Construction and application of citrus intelligent planting service platform based on WOFOST model

Abstract:【Objective】Chongqing is one of the dominant regions of citrus planting and one of the three late-maturing citrus producing areas in China. By the end of 2018, the citrus planting area in Chongqing was about 231 700 hm2, the output was over 3.2 million tons, and the total output value was nearly 30 billion yuan. However, the citrus planting management mode in Chongqing is still traditional and exten- sive. WOFOST model is a common crop growth model, which is often used to predict crop yield and  analyse the impact of different crop varieties, climate change, growth period changes and irrigation conditions on crop yield. It can provide theoretical basis for standardized, accurate and intelligent crop production management. In order to improve the intelligent management level of the whole process of cit- rus planting in Chongqing, this paper explores the construction of the citrus intelligent planting service platform (CIPSP) based on the WOFOST (World Food Studies) model, which is used to assist orange growers to make intelligent planting and management decisions, so as to improve the comprehensive economic benefits of citrus planting【. Methods】In this paper, a new citrus variety Orah Mandarin (Tem- ple Tengor × Dancy Tangerine) served as the experimental material, and the CIPSP system based on WOFOST model was designed and validated with five parts of system level software, including physi- cal server and virtual server operating system, virtualization software, open platform, large database and cloud platform management software. By referring to the methods introduced by Wan Qiuping, Dou Li- fang, Bai Qinfeng and Zhou long, the CIPSP system model knowledge base derived from WOFOST modeling idea was constructed; referring to Zhu Jiaqi’s method, the sensitivity analysis of the model pa- rameters was carried out by using the LH-OAT method to determine the CIPSP system model parameters; referring to the Jiang Haiyan’s method, an adaptive algorithm engine was designed to predict and correct the model parameters of CIPSP system. According to the needs of citrus intelligent planting service, CIPSP system structure was designed based on WOFOST model of data collection, screening, transmission, modeling and application; referring to Xu Xinliang’s method, the intelligent decisionmaking function of citrus planting (yield prediction simulation, soil moisture / soil quality and meteoro- logical real-time monitoring, etc.) was designed; referring to Lu Zhanjun’s method, we designed the de- tection function of common diseases and pests based on machine deep learning. At the same time, some functions of the system, such as yield prediction and simulation, real-time monitoring of soil moisture / soil quality and meteorology, and intelligent pest forecasting were verified by referring to the methods of Wan Qiuping, Zhang Xiangfei, Li Zhongliang and Shi Dongxu, the crop parameters of Orah Manda- rin in the experimental field were collected by using index measurement tool, and the simulated and measured values of Orah Mandarin yield were compared, in which 15 measuring points of three differ- ent growth years (2-3 years, 3-4 years and more than 5 years) were selected in Heyan village, Longhe Town, Changshou District, Chongqing【. Results】By designing crop, soil, climate and pest parameters, CIPSP system model knowledge base derived from WOFOST modeling idea of citrus growth cycle was constructed; LH-OAT method was used to analyze the sensitivity of CIPSP system model parameters based on WOFOST model, and the model parameter I with high sensitivity was determined quickly; based on the real-time observation data, the crop growth was simulated and compared. The adaptive al- gorithm was introduced to modify the model, and the selection determination coefficient (R2) and residu- al aggregation coefficient (CRM) were selected to evaluate the accuracy of the prediction results. Ac- cording to the process of data acquisition, screening, transmission, analysis and application, CIPSP sys- tem was designed into five independent layers: perception layer, edge layer, infrastructure layer, plat- form layer and operation layer; through CIPSP system to collect and analyze the real-time environmen- tal data, CIPSP system can realize the functions of citrus yield prediction and simulation, soil moisture / soil quality and meteorological real-time monitoring, intelligent pest forecasting and so on. The prelimi- nary verification and application of the system function in Heyan village, Longhe Town, Changshou District, Chongqing showed that the WOFOST model could well simulate and predict the yield of Orah Mandarin with the R2 value of 0.949, but the deviation between the measured value and the predicted value gradually increased with the increase of tree age; through the combination of the real-time infor- mation collected by solar weather station, soil moisture, soil monitor and crop growth model, the application of realtime environmental data collection and analysis of soil moisture / soil and meteorology was realized; the intelligent pest monitoring system was combined with remote sensing technology and hyperspectral UAV to realize on-site/remote detection, discrimination, early warning, prevention and control of common diseases and pests based on machine deep learning.【Conclusion】In this paper, CIP- SP system based on WOFOST model was constructed and successfully applied to yield prediction, planting environment monitoring, pest early warning and control in Orah Mandarin experimental field, which can improve the level of citrus intelligent planting management and realize technological innova- tion in software technology, production management and information system, but the model parameters and algorithms still need to be optimized.