Objectives An easy-to-operate fertilizer recommendation method in radish is needed in urgent for turning the low fertilizer use efficiency caused by excessive fertilization in China.

Methods The data of fertilizer input, yield and nutrient uptake of radish were collected from field experiments around the whole country. QUEFTS model was used to quantitatively analyze the nutrient requirement of radish, based on which yield response and fertilizer agronomic efficiency of radish were used for the construction of an intelligent Nutrient Expert system (NE) for radish. Then, a total of 63 field experiments including different fertilization method experiments and nitrogen gradient experiments were carried out to verigy the effect of the NE system. Six treatments in each fertilization method experiment were setup, including: farmers’ practices (FP), fertilizer recommendation based on soil testing (ST), on Nutrient Expert (NE), and the N, P and K omission from NE treatments. Six treatments in each nitrogen gradient experiment were setup, including: NE, N omission from NE (N0), 60% (40%NE) and 30% (70%NE) reduced N application rates from NE, 30% (130%NE) and 60% (160%NE) increased N application rates from NE. The yield, profit, plant NPK uptake, fertilizer efficiencies and soil NPK balances were analyzed at harvest of radish.

Results 1) QUEFTS model simulation indicated that nutrient requirements of radish increased linearly with the increase of fleshy root yield when the target yield was lower than 70% of potential yield. The average N, P and K requirements for producing 1 ton of fleshy root were 2.09, 0.53, and 2.80 kg, respectively. The yield responses to N, P and K fertilizer were 16.1, 8.9 and 9.1 t/hm², and the corresponding agronomic efficiency were 95.3, 96.7 and 57.7 kg/kg, respectively. 2) Field experiment showed that NE treatment significantly decreased N, P, and K fertilizer rates by 40.0%, 51.8%, and 16.6% than FP treatment, and decreased N and P fertilizer rates by 19.9% and 34.6% than ST. Compared with FP and ST, NE treatment increased fleshy root yield by 4.0 and 2.2 t/hm², net profit by 6509 and 1710 yuan/hm², agronomic efficiency of N, P and K fertilizer by 42.6 and 25.4 kg/kg, 66.7 and 41.5 kg/kg, and 17.5 and 9.8 kg/kg, recovery efficiency of N, P and K fertilizer by 12.4 and 7.1 percentage points, 12.0 and 6.6 percentage points, and 11.8 and 5.5 percentage points; NE treatment decreased N surplus in soil by 117.7 and 34.7 kg/hm², and P₂O₅ supplus by 94.7 and 42.0 kg/hm², respectively. For the maximum yield of spring and autumn redish, NE treatment decreased N application rate by 24.4% and 16.9%, but the yield responses reached 95.3% and 96.2% of the highest one, respectively.

Conclusions The main limiting factor of radish yield in China was N, followed by P and K. The yield response and agronomic efficiency were low, indicating that long-term excessive fertilization improved the indigenous soil fertility of vegetable fields. Based on yield response and agronomic efficiency, an intelligent fertilizer recommendation method for radish was es established. Its recommended rates of N and P significantly reduced by 40.0% and 51.8%, respectively, compared with local practices. Field verification experiments in major radish planting areas such as Beijing, Tianjin and Shandong achieved fertilizer reduction and efficiency increase. It also reduces the environmental risk caused by N and P surplus. Therefore, the application prospect of Nutrient Experts system for radish production in China is broad.