Increased resolution, complexity, and accuracy of Earth System Models (ESMs) has led to direct use of model projections to inform regional climate change impact assessments, including studies on changes in hydrology, water resources, and water security at sub-continental scales. However, ESMs often exhibit substantial regional biases in hydroclimate mean states, fluxes and – importantly – sensitivities. Biases in land model sensitivity project onto hydroclimate change signals under warming, undermining their accuracy and adding uncertainty to the already broad spread in climate change projections from model experiments.
It is therefore necessary to understand and reduce not just absolute biases, but also sensitivity biases, in order to increase the credibility and robustness of climate change impact assessments based on ESMs. Recent research on runoff sensitivity – the change in runoff as a function of changes in precipitation and temperature – showed a large potential for uncertainty reduction in model projections through the use of observational constraints. However, sensitivity biases remain inadequately measured and their causes poorly understood, thereby impeding model improvement, because process-oriented diagnostics (PODs) that target hydroclimate sensitivities and important features of their climatology, rather than simply mean states and fluxes, are missing from major diagnostics packages.
We propose to develop and refine a suite of PODs characterizing the fidelity of the hydrologic components – primarily runoff – of land surface models, to be applicable to models in the Coupled Model Intercomparison Project 5 and 6 (CMIP5/6). For the development of the PODs we will leverage several observational products and a range of existing and forthcoming simulations with the Community Earth System Model 2 (CESM2) and Community Land Model 5 (CLM5), including a Perturbed Parameter Ensemble, as well as Land Surface, Snow and Soil moisture MIP (LS3MIP) simulations as part of CMIP6. We will conduct additional sensitivity experiments with CLM5 to reveal model parameters relevant for runoff sensitivity and to refine the PODs.
The outcomes of this research will be a suite of PODs that provide process-based understanding of the origin of hydrologic sensitivity biases in ESMs, which will uncover opportunities for targeted model improvement. We will implement the new PODs in the NOAA Model Diagnostics Task Force (MDTF) Diagnostics Package and, with potential additional support from the Department of Energy, in the International Land Model Benchmarking (ILAMB) package (see Section 4.6 for more details).