangered or threatened), and |
|
the potential for mitigation. |
With the possible exception of the need to estimate direct collision mortality, these requirements are not unique to wind power proposals. The same considerations apply in predicting impacts of many other human developments. Thus, there is a broad range of experience in how to conduct this type of work. However, given the emphasis on estimating potential collision mortality at a wind plant, we need more specific data on local and migratory movements than would be required to assess impacts of most other types of developments.
Participants noted that the level of concern will vary depending on the types and numbers of birds present in the preferred area. If it is obvious that few birds are present, and no listed ones, pre-development studies can probably be less detailed and lengthy than would otherwise be needed. However, if the wind plant is subsequently built, this approach may not provide sufficient baseline data for a conclusive demonstration of "no effect".
Operational IssuesThe group identified the following as being among the key operational issues regarding the impacts of an operating wind plant:
|
Has the wind development affected birds? If so, how? Are the
species of concern at greater risk now? How much greater?
|
||||||
|
Do these measures differ statistically from corresponding pre-construction measures? Was the sampling adequate to detect a difference if it occurred? | ||||||
|
Does the actual impact match the impact predicted when the development was being planned? What are the discrepancies? | ||||||
|
If impacts are greater, less or otherwise different than predicted, what is the appropriate response, both in terms of regulatory action and research? | ||||||
|
Are imposed conditions effective in minimizing impacts? If not, what is the process for amending the operating conditions? | ||||||
|
How can monitoring results from one wind development be applied to subsequent developments? | ||||||
|
When a number of wind plants are being monitored, how can overlap in monitoring studies be minimized and efficiency maximized? |
To help answer these questions, workgroup participants suggested that standard protocols should be established and adopted to improve consistency.
Most participants felt that the best way to estimate impacts was to apply the same research protocols for operational monitoring as for pre-construction studies. To make this possible, a Before-After-Control-Impact (BACI) design needs to be established at an early stage, when pre-construction studies are implemented. This means that both the prospective wind plant and appropriate control sites need to be studied before construction, with consistent follow-up studies on the same sites after construction.
When pre-construction studies are included in the monitoring design, they establish baseline data for longer-term monitoring and they allow rigorous testing of treatment effects (impacts). When monitoring begins only in the post-construction period, impacts of wind developments can be assessed only by inference, not by rigorous statistical testing. Standardized data collection methods are desirable to allow researchers to combine data and analyses across time and also from different sites, thereby allowing stronger conclusions.
RecommendationsMethodological Guidebook.—The combined Workgroup on pre- and post-operational monitoring identified, as its highest priority recommendation, the need for development of a guidebook on study designs, protocols, recommended "metrics", and related statistical issues. Suggested approaches included
|
preparation of draft papers on these topics, |
|
review by technical experts and other stakeholders, |
|
a follow-up workshop to build consensus, and |
|
a series of regional workshops to discuss the application of the suggested methods in particular parts of the U.S.A. |
There also needs to be
|
a process for updating the recommendations based on experience. |
In developing the recommended protocols, the first step would be to compare methods currently in use and determine where there are differences. In these cases, it will be important to build consensus about whether to recommend a single standard procedure. If so, the best features of all existing protocols should be taken into account. In other cases it may be appropriate to suggest two or more alternative approaches, or to allow certain deviations from a protocol's recommended practice. When alternative approaches are listed as being acceptable, the guidelines should indicate what additional steps are needed to ensure comparability of the data obtained via alternative procedures.
The "metrics" to be recommended as standards need further discussion. There will probably be at least three categories of metrics: utilization measures, mortality measures, and population measures. In addition, there will often be a need for measures of ancillary variables, such as habitat, weather, and prey base. Recommended measures should be as simple as feasible. There is a need to develop specific guidelines about metrics and sampling designs to avoid inefficiencies, missing data, and non-comparable results.
Other Methodological Recommendations.—Other methodological recommendations identified by the Workgroups included the following:
|
Conduct studies to assess the effectiveness of techniques recommended in guidebook. |
|
Develop protocols for coordinated radar, visual and electro-optic observations of bird movements, taking advantage of the complementary strengths of these methods. |
|
Evaluate unbounded vs. fixed-distance point counts, and point counts vs. transect methods, for studying bird utilization of wind plants (pre- or post-construction). |
Workgroup
3.
Modeling, Forecasting
and Population Dynamics
This Workgroup was convened to discuss the following topics, and to make recommendations to the Avian Workgroup:
What research studies will help model or forecast where wind energy developments may conflict with priority species or with large numbers of species or individuals? Are population models helpful? What models ought to be developed and used?
