Abstract: Abstract: Land surface temperature (LST) is a key parameter in investigating environmental, ecological processes and climate change at various scales, and is also valuable in the studies of evapotranspiration, soil moisture conditions, surface energy balance, and urban heat islands. However, it is difficult to acquire satellite LSTs with both high spatial and temporal resolutions due to tradeoffs between these parameters. The Enhanced Spatial and Temporal Adaptive Reflectance Fusion Model (ESTARFM) algorithm was initially designed to predict surface reflectance and is based on the assumption that MODIS and Landsat surface reflectance are highly consistent over homogeneous areas. However, the ESTARFM method prediction results degrade somewhat when the method is used for heterogeneous fine-grained landscapes. This research extended the ESTARFM model from reflectivity range to thermal infrared for estimation of daily temperature at 90 m resolution combined MODIS and ASTER. The implementation of ESTARFM requires input of the search window size, selection of spectrally similar pixels, determination of the weight of similar pixels, and computation of the correction coefficient and temporal weight. The calculation of weights for spectrally similar pixels involves weighing the contribution of neighboring pixels to the computation of a central pixel. Using a local moving window, neighboring spectrally similar pixels were included for the computation of the LST corresponding to a central pixel with the temporal weights of the two dates. In this study, we used multiple bands, i.e., red, NIR, and LST bands, as the input variables and generated high spatial-temporal resolution land surface temperature, combining temporal change information from multi-temporal MODIS with high-resolution spatial resolution from ASTER. The objective of this paper was to evaluate the ESTARFM method using ground measurements coordination with ASTER LST products collected in an arid region of Northwest China during the first thematic Multi-Scale Observation Experiment on Evapotranspiration (MUSOEXE) over heterogeneous land surfaces in 2012, as part of the Heihe Watershed Allied Telemetry Experimental Research (HiWATER) .We didn't modified the model even if the linear hypothesis was directly applied in LST prediction, which may result in uncertainty and errors. The remote sensing data were acquired with from Jun to September in 2012. The results showed that ESTARFM was positively and linearly related with the actual measured TIR. The correlation coefficient values were all found to be greater than 0.71. The mean absolute error and root mean square error were all below 2.00 K and 2.60 K, respectively. From the feature of scattering plots between the predicted and observed LST, the data points fell close to the diagonal line in each panel, indicated that the predictions were all in good agreement with the observations. Overall, the values of mean absolute error and root mean square error between the predicted and the observed LST were quiet small; whereas the correlation coefficient values between the predicted LSTs and ASTER LST products were all found to be greater than 0.95.It should be noted that some pixels in the scatter plots showed differences between the predicted and observed LSTs. These discrepancies revealed a major limitation to the method; i.e., ESTARFM does not capture land cover that has been altered between two imaging dates. Thus, changes in land cover or other surface conditions can lead to prediction errors. In addition, the fusion results showed that value of correlation coefficient was better in non-vegetation area than vegetation and water area, and up to 0.91 especially in the August 27, 2012. However, the application of the ESTARFM and its variants to LST prediction is immature in terms of methodology. Many critical issues have not been solved, especially with respect to the determination of the search window size, conversion coefficient improvement, and thermal landscape heterogeneity.