Using Weighing Lysimeters to Link Vegetation Indices with Evapotranspiration Measurements

It is estimated that ~60% of all precipitation over land returns to the atmosphere through evapotranspiration (ET). Methods to estimate ET over large areas, such as at watershed scale, rely on land-surface models and increasingly the use of remote sensing data. Within modelling frameworks that use remote sensing data, ET estimation has focused on the use of vegetation indices (VIs) to represent the variability of vegetation and their ET response to changes within local water and energy budgets. The purpose of this study was to explore the relationship between common VIs (such as the normalized difference vegetation index [NDVI], leaf area index [LAI], and the soil adjusted vegetation index [SAVI]) derived from a hyperspectral sensor and ET over a growing season. Further we explore, using a Random Forest (RF) approach, the vegetation response at numerous spectral regions to ET. Using a network of large weighing lysimeters located at the University of Guelphs Elora Research Station we measured and recorded ET and VIs over multiple summers. Most tested VIs had an insignificant correlation to ET measurements, including LAI which was found to have a weak relationship with the ET measurements (R2 = 0.280). However, NDVI and SAVI (soil adjusted vegetation index) were found to have strong (R2 = 0.712 and 0.852 respectively), significant relationships to the measured ET amounts, proving to be a better representation than several commonly used VIs. Results from the RF model demonstrated several spectral regions sensitive to ET amounts, however when isolated their responses did not match the relationships observed using SAVI.