the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 06 Nov 2024
Coupling the ParFlow Integrated Hydrology Model within the NASA Land Information System: A case study over the Upper Colorado River Basin
Abstract. Understanding, observing, and simulating Earth's water cycle is imperative for effective water resource management in the face of a changing climate. NASA's Land Information System (LIS)/Noah-MP and the ParFlow groundwater model are the two widely used modeling platforms that enable studying the Earth's land surface and subsurface hydrologic processes, respectively. The integration of ParFlow and LIS/Noah-MP models and harnessing their strengths can provide an opportunity to simulate surface terrestrial water processes and groundwater dynamics together while enhancing the accuracy and scalability of hydrological modeling. This study introduces ParFlow-LIS/Noah-MP (PF-LIS/Noah-MP), which is an integrated, physically based hydrologic modeling framework. PF-LIS/Noah-MP enables the user to simulate land surface processes in conjunction with subsurface hydrologic processes while considering the interactions between the two. In this study, we compared the results of the coupled PF-LIS/Noah-MP and standalone LIS/Noah-MP models with a suite of in-situ and satellite observations over the Upper Colorado River Basin (UCRB) in the United States. This analysis confirmed that integrating ParFlow with LIS/Noah-MP not only enhances the capability of LIS/Noah-MP in estimating land surface processes over regions with complex topography but also enables it to accurately simulate subsurface hydrologic processes.
- Preprint
(7120 KB) - Metadata XML
-
Supplement
(3152 KB) - BibTeX
- EndNote
Status: open (until 26 Jan 2025)
-
RC1: 'Comment on hess-2024-280', Anonymous Referee #1, 22 Dec 2024
reply
This manuscript explores a case study of a new model coupling, the intregrated hydrologic code ParFlow and the land surface code LIS/Noah-MP, to simulate hydrologic processes across the Upper Colorado River Basin (UCRB). The authors compare results between the standalone land surface model and the coupled model to in situ and remote sensing observations of soil moisture, streamflow and groundwater levels. When published, this paper will make a valuable contribution to the literature. At the moment, several aspects of the manuscript require clarification and revision before publication. Notably, the details of the model coupling are not fully described in this paper and are instead referenced in a manuscript currently under review, which limits my ability to evaluate the robustness of the methods and results. In addition, it the introduction and conclusions could be revised to clarify the novelty of the manuscript. Therefore, I am recommending major revisions.
Major comments:
1) It is difficult to review this paper, as it presents a case study of a new model coupling, but the details of that coupling are described in a paper currently in review. Thus, I cannot evaluate the results presented here as the underling methods are not fully described. I recommend that the authors either (1) wait to publish this manuscript until the paper by Maida et al. (2024) is fully published, (2) publish a preprint of Maida et al. (2024), or (3) describe the ParFlow-LIS/Noah-MP coupling in depth.2) I recommend the authors edit the introduction to emphasize the motivation for and novelty of this manuscript. ParFlow has long been coupled to one land surface model within the LIS framework (the Community Land Model), so what additional functionality is provided by coupling ParFlow to LIS? There are certainly advantages provided by the data assimilation and uncertainty estimation tools within LIS, but those tools are not used in this manuscript. Perhaps, then, the novelty of this paper is the difference in process representation between CLM and NoahMP, but the comparison to ParFlow-CLM is not presented in this paper.
3) While reviewing the results, I'm not sure I come to the same conclusions as the authors. The abstract (lines 33-34) and conclusions (lines 556-557) find that coupling ParFlow to LIS/Noah-MP improves accuract in regions with complex topography. However, the metrics presented in figures 5, 6, 7, S2 and S3 show that root mean squared error and correlation coefficients are nearly identical between LIS/Noah-MP and ParFlow-LIS/Noah-MP. However, those metrics are averaged over the entire domain, but perhaps there's a difference when those metrics are averaged over areas with complex topography? Please clarify.
Minor comments:
Lines 61-78: This paragraph focuses on the importance of simulating irrigation, groundwater pumping and other water management infrastructure, but those processes are not included in the simulations in this paper. Thus, it would be helpful to clarify why this paragraph is included hereLines 105-108: The reference to WRF feels a little out-of-place since it isn't used in this study. It would be helpful to briefly introduce Noah-MP here instead, since it is mentioned in the abstract and in the next paragraph
Line 116: Should Fadji et al. (2024) be Maida et al. (2024)?
Line 111: It would be helpful to mention some examples of these couplings (ParFlow-CLM, ParFlow-WRF, etc)
Lines 131-150: Description of ParFlow could be more clear
Lines 148-149: "groundwater may take a longer time (for example compared to soil moisture) to reach a steady-state due to such a complicated subsurface configuration" Does "long time" here refer to simulation time or computational time? I'm unclear if this sentence is meant to describe the long time it takes to spin-up water content in the deep vadose zone due to slow rates of groundwater recharge, or if it refers to long computational time due to the difficulty of solving the Richards equation across a thicker vadose zone.
