Interpreting sensor data requires knowledge of sensor placement and the contextual environment surrounding the sensor. For a single sensor measurement, it is easy to document the context usually by visual observation. However, for millions of sensors reporting data back to a server, the contextual information needs to be automatically extracted from either data analysis or leveraging complimentary data sources. Data layers that overlap spatially or temporally with sensor locations, can be used to extract the context and validate the measurement. The second challenge is to minimize the amount of sensor data transmitted through the internet while preserving signal information content. Here we demonstrate two methods for communication bandwidth reduction: Computation at the edge and compressed sensing. We validate the above methods on wind and chemical sensor data to: (1) eliminate redundant measurement from wind sensors and (2) extract peak value of a chemical sensor measuring a methane plume. We present a general cloud based framework to validate sensor data based on statistical and physical modeling and contextual data extracted from geospatial data.