In assessing the impact of turbines on resident and migratory bird species, regulators may want to use statistical models that incorporate site-specific information to predict the answers to key questions about potential impacts on birds. Models are useful because they predict potential impacts based on defined assumptions, functional relationships, and available data. Also, they can provide indications of the confidence that can be placed in the resulting predictions. Regulators can use this information to help evaluate permit applications and to assess whether permitted sites are in compliance with their permits.
The Workgroup was composed primarily of scientists and regulators. It began by identifying some of the most important questions that are commonly posed by regulators, and then highlighted possible modeling efforts and discussed their statistical reliability. The group felt that the objective should be to develop models that help address questions important to regulators, and to do so with models that provide an adequate level of precision and confidence based on the minimum feasible number of parameters. The group distinguished between models that are primarily literature-based and those that are data-intensive.
Questions that Modeling Could Address
The workgroup focused on how models could help regulators answer their most important questions about past and potential avian impacts, including
|
How many birds and which species will be killed by a proposed wind farm development? |
|
Are any of these species potentially critical species? |
|
Are there potential sites where less impact on avian species is expected? |
|
How many deaths from wind turbines could an avian population sustain? |
|
Are there technological improvements that could reduce the deaths and impacts? |
The group discussed whether models exist that could answer these questions and, if not, whether model development was feasible.
It was recognized that, for each question, there is tension between what regulators need to know and what science can confidently predict. One problem, according to regulators, is that the desired accuracy of mortality predictions may change subjectively depending on the species in question or on local politics.
Literature-Based vs. Data-Intensive Models
Several participants stressed the need to have adequate data to develop and test the validity of the models. Without data, they said, modeling efforts would be an inefficient use of resources. Other participants noted that literature-based models could be useful even when there are important data gaps, for example in identifying and ranking the data gaps, and in guiding research design. Preliminary models might also be useful for preliminary evaluations of the likely severity of the problem in locations where few or no data on bird-wind turbine interactions are presently available.
Data-Intensive Models.—Data-intensive models are based on scientifically collected data, often collected specifically for use in the model. Group members said that these models would be useful because they would predict a particular species' mortality attributable to a wind farm development. These models would introduce scientific rigor into the prediction of avian mortality and its potential impact on a population. In addition, members commented, the model predictions would be bracketed by a confidence region so that regulators could take into account the potential uncertainty in the model prediction.
Examples of data-intensive models that the group discussed were population dynamics models and flight behavior models. (a) Population models would be useful for determining a species' population integrity threshold. These models could be used to assess whether a wind development might have a population effect. (b) A flight behavior model could simulate bird flight and turbine air flow to better understand how different species avoid turbine-related injury, and how various potential turbine modifications might affect predicted mortality.
Workgroup members briefly discussed the advantages and disadvantages of different designs for data-intensive models. One person suggested that equilibrium models should be developed initially to establish the relationships among parameters. Simulations could then be run to allow researchers to test the sensitivity to various perturbations and random effects. Others argued that disturbance models better reflect natural variation and are currently favored by most researchers.
One Workgroup member noted that the results from any model depend entirely on data and that, in the case of avian impacts, current data are insufficient to produce reliable models. From this viewpoint, the goal of data-intensive modeling efforts should be to determine what data to collect and then to collect those data, not to build a model that uses available data. In this light, the group discussed how data-intensive modeling may not answer the main questions in the short term. Potential first steps in modeling are resource selection models that anticipate avian mortality by identifying habitats that specific species are likely to utilize.
Literature-based Models.—The group discussed literature-based models as a way of starting to meet immediate information needs. These models would utilize information that already exists on bird distribution, habitat or resource selection, seasonal occurrence, migration, etc. These models could illustrate where impacts are most likely to occur and on what species, and could help determine the priorities that should be assigned to different data-collection efforts. Regulators and wind developers could use this information, along with wind resource data, to help direct baseline data collection. Also, regulators mentioned that these models would help identify the priority habitats for endangered species, and would assist in formulating appropriate recommendations to industry.
One participant noted that developing a model from these types of data involves combining several data sets, predicting outcomes, testing for accuracy, and making necessary adjustments. Some people might not view the model as "scientific", but it would help regulators around the country recognize and plan for potential conflicts. It could be helpful to organize the data via a GIS system.
