Earth Observation (EO) data has revolutionized the monitoring of global agricultural production. EO data, meaning imagery of the earth’s surface collected by satellites, allows scientists, policymakers, and extension officers to better track crop health across a growing season, provide warnings of drought and famine, and more accurately predict harvest yields.
Modeling frameworks that are able to input large amounts of EO data and output actionable results are critical in ensuring that EO’s capabilities are fully realized. NASA Harvest’s Crop Condition Co-Lead, Dr. Ritvik Sahajpal, has developed an EO-based model that is capable of forecasting crop yields months ahead of harvest time and has become a critical component of international crop monitoring efforts.
The model, Global Earth Observations for Crop Inventory Forecasting (GEOCIF), was highlighted at a recent training seminar hosted by members of NASA Harvest’s team at the University of Maryland. Over the course of three days, participants learned about the importance of EO data, its applications in agriculture, and how to use and interpret the GEOCIF model.
Led by Dr. Sahajpal, and Harvest’s Africa Program Lead, Dr. Catherine Nakalembe, the training was part of the Earth Observations for National Agriculture Monitoring Project funded by NASA SERVIR. NASA SERVIR works in partnership with leading regional organizations world-wide to help developing countries use information provided by Earth observing satellites and geospatial technologies. The project seeks to support Eastern and Southern African countries in developing their own national agriculture monitoring frameworks.
Organized in partnership with the Regional Center for Mapping of Resources for Development (RCMRD) the NASA SERVIR Eastern and Southern Africa Hub, the seminar had attendees from numerous organizations, including the Kenya Ministry of Agriculture, ICPAC, LocateIT, RCMRD, and SERVIR’s Science Coordination Office.
The training workshop brought together participants from a variety of private and public sector organizations.
GEOCIF is a critical part of the GEOGLAM Crop Monitor effort and is the model used to forecast yields of commodity crops like maize, soy, wheat, and rice in major producing countries. The model applies machine learning techniques to flexibly accommodate any number of EO data sources. Currently, GEOCIF inputs include measures of crop condition like NDVI, Leaf Area Index (LAI), and Evaporative Stress Index as well as abiotic factors including soil moisture, precipitation, and temperature data.
GEOCIF uses the best available crop-specific maps and crop calendars to forecast in-season crop yield forecasts and uses them to derive crop conditions by considering the varying response of each crop to abiotic factors, geography, and phenological growth stage.
The model has been applied to >80% of global maize and soybean producing areas; >60% of rice-producing regions; and >65% of wheat-producing regions. Its performance is assessed by comparing its yield forecasts to region-specific observed yields over the past 20 years. This flexibility of data inputs and accuracy refinement allows GEOCIF to produce crop yield forecasts 2-3 months before the harvest of key commodity crops.
Dr. Sahajpal showing the training participants graphs of different variables that go are input into the GEOCIF model, including NDVI, cumulative precipitation, maximum temperature, evaporative stress index, and soil moisture.
During the training, participants worked through interpretations of different EO products and analyzed data to predict crop yields.
Dr. Nakalembe hopes that the experience gained during this training will allow participants to better understand how EO data is important for agriculture and how it can be integrated into national monitoring programs, “Agricultural monitoring is especially important in food insecure regions. Countries in East and Southern Africa face challenges in understanding how crop conditions change throughout the growing season and can struggle to rapidly respond to negative conditions. EO data can really help close the gap between current capabilities and potential efficiency. I was really encouraged by the engagement of the participants and their enthusiasm to learn the tools.”
A training attendee who works for LocateIT, a Kenyan geospatial private company invited to the seminar, echoed those sentiments saying, “We are keen to engage and apply these methods and are now more aware of resolutions, as well as the steps involved in putting together the datasets required. The training was awesome and we hope to be part of future training.”
As threats to global food security increase, early and accurate forecasting continues to become more essential. The GEOCIF model is promising for the future of EO-based agriculture monitoring and the September training was successful in equipping more people to be able to understand and work with satellite models. Facilitating this technology transfer to an international, diverse group of professionals will expand the capacity of this tool to inform more vulnerable areas around the planet.