the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 29 Sep 2025
Investigation on the Impacts of Smart Curtailment for Bat Fatality Mitigation in Alberta
Abstract. As wind energy continues to expand in Canada, it is increasingly important to balance power generation with wildlife conservation. For migratory bat species, the risk of interactions with wind turbines varies throughout the year. In response to environmental conditions, curtailing turbine operation during periods of higher risk has been shown to reduce bat fatalities. This study models seasonal turbine curtailment scenarios across wind farms in the Canadian province of Alberta to estimate the resulting energy and economic impacts. High-resolution weather data were used to reconstruct complete wind speed records and simulate turbine output. The modeled power output was closely aligned with real production data reported by the province's energy operator. Results indicate that curtailment outcomes vary significantly depending on wind speed thresholds, seasonal wind conditions, and curtailment duration. Across all scenarios, smart curtailment reduced energy and financial losses by 20–40 % compared to blanket curtailment, highlighting the benefits of using meteorological and behavioral triggers. These findings provide practical insights for minimizing energy loss while supporting conservation goals.
- Preprint
(1193 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
- RC1: 'Comment on wes-2025-164', Anonymous Referee #1, 07 Nov 2025
-
RC2: 'Comment on wes-2025-164', Anonymous Referee #2, 05 Dec 2025
The pre-print ‘Investigation on the Impacts of Smart Curtailment for Bat Fatality Mitigation in Alberta’ attempts to investigate the energy and economic cost of curtailment to reduce bat fatalities based on windspeed only and windspeed plus temperature and precipitation.
While the manuscript attempts to address a widespread interest in understanding the costs of curtailment to reduce bat fatalities, it does not add to the understanding of curtailment costs and the results are questionable given the flawed data and methodology used. Furthermore, the data used and results lack the appropriate context or any discussion that places the results in context of our current understanding. The Conclusion section is highly inadequate and consists of only 18 lines with one reference.
The impact of each curtailment type on bat fatalities is not addressed. It is known that all curtailment is not created equal and that adding predictive parameters to curtailment (such as temperature and precipitation) can have varying effects on fatalities depending on the location. The rational for the choice of these parameters and how it impacts bat fatalities is not provided. Additionally, we are learning that these two variables are less useful than other predictors. Overall, the manuscript shows a lack of understanding of the wind energy sector, curtailment, and bat ecology (migratory bats are not impacted by White-nose syndrome).
A review addressing all of the concerns with manuscript would be longer than the manuscript itself – therefore I will focus discussion only on the data used and touch on the methodology. Both Thurber et al 2023 and Maclaurin et al 2022 demonstrate how to properly conduct this type of work and were cited, but these examples were not followed.
When selecting data, it is important to have a thorough understanding of how curtailment decisions are implemented and select data of appropriate spatial and temporal resolution. Curtailment is typically implemented in 10-minute periods based on the average windspeed observed at each individual turbine. In other words, curtailment is most often implement at a very fine spatial and temporal resolution and not across an entire wind facility at a time. The choice to use hourly data from the ‘nearest’ weather station is not adequate for this type of analysis. Furthermore, this decision leads to the authors conducting multiple spatial and temporal interpolation (the majority of the paper) – creating an unvalidated and potentially flawed dataset. I understand the need to make decisions and conduct work in light of uncertainty and data sparsity. However, that still requires using the best available data at that time. The best available data for wind that provides scales more closely aligned with curtailment decisions is modeled gridded weather data. Canda hosts the Climate Change Canada Wind Atlas and US based National Renewable energy laboratory have publicly available models wind speeds at various heights relevant to wind energy. The US based National Oceanic and Atmospheric Administration hosts a gridded RAP dataset with hourly interpolated data across North America. While each of these data sets also have flaws, they are well established and tested datasets that are far more appropriate for this work. In fact, these are the datasets used by references to prior work in the US and Ontario. Nonetheless, the best data for this work is actual turbine data.
The methodology is also mostly opaque with weird tangents providing far too much information (e.g., lines 205-217 which solely explain a boxplot). There was no information on the source of power curves, an extremely important part of the methodology.
The comparisons made are not statistical and there is no indication what the ranges presented represent (are they 95% CI?) or how the ‘more’ or ‘less’ amounts were calculated between curtailment types.
Overall this review just touches on the issues in the manuscript, which requires substantial revision from conceptualization to implementation to be acceptable science.
Citation: https://doi.org/10.5194/wes-2025-164-RC2 -
RC3: 'Comment on wes-2025-164', Anonymous Referee #3, 11 Dec 2025
General Comments:
While I appreciate the aims of this work, I see non-trivial methodological, interpretive, and editorial issues that collectively undermine the credibility and clarity of the manuscript.
The manuscript makes strong assertions about the accuracy and reliability of modeled data, particularly regarding the benefits of data correction and calibration. However, these claims are not adequately supported by robust evidence.
Several methodological steps, including gap-filling procedures, interpolation techniques, and the selection of reference weather stations, are either poorly described or lack justification. These gaps in methodological transparency compromise the reliability of the results and the reproducibility of the study.
The manuscript contains multiple overstatements, such as claims that modeled outputs “consistently aligned more closely” without addressing the magnitude of remaining discrepancy between modeled and actual measurements. Additionally, terms like “high-quality data” and “high-resolution” are used without clear definitions or criteria, leaving readers uncertain about the data’s robustness.
The comparison between smart and blanket curtailment is unsurprising and lacking novelty. Furthermore, the absence of direct ecological modeling or bat mortality assessment limits the manuscript’s relevance to policymakers and stakeholders concerned with balancing energy production and species protection.