Several members stated that population impacts may not need to be modeled explicitly if there is some procedure for temporary shutdown of turbine operations in circumstances with significant mortalities. They felt that much could be learned from hunting-control policies in states where regulators may limit the number of permits or may close hunting seasons if a single threatened species is killed. One group member said that upper limit "models" could be developed that represent the most conservative assessments of potential population impacts.
The group identified the following categories of models that could be helpful in this field:
|
Preliminary models to identify key data gaps and guide data collection; |
|
Geographic/habitat selection models that predict, for priority species, the numbers of birds in different areas during various seasons, overlain onto maps of wind resource potential; |
|
Population dynamics models that predict the effects of specified mortality levels on the populations of priority species; and |
|
Flight behavior models that might help identify beneficial technical changes in wind plant or turbine design. |
There is a general correspondence between these potential types of models and the previous list of key questions.
Currently, data collected at proposed and operating sites around the country are not always comparable, and may be difficult to combine or compare. This workgroup suggested establishing data collection protocols to standardize data for future use in developing and testing models. Improvement of data collection protocols and continuing basic research at various sites will eventually yield better models.
Workgroup 4. Avian Behavior and Mortality Reduction
IntroductionThis Workgroup was convened to discuss the following topic, and to make
recommendations to the Avian Workgroup:
What research should be conducted to better understand why birds are killed, and
whether technology can mitigate this impact, and if so what technology?
It is generally accepted that turbines can kill birds, but it is largely unknown Ø why certain birds that approach turbines are killed whereas others survive, and Ø how to lower the proportion killed. Scientists have many hypotheses about bird behavior near turbines, and regarding how visual and audio deterrents and siting plans could reduce mortality. However, as yet there has been little research to assess these effects statistically. One of the problems facing researchers is the infrequency of bird kills at turbines. This infrequency makes it quite difficult and costly to design a study that will collect sufficient data for a meaningful statistical analysis of the causes of fatalities, or of the comparative fatality rates with different turbine characteristics.
Interviews with stakeholders conducted by Resolve before the meeting (see paper by A. Arnold and C. Behr in these Proceedings) showed that stakeholders had three main types of questions in the area of avian behavior and mortality reduction:
|
questions about bird movements (migratory and local) and their interactions with wind plants, including collision risk by night and day; |
|
questions about surrogate variables that might be studied to supplement difficult-to-conduct studies of mortality; and |
|
questions about how to mitigate, including |
|
turbine locations vs. habitat, topography, etc., |
|
turbine design: tower design, perches, rotation rate, and |
|
deterrent measures: audio, visual, other. |
The Avian Workgroup scheduled this workgroup to discuss what areas of research should be given priority. The group consisted of regulators, industry representatives and researchers. The group discussed macro-level issues, e.g. routes of migratory and local movements; how to site and design wind plants to reduce mortality. However, more time was spent on research needed to address micro-level questions, e.g. what types of avian behavior are related to turbines; how do birds learn to avoid turbines; how can turbines be designed to reduce mortality. The group discussed the types of data that are needed, and how surrogate variables that are easier to quantify than fatality rates might be used to better address micro-level impacts.
Even with the limited time, many ideas were offered and some priorities for future work were suggested. These areas included
|
developing a basic framework for understanding the etiology of collisions, i.e. what are the relevant factors? |
|
understanding utilization of the zone of risk by resident and migratory species; |
|
researching whether resident and migratory birds adapt to turbines; |
|
examining the impact of different mitigation technologies on avian behavior near turbines; |
|
identifying applicable new technologies, e.g. through review and discussion with workers in related fields; and |
|
modeling mortality at wind resource areas using surrogate variables. |
During the discussions, workgroup members suggested research ideas, feasibility, and potential gains and analytical costs. Several also raised the importance of being cognizant of potential differences among different species; research should reflect this. In addition, several members mentioned that, because collisions occur rarely, it will always be difficult to fully understand the reasons for collisions. In response to this problem, surrogate measures of avian utilization and behavior were discussed.
Framework for Understanding Avian CollisionsOne of the major areas of discussion involved developing a preliminary model hypothesizing the factors that affect collision probability. While such a framework would initially rely on untested hypotheses, it could be reevaluated as data are analyzed and new research is conducted. This framework, based on the best available scientific judgment and information, would help all parties discuss and potentially agree on the most significant factors to research. The group suggested that the next steps in developing this framework would be to establish a group to draft a white paper on the topic, followed by a review process and possibly a workshop.
