| 23 Apr 2026
The HRRR Meteorology, Energy, and Transmission (MET) Toolkit: Advancing high-resolution atmospheric data for contiguous U.S. energy applications
Abstract. High-quality, multiyear atmospheric data are foundational for power system planning and grid integration. While the legacy Wind Integration National Dataset (WIND) Toolkit has long served as the industry standard, its historical record ends in 2013, leaving a critical gap in current modeling capabilities. Modern alternatives, such as the WIND Toolkit Long-term Ensemble Dataset (WTK-LED) and its Climate variant, offer extended coverage but exhibit higher wind speed biases and are computationally intensive to produce. This study introduces the High-Resolution Rapid Refresh Meteorology, Energy, and Transmission (HRRR MET) Toolkit, a repackaged version of the National Oceanic and Atmospheric Administration's native HRRR data. The HRRR MET Toolkit is designed to overcome the significant technical barriers associated with accessing native HRRR formats by providing a streamlined, user-friendly dataset with high vertical resolution at power generation-relevant heights. To ensure seamless continuity for long-term studies, the HRRR MET Toolkit is provided on the same uniform 2 km horizontal grid as the legacy WIND Toolkit, offering both modern accessibility and spatial consistency with the established historical record. To evaluate potential performance gains, we also assessed an experimental bias-corrected version using quantile mapping against the WIND Toolkit as a climatological reference. We provide a comprehensive validation of both HRRR variants alongside the WTK-LED, its Climate variant, and the 2023 National Offshore Wind (NOW-23) dataset against long-term observations across the contiguous United States. Results indicate that the HRRR MET Toolkit significantly outperforms the WTK-LED suite; for instance, it reduces hub-height average wind speed bias to 0.10 m s-1 (compared to 0.82 m s-1 for the WTK-LED) and achieves an hourly wind speed correlation of 0.82. Critically, the comparison between the native and bias-corrected HRRR variants reveals that the statistical correction offers marginal benefit and in some cases exacerbates positive wind speed biases in complex terrain. We conclude that the native HRRR physics are sufficiently robust for energy applications and therefore recommend the HRRR MET Toolkit as a highly accessible, accurate, and less complex standard for modern power system studies in the United States.
Review of “The HRRR Meteorology, Energy, and Transmission (MET) Toolkit: Advancing high-resolution atmospheric data for contiguous U.S. energy applications” by Bodini et al.
General comments:
This manuscript is extremely important as it demonstrates the very significant value that the operational HRRR model provides for wind resource characterization relative to other historical data sets, and moreover provides a tool to make the HRRR data set readily available to the wind energy community. It demonstrates this value through a series of well-conceived figures, displaying a wide variety of metrics. The manuscript if well-written, and easy to follow, and was a pleasure to read. My only two significant comments, are first, that the analysis appears to assume that the ERA5 is bias-free, which it is not. Second, and more importantly, the analysis does not appear to take into account tower shadowing effects in tall-tower data. If that is true, then I believe that tower shadowing can explain some, if not most, of the large differences in the bias behavior found between onshore tall tower sites and lidar sites. Ideally, the tall-tower analysis should be repeated removing data values potentially affected by tower shadowing.
Specific comments:
Line 38: What is the “climate variant” of the Wind-LED Toolkit?
Lines 38-40: Is there a reference to the analysis showing that the WIND-LED Toolkit has larger biases than does the WIND Toolkit?
Lines 68-70: A similar evaluation analysis as that of Dorenkamper et al. 2020 was provided by Wilczak et al, 2024, https://doi.org/10.3390/en17071667 for hub-height winds across the US.
Lines 150-154: Can the effective spatial resolution of the re-gridded data set be estimated? Due to the interpolation from four 3-km resolution grid points, I would expect this resolution to be on the order of 4-6 km. Can it be determined more precisely? In any case, the larger effective resolution should be mentioned.
Lines 172-176: For grid integration studies, having solar variables consistent with the HRRR-MET wind variables would be useful. Is there a reason why radiation variables were not included, and is there any plan to include solar variables in the future?
Figure 1. The color-scale here is a bit frustrating, as most mean windspeed values fall within a rather narrow range of 6-9 m/s, with very little color gradation present. As a result, all of the high plains states have an almost uniform color.
Figure 1. Is there any evidence in the multi-year data set of jumps in the CONUS-averaged mean wind speeds that occur at the transition dates between the various HRRR versions? Quantifying such jumps would be important for those concerned about the impact of the changes to the model physics between the model versions. – OK, I see that this has been addressed later on Lines 382-395. It would be good to foreshadow that later analysis here or in the Introduction.
