柴达木枸杞成熟期影响因子及集成模拟模型构建

Research on the Influencing Factors of Wolfberry Ripening Stage and Integrated Simulation Method in Qaidam

枸杞产业为劳动密集型产业,准确模拟枸杞成熟期对枸杞果实采摘等田间管理至关重要。研究采用相关分析法确定柴达木枸杞成熟期关键气象因子、灰色关联分析法确定与枸杞成熟期密切相关的物候期,建立多元回归线性模拟方程;基于两种模型模拟结果构建集成模拟模型并检验验证。结果表明:(1)三种方法所构建的模型中自变量间不存在多重共线性关系,方程具有统计学意义且均通过显著性检验。(2)基于气象因子构建的模拟模型中,老眼枝果实成熟始期和普期模拟效果最好,回代检验RA分别为63.16%和90.00%,RMSE为2.70d和2.08d,RE为1.39%和1.00%,模拟检验MAE分别为2.31d和1.35d;夏果成熟始期和普期模型模拟效果较差。(3)基于物候期构建的模拟模型中,夏果成熟始期和普期模拟效果最好,回代检验RA分别为89.47%和90.00%,RMSE为1.65d和1.60d,RE为0.75%和0.69%,模拟检验MAE分别为1.38d和1.00d。(4)枸杞集成模拟模型克服了单一模型局限性,模拟精度整体提高,模拟值与观测值误差均在3%以内。综合来看集成模拟模型适用于柴达木枸杞成熟期模拟。

The wolfberry industry is a labor-intensive industry.It is very important for fruit picking and management to accurate simulation wolfberry ripening stage.The meteorological factors significantly related to the maturity stage of wolfberry were determined by correlation analysis method.The development stage closely related to maturity were determined by grey correlation analysis.Then,linear regression equations were established respectively.Based on the simulated day sequence of two models,the integrated simulation model was established.All models had been validated.The results showed that(1)there was no multicollinearity between the independent variables for all models established by three methods,while the equation was statistically significant and passed the significance test.(2)In the simulation model based on meteorological factors,the best simulation effect was found in the starting ripening stage and the common ripening stage of Laoyan branch.RA was 63.16%and 90.00%,RMSE was 2.70 and 2.08,RE was 1.39% and 1.00% by the back substitution test,respectively.MAE of those was 2.31d and 1.35d by the simulation test.However,the results of the starting ripening stage and the common ripening stage of summer fruit branch did not have good performance.(3)In the simulation model based on other phenology,the starting ripening stage and the common ripening stage of summer fruit branch had the best simulation effect.RA of those was 89.47% and 90.00%,RMSE was 1.65 and 1.60,RE was 0.75% and 0.69%,by the back substitution test.MAE of those was 1.38d and 1.00d by the simulation test.(4)The limitations of the two models were overcome by the integrated simulation method.The accuracy of simulation results was improved,and the simulation errors of 6 models were less than 3%.It was concluded that the integrated simulation method had the highest accuracy,while suitable for the prediction of maturity stage of wolfberry.