the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 03 Apr 2025
Locating the Optimal Wind Resource within two Californian Offshore Wind Energy Areas
Abstract. The spatiotemporal variability of wind resource at two Californian offshore wind-energy areas (Humboldt and Morro Bay) is characterized using 23 years (2000–2022) of 2-km-resolution, hourly NOW-23 reanalysis data. Idealized power is estimated for the International Energy Agency 15 MW reference wind turbine. Wind speed and energy output are higher at Humboldt than Morro Bay and for summer months than winter months at both sites. Idealized daily energy output per turbine with one turbine within each model grid cell, peaks at 330 MWh in June for Humboldt and 300 MWh in May for Morro Bay. Energy output per turbine decreases from the oceanward to the coastward perimeter by ~20 %, dropping ~22 MWh across Humboldt and ~46 MWh across Morro Bay. Rotor-layer wind shear and veer exhibit strong seasonal variability, with summertime shear twice of wintertime shear at both sites. Daily wind resource variability is quantified through Fractional Variability (FV), defined as the ratio of the interquartile range of wind speeds/energy to the overall median value for that day of year. Locations and times with higher FV coincide with low wind-speeds (i.e., low output) for both sites. A linear optimization identifies the optimal wind resource locations (that maximizes energy output but minimizes output FV, wind shear, wind veer, and distance to shore) at the oceanward and coastward flanks of Humboldt and Morro Bay, respectively. The gradients in optimization scores are aligned parallel to the coast and are independent of the choice of power curves for rated powers of 8–16 MW.
- Preprint
(1942 KB) - Metadata XML
-
Supplement
(2020 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on wes-2025-55', Anonymous Referee #1, 05 May 2025
Review of WES-2025-55
The manuscript "Locating the Optimal Wind Resource within two Californian Offshore Wind Energy Areas" by Mitra et al. presents in general quite interesting and relevant research on offshore wind ressources in California. The further investigation of potential future offshore lease areas is from a scientific and industrial perspective very relevant.
However, I found several key deficiencies in the manuscript that lead to the conclusion that I can't recommend publication of the manuscript in wind energy science. Those points are - from my perspective - so severe that a revision of those points would lead to a completely different manuscript and results.
Specifically, I have identified the following shortcomings:
- The authors ignore wake effects in their design of the study. Thus, in principle, they are investigating the purely academic question of placing and moving around a single wind turbine in the two wind energy development areas. This results in a conclusion that the most optimal areas are for instance located at the southern edges of the wind farm, in areas with a very pronounced northerly wind component (see wind roses in supplementary material). In a real wind farm (and those are prospective real sites) those areas would be the ones with the strongest wake effects. Wake effects can lead to 10-30% yield loss, in extremely dense wind farm clusters even more. As the variability of the annual averaged daily energy output (Fig. 1) across the sites is in the same order of magnitude, the results would completely shift. There are layout optimization tools like floris, FOXES, pywake/TOPFARM that could be used for including wake effects in the optimization. Or at least using a geometric wind farm layout and an engineering wake model would help to increase the relevance of the results.
- The authors are making some random choices on data processing, which are not explained. For example: Why is there a resampling of the 2 km grid to a 0.02 degree reference grid? Why are there 25 random turbine power curves investigated by varying cut-in, cut-out and rated power where also other reference turbines could be used such as DTU 10 MW turbine, the IEA 22MW turbine, the IEA 5MW turbine or even publicly available power curves for some existing turbines (a web search can help) could be used?
- Why is the variability discussion not based on an annual energy production but on daily output variability? The first one would be an established measure from the wind energy community.
- There is no discussion at all on available reference measurements in the area, or how they went into the NOW-23 dataset, if they are sufficiently included there.
- The literature review refers at several points to a paper Mitra et al., 2024, which is not part of the list of references. There, an unpublished study (Mitra et al., 2025) is mentioned, which I couldn't find even in a web search. This also this part would need some very thorough revision.
