| 17 Jun 2026
Impact of wind farms on the occurrence and characteristics of low-level jets
Abstract. The low-level jet (LLJ) is an occurrence of super-geostrophic wind speeds in the lower regions of the atmosphere, typically below 1 km. LLJs are highly relevant for wind energy as they can substantially modulate power production, wake recovery, and loads of wind turbines and wind farms. However, it is not clear how the presence of wind farms affects the occurrence rate and characteristics of LLJs, with some simulations suggesting that the presence of farms leads to fewer LLJs and to LLJs occurring at higher elevations. In this work, we develop methods to take advantage of unique long-term wind profile measurements to experimentally assess the impact of wind farms on the occurrence rates and morphologies of LLJs. We use more than 27 years of profiling radar wind measurements collected for the period August 1998–April 2026 at the Atmospheric Radiation Measurement Southern Great Plains site in Lamont, Oklahoma, with over ten large wind farms (each comprising up to 100 turbines) installed near the measurement site at ranges from 17 km to 67 km, with the closest (and largest) installations just to the east, west, and south. We apply both traditional LLJ detection methods (i.e., drop-off and shear) and a data-driven agglomerative-clustering method to hourly-averaged wind speed profiles. The aggregate LLJ occurrence rates are essentially indistinguishable pre- and post-construction (approximately 40 % by the shear method and 36 % by the drop-off method), with a transient reduction of approximately 3 % (shear method) and 4.5 % (drop-off method) during the construction era (2012–2020). Data-driven analysis further resolves sector- and time-of-day-specific shifts: in all wind sectors examined, including a sector with no upstream farms, clusters with strong near-ground shear seem to be replaced by clusters with reduced near-ground shear. However, LLJ rates are not significantly affected, although the results suggest a stronger modification at night (local time 0–6, reductions of around 50 % for specific clusters). Given concerns about long-term background climatological variability, it is very difficult to ensure that these results are statistically significant. In contrast to previous simulations, we find no statistically significant evidence of wind farm modifications of LLJ rates.
Review of WES-26-98
Impact of wind farms on the occurrence and characteristics of low-level jets
By Moss et al.,
This manuscript reports on a study that investigates the change of low level jet characteristics between a pre-wind farm and a wind farm era in the region. Long term radar data are used to study wind profile statistics to answer this question. I have several concerns about this manuscript in terms of clearness which research question is precisely studied, how the applied methods serve the goal, the maturity of the text and especially maturity of figure captions. The text reads as a bumpy ride where the reader is not taken by the hand to learn about the outcomes, but it feels like a chain of detours. I gave up reading after Fig 12…… ( I think for the first time in 20 years of reviewing papers).
Recommendation: reject.
Major remarks:
Minor remarks:
Ln 1: is an occurrence of super-geostrophic wind speeds. This is only the case if it is an inertial oscillation, but there are other mechanisms to create wind maxima.
Figure 1: label the panels a,b,c, and refer to them in the fig captions.
Ln 47: Figure 1 illustrates schematically how this interaction might look. I must say I do not see any interaction in the graph, I only see a hypothesis how profiles and histograms look like before and after wind farm implementation. It would strengthen the paper if more of the mechanisms are being discussed. E.g. till this point in the paper I did not read about atmospheric stability. Wind turbines creative turbulence in their wakes which will mix the atmospheric profiles and thus make the flow less stably stratified, which should reduce or delay the de-coupling between surface and atmosphere.
Ln 59: is also considered how often turbines need to be switched off between cases, i.e. when LLJ and wind shear below the LLJ become too large/strong?
Ln 69: the LLJ is a local maximum wind speed. This needs to be reworded. If you mean locally in the vertical, then yes I agree, but horizontally I would disagree. Decoupling of the surface from the atmosphere occurs over large areas, and these areas are also advected by the mean wind. So you cannot think in vertical columns only. Please comment.
Ln 68-80: this paragraph needs to be tidied, since it is very unclear. I feel you try to explain how jet detection may occur, but at the same time basic references to works that explain and illustrate this detection as Baas et al. and Kalverla et al. and others are missing.
Ln 92: To achieve these goals. I did not read anything about goals or objectives before. Reword…
Ln 92-109: this piece can be drastically shortened to 3-4 sentences. It is a TOC that is too long. Better use the word count for formulating a proper and researchable research question.
