Evaluation of Multimodel and Multiscale NLDAS-2 Evapotranspiration Using Different Observations

Currently, NLDAS is an operational system that supports U.S. operational drought monitoring and seasonal hydrological prediction, in particular for the National Integrated Drought Information System, the U.S. Drought Monitor (USDM), and the NCEP Climate Prediction Center monthly drought briefings. Detailed information about NLDAS can be found at NOAA (http://www.emc.ncep.noaa.gov/mmb/nldas) and NASA (http://ldas.gsfc.nasa.gov/nldas/) websites. NLDAS consists of two phases: NLDAS-1 established the NLDAS infrastructure (2000-2005) and tested performance of the models; NLDAS-2 upgraded models and surface forcings to generate long-term NLDAS products, evaluated the quality and reliability of these long-term products using all available in situ observations and satellite-retrieved data, and applied these products in U.S drought analysis and monitoring (2006-present). In NLDAS-2, most products such as streamflow/total runoff, sensible and latent heat fluxes, ground heat flux, soil moisture, soil temperature, and land surface skin temperature have been comprehensively evaluated using either in-situ observations or remotely-sensed data. These evaluations cover different spatial scales from basin to continental and time scales from hourly to yearly. However, despite being one of the most important NLDAS-2 products, evapotranspiration (ET) has not yet been comprehensively evaluated.

This presentation summarizes the recent evaluation results for multi-model and multi-scale NLDAS ET. The evaluation uses the tools developed in the NLDAS test bed and various observational data sources such as hourly and daily flux tower measurements, 0.125-degree monthly MODIS ET retrieval, global 0.5-degree monthly FLUXNET ET, and mean annual basin budget-derived ET products (the difference between observed precipitation and streamflow). Based on this evaluation, we analyze some possible reasons for ET simulation errors. We also show preliminary results from recent efforts to further improve the performance of models and suggest some possible directions to improve NLDAS-2 land surface models in the future.