Exploring sensitivities of latent heat parameterizations using a coupled, integrated hydrologic model

Abstract

Evapotranspiration and its energy counterpart latent heat flux are critical components of the terrestrial water and energy balances. Precipitation, runoff, condensation and groundwater flow are also significant members of the hydrologic cycle, but it is evapotranspiration that links water in the subsurface to water on the land surface or in the atmosphere. Hydrologic models use parameterizations to represent the physical and physiological processes of evaporation and transpiration, respectively. These parameterizations assume some mathematical structure and require that values of certain parameters be assigned; however, questions remain regarding how changes in each of these two aspects translate to output quantities of interest. Quantifying the sensitivity of model outputs to structure and input value decisions is important to understand a model’s behavior and response to changes. This dissertation explores sensitivities associated with evaporation and transpiration parameterizations within the ParFlow-Common Land Model through systematic comparisons and the application of a new sensitivity analysis method called active subspaces. Results indicate that model sensitivities vary throughout the year. For bare ground evaporation, parameterization structure affects both the magnitude and behavior of annual estimates. Including complexities associated with atmospheric stability influences evaporation behavior the most under water-limited conditions. For a vegetated surface, important input parameters used to compute the rate of transpiration vary seasonally and diurnally. Further analysis uncovers unique relationships between input and output quantities associated with evaporation and transpiration. For example, ground temperature and bare ground evaporation exhibit a parabolic relationship where similar points group together based on seasonal values of subsurface pressure, specific humidity and air temperature. The input-output relationship discovered using the active subspace method changes in time and suggests that transpiration estimates are more sensitive to changes in important parameter values during midday summer time hours.