Publications

Cui, GT; Guo, WC; Goulden, M; Bales, R (2024). MODIS-based modeling of evapotranspiration from woody vegetation supported by root-zone water storage. REMOTE SENSING OF ENVIRONMENT, 303, 114000.

Abstract
As an important process regulating the water cycle, forest evapotranspiration (ET) also reflects vegetation and moisture conditions in response to a changing environment. For mountains in a Mediterranean climate and precipitation largely in winter, inconsistent seasonal variations between water stress and forest growth complicate remote-sensing-based ET modeling. Focusing on the Mediterranean-climate forests of California's Sierra Nevada, this regional study presents an NDWI-CWS (Normalized Difference Water Index-Canopy Water Stress) model to estimate daily ET in a globally important, representative region using MODIS NDWI and NDVI (normalized difference vegetation index) data, plus ground-based meteorological data. Specifically, to account for the important yet less-studied role of root-zone water storage in supporting the growth of deep-rooted woody vegetation (i.e. forest tree and shrub) during dry seasons and droughts, we proposed to use water availability based on NDWI in modeling woody ET. In contrast, given the predominant role of surface soil moisture for nonwoody vegetation (e.g. grass) with shallow roots, the model calculated non-woody ET using water availability based on the ratio of cumulative precipitation to potential ET (denoted as P/PET). With ET measurements from 18 flux towers and long-term water-balance measurements at 58 catchments in the relatively data-rich California study area, we compared ET estimates from four modeling experiments with different water availability combinations, and six global and regional ET products. Overall, our NDWI-CWS ET estimates generally agreed best with field measurements, with R2 of 0.74 for point-scale woody-ET comparison and R2 of 0.42 for catchmentscale comparison. The mean annual woody ET (634 mm) during water year 2003-2020 was larger than that (353 mm) in non-woody areas. Further, with NDWI-CWS ET estimates and snow data, we estimated that the mean root-zone water storage capacity (601 mm) in forests is double that (285 mm) in non-woody (e.g. grassland) vegetated areas in the Sierra Nevada. We found that remote-sensing NDWI-based water availability is highly correlated with deep root-zone water storage, and much so (correlation R of 0.74) during dry summers. However, P/PET-based water availability reflects surface soil moisture (R of 0.66) modulated by short-term precipitation. Together, explicitly accounting for the distinct roles of root-zone water storage and soil moisture for woody versus non-woody vegetated areas improves MODIS-based ET estimation, which is critical for regional water resources and forest management, in addressing water cycle in a warming climate.

DOI:
10.1016/j.rse.2024.114000

ISSN:
1879-0704