Global water classification data layers such as the European Joint Research Centre's Monthly Water History dataset allow for accurate and large scale analysis of trends in the extent of open water on Earth. Yet, on the local scale, there is opportunity to increase the accuracy and temporal density of such datasets. In this study, we have shown that a machine learning based technique can improve the sensitivity of water classification over small to medium sized water bodies. Our pipeline allows for complete land cover classification at every acquisition date for a region of interest after spatiotemporal imputation fills in gaps from clouds and other conditions in LANDSAT satellite imagery. An implementation using R and the Google Earth Engine computing platform is presented in this paper which collects, imputes, and classifies imagery along with identifying outliers and generating timeseries or spatial visualizations of local regions of interest for water-focused land cover studies.