Comparison of Root Water Uptake Depth Derived from Models and Observations: A Focus on the Amazon

Water uptake by deep roots is an important seasonal resilience mechanism in the Amazon, allowing vegetation to avoid water stress when dry season precipitation is scarce. Critically, most existing land-atmosphere modeling frameworks do not include the necessary structure or physical processes to represent deep roots, including deep hydrologically active soils, fine-scale topographic variations, and dynamic spatiotemporal groundwater variations. We address these shortcomings by implementing deep soils and a dynamic root water uptake scheme (known as DynaRoot) in the Noah-MultiParameterization (Noah-MP) land surface model, version 4.5. These modifications allow us to obtain a modeled estimate of root water uptake depth at each model grid point based on the local soil water profile, something that was not possible in previous versions of Noah-MP. We completed 20-year simulations of modified Noah-MP for a subregion of the Amazon. Validation of modeled root water uptake depth is a challenge due to the lack of available root depth observations, which are sparse in the Amazon. In this work, we describe a method for estimating root depth during drydown periods using remotely sensed data, including vegetation optical depth (VOD), evapotranspiration (ET), precipitation, and enhanced vegetation index (EVI). This method includes a simple model that relates rooting depth, ET, VOD, and soil moisture at field capacity. We compare output from this simple model with that from Noah-MP for the Amazon subdomain, and analyze observations-based rooting depth for the Amazon overall. In addition to validation of the Noah-MP results, our observations-based method provides a measure of rooting depth that can be used when in-situ observations are not available. Validation of the modified Noah-MP results can contribute to further model improvements. Future work will couple our modified version of Noah-MP with the Weather Research and Forecasting (WRF) model.