Gaining new insight into the spatiotemporal variability of soil moisture is vital for precision agriculture and efficient irrigation scheduling. Several satellites measure earth surface features that are related to soil moisture. However, the lack of field-measured data hinders successful monitoring of soil moisture dynamics in agricultural fields. Publicly available datasets from soil moisture monitoring networks have very few locations representing irrigated agricultural fields. Conversely, private soil moisture networks located in agricultural fields may be used to calibrate machine learning models for remote sensing of soil moisture. This study presents a random forest (RF) empirical surface soil moisture model using high-resolution remote sensing data (SMAP, Sentinel-1, Sentinel-2) and other land surface parameters such as soil texture, terrain, etc. Generally, microwave backscatter (from Sentinel-1) is expected to be influenced by soil moisture over bare surfaces; however, as the vegetation biomass increases, backscatter is expected to be more sensitive to vegetation than surface characteristics. This impedes the soil moisture retrieval because of the scattering and attenuation effects of vegetation on the backscatter. Therefore, empirical methods that relate the remote sensing data to the volumetric soil moisture content are necessary. A combined dataset with soil moisture data available from 275 public monitoring stations and ~10,000 agricultural fields was used in this study. RF model was trained using remote sensing and ground measurements from 2019 to 2021, available for 56 different crop types, using 5-fold testing. Preliminary results indicate a great overall performance of this model with RMSE = 0.043 cm3/cm3 and R2= 0.79. This research will additionally discuss the need for exploring other machine learning algorithms, and considerations about field-scale applicability of remote sensing of soil moisture.