Improving Drought and Flooding Predictions from Land Surface Models
To understand how water moves on the land and make predictions about the future, scientists use computer models called large-scale land surface models. One example is the Community Land Model version 5 (CLM5). However, these models have uncertainties, especially when it comes to how much water is in the ground and how it moves. To tackle this problem, we looked at how different hydrological parameters in CLM5 affect its predictions. We tested the model with different weather data from across the United States to see when these uncertainties matter the most. This information is important for predicting floods and droughts, which can help us make better decisions in the future.
This study has created a dataset that serves as a reference for evaluating the performance of CLM5. The dataset includes information on how sensitive the model is to different parameters, as well as outputs from a large number of simulations. This dataset can be used for various purposes, such as understanding the uncertainty in meteorological and hydrological parameters, guiding the calibration of the model, and helping the developers of CLM5 identify potential issues with the model's structure and parameters. The dataset can be used at different spatial scales, from local to regional to the entire United States.
Land surface models, like CLM5, are crucial for simulating how the Earth's terrestrial system behaves. However, there has been limited research on the uncertainties associated with CLM5's hydrological parameters and their impacts on water resource applications. To address this, we conducted a comprehensive study using five different meteorological datasets ((NLDAS-2, PRISM, Daymet, Livneh, and WRF-ERA5) to analyze the uncertainties in CLM5's hydrological parameters for 464 basins across the conterminous United States (CONUS). This study resulted in the creation of a benchmark dataset that captures CLM5's default hydrological performance, sensitivities of 28 hydrological metrics to different parameters, and a large ensemble of outputs for CLM5's hydrological predictions. The basins were grouped into seven clusters for regional analysis, which were then extended to cover the entire CONUS for guiding the use of the dataset in ungauged basins. These datasets are invaluable for calibrating CLM5 and have broad applications, including evaluating vulnerabilities to droughts and floods. They can help identify the hydroclimatological conditions under which parametric uncertainties have significant effects on hydrological predictions, and where further investigations are needed to understand how uncertainties in hydrological predictions interact with other Earth system processes.