Accurate estimation of daily evapotranspiration over large areas is important both for understanding hydrological processes on the earth and for water resources management. Remote sensing observations of land surface have been used to estimate evapotranspiration (ET) over large areas, when point measurements cannot provide such information efficiently because of insufficient coverage. Conventional methods to estimate regional daily ET are based on extrapolation of instantaneous ET estimates usually from polar-orbiting satellite observations at clear sky moments and assuming clear sky conditions prevailing throughout the day. However, such methods are unable to overcome uncertainties caused by eventual cloud interference duirng the day. The geostationary meteorological satellites have frequent temporal sampling and yet higher spatial resolution, carry the promise of solving the problem of time integration to estimate daily ET. Such observations at high temporal resolution are particularly helpful in capturing the diurnal variation of land surface temperature, the most critical land surface parameter in determining the energy partitioning between sensible heat flux and latent heat flux. However, cloud-free measurements during a day may be sparse and not simultaneous for different pixels. A time series analysis technique is therefore needed to fill the gaps in sparse satellite observations due to clouds contamination. In this research, instantaneous latent heat flux in turn the evapotranspiration is firstly calculated from an multi-scaled energy balance based model using a set of land surface parameters retrieved from observations of geostationary satellite sensors coupled with atmospheric fields from regional scale Numerical Weather Prediction (NWP) models. For clear days, the daily ET is calculated as the integration of all the instantaneous ET at the overpass moments of the geostationary satellite. For partially cloudy days, a time series analysis technique using Fourier transform analysis as implemeted in Harmonic Analyze of Time Series (HANTS) is used to reconstruct gap-filled time series of instantaneous ET along a day. In the end, daily evapotranspiration is calculated from the reconstructed gap-filled time series of instantaneous estimates of evapotranspiration. To preserve consistency between the satellite pixel scale and corresponding height of atmospheric variables, variables at the atmospheric blending layer height from NWP model output are used as atmospheric forcing in the remote sensing flux model in which a multi-scale resistance scheme is developed. The above algorithms were tested both in Europe and in China (northern Tibetan Plateau) using different NWP models and geostationary satellite data. For Europe case, land surface parameters were LandSAF products retrieved from SEVIRI (Spinning Enhanced Visible and Infrared Imager) onboard Meteosat Second Generation (MSG), the NWP model is RACMO-2 (the second version of Regional Atmospheric Climate Model). For China case, land surface parameters were retrieved from China's geostationary satellite FY-2, the NWP model is GRAPES. The algorithm validation was done using data from limited number of flux towers of CarboEurope sites in Europe and flux sites on northern Tibetan Plateau by comparing the estimated energy balance flux components with tower flux measurements. Analyses of daily variation of estimated surface heat fluxes show that the proposed method is able to generate large scale net radiation, sensible, latent, and soil heat fluxes that follow closely daily variation of the courses of these flux components as captured by ground measurements. In general, reconstruction of evapotranspiration time-series using HANTS algorithm is demonstrated to be a promising technique to overcome the interference of clouds and preserve inherent trends of evapotranspiration process over a day when applied to a large-scale. HANTS reconstruction is capable to maintain daily variation of evapotranspiration on less cloudy days by keeping good correlation with the ground measurements. However, the technique has proven to be limited to areas with partially cloud cover along a day course. The impact of cloud cover percentage on the accuracy of daily ET estimate was evaluated. Comparison of the estimated daily total ET from our algorithm with daily ET calculated using conventional method (using one-time measurements of a day and assuming clear sky throughout a day) has shown that the former works better and is less affected by the intensity of cloud interference along the day.