Detection of drought‐induced blue oak mortality in the Sierra Nevada Mountains, California

Abstract

AbstractFrom 2011 to 2016, California experienced a millennial‐intensity drought, generating high levels of tree mortality. Remote sensing has been used to monitor the long‐term impacts of drought; however, discriminating dead from live trees in arid and semiarid deciduous woodlands is challenging. The goals of this study were to assess and map the spatial patterns of drought‐induced tree mortality in a blue oak (Quercus douglasii) woodland, a highly drought‐tolerant species forming savannas along the lower foothills surrounding California's Central Valley. Airborne hyperspectral imagery was used to identify the most important wavelength regions predicting drought‐induced blue oak mortality. The best metric to predict canopy stress was a normalized ratio using the spectral bands 937.53 and 1100.08 nm with a correlation with tree mortality of R2 = 0.83. The image prediction of mortality for nine field plots found that 16 of 98 trees died (17.9%) during the drought. We further evaluated tree mortality in 82 independent plots, and we found the greatest image predictive accuracy for tree mortality between 1% and 10%. When applied at the landscape level, the regression‐based index found mortality ranged from 1% to more than 51% in oak stands with an average mortality of 10% over the entire study region. In addition, tree mortality at landscape level showed higher tree mortality in blue oaks on south‐facing aspects probably because of higher insolation rates and in sites with low potential for accumulated drainage.