Specific Comments:
Line 139-142 highlight critical flaws in this paper as currently written. The authors state “The consistent improvements across all metrics and years highlight the importance of robust data correction and calibration practices to ensure that modeled datasets accurately reflect real-world wind power generation. These results highlight the importance of data corrections and calibration adjustments, ensuring that the final modeled dataset provides a more reliable representation of actual wind power generation.” I did not find reasonable evidence in the manuscript to support these sentences. There isn’t presently evidence to support measure of accuracy of the modeled wind power data, that gap filling approaches significantly improved model precision, or that modeled and real power output was significantly enhanced – as asserted in lines 264-267. The comparisons conducted using MAPE and RMSE leverage values attributed to individual turbines using overly simplistic methods. I recommend more robust methods that can serve as strong evidence for these statements.
Smart curtailment resulting in lower estimated losses than blanket curtailment isn’t a surprising or novel outcome. All the subsequent quantified results comparing smart to blanket curtailment have a fair degree of overlap minimizing the significance of this takeaway.
Additionally, the authors state “While this study did not assess bat mortality directly, prior research has shown that shorter curtailment periods may increase risk to bats depending on migration timing (e.g., Arnett et al., 2016). Further ecological modeling would be needed to evaluate the trade-offs between energy conservation and species protection.” The absence of such an evaluation makes these results less likely to offer valuable insights for policymakers and industry stakeholders aiming to balance energy efficiency with environmental considerations (as asserted in lines 279-280).
Additional, comments are provided below.
Line 27: It is advisable to refrain from definitively stating that migratory bat populations are in decline. There isn’t published, data driven evidence that quantifies the range-wide populations growth rate through time.
Line 60: What makes the meteorological and production data you used “high-quality?” Suggest defining what constitutes high-quality and how the data you used meet these criteria in notable ways.
It is an overstatement that “modeled outputs consistently aligned more closely with the AESO values.” What is presented suggests trivial improvements at best. The remaining error between modelled versus actual is not trivial. The authors have not appropriately described this reality.
Line 72: Please provide evidence to support the statement that “ECCC data demonstrated strong alignment with operational wind power output data from the Alberta Electric System Operator (AESO)” and describe the criteria used to determine these data were a “suitable choice for estimating site-specific meteorological conditions.” As stated, it is ambiguous and it’s unknown to the reader whether this is based on the authors’ subjective discretion.
Line 78: All meteorological measurement instrumentation is subject to error. How do you know the accuracy of the ECCC weather station measurements? Is this proven? I suggest reviewing work by Jennifer King and others (e.g. https://www.osti.gov/biblio/1501672) and consider how a consensus measurement approach might benefit this work.
Line 80-81: The sentence “Then data gaps were filled in the primary station using corresponding records from the other two nearby stations, see Table A1” is insufficient. Precisely how were corresponding records from the other two nearby stations used to fill these gaps? How weather data gaps were filled are insufficiently described to determine whether the methods used to fill gaps inserted bias and error.
Line 82: Explain your choice for applied linear interpolation here. Is it because it’s demonstrated in literature to be the most appropriate choice? If so, include citations.
Line 83: The authors state “This multi-step approach resulted in complete, gap-free meteorological datasets for each of the 13 wind farms.” Gap-free isn’t inherently a good thing, especially if the method used for filling gaps is faulty. How these gaps were filled are insufficiently described to determine whether the methods used to fill gaps inserted bias and error. This is a significant vulnerability and compromises the reliability of the results. The authors accentuate this very point in the next sentence by stating that how the data are prepared influences the “accuracy and reliability of subsequent modeling calculations…”
Line 115: It is an overstatement that “modeled outputs consistently aligned more closely with the AESO values.” What is presented suggests trivial improvements at best. The remaining error between modelled versus actual is not trivial. The authors have not appropriately described this reality.
Line 121-125: The authors do not offer sufficient evidence that model fidelity was markedly improved. For example, I have not evidence that the overall results of this paper would have changed in any meaningful way if no gap filling was done.
Section 3.4.2: The methods described in this section to attribute specific outputs to individual turbines is to simply divide total wind farm output by the number of turbines. This is flawed because we know that wind speed is variable throughout a wind facility and production varies from turbine to turbine. As such, I cannot see a reasonable justification for using this as a validation and the comparison between the modeled and back-calculated wind speeds is mute.
Line 171-173: The authors state “that proximity alone is not sufficient to ensure meteorological interrelation. These results underscore the importance of evaluating both spatial and climatic alignment when selecting reference stations for wind modeling.” Yet, the authors selected weather stations based on proximity alone. As it currently reads, the authors have contradicted themselves.
Line 181: Please describe the alternative options for addressing these gaps and why inserting zero’s was the best option (supported by citations if applicable). Why not leave as N/A’s as opposed to zero’s? Or interpolate?
Technical Corrections:
Line 12: The authors state “As nations work toward net-zero emissions, renewable energy expansion is crucial.” The authors will be more impactful if they rephrase to make this a matter-of-fact statement. For example, “As nations work toward net-zero emissions, renewable energy installations are expanding dramatically.”
Line 15: There is a sentence that begins on line 15 that reads “However, the expansion of wind turbines poses ecological risks, particularly to migratory bat populations, which are vulnerable to collisions with turbine blades that interfere with their flight paths.” Suggest removing “with turbine blades that interfere with their flight paths” as it is superfluous and presumptive that the flight path is immovable and the blade is “interfering.”
Line 17: There is a sentence that begins on line 17 that reads “In addition to turbine-related fatalities, migratory bat populations are already declining due to habitat loss and white-nose syndrome (Frick et al., 2023).” Data driven evidence that populations of migratory bats are definitively declining. This is based primarily on expert opinion and anecdotal evidence. It isn’t fair to state definitively they are in decline until status and trends estimates have been produced supported by data from a significant portion of the species ranges. This should be rephrased. (Similar comment for Line 27)
Line 25: Suggest replacing “energy efficiency” with “electricity generation.”
