With an increasing global population and continuing climate change, food security has become a grand scientific and societal challenge. To tackle this challenge, it is critically important to obtain timely crop information such as yield potential and growing status as crop information is often time sensitive. With rapidly advancing remote sensing technologies including, for example, satellite- and UAV-based approaches to large-scale, fine-resolution, and continuous observation in visible bands, NIR, thermal, microwave, etc., timely crop knowledge discovery based on massive remote sensing data provides a promising means to tackle the food security challenge. Furthermore, to integrate remote sensing data of crop with related environmental data (e.g., temperature, precipitation, and radiation) can help understand crop changes in various environmental conditions.
How to harness such rich data sources to achieve timely crop knowledge discovery based on advanced computing and geospatial approaches such as deep learning and cyberGIS for multiple agricultural applications is the primary focus of this dissertation research. Specifically, several interrelated studies have been conducted to achieve high-performance and in-season crop type classification at both county and state scales in Illinois, USA; integrate climate and satellite data for wheat yield prediction in Australia; and detect in-season crop nitrogen stress using UAV- and CubeSat-based multispectral sensing at the field-level. These studies are enabled by cutting-edge machine learning methods (e.g., deep neural networks) and advanced cyberGIS capabilities (e.g. ROGER supercomputer). Collectively, findings and insights from the studies promise to transform data-intensive crop knowledge discovery.