SBIR-STTR Award

Soil moisture is an important parameter for AF weather measurements and forecasting. It impacts Army operations (off-road mobility, land operations) and Intelligence Communitys knowledge (agriculture: social unrest).AF has classified soil moisture measurements and data as a DoD space-based environmental monitoring gap. If unmitigated, the lack of soil moisture data will diminish our asymmetric advantage over adversaries. The Weather Company (TWC) ingests and fuses more than 100 terabytes of third-party data daily from more than 800 different data sources like pollen, radar, satellite imagery, traffic, personal weather stations, and agricultural equipment (tractors, sprayers, and soil sensors). NextGen Federal Systems is proposing to leverage TWCs services including their soil moisture data products, radar, satellite imagery, numerical models, and data fusion techniques combined with IBMs machine-learning services to construct a global soil moisture data product tailored to meet DoD requirements and mission utility. The effort will establish a supervised machine-learning methodology for imagery classification trained on ground-truth soil moisture measurements from in situ, remote sensed, and model data. Big Data, Data Fusion, machine-learning, Forecasting algorithms, Soil moisture, weather

The proposed Soil Moisture Advancements from Research to Transition (SMART) solution will adapt IBM's Physical Analytics Integrated Repository and Services (PAIRS) Geoscope and the Watson Agriculture Package commercial technologies and use them as the basis for a novel Machine-learning (ML) system. SMART will establish a containerized service to build and deliver machine-learned soil moisture datasets and implement a data science workflow to accelerate R&D of new ML-based weather products and tools. The objectives of this effort are to (1) Establish a Weather Data Science Lab (WxDSL) that will provide a domain for researchers to share data, algorithms, and results; (2) Provide a sustainable soil moisture product that is at least as good as SMAP Level 4 data products; (3) Provide trafficability relationships to soil moisture products using the WxDSL to work out new approaches to deriving trafficability; (4) Enable Independent validation and verification (V&V) and provide results.