Real-time and multi-stage recommendations for nitrogen fertilizer topdressing rates in winter oilseed rape based on canopy hyperspectral data

Abstract

Due to the intensive rotation and environmental concerns related with the over-fertilization, optimizing nitrogen (N) fertilization is crucial to winter oilseed rape cultivation in the Yangtze River Basin in China. Although remote sensing techniques have been applied to aid the site-specific crop N management, most studies are conducted for a specific growth stage, ignoring that multi-stage topdressing are commonly applied for crops. This study aimed to determine the real-time and multi-stage N fertilizer topdressing recommendations for winter oilseed rape by developing site-specific topdressing recommendation (SSTR) models using hyperspectral canopy reflectance data. The in-situ hyperspectral data were collected from five N fertilization experiments conducted at two sites in Hubei province, China during 2013–2016. A ratio vegetation index (RVI) based on reflectance at 764 and 657 nm wavelengths (RVI(764,657)) was proposed as the effective indicator of the N status in winter oilseed rape. For each critical growth stage, the empirical relationships between the RVI(764,657), the leaf N concentration (LNC), and the yield were used to develop SSTR models and determine the threshold below which topdressing was needed. In the growth season of 2016–2017, the performance of the SSTR models were evaluated by conducting the experiment of the real-time and multi-stage topdressing recommendation. This experiment included 54 multi-stage topdressing strategies. At each growth stage, whether the topdressing was needed and the corresponding topdressing rates were determined with the canopy spectra measurements and the SSTR model at this stage. The site-specific N topdressing was then applied on the next day after spectra measurements were taken. The LNC, the dry mass (DM), and the yield of this experiment were compared with the reference experiment of the local farmer’s topdressing strategy. Results showed that the maximal yield was achieved by topdressing either in the eight-leaf stage (yield = 2983 kg/hm2) or the over-wintering stage (yield = 2902 kg/hm2) for the N120 treatment. The yield of 2983 kg/hm2 was 9.8 % lower than the theoretical maximal yield of 3306 kg/hm2, but 8.3 % higher than the maximal yield in the reference experiment. This maximal yield was achieved by reducing 25 % of the theoretically optimal total N rates. Moreover, we found that neither the LNC nor the DM and the yield were linearly related with the total N fertilizer application rates. The timing of the N fertilization had more important impacts on the DM and the yield than the total N fertilizer rates. Postponing the timing of topdressing could decrease the DM and the yield. Results from this study suggested that the newly developed SSTR models could help effectively distribute the N fertilizer based on the real-time demand of winter oilseed rape in critical growth stages.