Line 155: Which variables are included in the initial conditions? Soil moisture, surface temperature, what else?
Lines 168-171: I recommend mentioning that the LIS data assimilation framework is not used in this study
Lines 181-182: Is the coupling specific to NoahMP? Or could someone use this same code to use VIC or HySSIB instead of NoahMP?
Section 4: The description of this coupling could be more detailed. How is transpiration from the root zone handled? Is transpiration only from the top soil layer or does NoahMP draw water from deeper layers as well? Also, how is overland flow handled? In this description (and the image in Fig 1), it appears as though ParFlow only simulates subsurface flow.
Line 254: Should be "USGS stream stations"
Line 255: I'm surprised at how few monitoring wells there are. Have the authors considered adding water level measurments from either the Colorado Water Conservation Board or the Utah ? From a cursory glance (https://dwr.state.co.us/Tools/GroundWater/WaterLevels), it seems like there are many water level measurements not included in the USGS database. Also, what are the screened intervals for each well? If a well screen extends across multiple model cells, how are modeled and observed values compared?
Fig. 3: I recommend clarifying that WTD corresponds to "water table depth". Also, do all of these monitoring wells truly represent the depth to the water table? Or do they represent groundwater head? It's unclear whether these wells are screened across the water table.
Line 283: The manuscript could be improved by expanding this section and including additional details on model set up, such as boundary conditions and the extent and discretization of the domain. What are the lateral boundary conditions for the PF-LIS model? I assume that cells outside the UCRB would be inactive, but results for those cells are shown in Fig 4, 5, etc. Similarly, what is the extent of the LIS/Noah-MP domain? What are the lateral boundary conditions?
Line 293: 1 km lateral resolution?
Line 296: Are these depths or thicknesses?
Line 317: Were these three 20-year periods run sequentially? Also, how was 60 years determined to be an adequate spin-up period? Are there metrics to determine whether the system is at dynamic steady state?
Lines 320-321: How different were the initial conditions across the shared portions of the PF-LIS/NoahMP domain? How were differences in the two soil moisture fields reconciled before starting the first coupled simulation?
Line 323: What metrics were used to determine that the system was at quasi-equilibrium?
Line 329: What size time step was used for input forcing and the output analysis for these simulations? Hourly meteorological forcing? Daily pressure/saturation output?
Line 349: What is the difference in input forcing that provides this finer spatial resolution in PF-LIS/Noah-MP than in LIS/Noah-MP alone? Weren't both codes run using the same lateral resolution?
Fig. 4: What are the values outside of the UCRB watershed boundary and why does the resolution appear to be lower beyond that boundary in PF-LIS? Are those cells identical between the two simulations?
Line 360: How does the vertical resolution of the simulations compare to the SMAP penetration depth?
Fig. 5: In the caption, it could be useful to clarify the depth interval for the simulated soil moisture values.
Line 384: Could the difference in overland flow between PF-LIS/Noah-MP and LIS/Noah-MP also contribute to the increased spatial heterogeneity observed in PF-LIS/Noah-MP simulations?
Lines 399-400: It's not immediately clear from the figures that PF-LIS/Noah-MP improves the accuracy of soil moisture in high altitude regions. Is there an alternate figure that more clearly shows this result?
Line 455: Why do you think PF-LIS/Noah-MP is unable to capture the timing of runoff? Is this due to errors in hydraulic conductivity, which cause inaccurate estimates of the timing of the rainfall-runoff response?
Line 471: A minor point, but it could be useful to add to this diagram the number of points that are in each quadrant/category.
Line 492: How were groundwater heads compared between simulated and observed values? Given that some cells are up to 200 m thick and PF-LIS/Noah-MP reports a single pressure value per cell, do these calculations assume hydrostatic equilibrium within a given cell to calculate the exact water table depth within that cell? Similarly, for wells that have a long screen length and are screened entirely below the water table, the reported water level measurements integrate pressure across the length of the screen.
Lines 495-497: "Stations located in topographically complex surroundings tend to yield lower model performance compared to those in areas with smoother and flatter environments." Would it be possible to include a figure in the supplement to support this statement? It might be more clear to show this relationship in a map rather than in a table of latitude and longitude values.
Lines 528-530: This is an interesting result! Does this discrepancy also suggest that PF-LIS/Noah-MP underestimates evapotranspiration because croplands in the simulations do not receive any groundwater-fed irrigation? Another option for future work would be to compare remote-sensing-based estimates of ET with estimates from both PF-LIS/Noah-MP and LIS/Noah-MP.
Citation: https://doi.org/10.5194/hess-2024-280-RC1
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 179 | 52 | 9 | 240 | 30 | 6 | 7 |
- HTML: 179
- PDF: 52
- XML: 9
- Total: 240
- Supplement: 30
- BibTeX: 6
- EndNote: 7
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1