Line 200. After applying the quantile bias correction to the HRRR, are the biases exactly zero at each grid point?
Line 204-205. However, the ERA5 has been shown to have its’ own biases across the CONUS (Wilczak et al, 2024; Pronk et al, 2022). Don’t those biases need to be removed? See also comments on manuscript lines 463-465 below.
Line 218-220. I don’t understand what is meant by “methodological consistency” in requiring that the exact same time periods are used to compare the biases between the different products. Previously, the Wind Tool Kit (2007-2013) was used to bias correct the HRRR-MET (2015-present) event though the two model data sets have no period of overlap whatsoever. Wouldn’t the same methodological (in)consistency apply in that case?
Lines 250-260: Tower wind speeds can be significantly impacted by tower shadowing effects when the cup anemometers or sonic anemometers are downwind of the tower structure. Was anything done to minimize these effects? If not, why not? No mention is made whether the QC flags remove all downstream values affected by tower shadowing. See also comments for Line 295 below and Figure 9 below.
Also, I believe that many of the “tall towers” listed in the Gulf of Mexico are in fact short towers mounted on top of huge offshore drilling platforms, in which case flow distortion effects from the platforms can be very large, even larger than the tower shadowing effects. Was anything done to account for those flow distortion effects? At a minimum, the heights above the tops of the drilling structures should be mentioned, and the possibility for flow distortion effects should be acknowledged.
Line 276: Is there a reference for the Wasserstein distance?
Line 295: I suspect that much of the overall positive biases found in all of the models, and especially when compared against onshore tall towers, is due to the underestimation of the observed wind speeds due to tower shadowing.
Lines 312-313: I don’t understand the sentence “This suggests that while quantile mapping effectively corrects offshore under-speeding, it may exacerbate positive biases onshore where local terrain effects dominate”. This seems to put the blame on the quantile mapping procedure, when it is probably due to inadequacies in the Wind ToolKit data, specifically to differences in the onshore and offshore Wind Toolkit biases relative to the truth.
Figure 9: I note here that the models all tend to closely follow the observed PDFs when measured by lidars, whereas in Fig.7 for tall towers, larger differences exist, which are consistent with tower shadowing effects in which observed high wind speeds are shifted to lower wind speed bins,
Figure 18: This is a very nice and important figure showing the time evolution of the HRRR model as it has improved over time. It may be worth noting here that at least some of these improvements occurred because of the concerted, collaborative effort between DOE, NOAA, and the private sector at finding and correcting HRRR model errors through the WFIP field campaigns. This demonstrates the value of those collaborations and in the investments made in them. Also, I’m curious if the authors can speculate on why the bias increases pretty sharply towards the end of the timeseries, while the other metrics show continued improvement?
Line 405: The impact of data assimilation is an interesting and important question. To some degree, one could test whether the higher skill of the HRRR comes from DA, or from better model physics, by looking at HRRR skill at hour 18, or even hour 48 when those longer forecasts are available. At those longer forecast horizons, the value of DA will diminish, and higher skill in the HRRR relative to the other models will depend more on more physics. This question probably is beyond the scope of this paper, but it could be mentioned as an area for future research.
Line 459: I find it strange that 10m ERA5 data are used in this comparison, even though 100m data are readily available. The rationale for this seems to be that monthly values are pre-averaged at 10 m, but it would be trivial to average hourly values to monthly values, and would make the results much more valuable for wind energy purposes.
Line 461. If I interpret this sentence correctly, ERA5 data only from 1991-2020 are used to determine the long-term conditions, even though the ERA5 data set spans a much longer period. If so, why restrict it to only these 30 years?
Line 464: “… relative to observations …”. I’m confused here, because I thought this comparison was against the ERA5, not observations. The ERA5 values are not observations, but are highly dependent on the IFS model.
Lines 463-465. A significant portion of the Wind Toolkit positive bias relative to the ERA5 shown in Figure B1 is likely due to the known low-bias of the ERA5 (Wilczak et al, 2024). It may be beneficial to acknowledge this, otherwise readers will mistakenly interpret the bias in this figure as a shortcoming of the Wind Toolkit, when in fact it is more likely a shortcoming in the ERA5.
Line 467: At selected observation sites?
Figure C1. The plots say WTK-LED but the caption says Wind Toolkit (no LED). Aren’t those different models?