Prediction of tea yield and picking date based on the AquaCrop model

This study aimed to verify the applicability of crop models for the yield and picking data prediction of different tea varieties in Zhejiang Province and to fill the gap in tea simulations of crop models. Three tea varieties ('White Leaf 1' 'Longjing43', and 'Longjingqunti') were selected in the typical tea planting areas of Zhejiang Province (Anji County, Songyang County and Hangzhou City). Based on the AquaCrop model recommendations and the tea plant conservative parameters provided by the Food and Agriculture Organization (FAO), the parameters of the AquaCrop model were obtained from field experiments, data collection, and parameter sensitivity analysis, and the model was localized and corrected using data from previous years. The yield forecast models for Anji and Songyang Counties and prediction models of the spring tea picking data for the three tea varieties were established based on the AquaCrop. The AquaCrop model simulated the average total annual tea output of Songyang County from 2013 to 2017 to be 1.497 t·hm^-2, with a relative error of 1.98%. The average annual output of spring tea in Anji County from 2014 to 2018 was 0.164 t·hm^-2, with a relative error of 0.99%. The normalized root mean square error of the AquaCrop model for tea yield simulation in Songyang and Anji Counties was 2.20% and 1.10%, respectively; and the root mean square error was 0.0325 t·hm^-2and 0.0018 t·hm^-2, respectively; the conformity index was 0.84 and 0.88, respectively. The prediction standard of the tea growing degree-days (GDDs) for 'White Leaf 1' 'Longjing43', and 'Longjingqunti' were determined, and the prediction formulas of the three tea GDDs were obtained via the stepwise regression method. The mean absolute errors (MAE) of AquaCrop model based GDDs prediction of three tea varieties were 1.1 d, 2.1 d, and 1.1 d, respectively. The AquaCrop model based on stepwise regression prediction of the tea picking data was significant (P < 0.01), with the simulated MAEs for three tea varieties were 0.7 d, 0.7 d, and 0.9 d, respectively. The results show that the AquaCrop model has good adaptability to different tea varieties in Zhejiang Province after correction. Localized AquaCrop models can be used to study the water management and yield potential of tea gardens. Both prediction models the AquaCrop model of the tea picking data based on GDDs prediction and stepwise regression prediction have applicable value, and the predictions based on the stepwise regression analysis model is more accurate with higher practical value than the GGDs forecasting model.