Benefits and Pitfalls of GRACE and Streamflow Assimilation for Improving the Streamflow Simulations of the WaterGAP Global Hydrology Model. Schulze, K., Kusche, J., Gerdener, H., Döll, P., & Müller Schmied, H. Journal of Advances in Modeling Earth Systems, 16(10):e2023MS004092, October, 2024.
Benefits and Pitfalls of GRACE and Streamflow Assimilation for Improving the Streamflow Simulations of the WaterGAP Global Hydrology Model [link]Paper  doi  abstract   bibtex   
Abstract Distribution and change of freshwater resources is often simulated with global hydrological models. However, owing to process representation limitations and forcing data uncertainties, these model simulations have shortcomings. Combining them with observations via data assimilation, for example, with data from the Gravity Recovery and Climate Experiment (GRACE) mission or streamflow measured at in situ stations is considered to improve the realism of the simulations. We assimilate gridded total water storage anomaly (TWSA) from GRACE into the WaterGAP Global Hydrology Model (WGHM) over the Mississippi River basin via an Ensemble Kalman Filter. Our results agree with previous studies where assimilating GRACE observations nudges TWSA simulations closer to the observations, reducing the root mean square error (RMSE) by 21% compared to an uncalibrated model. However, simulations of streamflow show degeneration at more than 90% of all gauge stations for metrics such as RMSE and correlations; only the annual phase of simulated streamflow improves at half the stations. Therefore, for the first time, we instead assimilated streamflow observations into the WGHM, which improved simulated streamflow at up to nearly 80% of the stations, with normalized RMSE showing improvements of up to 0.1, while TWSA was well‐simulated in all metrics. Combining both approaches, that is, jointly assimilating GRACE‐derived TWSA and streamflow observations, leads to a trade‐off between a good fit of both variables albeit skewed to the GRACE observations. Overall, we speculate that our findings point to limitations of process representation in WGHM hindering consistent flux simulation from the storage history, especially in dry regions. , Plain Language Summary The distribution of freshwater on Earth can be simulated by hydrological models like the WaterGAP Global Hydrology Model (WGHM). Changes of the total water storages can also be derived from satellite gravimetry, for example, with the Gravity Recovery and Climate Experiment (GRACE) mission. Model and observations do not necessarily agree and are both prone to errors. We combined the model and observations over the Mississippi River basin taking the errors into account by applying a data assimilation technique. This led to a more realistic simulation of the total water storage changes. Since simulated streamflow was found to degenerate at nearly all gauge stations, we assimilated streamflow observations (instead of GRACE data) into the WGHM. This turned out to improve simulated streamflow at up to nearly 80% of the stations. The assimilation of both data sets (GRACE and streamflow) together leads to a compromise between the TWSA and the streamflow simulations, with the GRACE data affecting the simulations more than the streamflow data. Our findings are expected to aid reducing the weaknesses of the WGHM model equations and thus contribute to improved quantification of Earth's freshwater distribution in the future. , Key Points GRACE assimilation improves the WGHM storage representation, but degrades the simulated streamflow at up to 99% of the gauge stations Assimilating streamflow data instead improves the streamflow simulations at up to 79% of the validation stations Joint assimilation leads to a trade‐off between a good fit of simulated storages and streamflow albeit skewed to the GRACE observations
@article{schulze_benefits_2024,
	title = {Benefits and {Pitfalls} of {GRACE} and {Streamflow} {Assimilation} for {Improving} the {Streamflow} {Simulations} of the {WaterGAP} {Global} {Hydrology} {Model}},
	volume = {16},
	issn = {1942-2466, 1942-2466},
	url = {https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2023MS004092},
	doi = {10.1029/2023MS004092},
	abstract = {Abstract
            Distribution and change of freshwater resources is often simulated with global hydrological models. However, owing to process representation limitations and forcing data uncertainties, these model simulations have shortcomings. Combining them with observations via data assimilation, for example, with data from the Gravity Recovery and Climate Experiment (GRACE) mission or streamflow measured at in situ stations is considered to improve the realism of the simulations. We assimilate gridded total water storage anomaly (TWSA) from GRACE into the WaterGAP Global Hydrology Model (WGHM) over the Mississippi River basin via an Ensemble Kalman Filter. Our results agree with previous studies where assimilating GRACE observations nudges TWSA simulations closer to the observations, reducing the root mean square error (RMSE) by 21\% compared to an uncalibrated model. However, simulations of streamflow show degeneration at more than 90\% of all gauge stations for metrics such as RMSE and correlations; only the annual phase of simulated streamflow improves at half the stations. Therefore, for the first time, we instead assimilated streamflow observations into the WGHM, which improved simulated streamflow at up to nearly 80\% of the stations, with normalized RMSE showing improvements of up to 0.1, while TWSA was well‐simulated in all metrics. Combining both approaches, that is, jointly assimilating GRACE‐derived TWSA and streamflow observations, leads to a trade‐off between a good fit of both variables albeit skewed to the GRACE observations. Overall, we speculate that our findings point to limitations of process representation in WGHM hindering consistent flux simulation from the storage history, especially in dry regions.
          , 
            Plain Language Summary
            The distribution of freshwater on Earth can be simulated by hydrological models like the WaterGAP Global Hydrology Model (WGHM). Changes of the total water storages can also be derived from satellite gravimetry, for example, with the Gravity Recovery and Climate Experiment (GRACE) mission. Model and observations do not necessarily agree and are both prone to errors. We combined the model and observations over the Mississippi River basin taking the errors into account by applying a data assimilation technique. This led to a more realistic simulation of the total water storage changes. Since simulated streamflow was found to degenerate at nearly all gauge stations, we assimilated streamflow observations (instead of GRACE data) into the WGHM. This turned out to improve simulated streamflow at up to nearly 80\% of the stations. The assimilation of both data sets (GRACE and streamflow) together leads to a compromise between the TWSA and the streamflow simulations, with the GRACE data affecting the simulations more than the streamflow data. Our findings are expected to aid reducing the weaknesses of the WGHM model equations and thus contribute to improved quantification of Earth's freshwater distribution in the future.
          , 
            Key Points
            
              
                
                  GRACE assimilation improves the WGHM storage representation, but degrades the simulated streamflow at up to 99\% of the gauge stations
                
                
                  Assimilating streamflow data instead improves the streamflow simulations at up to 79\% of the validation stations
                
                
                  Joint assimilation leads to a trade‐off between a good fit of simulated storages and streamflow albeit skewed to the GRACE observations},
	language = {en},
	number = {10},
	urldate = {2024-10-23},
	journal = {Journal of Advances in Modeling Earth Systems},
	author = {Schulze, K. and Kusche, J. and Gerdener, H. and Döll, P. and Müller Schmied, H.},
	month = oct,
	year = {2024},
	pages = {e2023MS004092},
}
Downloads: 0
About
Documentation
Keywords
Search
Users
Terms and Conditions of Use
Privacy Policy
Sitemap
© BibBase.org 2021