Quantifying the uncertainties of model physical processes in affecting precipitation, surface fluxes and land-air coupling strength over the Amazon region

Description

This study aims to quantify the relative importance and uncertainties of different physical processes in climate models, including cloud microphysics, convection, boundary layer and surface layer, and land surface processes, in affecting precipitation, surface fluxes and land-air coupling strength over the Amazon region. We performed 120 ensemble simulations with the WRF model using a stratified experimental design including 6 cloud microphysics, 3 convection, 6 PBL and surface layer, and 3 land surface schemes. A multiple-way analysis of variance (ANOVA) approach is used to quantitatively analyze the differences among group (scheme) means and their associated variations among and between groups (schemes). We used two-legged coupling metrics, which included both terrestrial (soil state to surface fluxes) and atmospheric (surface fluxes to atmospheric states or precipitation) legs, to diagnose the land-air interaction and coupling strength. ARM measurements collected from the GoAmazon field campaign together with satellite and reanalysis data are used to evaluate model simulated precipitation, surface heat flux and land-air coupling strength. Results show that most of the simulations capture the magnitude and diurnal cycle of surface air temperature well but overestimate the daytime surface sensible and latent heat fluxes. The model captures well the diurnal cycles of PBL height and surface humidity, but most simulations underestimate their magnitudes. Among 120 ensemble simulations we identified an optimal model configuration and setting parameters that generates the overall best model performance against observations. Our sensitivity analysis based on the quantitative ANOVA approach suggests that relative contributions of different physical processes to the total variance vary with the objective variables and locations we analyzed. For example, the land surface process contributes to most of the variance for sensible heat flux and surface air temperature. Land surface, boundary-layer processes and convection are equally important for PBL height. The contributions to total variance from interactions between any two processes are relatively small comparing to that from the dominant individual process. Results of this uncertainty quantification study can help us better understand the roles of different physical processes in land-air interaction, quantify the model uncertainties from various sources such as physical processes, parameters and structural errors, and provide insights for improving the model physics parameterizations.