Toward merging MOPEX and CAMELS hydrometeorological datasets: Compatibility and statistical comparison

Abstract

This study compares two large hydrometeorological datasets, the Model Parameter Estimation Experiment (MOPEX) and the Catchment Attributes and Meteorology for Large-sample Studies (CAMELS), with the aim of quantifying differences that might impact their mergers. This comparison focuses on 47 shared watersheds within the continental United States spanning daily, monthly, seasonal, and annual scales for the overlapping water years of 1981 to 2000. Results indicate significant differences between the datasets at daily time steps, highlighting the challenge of high-temporal-resolution data reconciliation; however, compatibility markedly improves with temporal aggregation at monthly, seasonal, and annual scales. Systematic biases are evident, with MOPEX showing a warm bias for temperature and CAMELS displaying a wet bias for precipitation. For future studies analyzing monthly or annual runoff trends, no corrections to the raw data are necessary, as the biases do not significantly affect large-scale temporal analyses. Studies focusing on fine-scale hydrological characteristics, such as daily precipitation events, the frequency of wet and dry days per month, or single-basin dynamics, may require a statistical bias correction to ensure accuracy. Uncertainty is inherent in all climate datasets due to differences in data sources, interpolation methods, and spatial coverage. The transition from MOPEX to CAMELS does not notably introduce additional uncertainty beyond what is already present in the original datasets. The variability between the datasets is comparable to the inherent variability within each individual dataset and is neither a useful criterion for dataset selection nor a barrier to potential merger. As a result, the overall uncertainty in annual or decadal modeling outcomes remains essentially the same, regardless of which dataset is used. That said, model outputs should be calibrated against observational reference data to account for systematic errors. Statistical analyses demonstrate that both datasets are representative of climatic conditions, trends, and extreme events. Our findings validate the results of previous research employing either dataset. Furthermore, this study serves as a foundation for the merging and extension of MOPEX and CAMELS datasets without any alterations, providing a comprehensive, long-term dataset suitable for hydrological modeling and climate analyses while maintaining comparability across basin and temporal scales.