Some of the factors that would be considered in this framework include learning and adaptation, prey base near wind farms, species' behavior with regard to collision probability, migration routes, night activity, and weather effects on visibility and avian behavior. The framework could be useful in at least two ways: (1) It would help direct research design and data collection efforts at particular sites. (2) As additional research is conducted to evaluate the major assumptions built into the framework, the framework would evolve and become more realistic. Eventually, the framework could develop into a predictive model when the key assumptions are validated. It was noted that Dr. Vance Tucker of Duke University had begun developing a model of bird/wind turbine interactions that may be start in this direction.
One illustration of the value of such a framework in helping plan field research is as follows: If we expect that collisions would occur most frequently at night, then research should concentrate on utilization and abatement strategies for nocturnal species. Likewise, if migratory birds are held to be at the greatest risk, research should focus on their use of habitat near the wind resource area, emphasizing their behavior in close proximity to turbines. Data collected during this research would be useful both on a site-specific basis and in improving the framework model for future applications.
Migratory Bird Utilization of Zone of RiskThe need to understand the specific impact for migratory birds was discussed. Some of the issues are different for migratory birds than for residents. An important hypothesis that remains untested is whether migratory birds are more or less likely be killed than are resident birds. Migrants may have briefer exposure to the wind plant, but are likely to be unfamiliar with it and with the risks. One member of the group asked whether wind resource areas affect migration routes by altering habitats that might be used by migrants.
Collisions of migrants with turbines are expected to be more common in some areas than in others. Experience in California is unlikely to be representative for some other parts of the U.S.A. In corridors heavily used by migrants, higher collision rates are possible, as shown in Europe (e.g., Winkelman 1995).
Several members noted that a starting point for anticipating impacts could be gained from existing data on migration routes. Airport or weather radars were also mentioned as a possible source of new information about general migration routes. Several participants questioned whether these data would be useful given the limitations of long-range radars (see the paper by B. Cooper in these Proceedings) and suggestions that migration routes can shift between years. However, the consensus was that broad-scale radar data from long-range radars could provide information on relative utilization of different migration routes, although not discriminating low from higher altitude migration. Once one or more specific sites of interest to wind developers are identified, higher resolution shorter-range techniques are available for site-specific studies of migration patterns (see the papers by B. Cooper and S. Gauthreaux in these Proceedings).
Several members discussed neotropical migrants. Since these species often migrate at night, their migrations are most readily studied by radar. However, radars cannot discriminate different species of nocturnal passerine migrants. It was also noted that estimating collision mortality of small birds can be difficult because the bodies are often scavenged quickly, as well as being more difficult to locate than bodies of larger species. As a surrogate for understanding the behavior of migrating neotropical migrants around turbines, members suggested examining the voluminous literature on their collisions with (or avoidance of) other tall structures such as television, radio, and water towers.
One member commented that it was important to distinguish between macro- and micro-level analysis. Macro-level analysis of migration examines siting issues and may use long-range radar and published studies as data. Micro-level analysis focuses on the causes of collisions and technological factors that could reduce the number of collisions. Different types of data are needed for these two levels of analysis, and researchers need to collect data of the type appropriate to the questions being asked.
Potential Adaptation of Resident Species to TurbinesResident species would have more exposure to turbines than would migratory species. Although this increased exposure could lead to greater risk of collision, it could also provide an opportunity for resident birds to learn about turbine layout and dangers. Some members of the group questioned whether adaptation by resident birds confounds experimental design. Members hypothesized that observations of near-fatal encounters and of reactions to "unnatural" stimuli, as investigated through deterrent experiments, may yield important findings. Others mentioned that the effects of natural selection and adaptation (learning) on bird responses to wind plants and to specific turbines are largely unknown.
One suggested approach to measuring adaptive effects would be to reverse technological treatments part way through a study (initial control units become treated units; initial treatment units become controls). There were suggestions that results would probably be species-specific. Planned perching studies and Kenetech Windpower's experiments with pigeons may provide useful data. To examine the hypothesis that different species react differently to turbines, one participant suggested that a first step could be to examine "death lists" to begin to identify commonly- vs. rarely-struck taxa and behavioral guilds.
Effect of Different Mitigation Technologies on Behavior and MortalityParticipants suggested that the aforementioned "framework" could help identify and assign priorities to factors meriting tests or studies, and to research that could help answer the relevant questions. Elements that should be considered for inclusion on the priority list include turbine design (e.g., tower type, rotation rate, blade painting, and audio deterrents) and wind plant design (e.g. turbine layout, topography, prey base, and vegetation/habitat). The group noted that other types of bird deterrent measures should also be considered for efficacy testing, e.g. the possible use of strobe lights at wind resource areas to alert but not attract nighttime migrants.
