This paper applies statistical methods to interpolate missing values in a dataset of radiative energy fluxes at the surface of Earth. We apply Random Forest (RF) and seven other conventional spatial interpolation models to a global Surface Solar Radiation (SSR) dataset. We apply three categories of predictors; climatic, spatial, and time series variables. Although the first category is the most common in research, our study shows that it is actually the last two categories that are best suited to predict the response. In fact, the best neighboring variable is almost 40 times better than the best climatic variable in predicting SSR. Furthermore, our analysis shows that the Mean Absolute Error is 10.2 on average using RF, with a standard deviation of 1.5. Conventional methods have an average MAE of 21.3, with an average standard deviation of 6.4. This highlights the benefits of using machine learning in environmental research.