GEO-Amazon Earth Observation Cloud Credits Awarded to NASA Harvest Partners

In an effort led by NASA Harvest in partnership with the Buenos Aires Grain Exchange, the Kenya Ministry of Agriculture, and Lutheran World Relief, the GEO-Amazon Earth Observation Cloud Credits Program has awarded grant funding to a new project for mapping national-scale agricultural practices. Satellite Earth observations (EO) provide frequent and timely data that are vital to monitoring complex and highly-variable agricultural systems. This large volume of data can give insight to crop type, crop condition, and potential crop yield for a multitude of regions that have diverse socio-economic levels, cultural practices, and climates. The key to enacting successful agricultural policy is putting this valuable information into the hands of critical stakeholders across governments (at all scales), international aid organizations, market analysts, agricultural insurance agencies, and organizations throughout the supply chain in a meaningful way. 

With this in mind, the team plans to develop medium spatial resolution maps of both national cropland area and crop type using the GEO implementation of Sen2Agri-AWS throughout Argentina, Ukraine, Mali, Kenya, Sudan, and the United States with the help of in-country NASA Harvest partners. The resulting maps aim to provide stakeholders with information on the spatial extent of local cropland area, historical crop types, and in-season crop type maps to help inform crop condition assessments and planted area estimates. Additionally, an inherent and aligned benefit of this project is an opportunity to evaluate the usability, cost, and accuracy of the Sen2Agri system. This evaluation can help inform users at the national level about the return on investment of the Sen2Agri system. The team plans to compare results generated using this system to results that have been compiled from other in-house machine learning methods and commonly used software packages.

Unfortunately, most peer-reviewed crop mapping methods are not openly available to the public, therefore an emphasis of this project is to make all data inputs and outputs transparent to end users. By leveraging partnerships throughout the consortium and through widespread sharing of experience, data, and analysis, the team has created a publicly available data-sharing plan that ensures the transparency of the methods and reproducibility of the results for future use. The team plans to release all relevant data and software documentation used in the creation of the cropland area and cropland type maps in an effort to provide users and decision-makers with accurate and informative resources. Robust methods for identifying spatiotemporal patterns in crop distributions from Earth Observation data are necessary to ensure that downstream analyses of crop condition, crop yield, and agricultural sustainability variables correctly capture agricultural landscapes.