Random Forests modelling for the estimation of mango (Mangifera indica L. cv. Chok Anan) fruit yields under different irrigation regimes

‘Chok Anan’ mangoes, mainly produced in Northern Thailand, are appreciated for their light to bright yellow colour and sweet taste. Because fruit development of the on-season mangoes occurs during the dry season, farmers have to irrigate mango trees to ensure high yields and good quality. Therefore, it is important to understand the effects of water supply on the yield of mango fruit for better control and effective use of limited water resources. In this study, we aim to demonstrate the applicability of Random Forests (RF) for estimating mango fruit yields in response to water supply under different irrigation regimes. To cope with the variability of mango fruit yields observed in the field, a set of RF models was developed to estimate the minimum, mean and maximum values for each of the mango fruit yields, namely “total yield” and “number of marketable mango fruit”. In RF modelling, a combination of 10-day rainfall and irrigation data was used as model input in order to evaluate the effects of water sources on the mango fruit yields. The RF models accurately estimated the maximum and mean values of mango fruit yields, and showed moderate accuracy for the minimum mango fruit yields. The variable importance measure computed in the RF calculation suggested that the timing of water supply affects the mango fruit yields whereby rainfall and irrigation have different effects on the mango fruit yields. This case study on the estimation of mango fruit yields demonstrates the applicability of RF in the field of agricultural engineering, with a specific focus on water management. The model performance and the information retrieved from the RF models allow for precise modelling and the development of improved management practices in target agricultural systems.