The effect of pixel heterogeneity on surface heat and water vapor flux estimated by the remote sensing-based model coupled with deep learning

•The DNN have an ability to generate LST at multiple pixel sizes.•The DNN and TSEB methods are used to construct the hybrid model.•The hybrid model can obtain accurate high-spatial-resolution H and LE.•The effect of pixel size on modeled H and LE estimates are quantified.•The deviation of modeled fluxes increases as the pixel size become coarser.