Line 27: The use of “growing scale in the first sentence of this paragraph is awkward phrasing. The way the rest of the sentence is phrased, “growth in deployment” or expansion of wind energy infrastructure” would be more appropriate.
Line 27: I suggest caution in the use of the word “risk.” The definition of risk can vary from author to author and reader to reader. Most plainly put, more wind infrastructure equates to more exposure. Based on encounter theory, this results in an increased likelihood of interactions between populations and infrastructure.
Line 27: It is advisable to refrain from definitively stating that migratory bat populations are in decline. There isn’t published, data driven evidence that quantifies the range-wide populations growth rate through time.
Lines 29 and 31: Common names do not require capitalization unless they include a proper noun (e.g., Indiana bat).
Line 38: There are many different types of “smart-curtailment” that utilized different methods to inform curtailment. Not all use meteorological-based triggers.
Line 38-39: Suggest rephrasing “reducing energy losses while still mitigating bat fatalities” to something like “in an attempt to minimize energy losses and bat fatalities.”
Line 39: There is a sentence that reads “These triggers are typically based on atmospheric conditions such as wind speed, temperature, and precipitation.” Suggest adding some citations.
Line 40: Please clarify how CE-O “implemented a curtailment framework.” Do you mean in the modeling exercise or on the ground implementation?
Line 43: Suggest replacing Maclaurin et al. with a stronger citation. Maclaurin et al. modeled different scenarios using these criteria but citing work that demonstrated bats are less likely to be active in cold or rainy conditions would be stronger.
Line 47: Seemingly awkward break in text.
Line 60: What makes the meteorological and production data you used “high-quality?” Suggest defining what constitutes high-quality and how the data you used meet these criteria in notable ways.
Line 61: Suggest noting for the reader here whether the AESO production data are wind facility specific and the temporal resolution of these data.
Line 63: Suggest editing here (and in all instances) changing “total capacity” to “total rated capacity.”
Line 69: Please define what you mean by “high-resolution.” Are you referring to spatial or temporal? Or both? In any case, “high” is relative so it’s advised you provide the reader with specifics on what the resolution is. Additionally, please clarify whether these are direct measurements only or whether any of the original data are derived (e.g., interpolated measurements).
Line 76: Please include in text summary statistics on distances between wind facilities and ECCC weather stations and height above ground the sensors are measuring at.
Line 79: The reader would benefit from examples of what commonly causes gaps in the ECCC data. Insight into how and why this occurs may help the reader have confidence in the data despite gaps.
Line 80: Strongly suggest adding descriptive statistics in text to give the reader quick insight into what these distances generally are.
Line 82: Explain your choice for applied linear interpolation here. Is it because it’s demonstrated in literature to be the most appropriate choice? If so, include citations.
Line 88-89: The authors state “the wind speeds from the completed datasets were adjusted to the hub heights of the turbines.” The reader would benefit from information on the height original measurements were collected at.
Line 92: Suggest defining “surface roughness length” for the reader at this first instance when the term is used.
Line 96-97: Please explain where site-specific air density and standard air density came from.
Line 97-101: Consider consolidating these sentences and making more concise. It’s currently somewhat of an aside that is taking up more real estate than necessary.
Line 103: Please define the timestep referred to here.
Line 111: Add a citation for NLCD.
Line 158-159: Citations should be added that corroborate that between 4 and 11m/s “correspond to the most sensitive portion of the turbine power curve with respect to the application of curtailment to reduce harm to bats.”
Section 3.5: It would be valuable to include generation losses as a lead in to the description of financial losses.
Line 176: Please describe how average pool price corresponds or relates to net present value. Additional descriptions of how data on price is pooled is generated would be useful to the reader. Presently it is unclear where these come from, how and why they are pooled, etc.
196-204: Descriptions of how values were calculated and the corresponding equations should be moved to methods.
Line 203: Where did potential production come from? This is the only time in the manuscript the authors refer to potential production and there is no description of data source or how the value was derived.
Line 269: In some instances in the paper the authors use a different date format than is used hear. Please check for consistency.
Tables and figures: All captions for all tables and figures should stand alone. All acronyms should be defined, and source information cited. Headers referring to the same thing should be parallel. Please scrutinize each to ensure the material within each is described, defined, and cited appropriately. Any additional comments specific to given table or figure is offered below.
Table 2:
- Are the latitude and longitude meant to depict the centroid of the wind facility? Please clearly state.
- Are 100% of the turbines for each wind facility the same model and hub height? This isn’t always the case so best to be clear.
- Suggest changing “Total capacity” to “Total rated capacity”
Figure 1:
- Presentation between 2020 and 2021 is not parallel.
Table A1:
- Here asset name is used where farm name is used in an earlier table. Please check for consistency.
Table A2:
- Please include a citation where the reader can discover the definitions and more information about CLC and NLCD values and descriptions.
Citation: https://doi.org/10.5194/wes-2025-164-RC3
Status: closed
-
RC1: 'Comment on wes-2025-164', Anonymous Referee #1, 07 Nov 2025
General comments to the manuscript:
abstract and introduction were well and clearly written. Overall, the aim of the study is also well-aligned with the scope of WES. The manuscript present results of applied relevance with the potential to accelerate the field of wind development through adaptive control of wind turbine operation to address sustainability, environmental and socioeconomic considerations.
