Earth System Modeling Tested for CLM4.5 in a Costa Rican Tropical Montane Rainforest

Terrestrial ecosystems in the tropics are important for global carbon and water cycling, which makes modeling of their land-surface processes essential for accurate understanding of land-atmosphere interactions. However, modeling of tropical regions, especially mountainous ones, is known to be subject to significant errors in the prediction of evapotranspiration. Our previous work has highlighted the effects of the prolonged wetness experienced by such sites, focusing on carbon and water exchange at the leaf/stand level. Here, we explore the implications these findings have for modeling at the stand/canopy scale. This study examined the performance of the Community Land Model (CLM4.5) against measurements from a tropical montane rainforest in Costa Rica. The study site receives over 4,000 mm of mean annual precipitation. Measurements include leaf temperatures, transpiration (sap flows), fluxes via eddy-covariance, and vertical profiles of H2O and CO2 concentrations, micrometeorological variables, and leaf wetness. In this work, results from point-scale CLM4.5 were compared to canopy data. The model fails to capture the effects of frequent rainfall events and mountainous topography on the variables of interest (temperatures, leaf wetness, and fluxes). We found that soil and leaf temperatures were overestimated (≈ +2°C) at noon and underestimated (≈ -1°C) during the night; daily transpiration was approximately double than that observed. Simulated leaf wetness deviated significantly from the measurements, both in timing and extent, which affected temperatures and evapotranspiration partitioning. Slope effects appeared in the average diurnal variations of surface albedo and carbon flux from actual data but were not captured in CLM. Our investigation indicated that interception and aerodynamic resistance models contribute to model errors, suggesting potential improvements for modeling in very wet and/or mountainous regions.