Estimating the yield of Parkia biglobosa fruits from field data and image-based predictive models

Accurate estimates of tropical tree fruit yields are essential for assessing individual tree productivity and landscape-scale applications. This study examines the productivity of Parkia biglobosa pods by integrating tree morphology, vegetation health indices from drone and satellite images, and soil properties into a mathematical model. The model evaluates interactions between these factors and their contribution to yield. Fruits were harvested from 105 mature trees within three study sites in the savanna ecological zones of Ghana and weighed. The crown radius of selected trees was measured in meters from mosaics derived from drone images in QGIS. Satellite images and data on the physical and chemical soil properties were obtained from the Norway International Climate and Forest Initiative (NICFI) and the Soil Reference and Information Centre (ISRIC), respectively. The yield was estimated by an exponential relationship involving the tree crown radius, average blue band reflectance from satellite images, and interaction with the locality. The model explained 75 % of the variation in yield. Excluding interaction terms, the best model estimated yield from tree crown radius, average near infrared reflectance of the crown, and soil pH at 30 – 60 cm depth. It indicated that crown radius, soil pH, Blue and Near-infrared spectral characteristics, along with variations in locality, such as soil and tree management practices, influenced the yield of Parkia biglobosa. Collectively, they explained 67.8 % of the variation in the yield. Policies to control tree stand undergrowth, prevent bushfires, promote less invasive harvesting techniques, and monitor soil nitrogen and pH levels are recommended for localities to enhance tree health and productivity.