Noah-MP Parameter Optimization at Southern Great Plains Using Bayesian Optimization

Abstract

Understanding the key mechanisms that govern land-surface processes is crucial for accurately characterizing the energy, mass, and momentum exchanges between the atmospheric boundary layer, the land surface, and the subsurface across various atmosphere-soil-vegetation systems. To enhance our understanding of the mechanisms of land surface-atmosphere interactions and reduce the mismatches between Noah-Multiparameterization Land Surface (Noah-MP) simulations and observational data, we optimized the seven most sensitive parameters in the Noah-MP for a 9-day period (2–10 April 2016) at the Southern Great Plains site using Bayesian Optimization (BO). The implementation of Noah-MP in the single-column Weather Research and Forecasting (WRF) model and the application of BO allow for site-level parameter estimation. The Noah-MP shows an improvement in the simulation of sensible heat flux and latent heat flux when we use optimized parameters instead of the default parameters according to data from flux tower observations. The optimized parameter values indicate that the top layer (0–20 cm below the ground surface) of soil at the Southern Great Plains site may contain a higher sand content than indicated by the silty-clay-loam soil classification provided by AmeriFlux agricultural records. These optimized parameters also improve the simulation of atmospheric state variables, especially the 2-m specific humidity, and result in improved top-layer soil moisture and temperature simulations. The optimized parameters not only improve the simulation during the selected 9 days in April 2016 but also lead to better simulation results throughout the entire alfalfa growing season in 2016 (April and May) at the Southern Great Plains site.