- The manuscript has quite a long document of supplementary material, which is at some points very relevant for the interpretation of the results, such as the used power curve, or the wind roses. The choice of what went into supplementary material seems quite random to me. I would recommend putting them rather into an appendix or at some points even in the main part of the document.
Citation: https://doi.org/10.5194/wes-2025-55-RC1 -
AC1: 'Reply on RC1', Arka Mitra, 05 Jun 2025
We thank the reviewer for their comments. We are presently working on a more detailed point-by-point response to address the various issues raised by the reviewer. In this response, we seek to highlight our approach in addressing the concerns and to demonstrate that addressing the problems will not lead to a “completely different manuscript and results”. Rather, we believe that incorporating the reviewer’s suggestions into our existing technique will only strengthen and enhance our existing syudy. The detailed point-by-point response and the updated manuscript will follow in a separate response.
We appreciate your feedback and want to share that we have already completed the analysis on several of your key suggestions. Integrating these new findings would not necessitate a resubmission of the current article, in our opinion. Our initial goal was to develop a subsequent article focusing on additional analyses, particularly concerning wake effects. However, we acknowledge that incorporating these results would indeed enhance the current article.
Here's how we will address the Reviewer's major concerns:
1) Wake Modeling: As mentioned, within the manuscript (e.g., Lines 70-71 and 387-390), the specific choice to use “idealized power” was a calculated choice on our because we believed the true offerings of this study were the Fractional variability (FV) metric for variability and the linear optimization method to determine the location of the optimal wind-resource within a pre-construction wind energy area. We felt that including the absolute impacts of wind shear/veer and wind turbine wakes would overcomplicate the narrative flow of the current study and draw attention away from the key offerings. Hence, we had repeatedly addressed the “idealized power” consideration as a shortcoming of the current study, which (as mentioned in the Conclusion), we sought to address in our next paper.We are pleased to inform you that we had already determined the relative percentages of power output lost to wind turbine wakes using an engineering wake model (PyWake) even before this manuscript was submitted, and their spatial patterns were, in fact, similar to what the reviewer suggests. However, on the reviewer’s suggestion, we reconsidered our approach, and we believe that the results of that wake analysis and its impact on our optimization can be easily folded into the current manuscript. We will discuss the impact of wakes on energy production in a new Section 3.3 immediately following the discussion on “Spatiotemporal Variability of Daily Aggregate Energy”. The current Section 3.3 will be moved to an updated Section 3.4, which will now include an updated linear optimization score which now also minimizes wind turbine wake losses (apart from the other considerations).
2) Choices on Data Processing and Wind Turbine Power Curves: We shall justify the choice of re-gridding within the Data and Methods section. Randomly choosing power curves allowed us the flexibility in demonstrating that our optimization is stable, not just on existing reference wind turbines, but also on possible future designs in this range. However, we appreciate the reviewer’s suggestion and shall also redo the same sensitivity analysis with other existing wind turbine power curves.
3) Not using AEP but daily power variability: This was addressed in the introduction:“Wind speeds and the electric power generated from wind farms can vary across different spatiotemporal scales. Hence, in recent years, a lot of emphasis has been put on the ready availability of short-term (sub-hourly to daily) wind speed forecasts at or near wind-turbine hub heights and wind turbine rotor layers (Wilczak et al., 2019; James et al., 2022). As a result, such forecasts are now readily available from direct numerical simulation and statistical post-processing. Variability within these forecasts on hourly and daily timescales are largely driven by the inherent diurnal variability of the Earth’s climate (occasionally disrupted by random weather events, such as surface low-pressure systems). Hence, characterizing this natural daily variability and the deviations from the same is crucial in designing the layout of wind turbines, estimating the electricity produced by wind farms over their lifetime and in devising strategies to improve the lifetime of the farms in question.”
We did not seek to quantify the interannual variability, but designed a metric to quantify small-scale variability that would be relevant to a more holistic decision-making for future wind farm designs. This point will be emphasized further in the updated manuscript.
