Investigating drought-induced tree mortality using remotely sensed ET data in a temperate forest in the Central US

Drought can have large impacts on forested ecosystems. Short-term drought can make forests more susceptible to insect attack, infection and wildfire, while long-term severe drought can directly cause catastrophic hydraulic disconnection that induces tree mortality. Evapotranspiration (ET) is a key parameter that links the hydrological and ecological processes and is an effective indicator of vegetation health. While ET can be measured in-situ using weighing lysimeters, sap flow, plant chambers, eddy covariance and so on, these measurements are normally carried out at a point level or over a relatively small footprint area. Mapping ET using satellite remote sensing data has been widely applied to study the spatial dynamics of water use at regional scales. Surface energy balance methods, based on remotely sensed land surface temperature retrieved from thermal infrared imaging systems, are commonly used to map ET. In this study, we used the ALEXI/DisALEXI (Atmosphere-Land Exchange Inverse model and associated flux disaggregation algorithm) energy balance modeling system, which fuses ET retrievals from multi-scale satellite sensors including Landsat and MODIS (Moderate Resolution Imaging Spectroradiometer). ALEXI/DisALEXI was applied over a drought-sensitive temperate forest in the Central US. to estimate 30 m daily ET from 2010 to 2014, including an extreme drought event that occurred in 2012. This study area contains an AmeriFlux site (US-MOz), and has a good record of tree mortality data, especially after the 2012 drought. The model results were evaluated at the flux tower site and showed good agreement with observed fluxes. The actual-to-reference ET ratio (f_RET) was also estimated and used as a vegetation stress indicator to investigate the relationship between evaporative stress and drought-induced tree mortality. We found a significant correlation (R² = 0.52) between tree mortality data in 2013 and f_RET difference between pre-drought year (2010) and drought year (2011, 2012), which suggests a high correlation between drought-induced tree mortality and pre-drought forest health condition at the US-MOz site. This study demonstrated the capability of using remotely sensed ET metrics at high spatiotemporal resolution for monitoring forest water use and drought-induced tree mortality.