CONICET | Buscador de Institutos y Recursos Humanos

The ability to monitor latent heat flux (LE) is relevant for applications requiring spatially-resolved estimates of moisture availability over large areas. Recently, a number of studies have focused on estimating surface energy fluxes via assimilation of land surface temperature (LST) observations into variational data assimilation (VDA) schemes. In this study, the performance of the combined-source variational data assimilation (CS-VDA) framework is assessed in detail using surface multitemporal heat fluxes collected at a Eucalypt forest savanna of Northern Australia. The CS VDA model treats the soil and vegetation as one medium. We extended previous studies, to a semi-arid ecosystem and included 1and 3 hour global meteorological forcing data (GMD) rather than site-specific observations to drive CS VDA model. To implement the VDA model we relied on hourly water fluxes and meteorological measurements from an eddy covariance (EC) site located at our Australian study site. Using 1 and 3-hour average in-situ measurements, the mean difference between estimated and observed LE was ∼30%, which agreed with errors reported in the literature. We found that at least an 3 hour average meteorological data was required as input to the CS-VDA model, so 1-3-hourly temporal resolution MERRA-GMA and GLDAS, respectively, were used. We found that replacing local meteorological data with GMD reduced the performance of the LE estimation in comparison to in-situ measurements (GLDAS: RMSEdaily=98.52 W/m2, GMA: RMSEdaily=82.02 W/m2). Despite this, the model LE RMSE values at 8-day temporal scale (GMA: RMSE8-days=32.20 W/m2) was similar of those reported at this area by others models. This study provides the basis to produce an operational daily LE product using GMD dataset and LST remote sensing data within the CS-VDA algorithm.