Are canonical relations to partition evapotranspiration valid for managed systems?

Partitioning Evapotranspiration (ET) flux into its primary components; direct evaporation (E) and plant transpiration (T) is important to understand biophysical feedbacks of the ecosystem. E and T respond to the environmental variables with very distinct processes. In general, E is controlled by available energy and surface soil water availability and T is additionally modulated by plant physiological processes. As traditional methods for measuring E or T (e.g., lysimeters, sap flow measurements, eddy covariance, etc.) cannot provide information on the ratio of T and ET (called T/ET, hereafter) over large scales, studies have developed simplistic canonical relations between T/ET and vegetation indices (e.g., Leaf Area Index (LAI), Enhanced Vegetation Index (EVI), etc.) to upscale point estimates to large scales. This study assesses the validity of these relations in managed systems. To this end, using Flux Variance Similarity (FVS) theory, we partitioned total ET fluxes at three wheat-based cropping systems under different grazing management, viz. Grain-only, Graze-grain, and Graze-out. Partitioned water fluxes were then used to develop relations between T/ET and Landsat-derived EVI for each field. We found a strong T/ET-EVI relationship along the lines of previous reported studies but only in Grain-only case (unmanaged system). T/ET-EVI relation was not reliable in managed systems (both Graze-grain and Graze-out). Furthermore, we show that using canonical relations in managed system produces relatively larger errors in estimation of T/ET at weekly to monthly time scales. These results highlight the need for caution in usage of canonical relations for T/ET partitioning in managed systems.