I did, however, find the methodology and results sections to be somewhat scrambled and less well-structured, as such affecting the presentation of the study negatively with several details lacking (see specific comments) that I would suggest including for improved transparency of the study design and reproducibility.
The authors quantify power production losses based on modelled blanket and smart curtailment schemes (based on temperature and precipitation criteria) across different cut-in speeds relevant for bat activity and find that the proposed smart curtailment scenario reduces power losses and, consequently, costs. They also include the realistic consideration that the costs of developing, implementing and maintaining a smart curtailment system are not factored into the analysis but it would be relevant to relate a prospected cost of such a system to the predicted revenue gained from smart curtailment.
A substantial effort is put into the generation of gap-free meteorological data series based on data from three weather stations in the surrounding area of each wind farm, but some are at considerable distance (up to >50 km away although it is not specified if distance from a given weather station is measured to the center of each wind farm (WF) or the closest edge of the WF and the WF area was not included. I therefore question whether the models would be more accurate and support the aim of the study more clearly if weather parameters had been extracted for the actual wind turbines/WF site coordinates from the global Copernicus ERA5 dataset. There could be perfectly good arguments against it but perhaps it would be an idea to discuss why the chosen approach was used instead.
The manuscript also lacks a comprehensive discussion section (none is currently included) at present it merely includes half a page of concluding remarks with a single reference. Consequently, the outcomes are not linked to previous research and results and does not consider the importance of smart versus blanket curtailment from an ecology perspective, even though it had likely implications for bat management and conservation as well.
Specific comments (and a disclaimer: I am not a statistics expert, which may also be reflected in some of my comments, and would therefore recommend that the manuscript also be reviewed by someone with more insight into the modelling approach used):
Background l. 27, regarding white-nose syndrome: white-nose is not a stressor at global level, consider specifying either which migratory bat populations (e.g., North-American) or (my preference) broaden the sentence overall to 'increasing global risks to migratory bats' but specify 'on a continental scale' for white-nose.
Table 1. temporal window: what is the reasoning behind the two different nightly schedules? Suggest including this in the otherwise thorough description of the scenarios in the paragraphs above the table. Does the shorter temporal window used for peak season not impact the outcome?
Methodology l. 60-61: which criteria are behind 'consistent' and 'high-quality'?
Table 2: suggest including overall area of each wind farm
l. 70, following ‘weather stations’: Consider including reference to table A1 and info about the height of weather station measurements already here. I realize that both are included further below but was missing the information when reading this section.
l. 72, ‘strong alignment’: would be good to show this alignment graphically or back it with results of data analysis. How where they compared?
l. 82: ‘surface roughness length’ was an unfamiliar term to me, recommend to include a short definition
l. 100-101: I understand this argument but for the applied scenario, losses would have to be considered as well. Would suggest, as minimum, to provide a range (min/max) of expected/typical losses and indicate how much variation is expected between wind farms and whether this is expected to influence results, or if not, then why not. This ties to line 104-105 also, if any of the parameters mentioned here are covariates of the losses component, then I am unsure how it can just be treated as a constant. Maybe in other words, as a non-technical outsider, I have no way of evaluating how much this simplification might influence the outcome of the study.
l. 103: timesteps not explained/defined previously, what are the timesteps?
l. 103: ‘a linear interpolation’, should it be the linear interpolation or if this is a reference to a specific type of such then specify.
l. 111, NLCD: if already in reference list, then useful to add abbreviation there to make it easier to find, if not, then consider adding link here.
l. 113: This paragraph and figure 1 should be part of the results section
l. 116: there are quite a few abbreviations in play, so perhaps remind the reader what AESO is (e.g. actual power output data (AESO))
l. 116: initial versus gap-filled: is the only difference that 'initial' has gaps in weather data? Suggest to make this a bit more clear or introduce earlier in methods how these differ. In Figure 1, the legend says modelled and initial but are they not both modelled? Using initial and gap-filled would perhaps be more instructive, provided that 'initial' is explained in the text above.
L. 115, ‘consistently aligned more closely’: I only detect a minute difference for July and August 2022 and for all three months in 2023. Suggest changing Y-axis to show the difference more clearly plus include deviation measures and/or include and refer to actual values and uncertainty estimates in the appendix/supplementary materials.
Figure 1: It seems to me that the results per wind farm would be useful as well, consider adding, if feasible. If not, then an idea for a follow up study, perhaps?
l. 122-124: How relevant are first and second decimals for percentages a) relative to realistic uncertainty and b) to the broad scope in terms of costs? Is it warranted to include decimals and is the difference significant?
l. 138-142: suggest moving this: first sentence to results, the rest to discussion.
1. 138, ‘calibration methods’: where are these described? Use the same term there to make more apparent what was considered calibration methods.
l. 155, ‘power curves…’: the equation does not show the plots, perhaps rephrase to the power curves arising from eq. 7? It might be instructive to show a graphical example of the/a power curve(s).
l. 162, ‘distributional metrics’: please explain
1. 162-164: sentence should be part of results.
l. 167, ‘binned into 4 – 11 m/s intervals’: Unclear to me what these means, what are the intervals?
l. 169-173: recommend to move this to results and please review my suggestion about using Copernicus weather data.
l. 183: perhaps useful to indicate how often/how many hours/how large a fraction this was relevant for to support the statement 'slight underestimate'
l. 187, ‘meteorological sensors’: where there no meteorological sensors on the turbines on-site? Suggest to include in methods why such were not considered for the analysis.
l. 194-215: This all belongs to methods from my perspective and takes away from the aim of the results; to summarize main outcomes. Please review carefully and re-structure methods and results accordingly.