Ln 116: it is not clear here how a radar is a suitable instrument to detect wind speed profiles, and specifically wind maxima as LLJs. I would say a radar needs larger particles in the atmosphere to detect displacements than that are available in the atmosphere. The authors should also make clear why a radar is better for this purpose than a radio sounding or a wind profiler.
Ln 122: one should indicate what is the vertical resolution of the radar output? Is it fine enough to detect wind maxima? Ln 124 says it is 60 m vertical spacing. In my view that is way too coarse to detect LLJs. I would like to see a comparison of LLJ detection between high-resolution tower data or radio sounding data on the one hand and the radar at 60 m on the other. I expect that when the first two methods are degraded to 60 m, then the LLJ climatology will alter to less LLJs, more vertically diluted LLJs and weaker LLJs.
Ln 122, same section please add what is the accuracy of the wind speed profile and its shear as detected with the radar, and for the jet nose in particular.
Ln 130: 1-hour profiles. Please justify why 1h profiles are suitable for studying LLJ dynamics. LLJs can evolve relatively quickly, i.e. they originate from a sudden decoupling of the surface from the atmosphere. The fact it happens suddenly already means short time scales are relevant.
Ln 139-140: We restrict the data to heights between 125m and 1 km. This is tricky in the sense that earlier studies have shown that substantial number of LLJs occur below the 125 m level (e.g. Baas et al. 2009), while Kalverla et al, (2017) remarks that “It is noteworthy that most LLJs are observed at or around 100 m altitude”. It seems the current authors sample only a subset of LLJ potential. Please comment.
Ln 142: the PCA appears completely out of the blue here. The reader is not guided why this analysis is performed, neither what is the precise purpose of it. Need to be revised.
Figure 5, caption: refer to each panel in the caption. Als expand MBB, so the reader does not need to go back to the main text for clarification. Also be more precise in the caption about the precise location of these data.
Ln 222-225: justify why the mentioned threshold values are suitable for your specific study location.
Figure 7: “Populations of the four quadrants” does not say much to the reader, who is forced to go back to the text. And then learns that the analysis is not between obs and model, for which MCC is usually used, but between two models. What is exactly the purpose of this aspect of the study?
Table 2, and related text: it is completely unclear to me why this analysis is performed. Why do we need to know how often and when a case A transitions into case B or vv. I.e. we talk about nothing that is related to the physics, only about whether two detection methods agree or not. I do not see what we learn about this analysis in answering whether LLJ statistics change with the implementation of a wind farm or not. It is a big de-tour.
Section 4 till ln 361: this section is very wordy and not necessarily clarifying. Please add more context why this analysis is needed to answer the research question. Of course you can enter all data into such an algorithm, and get output, but why would one perform such an analysis if it does not contribute to the research question. It also reconfirms that the research question in the Introduction is not well stated.
Figure 10: is discussed in the text in only 4 lines. So is it really necessary to be included?
Figure 10, caption: … of subclustered results. Of what? Be more precise in the formulation of the fig captions. The caption lacks the time range of the data involved, and the location where data have been taken. The difference between green and red boxes is not indicated in the caption. Also red/green choice makes the figure not very colourblind proof. Meaning of C1…C9 is not indicated in the caption either.
Figure 11: x and y axes should be swapped since the original cluster is the independent variable in this analysis and the reclustered one the dependent variable.
Figure 11, caption: Confusion matrix showing initial agglomerative labels and relabelled results from the final k-means pass. The caption does not say much. I read it several times, and tried to find it in the main text what I should learn from this Fig, but it is unclear. If it is the % of data that stick to their original cluster after implementation of wind farm, then please say it more clearly and more straightforward. If this is not the message, then I do not see why this “statistics sauce” (sorry to say) is needed.
Figure 12: In many panels text/legend is plot over the key lines, which is not a proper way of doing.
Figure 12, caption: Color indicates the period. Yes they do, but as reader I first need to go back to earlier figures to find which color relates to which period.
I am sorry to say but I gave up reading from here, since it is unclear what the paper heads towards, what is the precise goal, how the applied methods serve that goal, why 9 clusters have been chosen without a physical basis….