Eddy covariance data from Xilinhaote National Climatological Observatory in Xilin Gol League during growing seasons of 2010—2013 as well as MODIS data were used to validate an ecosystem respiration model based on enhanced vegetation index (EVI), land surface water index (LSWI) and land surface temperature (LST) in a semi-arid grassland of Inner Mongolia. The limitations of this remote sensing respiration model were also discussed. The results indicate that this model can successfully simulate the variations of nocturnal ecosystem respiration (R_eco) in the growing seasons and between different years. The simulated nocturnal R_eco also agreed remarkably with the observed R_eco (R²=0.90, RMSE=0.02 mgCO₂/(m²·s)). Moreover, the observed nocturnal R_eco showed a good linear correlation with EVI_s×W_s (R²=0.63), in which EVI_s and W_s are response functions of EVI and LSWI on photosynthesis, respectively. The response of nocturnal R_eco to LST was also found following the L-T equation (R²=0.39). In addition, the difference between responses of nocturnal Reco to EVI_s×W_s and LST in the early, middle and late stages of the growing season is indicated as one principal source of the deviations of model results.