l. 220: suggest to also include before 5.5 m/s: ‘at the lowest cut-in speed of 5.5 m/s…
Figure 4-6: explain in figure text what grey point are (e.g. excluded outliers)
l. 229: ‘broadly consistent’: this statement could be a bit more specific, e.g. modelled production losses are consistently slightly higher than those estimated from back-calculated hub-height wind speeds but smart curtailment consistently leads to lower losses for both datasets.
l. 271: ‘significant’: how was significance tested?Technical corrections:
l. 83: delete ‘s’ from theses
l. 238, ‘Peak Season’: check for consistency in capitalization of categories throughout sectionCitation: https://doi.org/10.5194/wes-2025-164-RC1 -
RC2: 'Comment on wes-2025-164', Anonymous Referee #2, 05 Dec 2025
The pre-print ‘Investigation on the Impacts of Smart Curtailment for Bat Fatality Mitigation in Alberta’ attempts to investigate the energy and economic cost of curtailment to reduce bat fatalities based on windspeed only and windspeed plus temperature and precipitation.
While the manuscript attempts to address a widespread interest in understanding the costs of curtailment to reduce bat fatalities, it does not add to the understanding of curtailment costs and the results are questionable given the flawed data and methodology used. Furthermore, the data used and results lack the appropriate context or any discussion that places the results in context of our current understanding. The Conclusion section is highly inadequate and consists of only 18 lines with one reference.
The impact of each curtailment type on bat fatalities is not addressed. It is known that all curtailment is not created equal and that adding predictive parameters to curtailment (such as temperature and precipitation) can have varying effects on fatalities depending on the location. The rational for the choice of these parameters and how it impacts bat fatalities is not provided. Additionally, we are learning that these two variables are less useful than other predictors. Overall, the manuscript shows a lack of understanding of the wind energy sector, curtailment, and bat ecology (migratory bats are not impacted by White-nose syndrome).
A review addressing all of the concerns with manuscript would be longer than the manuscript itself – therefore I will focus discussion only on the data used and touch on the methodology. Both Thurber et al 2023 and Maclaurin et al 2022 demonstrate how to properly conduct this type of work and were cited, but these examples were not followed.
When selecting data, it is important to have a thorough understanding of how curtailment decisions are implemented and select data of appropriate spatial and temporal resolution. Curtailment is typically implemented in 10-minute periods based on the average windspeed observed at each individual turbine. In other words, curtailment is most often implement at a very fine spatial and temporal resolution and not across an entire wind facility at a time. The choice to use hourly data from the ‘nearest’ weather station is not adequate for this type of analysis. Furthermore, this decision leads to the authors conducting multiple spatial and temporal interpolation (the majority of the paper) – creating an unvalidated and potentially flawed dataset. I understand the need to make decisions and conduct work in light of uncertainty and data sparsity. However, that still requires using the best available data at that time. The best available data for wind that provides scales more closely aligned with curtailment decisions is modeled gridded weather data. Canda hosts the Climate Change Canada Wind Atlas and US based National Renewable energy laboratory have publicly available models wind speeds at various heights relevant to wind energy. The US based National Oceanic and Atmospheric Administration hosts a gridded RAP dataset with hourly interpolated data across North America. While each of these data sets also have flaws, they are well established and tested datasets that are far more appropriate for this work. In fact, these are the datasets used by references to prior work in the US and Ontario. Nonetheless, the best data for this work is actual turbine data.
The methodology is also mostly opaque with weird tangents providing far too much information (e.g., lines 205-217 which solely explain a boxplot). There was no information on the source of power curves, an extremely important part of the methodology.
The comparisons made are not statistical and there is no indication what the ranges presented represent (are they 95% CI?) or how the ‘more’ or ‘less’ amounts were calculated between curtailment types.
Overall this review just touches on the issues in the manuscript, which requires substantial revision from conceptualization to implementation to be acceptable science.
Citation: https://doi.org/10.5194/wes-2025-164-RC2 -
RC3: 'Comment on wes-2025-164', Anonymous Referee #3, 11 Dec 2025
General Comments:
While I appreciate the aims of this work, I see non-trivial methodological, interpretive, and editorial issues that collectively undermine the credibility and clarity of the manuscript.
The manuscript makes strong assertions about the accuracy and reliability of modeled data, particularly regarding the benefits of data correction and calibration. However, these claims are not adequately supported by robust evidence.
Several methodological steps, including gap-filling procedures, interpolation techniques, and the selection of reference weather stations, are either poorly described or lack justification. These gaps in methodological transparency compromise the reliability of the results and the reproducibility of the study.
The manuscript contains multiple overstatements, such as claims that modeled outputs “consistently aligned more closely” without addressing the magnitude of remaining discrepancy between modeled and actual measurements. Additionally, terms like “high-quality data” and “high-resolution” are used without clear definitions or criteria, leaving readers uncertain about the data’s robustness.
The comparison between smart and blanket curtailment is unsurprising and lacking novelty. Furthermore, the absence of direct ecological modeling or bat mortality assessment limits the manuscript’s relevance to policymakers and stakeholders concerned with balancing energy production and species protection.
Specific Comments:
Line 139-142 highlight critical flaws in this paper as currently written. The authors state “The consistent improvements across all metrics and years highlight the importance of robust data correction and calibration practices to ensure that modeled datasets accurately reflect real-world wind power generation. These results highlight the importance of data corrections and calibration adjustments, ensuring that the final modeled dataset provides a more reliable representation of actual wind power generation.” I did not find reasonable evidence in the manuscript to support these sentences. There isn’t presently evidence to support measure of accuracy of the modeled wind power data, that gap filling approaches significantly improved model precision, or that modeled and real power output was significantly enhanced – as asserted in lines 264-267. The comparisons conducted using MAPE and RMSE leverage values attributed to individual turbines using overly simplistic methods. I recommend more robust methods that can serve as strong evidence for these statements.