Several group members expressed interest in having more meeting time to fully discuss specific recommendations for mitigation research. In general, it was noted that hypothesis tests would involve using data collected via survey methods and statistical designs proposed earlier in the workshop. These approaches would be used to examine the effects (on mortality, surrogate variables, or both) of changes in individual turbines and in overall wind plant design. Several participants also suggested that much could be learned from past experience in assessing collisions with other structures. Results from water towers may be of special relevance because they often have a similar height and outline as a turbine and its rotor plane.
Identifying Applicable New TechnologiesWorkgroup members felt that much could be learned from people designing and conducting studies in other related fields. Relevant work in other fields could include
|
statistical methods useful in work on other kinds of rare events (not necessarily involving birds or collisions; e.g. Green and Young 1993); |
|
studies of bird collisions with other types of structures (reviewed by Weir 1976; Avery et al. 1980; EPRI 1993; Bevanger 1994; Hebert et al. 1995); |
|
deterrent measures used in attempts to keep birds away from other areas (airports, aircraft, crops, tailings ponds, oil spills, etc., as reviewed by Lucid and Slack 1980; DeFusco and Nagy 1983; Payson and Vance 1984; B.S.C.E. 1988; Knittle and Porter 1988; Marsh et al. 1991; Koski et al. 1993); and |
|
remote data recording and triggering devices useful in documenting rare events and/or triggering deterrent devices (Kenetech Windpower, NREL, and EPRI have cooperated in some preliminary work on these techniques). |
It was suggested that a workshop bringing together appropriate technical "problem-solvers" from different fields might be the best approach to identify relevant technologies used in other fields.
Modeling Behavior and Mortality Using Surrogate VariablesA recurring discussion topic was the need to identify surrogate variables that are easier to study than mortality of birds at turbines. To be useful, a surrogate variable must be strongly correlated with the variable of interest (here mortality), and there must be a way to confirm that they are strongly correlated. The latter is difficult for the same reason that surrogate variables are desirable—because of the rarity of mortality. Variables that were mentioned as possible surrogate variables included occurrence and circumstances of perching on or near turbines, distance to active nests, measures of prey base, time spent within the zone of influence, and number of flights through or near the rotor plane by day or by night, by resident or by migratory birds.
Researchers need to know the circumstances that surround collisions and information on the collisions themselves. To collect baseline data, several participants suggested video systems. Ideally, these would be coupled with image analysis, audio, vibration, or other sensors Ø to avoid recording during most periods of "no bird activity" and Ø to provide more information on bird approaches to turbines and on collision events. The possibility of modeling the physics of bird flight and wind streams near turbines to better understand collisions was again mentioned. (This suggestion was also mentioned in Workgroup 3, and during the "Framework" discussion by the present Workgroup.
One of the problems with identifying a surrogate variable for mortality is that, since collision deaths occur infrequently, it will not be easy to demonstrate that there is a reliable correlation. Demonstration of good correlation is important because variables that might initially seem to be closely associated with collision risk may involve various complications. It was noted that research to demonstrate the reliability of surrogate variables could be costly, with no guarantee that the variable being considered as a candidate surrogate would prove to be suitable.
RecommendationsThe group recommended, as a first priority, the development of a framework or conceptual model to help understand the factors affecting the links between wind plant and turbine design, avian behavior, and collision risk. This framework would discuss the scientific opinions on possible factors and would identify the hypotheses deserving most immediate attention. To develop this framework, the group proposed that several people should draft a paper, distribute it, and then attempt to reach consensus. A follow-up workshop might be useful. The initial goal would be to develop a framework useful for focusing thinking and for planning data collection and research. As data are collected, the framework would be refined and would eventually evolve into a more specific model. The framework would help regulators identify the specific statistics that they should rely on, and the data that should be collected to evaluate a proposed site and a proposed wind plant design.
Another level of research would be to continue studying avian mortality at actual or proposed wind plants, and identifying and testing potential surrogate variables. This type of work is often done when evaluating a site for potential development or the mortality at a currently operating site. The group felt that the proposed framework could be useful in helping to design research on the effects of wind plant and turbine design on mortality and related surrogate variables. After the key variables are identified, this work could be implemented via procedures such as those discussed under "Protocol II" in the paper by M. Morrison and H. Davis.
A third proposed activity would be to sponsor a workshop to bring together appropriate technical "problem-solvers" from different fields to discuss their experience in using technologies to overcome other related problems in research and data collection.
All three of these issues were raised in the final plenary as important areas for further research.
