Parameter Sensitivity and Transferability Study Across Major US Watersheds

Abstract:

In this study, the sensitivity of Community Land Model (CLM)-simulated water and energy fluxes to hydrological parameters across 431 Model Parameter Estimation Experiment (MOPEX) basins are examined using a stochastic sampling-based sensitivity analysis approach. The basins are then classified according to their parameter sensitivity patterns, using Principal component analysis (PCA) and expectation-maximization (EM) –based clustering approach. Similarities and differences among the parameter sensitivity-based classification system and existing climate classification systems are discussed. Within each class, the same inversion modeling setups, with a unique subset of unknown parameters, can be used for parameter calibration. By reducing the parameter dimensionality to a reasonably low number, the classification makes the inverse modeling possible and less ill-posed. Surrogate models are then developed and validated as computationally efficient alternative to the CLM numerical simulators. A set of experiments of surrogate-based model calibration, including deterministic optimization and Markov chain Monte Carlo (MCMC)–Bayesian approaches, were conducted to evaluate the transferability of parameter values within and between the watershed classes.