Evaluation of a New Observationally Based Channel Parameterization for the National Water Model
- 1Department of Civil, Environmental, and Architectural Engineering, University of Colorado Boulder, Boulder, 80309, USA
- 2Cooperative Institute for Research in Environmental Sciences, University of Colorado Boulder, Boulder, 80309, USA
- 3National Center for Atmospheric Research, Boulder, 80305, USA
- 1Department of Civil, Environmental, and Architectural Engineering, University of Colorado Boulder, Boulder, 80309, USA
- 2Cooperative Institute for Research in Environmental Sciences, University of Colorado Boulder, Boulder, 80309, USA
- 3National Center for Atmospheric Research, Boulder, 80305, USA
Abstract. Accurate representation of channel properties is important for forecasting in hydrologic models, as it affects height, celerity, and attenuation of flood waves. Yet, considerable uncertainty in the parameterization of channel geometry and hydraulic roughness (Manning’s n) exists within the NOAA National Water Model (NWM), due largely to data scarcity: only ~2,800 out of the 2.7 million river reach segments in the NWM have measured channel properties. In this study, we seek to improve channel representativeness by updating channel geometry and roughness parameters using a large, previously unpublished hydraulic geometry (HyG) dataset of approximately 48,000 gages. We begin with a Sobol’ sensitivity analysis of channel geometry parameters for 12 small semi-natural basins across the continental U.S., which reveals an outsized sensitivity of simulated flow to Manning’s n relative to channel geometry parameters. We then develop and evaluate a set of regression-based regionalizations of channel parameters estimated using the HyG dataset. Finally, we compare the model output generated from updated channel parameter sets to observations and the current NWM v2.1 parameterization. We find that, while the NWM land surface model holds the most influence over flow given its control over total volume, the updated channel parameterization leads to improvements in simulated streamflow performance relative to observed flows, with a statistically significant mean R2 increase from 0.479 to 0.494 across approximately 7,400 gage locations. HyG-based channel geometry and roughness provide a substantial overall improvement in channel representation over the default parameterization, updating the previous set value for most reaches of Manning’s n = 0.060 to a new range between 0.006 and 0.537 (median 0.077). This research provides a more representative, observationally based channel parameter dataset for the NWM routing module, as well as new insight into the influence of the routing module within the overall modeling framework.
Aaron Heldmyer et al.
Status: final response (author comments only)
-
RC1: 'Comment on hess-2021-552', Anonymous Referee #1, 03 Jan 2022
Review for paper: “Evaluation of a New Observationally Based Channel Parameterization for the National Water Model”
In this paper, the authors explored the effects of modifying channel routing parameters in the National Water Model (NWM) streamflow simulations using a regionalized channel geometry and Manning’s roughness dataset.
The study is an important contribution to the possibility of improving the NWM in order to provide a better quality of the results, focusing especially on those areas where significant differences were found.
This reviewer considers the paper is suitable for publication in Hydrology and Earth System Sciences” Journal, after the authors address the following suggestions and comments.
Line 22. An increase in mean R2 by just over one hundredth (from 0.479 to 0.494) is not a significant variation. I suggest changing the word “significant” for “modest” as you do in the Conclusions (Line 454).
Line 40. Here you must define the abbreviation “LSM” (Land Surface Model), since is the first time in the document that you are using it.
Line 130. I suggest including a brief discussion on the dependence and variation of the parameter ncc as a function of the extension of the floodplain (dxcc variation). Same for the dependence and variation of n as a function of the channel depth (d variation).
Line 171. Explain more details about the criteria that you considered to define the 12 basins of study. It could be summarized in a table that list the climate, land cover and terrain characteristics for each basin.
Line 201. One of my main concerns is that the authors didn’t explore the uncertainty of the longitudinal slope (S) in Equation 10 to obtain Manning’s n. The authors made two strong assumptions that should to be better discussed and justified: 1) Authors used the terrain slope instead of the hydraulic grade line in the Manning equation, and 2) the slope was not measured but obtained from a terrain model that could include errors of the DEM and those errors propagate in the obtained Manning’s n.
Line 206. I suggest renaming the parameter “b” in the linear regression Equation 11, as it might be confusing with the exponent “b” in Equation 6.