Smart curtailment resulting in lower estimated losses than blanket curtailment isn’t a surprising or novel outcome. All the subsequent quantified results comparing smart to blanket curtailment have a fair degree of overlap minimizing the significance of this takeaway.
Additionally, the authors state “While this study did not assess bat mortality directly, prior research has shown that shorter curtailment periods may increase risk to bats depending on migration timing (e.g., Arnett et al., 2016). Further ecological modeling would be needed to evaluate the trade-offs between energy conservation and species protection.” The absence of such an evaluation makes these results less likely to offer valuable insights for policymakers and industry stakeholders aiming to balance energy efficiency with environmental considerations (as asserted in lines 279-280).
Additional, comments are provided below.
Line 27: It is advisable to refrain from definitively stating that migratory bat populations are in decline. There isn’t published, data driven evidence that quantifies the range-wide populations growth rate through time.
Line 60: What makes the meteorological and production data you used “high-quality?” Suggest defining what constitutes high-quality and how the data you used meet these criteria in notable ways.
It is an overstatement that “modeled outputs consistently aligned more closely with the AESO values.” What is presented suggests trivial improvements at best. The remaining error between modelled versus actual is not trivial. The authors have not appropriately described this reality.
Line 72: Please provide evidence to support the statement that “ECCC data demonstrated strong alignment with operational wind power output data from the Alberta Electric System Operator (AESO)” and describe the criteria used to determine these data were a “suitable choice for estimating site-specific meteorological conditions.” As stated, it is ambiguous and it’s unknown to the reader whether this is based on the authors’ subjective discretion.
Line 78: All meteorological measurement instrumentation is subject to error. How do you know the accuracy of the ECCC weather station measurements? Is this proven? I suggest reviewing work by Jennifer King and others (e.g. https://www.osti.gov/biblio/1501672) and consider how a consensus measurement approach might benefit this work.
Line 80-81: The sentence “Then data gaps were filled in the primary station using corresponding records from the other two nearby stations, see Table A1” is insufficient. Precisely how were corresponding records from the other two nearby stations used to fill these gaps? How weather data gaps were filled are insufficiently described to determine whether the methods used to fill gaps inserted bias and error.
Line 82: Explain your choice for applied linear interpolation here. Is it because it’s demonstrated in literature to be the most appropriate choice? If so, include citations.
Line 83: The authors state “This multi-step approach resulted in complete, gap-free meteorological datasets for each of the 13 wind farms.” Gap-free isn’t inherently a good thing, especially if the method used for filling gaps is faulty. How these gaps were filled are insufficiently described to determine whether the methods used to fill gaps inserted bias and error. This is a significant vulnerability and compromises the reliability of the results. The authors accentuate this very point in the next sentence by stating that how the data are prepared influences the “accuracy and reliability of subsequent modeling calculations…”
Line 115: It is an overstatement that “modeled outputs consistently aligned more closely with the AESO values.” What is presented suggests trivial improvements at best. The remaining error between modelled versus actual is not trivial. The authors have not appropriately described this reality.
Line 121-125: The authors do not offer sufficient evidence that model fidelity was markedly improved. For example, I have not evidence that the overall results of this paper would have changed in any meaningful way if no gap filling was done.
Section 3.4.2: The methods described in this section to attribute specific outputs to individual turbines is to simply divide total wind farm output by the number of turbines. This is flawed because we know that wind speed is variable throughout a wind facility and production varies from turbine to turbine. As such, I cannot see a reasonable justification for using this as a validation and the comparison between the modeled and back-calculated wind speeds is mute.
Line 171-173: The authors state “that proximity alone is not sufficient to ensure meteorological interrelation. These results underscore the importance of evaluating both spatial and climatic alignment when selecting reference stations for wind modeling.” Yet, the authors selected weather stations based on proximity alone. As it currently reads, the authors have contradicted themselves.
Line 181: Please describe the alternative options for addressing these gaps and why inserting zero’s was the best option (supported by citations if applicable). Why not leave as N/A’s as opposed to zero’s? Or interpolate?
Technical Corrections:
Line 12: The authors state “As nations work toward net-zero emissions, renewable energy expansion is crucial.” The authors will be more impactful if they rephrase to make this a matter-of-fact statement. For example, “As nations work toward net-zero emissions, renewable energy installations are expanding dramatically.”
Line 15: There is a sentence that begins on line 15 that reads “However, the expansion of wind turbines poses ecological risks, particularly to migratory bat populations, which are vulnerable to collisions with turbine blades that interfere with their flight paths.” Suggest removing “with turbine blades that interfere with their flight paths” as it is superfluous and presumptive that the flight path is immovable and the blade is “interfering.”
Line 17: There is a sentence that begins on line 17 that reads “In addition to turbine-related fatalities, migratory bat populations are already declining due to habitat loss and white-nose syndrome (Frick et al., 2023).” Data driven evidence that populations of migratory bats are definitively declining. This is based primarily on expert opinion and anecdotal evidence. It isn’t fair to state definitively they are in decline until status and trends estimates have been produced supported by data from a significant portion of the species ranges. This should be rephrased. (Similar comment for Line 27)
Line 25: Suggest replacing “energy efficiency” with “electricity generation.”
Line 27: The use of “growing scale in the first sentence of this paragraph is awkward phrasing. The way the rest of the sentence is phrased, “growth in deployment” or expansion of wind energy infrastructure” would be more appropriate.
