Estimating Crop Yields in Uruguay

Maize field in Uruguay Photo: Instituto Nacional de Investigación Agropecuaria

By Sunny Ng, Climate and Society ’13

Ministries of agriculture generally want to estimate crop yield to plan for food security and crop exports. Ever wonder how it’s estimated?

In Uruguay, the Enhanced Vegetation Index (EVI), an index used to measure vegetation, is used along with GPS data of crop fields to estimate crop yield. The EVI method has two major disadvantages. First, as with any index, there is a degree of uncertainty. For example, at what threshold can you decipher a crop from grass? Second, by relying on GPS crop field data, estimates of crop yield cannot be determined until late in the season.

This summer, I worked with Yifang Yang, a Climate and Society colleague, on this NASA DEVELOP project to improve the method of detecting crops at an early stage for our case study in Uruguay. Using imagery from NASA’s Landsat 7 satellite available at 30 meter resolution, we explored two techniques. Conceptually we know that from month to month, the land cover of crop fields evolve from soil, before seeds are planted, to vegetated land to dense vegetation. Using this prior knowledge, we also accounted for unknown land cover classes covering crop fields such as clouds and water. While this method detected crops fairly well, small areas of noise and elongated vegetation strips growing along rivers were also detected. To eliminate non-crop vegetation, geometric data of crop fields was used to screen out elongated shapes and small pixels not representative of crop fields.

Land cover classification using the Satellite Image Automatic Mapper (left) and detection of soybean crops in black outline (right)

The research done this summer significantly improved the detection of crops. One key advantage of using a time series of satellite imagery is that it allows us to detect crops at early stages, within three months of planting. Another advantage of using crop field geometries is the identification and elimination of specific shapes.

So what’s next? We need to apply this method to eliminate other crop fields detected such as wheat fields, which are not planted during the same season as soybeans and maize. This case study can be applied to other years and other regions and can be verified with GPS crop field data.

If you are curious about how satellite imagery can be used to solve other real world issues, check out other NASA DEVELOP projects that were hosted at the International Research Institute for Climate and Society, a new node of the NASA DEVELOP National Program during Spring 2013. This summer, the IRI and Climate and Society students worked on two other research projects with NASA DEVELOP: Uruguay Agriculture, Sudan Health and Air Quality: Leishmaniasis in Gedaref and Examining Meningococcal Meningitis and Climate in the Sahel.