A remote sensing and modeling integrated approach for constructing continuous time series of daily actual evapotranspiration

Satellite remote sensing-based surface energy balance (SEB) techniques have emerged as useful tools for quantifying spatialized actual evapotranspiration at various temporal and spatial scales. However, discontinuous data acquisitions and/or gaps in image acquisition due to cloud cover can limit the applicability of satellite remote sensing (RS) in agriculture water management where continuous time series of daily crop actual evapotranspiration (ETc act) are more valued. The aim of the research is to construct continuous time series of daily ETc act starting from temporal estimates of actual evapotranspiration obtained by SEB modelling (ETa eb) on Landsat-TM images. SEBAL model was integrated with the FAO 56 evaporation model, RS-retrieved vegetative biomass dynamics (by NDVI) and on-field measurements of soil moisture and potential evapotranspiration. The procedure was validated by an eddy covariance tower on a vineyard with partial soil coverage in the south of Sardinia Island, Italy. The integrated modeling approach showed a good reproduction of the time series dynamics of observed ETc act (R2 =0.71, MAE=0.54 mm d-1, RMSE=0.73 mm d-1). A daily and a cumulative monthly temporal analysis showed the importance of integrating parameters that capture changes in the soil-plant-atmosphere (SPA) continuum between Landsat acquisitions. The comparison with daily ETc act obtained by the referenced ET fraction (ETrF) method that considers only weather variability (by ETo) confirmed the lead of the proposed procedure in the spring/early summer periods when vegetation biomass changes and soil water evaporation have a significant weight in the ET process. The applied modelling approach was also robust in constructing the missing ETc act data under scenarios of limited cloud-free Landsat acquisitions. The presented integrated approach has a great potential for the near real time monitoring and scheduling of irrigation practices. Further testing of this approach with diverse dataset and the integration with the soil water modeling is to be analyzed in future work.