Line 27: I suggest caution in the use of the word “risk.” The definition of risk can vary from author to author and reader to reader. Most plainly put, more wind infrastructure equates to more exposure. Based on encounter theory, this results in an increased likelihood of interactions between populations and infrastructure.
Line 27: It is advisable to refrain from definitively stating that migratory bat populations are in decline. There isn’t published, data driven evidence that quantifies the range-wide populations growth rate through time.
Lines 29 and 31: Common names do not require capitalization unless they include a proper noun (e.g., Indiana bat).
Line 38: There are many different types of “smart-curtailment” that utilized different methods to inform curtailment. Not all use meteorological-based triggers.
Line 38-39: Suggest rephrasing “reducing energy losses while still mitigating bat fatalities” to something like “in an attempt to minimize energy losses and bat fatalities.”
Line 39: There is a sentence that reads “These triggers are typically based on atmospheric conditions such as wind speed, temperature, and precipitation.” Suggest adding some citations.
Line 40: Please clarify how CE-O “implemented a curtailment framework.” Do you mean in the modeling exercise or on the ground implementation?
Line 43: Suggest replacing Maclaurin et al. with a stronger citation. Maclaurin et al. modeled different scenarios using these criteria but citing work that demonstrated bats are less likely to be active in cold or rainy conditions would be stronger.
Line 47: Seemingly awkward break in text.
Line 60: What makes the meteorological and production data you used “high-quality?” Suggest defining what constitutes high-quality and how the data you used meet these criteria in notable ways.
Line 61: Suggest noting for the reader here whether the AESO production data are wind facility specific and the temporal resolution of these data.
Line 63: Suggest editing here (and in all instances) changing “total capacity” to “total rated capacity.”
Line 69: Please define what you mean by “high-resolution.” Are you referring to spatial or temporal? Or both? In any case, “high” is relative so it’s advised you provide the reader with specifics on what the resolution is. Additionally, please clarify whether these are direct measurements only or whether any of the original data are derived (e.g., interpolated measurements).
Line 76: Please include in text summary statistics on distances between wind facilities and ECCC weather stations and height above ground the sensors are measuring at.
Line 79: The reader would benefit from examples of what commonly causes gaps in the ECCC data. Insight into how and why this occurs may help the reader have confidence in the data despite gaps.
Line 80: Strongly suggest adding descriptive statistics in text to give the reader quick insight into what these distances generally are.
Line 82: Explain your choice for applied linear interpolation here. Is it because it’s demonstrated in literature to be the most appropriate choice? If so, include citations.
Line 88-89: The authors state “the wind speeds from the completed datasets were adjusted to the hub heights of the turbines.” The reader would benefit from information on the height original measurements were collected at.
Line 92: Suggest defining “surface roughness length” for the reader at this first instance when the term is used.
Line 96-97: Please explain where site-specific air density and standard air density came from.
Line 97-101: Consider consolidating these sentences and making more concise. It’s currently somewhat of an aside that is taking up more real estate than necessary.
Line 103: Please define the timestep referred to here.
Line 111: Add a citation for NLCD.
Line 158-159: Citations should be added that corroborate that between 4 and 11m/s “correspond to the most sensitive portion of the turbine power curve with respect to the application of curtailment to reduce harm to bats.”
Section 3.5: It would be valuable to include generation losses as a lead in to the description of financial losses.
Line 176: Please describe how average pool price corresponds or relates to net present value. Additional descriptions of how data on price is pooled is generated would be useful to the reader. Presently it is unclear where these come from, how and why they are pooled, etc.
196-204: Descriptions of how values were calculated and the corresponding equations should be moved to methods.
Line 203: Where did potential production come from? This is the only time in the manuscript the authors refer to potential production and there is no description of data source or how the value was derived.
Line 269: In some instances in the paper the authors use a different date format than is used hear. Please check for consistency.
Tables and figures: All captions for all tables and figures should stand alone. All acronyms should be defined, and source information cited. Headers referring to the same thing should be parallel. Please scrutinize each to ensure the material within each is described, defined, and cited appropriately. Any additional comments specific to given table or figure is offered below.
Table 2:
- Are the latitude and longitude meant to depict the centroid of the wind facility? Please clearly state.
- Are 100% of the turbines for each wind facility the same model and hub height? This isn’t always the case so best to be clear.
- Suggest changing “Total capacity” to “Total rated capacity”
Figure 1:
- Presentation between 2020 and 2021 is not parallel.
Table A1:
- Here asset name is used where farm name is used in an earlier table. Please check for consistency.
Table A2:
- Please include a citation where the reader can discover the definitions and more information about CLC and NLCD values and descriptions.
Citation: https://doi.org/10.5194/wes-2025-164-RC3
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 284 | 185 | 27 | 496 | 24 | 32 |
- HTML: 284
- PDF: 185
- XML: 27
- Total: 496
- BibTeX: 24
- EndNote: 32
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
General comments to the manuscript:
abstract and introduction were well and clearly written. Overall, the aim of the study is also well-aligned with the scope of WES. The manuscript present results of applied relevance with the potential to accelerate the field of wind development through adaptive control of wind turbine operation to address sustainability, environmental and socioeconomic considerations.
I did, however, find the methodology and results sections to be somewhat scrambled and less well-structured, as such affecting the presentation of the study negatively with several details lacking (see specific comments) that I would suggest including for improved transparency of the study design and reproducibility.
The authors quantify power production losses based on modelled blanket and smart curtailment schemes (based on temperature and precipitation criteria) across different cut-in speeds relevant for bat activity and find that the proposed smart curtailment scenario reduces power losses and, consequently, costs. They also include the realistic consideration that the costs of developing, implementing and maintaining a smart curtailment system are not factored into the analysis but it would be relevant to relate a prospected cost of such a system to the predicted revenue gained from smart curtailment.
