Understanding the impact of vegetation dynamics on the water cycle in the Noah-MP model

The impact of extreme climate events, especially prolonged drought, on ecosystem response, can influence the land-atmosphere interactions and modify local and regional weather and climate. To investigate the impact of vegetation dynamics on the simulation of energy, water, and carbon exchange at the land surface and streamflow, especially during drought conditions, we compared the performance of multiple versions of the Noah- multiparameterization (MP) land surface model (both Noah-MP LSM, version 3.6 and 4.0.1) with default configurations as well as various vegetation physics options, including the dynamic or input leaf area index (LAI) and the fractional vegetated area (FVEG). At the site level, simulated water and energy fluxes from each version were compared to eddy covariance (EC) flux tower measurements and remote sensing data from Moderate-Resolution Imaging Spectroradiometer (MODIS) at well-characterized natural grassland sites in Kansas from 2008 to 2018. The ability of each version to reproduce annual mean river flows was compared to gauged observations at United States Geological Survey (USGS) stations over 11 years (2008-2018). Model performance in replicating spatial patterns during extreme events was assessed by comparing simulated soil moisture (SM) percentiles over the state of Kansas to the U.S. Drought Monitor (USDM). Results from these comparisons indicate that (a) even though there were differences in the latent heat (LE) components (i.e., transpiration, canopy evaporation, and soil evaporation), the total LE is mostly insensitive to variations in LAI across all model versions. This indicates that the incoming net radiation limits the total evaporation, as the presence of adequate soil moisture allows for higher soil evaporation when LAI limits transpiration; (b) regardless of the model version, the force of the precipitation largely dictates the accuracy of evapotranspiration (ET) simulation; (c) Overestimation of LE resulted in underestimation of streamflow, particularly over the land surface type dominated by a combination of grasses and cropland in the western and central part of the state; (d) all of the tested Noah-MP 4.0.1 vegetation physics produced spatial patterns of drought that more closely matched the USDM as compared to version 3.6. These findings have important relevance for applications of large-scale ecosystem-atmosphere feedbacks in water, carbon, and energy exchange.