Assessing the Potential of Multispectral UAV-Derived Vegetation Indices for Estimating Water Use of Taro (

Water availability is typically a limiting factor in rainfed sub-Saharan farming systems. Smallholder farmers in this region grow neglected and underutilized crops (NUCs) such as taro to supplement food shortages. Understanding the water use of NUCs could contribute to improving productivity. Traditional methods of monitoring actual evapotranspiration (ETa) are point-based and do not capture the variability in smallholder fields. In contrast, empirical models that use vegetation indices derived from multispectral unmanned aerial vehicles (UAVs) have demonstrated the potential to provide reliable spatial variability of ETa. This study evaluated the viability of a VI-based empirical model to estimate the ETa of taro under different weed management practices in a smallholder farm using multispectral UAV imagery. Three treatments were tested: a consistently weeded plot, a plot that remained unweeded throughout the season, and an intermediate field that was partially managed for a short period after planting. A validated VI-based model that used the enhanced vegetation index (EVI2) as a proxy for the crop coefficient (Kc) was combined with reference evapotranspiration (ETo) to determine ETa on the three taro plots. The results demonstrated that the weeded plot used 27.0% less water, while the unweeded field used 35.5% more water than the intermediate field (representing the farmer's typical integrated weed management practices throughout the growing season). Although ETa in the unweeded plot represents combined evapotranspiration from taro, weeds, and the soil surface rather than taro water use alone, the findings support an association between reduced crop-weed competition, lower evapotranspiration, and increased taro yield. The VI-based empirical models derived from multispectral UAV imagery showed great potential for promoting water use efficiency through integrated weed management (IWM) in smallholder farms.