Soil moisture is an important parameter in various applications such as climatology, meteorology, hydrology, and water resource management, so it is quintessential to have a product with high spatial resolution. Due to the fact that soil moisture data with high spatial resolution is not currently available, one of the goals of this article is to downscale the existing soil moisture products and improve their spatial resolution to 1 square kilometer. For this purpose, two methods have been used based on regression and neural network along with other available satellite data and products including different combinations of land surface temperature (LST), normalized difference vegetation index (NDVI), passive microwave sensor data including brightness temperature in different polarizations (TBH and TBV), digital elevation model (DEM) and short-wavelength infrared (SWIR) data from MODIS to downscale the 3 km SMAP satellite soil moisture products. The innovation of this study includes investigating the effect of window size on the accuracy of downscaling, the effect of interpolation type and the use of Sentinel-3 satellite. The evaluation results have shown that the soil moisture of Fars and Golestan provinces, respectively, have a correlation coefficient (R) of 0.82 to 0.93 and 0.72 to 0.77, the mean absolute percent error (MAPE) in both regression and neural network methods less than 21 to 30 and 42 to 46 percent, and the lowest root mean square error (RMSE) equal to 0.0448 and 0.0445 in the neural network method. Also, in the area of Fars province, the regression modeling results of the plain area are more satisfactory than those of the mountain area.