Line 235. Explain more details of the reasons you considered for choosing the 99th and 99.9th percentile flows to calculate TW.
Line 277. In this line you mentioned Figure S1 but in the Appendix A the Figure is called Figure A1. This figure also has a poor resolution and is difficult to read it. Improve the quality of the figure.
Line 287 - 289. The analysis written in these lines does not correspond to what is shown in Figure 6. The caption of Figure 6 says “…and boxes with text indicate the combination that resulted in the lowest error, which is shown within the box”. However, you mention “For example, the regression determined from 90th percentile flow yielded the smallest Manning’s n error in the California region (18), whereas the smallest error in the Tennessee region (06) was achieved at the full CONUS-wide regionalization scale” but according to Figure 6 for region 18 the smallest Manning’s n error is in the 75th percentile and for region 6 the smallest error is for the HU4 scale.
Review both the figure and the discussion for consistency.
Line 302. In this line you mentioned Figure S2 but in the Appendix A the Figure is called Figure A2. Improve the quality of the figure.
Line 344. In this line you mentioned Figure S3 but in the Appendix A the Figure is called Figure A3.
Line 345. Improve the quality of Figure 9
Line 380. Improve the quality of Figure 10
Line 405. How can you ensure that a smaller scale typically results in the lowest error, if according to Figure 6, only 9 out 18 regions show the smallest Manning’s error for HUC4 which is equivalent to 50% of the study regions, and 8 regions show the smallest error for HUC2 which is around 45%? I do not see much of difference here to affirm that sentence.
Line 410. In your analysis you mention “Furthermore, the strong performance of the HUC4 regionalization scale relative to HUC2 in the Missouri Region (10) speaks to the diversity of terrain conditions…” However, according to Figure 6, in region 10 the smaller Manning’s error was found in HUC2 which means the strong performance here is not for HUC4 regionalization scale.
- AC1: 'Reply on RC1', Aaron Heldmyer, 27 May 2022
-
RC2: 'Comment on hess-2021-552', Anonymous Referee #2, 04 May 2022
Overall the paper follows sound and well known techniques to estimate, regionalize, and calibrate channel geometric and friction parameters in applications used in continental scale water modeling. Due to the large spatial scales of the National Water Model, several simplifying assumptions are understandibly employed to come up with regionalized estimates of the most important parameters as determined by a sensitivity analysis. These assumptions are well characterized and the limitations of them are properly described in the assumptions. Although the overall skill improvement of the aggregated response in time and space is minor, several knowledge contributions are made in the process. An assessment of the spatial variance of the methods is properly documented and some of the major contributing factors to skill performance outside of the scope of the paper are discussed.
One of the principle questions that is a bit unclear to me is if the Land Surface Model wasn't ran in the 8 year simulation then how were the inflows and subsurface fluxes determined? Since stream flows are highly dependent on magnitude to determine their optimal, respective parameter settings, it's important to have more clarity as to where these data points were obtained. Otherwise, the limitations of the study are clearly outlined especially including the coarseness of the regionalization, lack of additional predictors in the regression, lack of spatial relationships, and lack of consideration for compound friction values. A fair survey is conducted of more robust techniques that can be employed in the future to possibly obtain better results. An assessment of the spatial variance of the performance of the methods is properly documented and some of the major contributing factors to skill performance outside of the scope of the paper are discussed. The study's conclusions are fair given the methodologies employed and the results obtained. Overall, the motivation for more work into calibrating continental scale hydrologic models is well argued for.
Specific technical corrections hover around the ambiguous use of variable symbols including but limited to m, b, i, S, and w. The authors should strive to use unique notations for each variable across equations to avoid unnecessary ambiguities. More clarity can be provided when discussing the different samples of stream gages used. Please see the attached file for more technical comments.
- AC2: 'Reply on RC2', Aaron Heldmyer, 27 May 2022
Aaron Heldmyer et al.
Aaron Heldmyer et al.
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
427 | 105 | 15 | 547 | 6 | 8 |
- HTML: 427
- PDF: 105
- XML: 15
- Total: 547
- BibTeX: 6
- EndNote: 8
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1