We propose to demonstrate a fully automatic capability for generating a global, high resolution (30 m) land classification map, with continuous updates from satellite imagery and crowdsourced ground truth, and an annual publication cycle. The resulting map will be fully extensible, supporting many layers including Anderson type land cover categories, specific crop types in agricultural areas, and even continuous variables such as crop yield estimates. Our methodology involves fusing imagery with existing raster and vector ground truth data sets, applying a novel classification technique to enable vastly improved spatial/temporal model portability, and continuous updating as additional imagery and crowdsourced ground truth data becomes available. By employing a systematic crowdsourcing strategy that is responsive to local and regional market incentives, the map's accuracy will improve most over time in those geographic regions where higher accuracy is most needed. The resulting map product will find immediate use for a multitude of applications in agriculture, commodities trading, ecological forecasting, wildlife conservation and climate change studies. We foresee monetizing the continuous production effort by obtaining stakeholder funding to make a standard product freely available to the global public, but we also intend to develop higher precision, commercial products for lucrative applications such as precision agriculture in targeted locations worldwide.