Multi-stage fertilizer recommendation for spring maize at the field scale based on narrowband vegetation indices

Abstract

Crop fertilizer recommendations that are site-specific and can be adjusted in multiple growth stages can result in improved nutrient management, fertilizer use efficiency, and grain yield. There are considerable challenges to collecting a large number of soil and crop samples using the traditional methods, so a modern approach to developing fertilizer recommendations is urgently needed. In this study, we used the topdressing recommendation method based on unmanned aerial vehicle-based hyperspectral images and conducted a long-term nitrogen–phosphorus–potassium (NPK) combined fertilizer experiment in a spring maize cropping system. The hyperspectral imagery was acquired at multiple critical stages (jointing, silking, and maturity) in the experimental field containing 120 microplots during 2019–2020. The leaf nitrogen content (LNC) and yield measurement data for each microplot were also collected in multiple stages. Ten narrowband vegetation indices (VIs) were derived from the hyperspectral imagery as effective indicators for the estimation of the crop N status and grain yield. To develop topdressing recommendations for multiple stages, first, the optimal NPK fertilizer amounts, which were applied as the total fertilizer requirement to achieve the highest yield, were obtained. Second, the thresholds of the narrowband VIs, which justify if topdressing N fertilizer is needed, were calculated using the regression model established based on the relative yield and VIs. Finally, the N fertilizer amounts already applied in the jointing and silking stages were obtained based on the response of the narrowband VIs to the N fertilizer amounts. The results reveal that the optimal amount of N fertilizer is 180 kg ha−1, and the grain yield can be improved by using a combination of P and K fertilizers. Five VIs were significantly correlated with both the LNC and yield in the jointing and silking stages, and the thresholds of these VIs, below which topdressing fertilizer is needed, were identified at 95% of relative yield. The five VIs all presented high fitting accuracies with the amount of N fertilizer applied. However, there were noticeable differences in the sensitivities of the VIs in the jointing and silking stages. Therefore, we developed a cost-effective approach and provide multi-stage fertilizer recommendations at the field scale for the spring maize cropping system. This method can also benefit other precision agriculture applications if incorporated with remote sensing techniques.