5) Manuscript cited as Mitra et al. (2025): We had hoped to have that manuscript submitted to Wind Energy Science, however, it was found not to fully adhere to the WES scope. So, we have published the study by Mitra et al. (2025) as an Argonne Technical Report with an OSTI DOI number. The updated citation will be in the updated manuscript.
6) Restructuring the supplementary material: The suggested changes will be incorporated in the updated manuscript.We intend to implement these changes as quickly as possible and come up with a more detailed point-by-point response to each of the comments and the updated manuscript. Such changes do not merit the withdrawal of the current manuscript and resubmission.
Citation: https://doi.org/10.5194/wes-2025-55-AC1
-
AC1: 'Reply on RC1', Arka Mitra, 05 Jun 2025
-
RC2: 'Comment on wes-2025-55', Anonymous Referee #2, 05 Jun 2025
This manuscript addresses an important and timely topic in offshore wind energy by examining the spatiotemporal variability of wind resources at two proposed Californian lease areas using long-term, high-resolution reanalysis data. The effort to integrate wind variability, shear, and veer into a spatial optimization framework is commendable, and the analysis benefits from a rich dataset and a relevant application to real-world siting challenges. However, despite these strengths, the manuscript in its current form is not suitable for publication. The methodological foundation is underdeveloped, key concepts are either misapplied or oversimplified, and the paper lacks the necessary context and rigor expected for a scientific contribution in this area. Major revisions would not be sufficient to address the fundamental flaws. Rather, a substantial reworking of the manuscript, both in terms of technical approach and scientific framing, is required. What is needed is effectively a new paper: one that clearly engages with the existing literature, justifies methodological choices, incorporates essential physical processes such as wake interactions, and includes a proper discussion of limitations and implications.
General remarks
The manuscript relies heavily on supplementary material to present several key results, including figures that are essential to understanding and evaluating the methodology and its outcomes. While supplementary materials can be appropriate for secondary or supporting analyses, critical figures, such as those demonstrating the sensitivity of optimization results to turbine characteristics or illustrating variability patterns, should be included directly in the main manuscript.
Methodology
The methodology presented in the preprint suffers from several critical limitations that undermine the robustness and practical applicability of the results. First and foremost, the choice of weights in the multi-objective optimization function is arbitrary and lacks any empirical or stakeholder-driven justification. All weights are set to one (with a slightly higher value of 1.5 for distance to shore), implying equal importance of energy output, variability, shear, veer, and cost, which is rarely the case in real-world wind farm planning. Because the optimization score is highly sensitive to these weights, the interpretation of optimal locations is equally sensitive and thus not reliable without a thorough sensitivity analysis or justification of trade-offs. Compounding this issue is the misuse of terminology: the authors refer to their method as "gradient descent optimization," but no optimization is performed in the classical sense. The weights are fixed, and the scoring is applied statically to each grid cell, more akin to a composite ranking or multi-criteria evaluation than an iterative optimization process.
A far more serious flaw is the complete omission of wake effects, which are among the most important considerations in wind farm layout optimization. By treating each grid cell in isolation and assuming one turbine per cell, the methodology neglects the performance losses caused by turbine-turbine interactions, which can dramatically alter the spatial efficiency of a layout. As a result, the identified “optimal” locations may not remain optimal when array-level effects are considered. Finally, the normalization of all variables to a 0–1 range without accounting for their relative physical or economic significance can distort the true influence of each factor on the final score. Taken together, these issues suggest that while the intent of the optimization framework is commendable, its execution lacks the methodological rigor and realism required for meaningful conclusions in the context of offshore wind development.
