507 Why Do Land Surface Models Produce a Lower Ratio of Transpiration to Evapotranspiration?

Tuesday, 9 January 2018
Exhibit Hall 3 (ACC) (Austin, Texas)
Liling Chang, Univ. of Arizona, Tucson, AZ

An estimation of the transpiration/evapotranspiration ratio (T/ET) is essential for understanding how plants use water, affect the climate, and interact with water and carbon cycles. Generally, Land Surface Models (LSMs) produce lower values of global terrestrial T/ET than estimates through field isotopic analysis by about 20%. We assume that LSM parameterization schemes and subgrid topographic effects on water transport and solar energy may result in the underestimated T/ET ratio. To test our hypothesis, we initially conducted two experiments with a 3-dimensional, physically-based ecohydrological model in the Marshall Gulch, subhumid mountainous catchment in Arizona: 1) an experiment (CTRL) with an improved, more process-based scheme of soil surface resistance, topographic shading effects, dynamic vegetation, and local slopes derived from the Digital Elevation Model (DEM) and 2) an LSM-like experiment using an empirical soil surface resistance scheme, excluding topographic shading effects and dynamic vegetation, and reducing local slopes to 1/10 of the DEM. The calculated T/ET ratios for the CTRL and LSM-like experiments are 72% and 55%, respectively. To discern how individual parameterization and topography factors contribute to the difference in the T/ET ratio, we further evaluated the influence of each factor on the ratio. After comparison, we concluded that the elevated T/ET ratio is caused primarily by the improved soil surface resistance scheme and lateral subsurface flow driven by topography. In particular, local slopes in the Marshall Gulch catchment is vital for redistributing precipitation, generating lateral subsurface flow, replenishing water to the convergence zone, and enhancing partitioning of evapotranspiration to transpiration.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner