Impact of Vegetation Assimilation on Flash Drought Characteristics Across the Continental United States

Abstract

Predicting and managing the impacts of flash droughts is difficult owing to their rapid onset and intensification. Flash drought monitoring often relies on assessing changes in root-zone soil moisture. However, the lack of widespread soil moisture measurements means that flash drought assessments often use process-based model data like that from the North American Land Data Assimilation System (NLDAS). Such reliance opens flash drought assessment to model biases, particularly from vegetation processes. Here we examine the influence of vegetation on NLDAS-simulated flash drought characteristics by comparing two experiments covering 1981-2017: open loop, (OL) which uses NLDAS surface meteorological forcing to drive a land-surface model using prognostic vegetation, and data assimilation (DA), which instead assimilates near-real-time satellite-derived leaf area index (LAI) into the land-surface model. The OL simulation consistently underestimates LAI across the U.S., causing relatively high soil moisture values. Both experiments produce similar geographic patterns of flash droughts, but OL produces shorter duration events and regional trends in flash drought occurrence that are sometimes opposite to those in DA. Across the Midwest and Southern U.S., flash droughts are four weeks (about 70%) longer on average in DA than OL. Moreover, across much of the Great Plains, flash drought occurrence has trended upward according to the DA experiment, opposite to the trend in OL. This sensitivity of flash drought to the representation of vegetation suggests that representing plants with greater fidelity could aid in monitoring flash droughts and improve the prediction of flash drought transitions to more persistent and damaging long-term droughts.