the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 18 Nov 2024
A WRF-Hydro-based retrospective simulation of water resources for US integrated water availability assessment
Abstract. A systematic and periodic evaluation of water supply across the United States is critical for gaining comprehensive insights into the present state of the nation's water resources and strategically planning for the future. The U.S. Geological Survey (USGS) Integrated Water Availability Assessments (IWAAs) is a national initiative designed to characterize past, present, and future water availability at selected basins in the United States. The Weather Research and Forecasting model hydrological modeling extension package (WRF-Hydro) is one of the selected hydrologic models used to generate an estimate of national hydrological fluxes and storage across the conterminous United States (CONUS). The WRF-Hydro application is being forced using the state-of-the-art CONUS404 dataset, a regional hydroclimate dataset over the CONUS, and evaluated over water years 2010–2021. Calibration leads to substantial improvements in simulated streamflow across most of the CONUS. Following parameter regionalization, streamflow performance is reasonable at USGS gages, particularly in the eastern and western regions. However, certain challenges arise in the central US, Arizona, and south Florida, where the model exhibits poor performance. The observed shortcomings in these regions can be attributed to a combination of deficiencies within the framework of the model code, its configuration and atmospheric forcing errors, with a specific emphasis on temporal accuracy issues.
Throughout the CONUS, WRF-Hydro IWAAs based simulations of snow water equivalent closely align with the Snow Data Assimilation System (SNODAS) during the snow accumulation season but show low biases during the snow ablation season. WRF–Hydro IWAAs based actual evapotranspiration (ET) simulations generally exhibit close agreement with Global Land Evaporation Amsterdam Model (GLEAM) ET estimates when comparing cumulative distribution functions across CONUS. Despite this overall agreement, simulated WRF-Hydro IWAAs ET is higher in parts of the central US and lower in parts of the northeast, southeast, and northwest regions of the US, and in urban areas when compared to GLEAM. There is a strong agreement between WRF-Hydro IWAAs based simulations and GLEAM surface soil moisture (top 10 cm) values, with the WRF-Hydro IWAAs model simulating some higher estimates particularly over the eastern US. Similarly, simulated WRF-Hydro IWAAs root-zone soil moisture is underestimated in the southeast US while there are positive biases observed in the western US, relative to the GLEAM simulations. These comparisons to independent datasets indicate the WRF-Hydro application developed for the USGS IWAAs is producing reasonable simulations in many locations across CONUS but is over- or underestimating model variables in some regions.
- Preprint
(42528 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on hess-2024-262', Anonymous Referee #1, 10 Dec 2024
The comment was uploaded in the form of a supplement: https://hess.copernicus.org/preprints/hess-2024-262/hess-2024-262-RC1-supplement.zip
-
RC2: 'Comment on hess-2024-262', Anonymous Referee #2, 20 Dec 2024
The manuscript titled "A WRF-Hydro based retrospective simulation of water resources for US integrated water availability assessment" by Rafieeinsasab et al. discusses a nationwide setup of the WRF-Hydro modeling system aimed at assessing water resources and availability in the United States. The paper is well written and clearly structured; however, it seems more like a report than an exploration of scientific questions. There are already numerous studies available that describe the calibration of WRF-Hydro. What's new here is the parameter regionalization approach but it seems to have a limited applicability for some of the regions. The discussion of the calibration results doesn't highlight potential starting points for improvements of the model (or does touch the sore spots). On the other hand, the water budget study remains quite general with only long-term average analyses and no in depth discussion of the model's applicability to the different climatic conditions of the CONUS. Thus, the study's objective is ambiguous as it is neither a proper physically based analysis of the model itself nor an detailed examination of the model's capability to simulate water budgets under various climatic conditions. Therefore, I'd like to invite the authors to focus more on one of these aspects and I would recommend to do this in favor of a more nuanced analysis of the model's capability and shortcomings to simulate water budgets and resources.
Specific comments:
Here the model was used in a similar setup as it has been developed for the NWM which primarily addresses discharge prediction. However, since this study aims to examine water budgets – also on a longer term perspective – it is worth questioning whether the chosen NWM-based setup is appropriate for the task. Groundwater, for instance, is considered only in terms of a discharge contribution, approximated by a conceptual linear reservoir model. However, for water resources assessment Groundwater represents an important storage body. As shown in Rummler et al. (2022, https://doi.org/10.1002/hyp.14510) a more sophisticated description of groundwater processes WRF-Hydro can lead to significantly improved discharge estimates. Further improvements to the model were demonstrated in a study focused on semi-arid environments, in which even some of the co-authors were involved (Lahmers et al. 2019, https://doi.org/10.1175/JHM-D-18-0064.1). However, these findings were not considered here even though they could have enhanced simulations for the southwestern regions.
For the bias correction, what is the reasoning for using a "day-of-the-year" approach? Is it that you want to do a kind of climatological bias correction? You stated that Daymet is only available until 2017. Looking to the data portal, one can find data until 2023. So data including 2021 should have been available at the time of your analysis. Nevertheless, using a long term climatological correction does not account for inter-annual variability (e.g., enduring extraordinary drought periods) and may further disregard fundamental changes in climate between the overall averaging period (1980-2017) and the study period (2010-2021). With the available data you could have pursued a more time-variant approach.
Concerning the selection of relevant parameters (Table 1) it is mentioned that it also relies also on literature review; however, the extensive Noah MP parameter study by Cuntz et al. (2016, https://doi.org/10.1002/2016JD025097) is not considered or at least not mentioned. How has the relevance of the parameters been assessed in general? In terms of calibration results it would be interesting to learn about the ranges, patterns, and relationships of the optimized values but this relates to my initial comment about the focus of this study –whether it should concentrate more on the model itself or on the budget analysis.
Why do you consider hourly measures for the evaluation of the calibration basins and daily measures for the regionalized stream flow?
It is anticipated that for the basins with regionalized parameter sets performance will be reduced. Have you considered employing also other regionalization approaches, such as proposed by Schweppe et al. (2022, https://doi.org/10.5194/gmd-15-859-2022)?
In section 5.2 it would be very interesting to read more about the physical reasons for the mismatch of the simulations. Perhaps this could also be discussed in a final overarching interpretation of the results that includes all the different water budgets analyzed in this study.
The rest of the analysis for snow, evaporation, soil moisture only shows that the model can reproduce a 10-year climatological mean. This may also be achievable from an reanalysis. Here I would expect a more in-depth analysis of the water budgets, regional features, the shortcomings of the model and potential for improvement, as well as the benefits of simulating the water budgets at such high resolution and over this large spatial extent. Additionally, the general suitability of the model in its current configuration for longer-term trend analysis and future water availability projections as is intended by the IWAA initiative should be discussed.
Finally, my last point is about long term storage, i.e. groundwater. There is no evaluation of the changes in terrestrial water storage which I assume is important for water budget analysis and water balance closure. GRACE derived deviations should be well suited for such an analysis given the large spatial extent of the study region.
Minor comments:L263: There are several studies that point out the ineptitude of NSE as a goodness of fit measure for daily to subdaily hydrographs.
L263-275: There's no information about the temporal resolution used for model validation.
Figs. 7 & 8: Captions doe not include information about the time period used for the analysis.
Figs. 14–19: Captions doe not include information about the time period used for the analysis.Citation: https://doi.org/10.5194/hess-2024-262-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 215 | 60 | 7 | 282 | 5 | 4 |
- HTML: 215
- PDF: 60
- XML: 7
- Total: 282
- BibTeX: 5
- EndNote: 4
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1