Introduction and discussion
Another significant limitation of the manuscript lies in the introduction and discussion sections, which fail to adequately engage with the existing body of literature on methodologies for locating optimal wind resource areas. The introduction does not provide a review of current best practices or established benchmarks in wind farm sitting, such as multi-objective layout optimization frameworks that incorporate wake losses, cost models, or environmental constraints. This omission makes it difficult for readers to understand how the proposed method compares to state-of-the-art approaches in the field. A proper scientific contribution must situate itself within the existing landscape of methodologies, identify what gaps it intends to fill, and justify why its approach is advantageous or innovative. Moreover, no comparison is made between their simplistic scoring framework and more sophisticated optimization methods commonly used in industry and academia—such as genetic algorithms, greedy heuristics with wake modeling, or levelized cost of energy (LCOE)-based evaluations. The manuscript lacks any discussion of the limitations of their approach in relation to these established methods. The absence of such comparative analysis severely weakens the manuscript's credibility and relevance. A meaningful discussion section should critically reflect on the strengths and shortcomings of the proposed methodology considering existing literature, but this is entirely missing. As a result, the current discussion remains superficial and does not offer the reader a clear sense of the method's practical utility or scientific value.
Citation: https://doi.org/10.5194/wes-2025-55-RC2 -
AC2: 'Reply on RC2', Arka Mitra, 05 Jun 2025
We thank the reviewer for their comments. We are presently working on a more detailed point-by-point response to address the various issues raised by the reviewer and thereby produce an updated manuscript. In this response, we address the reviewer’s major concerns and demonstrate that addressing them will not lead to “effectively a new paper”. Incorporating the reviewer’s recommendations into our existing technique will strengthen and enhance our proposed methodology. The detailed point-by-point response and the updated manuscript will follow in a separate response.
Here's how we shall address the broad issues raised by Reviewer-2:
1) Supplementary Material: This point was also raised by Reviewer-1 and will be appropriately addressed in the revised manuscript.2) Issues regarding weights & Gradient Descent Optimization: The choice of initial weights is indeed arbitrary, as with any optimization model. But these are only the initial weights and not their final values. However, the weights are not fixed but rather updated during the gradient descent operation to find the minima of the cost function. As a result, the operation is indeed, in a classical sense, a “gradient descent optimization”. These points will be made clearer in the resubmission, through clarifications in the updated Section “Location of Optimal Wind Resource”.
3) Wind Turbine Wakes: This point about wakes was also raised by Reviewer-1, and we will include wake effects in our methods in our revised manuscript. An engineering wake model (PyWake) was used to estimate wind turbine wake losses and to revise our optimization score (more details in the response to Reviewer 1). This discussion will be added to our manuscript in a new Section 3.3, and Section 3.3 in our previous submission would be updated to a new Section 3.4.
4) Normalization of Score to 0-1 and Applicability to real-world scenarios: About “the normalization of all variables to a 0–1 range without accounting for their relative physical or economic significance”. The proposed optimization score already considers a linear superposition of both physical (variability, shear, veer) and economic significance (energy output and distance to coast as a simplistic cost proxy). The goal of this study is to propose a generalized method to simply leverage historical reanalysis data to address the question: “Where would be the most ideal location within a proposed site to place a wind turbine of given characteristics, where we do not have an extensive observational record?” Through further discussions on the applicability of this method, we shall highlight how this approach can be leveraged in real world scenarios.
5) Comparison to Existing Literature: We shall introduce a more detailed discussion in the “Introduction” and “Summary & Discussion” sections to address these issues.
We intend to implement these changes as quickly as possible and provide a more detailed point-by-point response to each of the comments and the updated manuscript. As already noted, many of the issues raised by the Reviewer echo concerns raised by Reviewer-1, and such would be addressed jointly. Such changes do not merit the withdrawal of the current manuscript and resubmission.Citation: https://doi.org/10.5194/wes-2025-55-AC2
-
AC2: 'Reply on RC2', Arka Mitra, 05 Jun 2025
Data sets
NOW-23 National Renewable Energy Laboratory (NREL), USA https://data.openei.org/submissions/4500
15 MW Reference Wind Turbine (RWT) Power Curve National Renewable Energy Laboratory (NREL), USA https://nrel.github.io/turbine-models/IEA_15MW_240_RWT.html
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 155 | 33 | 13 | 201 | 23 | 15 | 21 |
- HTML: 155
- PDF: 33
- XML: 13
- Total: 201
- Supplement: 23
- BibTeX: 15
- EndNote: 21
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1