Optimizing phosphorus fertilizer application to balance potato yield and quality: A neural network analysis

• P supply has different effects on tuber yield and various quality indicators of potato. • Artificial neural networks can be used to balance potato yield and quality. • The GA-BP model can customize P plans based on yield and quality requirements. • The GA–BP model demonstrated high predictive accuracy (relative error < 10 %). Phosphorus (P) is an important factor determining both potato tuber yield and tuber quality. However, conventional P fertilizer management has typically focused on increasing tuber yield, often neglecting the impact on tuber quality attributes such as starch content, dry matter (DM) content, and reducing sugar (RS) content. As a result, increased yields are often obtained at the expense of quality, despite the latter being a key determinant of consumer satisfaction. This study aimed to establish optimal P fertilization strategy for achieving yield-quality synergy in potato tubers through artificial neural network modeling. The results of 3 years of field experiments showed that P application rates significantly affect potato tuber yield and various quality indicators, including marketable tuber rate, DM content, starch content, RS content, soluble sugar content, and crude protein content. Based on the data collected from the field experiments, we developed a backpropagation neural network model optimized by a genetic algorithm (GA-BP). The model demonstrated remarkable prediction accuracy, with a R² of 0.9877, a RMSE of 12.63, and a MAE of 10.37. Therefore, the genetic algorithm–backpropagation model can be effectively applied to ascertain the optimal P application rates, thereby achieving a balance between tuber yield and tuber quality in potato production.