ESTIMATION OF MAIZE GRAIN YIELD USING MULTISPECTRAL SATELLITE DATA SETS (SPOT 5) AND THE RANDOM FOREST ALGORITHM

Abstract

Cropyield estimation is a very important aspect in food production as it providesinformation to policy and decision makers that can guide food supply not onlyto a nation but also influence its import and export dynamics. Remote sensinghas the ability to provide the given tool for crop yield predictions beforeharvesting. This study utilised canopy reflectance from a multispectral sensorto develop vegetation indices that serve as input variables into an empiricalpre-harvest maize (Zea mays) yield prediction model in the north easternsection in Free State province of South Africa. Selected fields in this regiongrown of maize under rain-fed conditions were monitored and the grain harvestedafter 7-8 months with actual yields measured. The acquisition of suitable mediumresolution SPOT 5 images over this area in 2014 was in March and June beforethe grains were harvested in July. A number of spectral indices were developedusing the visible and near infrared bands. Through the random forest algorithmpredictive models, maize grain yields were estimated successfully with an R2of 0.92 (RMSEP = 0.11, MBE = -0.08) for the Agnes field and for Cairo the R2was 0.9 (RMSEP = 0.03, MBE = 0.004) from the March images. These results wereproduced by the SAVI and NDVI respectively for both fields. It was thereforeevident that the predictive model applied in this study was site specific andwould be interesting to be tested for other maize cultivars as well as testingfor an optimal period during the plant life cycle to predict grain yields ofmaize in South Africa.