Towards Modelling of Maize Crop Yield Using UAV and Earth Observation Spectral Fusion in the Kingdom of Eswatini

The increasing worlds population coupled with climate change has steadily increased demand for food. Consequently, countries have been faced with the challenge ensuring food security to their citizens. To meet this mandate, governments require agricultural production information to aid resource allocation and sustainable farming practices so as to attain improved food security for the developing world. Information on spatial distribution of crops such as maize is therefore an important step towards sustainable agricultural production. Earth observation techniques offer an effective and efficient platform for mapping crops due to improved temporal and spatial resolutions. This capability provides a reliable system with near-term potential to provide stakeholders with timely information on crop distribution, status, and predicted yield. The advent of light Unmanned Aerial Vehicles (UAVs) has also pushed the frontiers of data acquisition and phenomena monitoring. Here, we demonstrate how we coupled data from UAVs and Sentinel-2 images to map crops in four pilot locations in the Kingdom of Eswatini. Multispectral drone imageries containing the Red, Green, Blue, Red-Edge and Near InfraRed spectral bands were acquired using the Wingra One fixed wing drone. Classification algorithms using random forests (RF) and maximum likelihood classification (MLC) algorithms were applied on the S2, UAV and fused UAV+S2 images in order to spectrally distinguish maize crops from other crop and vegetation cover types. Our results indicate that RF gives a better accuracy on the UAV+S2 fused product compared to RF classification of un-fused images (UAV and S2). The methodology used in this study could be replicated and up-scaled to other maize growing countries in Africa to fully obtain accurate and detailed maize crop masks that would be critical inputs in modelling of maize crop yield. Keywords: Unmanned Aerial Vehicles, Image Fusion, maize crop mapping, maize crop yield