Developing crop yield forecasting models for four major Ethiopian agricultural commodities

Abstract

Crop yield estimation in many developing countries (including Ethiopia) depends on conventional techniques of data collection and ground-based field reports. However, these methods are costly, time consuming, and prone to large errors because of incomplete ground observations, thus resulting in poor crop yield assessments. Moreover, the information usually becomes available too late for appropriate actions to be taken to avert food shortages and related problems. The goals of this research were to develop an information system that supports crop yield forecasting of major agricultural commodities in Ethiopia and to improve the accuracy and spatial and temporal limitations of existing systems. We conducted a ground survey to measure the actual yields of selected major crops (teff, maize, wheat, and rice) that are dominantly consumed in most of Ethiopia. We also used satellite data from Sentinel-2, which has a relatively high spatial (10 m) and temporal (5 day) resolution. Three indexes were produced and the results were correlated with ground measured crop yields. Two of the three indexes were significantly positively correlated with yield for three of the four crops. Maize yield fit best with the Enhanced Vegetation Index (EVI: R2 = 0.78; P < 0.01). The Normalized Difference Vegetation Index (NDVI) and EVI were the best fits (both R2 = 0.88; P < 0.01) for wheat yield, and teff yield was most strongly correlated with EVI (R2 = 0.90; P < 0.001). Rice had the lowest correlation between yield and all of the indexes. Thus, for three of the four crops, upscaling of the models to predict yields from farmlands by using satellite images is promising. In agro-ecologies similar to those for which the index models were developed, generally good levels of prediction are expected, but the results may vary in different topographies and agro-ecologies. Hence, additional testing of the model and the indexes in other agro-ecologies and with other crops is highly recommended.