A substantial effort is put into the generation of gap-free meteorological data series based on data from three weather stations in the surrounding area of each wind farm, but some are at considerable distance (up to >50 km away although it is not specified if distance from a given weather station is measured to the center of each wind farm (WF) or the closest edge of the WF and the WF area was not included. I therefore question whether the models would be more accurate and support the aim of the study more clearly if weather parameters had been extracted for the actual wind turbines/WF site coordinates from the global Copernicus ERA5 dataset. There could be perfectly good arguments against it but perhaps it would be an idea to discuss why the chosen approach was used instead.
The manuscript also lacks a comprehensive discussion section (none is currently included) at present it merely includes half a page of concluding remarks with a single reference. Consequently, the outcomes are not linked to previous research and results and does not consider the importance of smart versus blanket curtailment from an ecology perspective, even though it had likely implications for bat management and conservation as well.
Specific comments (and a disclaimer: I am not a statistics expert, which may also be reflected in some of my comments, and would therefore recommend that the manuscript also be reviewed by someone with more insight into the modelling approach used):
Background l. 27, regarding white-nose syndrome: white-nose is not a stressor at global level, consider specifying either which migratory bat populations (e.g., North-American) or (my preference) broaden the sentence overall to 'increasing global risks to migratory bats' but specify 'on a continental scale' for white-nose.
Table 1. temporal window: what is the reasoning behind the two different nightly schedules? Suggest including this in the otherwise thorough description of the scenarios in the paragraphs above the table. Does the shorter temporal window used for peak season not impact the outcome?
Methodology l. 60-61: which criteria are behind 'consistent' and 'high-quality'?
Table 2: suggest including overall area of each wind farm
l. 70, following ‘weather stations’: Consider including reference to table A1 and info about the height of weather station measurements already here. I realize that both are included further below but was missing the information when reading this section.
l. 72, ‘strong alignment’: would be good to show this alignment graphically or back it with results of data analysis. How where they compared?
l. 82: ‘surface roughness length’ was an unfamiliar term to me, recommend to include a short definition
l. 100-101: I understand this argument but for the applied scenario, losses would have to be considered as well. Would suggest, as minimum, to provide a range (min/max) of expected/typical losses and indicate how much variation is expected between wind farms and whether this is expected to influence results, or if not, then why not. This ties to line 104-105 also, if any of the parameters mentioned here are covariates of the losses component, then I am unsure how it can just be treated as a constant. Maybe in other words, as a non-technical outsider, I have no way of evaluating how much this simplification might influence the outcome of the study.
l. 103: timesteps not explained/defined previously, what are the timesteps?
l. 103: ‘a linear interpolation’, should it be the linear interpolation or if this is a reference to a specific type of such then specify.
l. 111, NLCD: if already in reference list, then useful to add abbreviation there to make it easier to find, if not, then consider adding link here.
l. 113: This paragraph and figure 1 should be part of the results section
l. 116: there are quite a few abbreviations in play, so perhaps remind the reader what AESO is (e.g. actual power output data (AESO))
l. 116: initial versus gap-filled: is the only difference that 'initial' has gaps in weather data? Suggest to make this a bit more clear or introduce earlier in methods how these differ. In Figure 1, the legend says modelled and initial but are they not both modelled? Using initial and gap-filled would perhaps be more instructive, provided that 'initial' is explained in the text above.
L. 115, ‘consistently aligned more closely’: I only detect a minute difference for July and August 2022 and for all three months in 2023. Suggest changing Y-axis to show the difference more clearly plus include deviation measures and/or include and refer to actual values and uncertainty estimates in the appendix/supplementary materials.
Figure 1: It seems to me that the results per wind farm would be useful as well, consider adding, if feasible. If not, then an idea for a follow up study, perhaps?
l. 122-124: How relevant are first and second decimals for percentages a) relative to realistic uncertainty and b) to the broad scope in terms of costs? Is it warranted to include decimals and is the difference significant?
l. 138-142: suggest moving this: first sentence to results, the rest to discussion.
1. 138, ‘calibration methods’: where are these described? Use the same term there to make more apparent what was considered calibration methods.
l. 155, ‘power curves…’: the equation does not show the plots, perhaps rephrase to the power curves arising from eq. 7? It might be instructive to show a graphical example of the/a power curve(s).
l. 162, ‘distributional metrics’: please explain
1. 162-164: sentence should be part of results.
l. 167, ‘binned into 4 – 11 m/s intervals’: Unclear to me what these means, what are the intervals?
l. 169-173: recommend to move this to results and please review my suggestion about using Copernicus weather data.
l. 183: perhaps useful to indicate how often/how many hours/how large a fraction this was relevant for to support the statement 'slight underestimate'
l. 187, ‘meteorological sensors’: where there no meteorological sensors on the turbines on-site? Suggest to include in methods why such were not considered for the analysis.
l. 194-215: This all belongs to methods from my perspective and takes away from the aim of the results; to summarize main outcomes. Please review carefully and re-structure methods and results accordingly.
l. 220: suggest to also include before 5.5 m/s: ‘at the lowest cut-in speed of 5.5 m/s…
Figure 4-6: explain in figure text what grey point are (e.g. excluded outliers)
l. 229: ‘broadly consistent’: this statement could be a bit more specific, e.g. modelled production losses are consistently slightly higher than those estimated from back-calculated hub-height wind speeds but smart curtailment consistently leads to lower losses for both datasets.
l. 271: ‘significant’: how was significance tested?
Technical corrections:
l. 83: delete ‘s’ from theses
l. 238, ‘Peak Season’: check for consistency in capitalization of categories throughout section