Accurate and timely crop yield forecasts are critical for making informed agricultural policies and investments, as well as increasing market efficiency and stability. Earth observation data from space can contribute to agricultural monitoring, including crop yield assessment and forecasting. A recent study by University of Maryland NASA Harvest Hub researchers (Belen Franch, Eric Vermote, Sergii Skakun, Jean-Claude Roger, Inbal Becker-Reshef, Emile Murphy and Chris Justice), "Remote sensing based yield monitoring: Application to winter wheat in United States and Ukraine" presents a new crop yield model based on the Difference Vegetation Index (DVI) extracted from Moderate Resolution Imaging Spectroradiometer (MODIS) data at 1 km resolution and the un-mixing of DVI at coarse resolution to a pure wheat signal (100% wheat within the pixel). The MODIS dataset of 17+ years allows the researchers to build a strong empirical model based on a large number of statistics and covering a wide range of crop conditions relative to other available satellite data.
A new method, recently titled the Agriculture Remotely-sensed Yield Algorithm (or ARYA for short, a play on the Game of Thrones character Arya Stark, whose family is associated with the words "Winter is Coming"), helped to determine the total DVI signal from each pixel. This method was developed specifically to work at a coarse resolution, where the signal from the targeted crop is mixed with other surfaces. This method unmixes the signal coming from the crops, using the crop type masks as a reference. Using that signal, the authors use a linear regression to relate three regressors with the yield: the peak (amplitude) of the seasonal evolution of the DVI of the crop (once unmixed), the length of time of that peak, and the average of the evoporative fraction 30 days after the peak, to account for any stress condition after the peak. Six linear combinations of these three variables are considered, and where there are signficant regressors for each country, the corresponding equation is applied.
The model was applied to estimate the national and subnational winter wheat yield in the United States and Ukraine from 2001 to 2017. At the subnational level, the model performed very well for both countries, and at the national level for the United States (US) and Ukraine the model provides a strong coefficient of determination of 0.81 and 0.86, respectively, demonstrating good performance at this scale. The model was also able to capture low winter wheat yields during years with extreme weather events, for example 2002 in US and 2003 in Ukraine. The RMSE of the model for the US at the national scale is 0.11 t/ha (3.7%) while for Ukraine it is 0.27 t/ha (8.4%).
This model is based on the study of different patterns of the wheat DVI evolution through the growing season. Some yield models use other vegetation indexes such as the Normalized Difference Vegetation Index (NDVI) as a reference. However, the MODIS Surface Reflectance product is now well established and widely used by the science community as a reference. Given the limitations of using the NDVI on dense vegetation, this simpler vegetation index, originally discarded for not minimizing for atmospheric or BRDF effects, can now be used. Results obtained using this index demonstrate that it is well correlated to crop yields and is responsive to high yield values.