Abstract: Climate yield simulation is critical to the insurance design and benefit evaluation of meteorological service for agriculture. Based on the peach yield data in Fenghua from 1995 to 2018, Fourier transform, moving average, Logistic, HP filter and exponent methods were used to separate the trend yield and climate yield of juicy peach. The common meteorological disaster index of juicy peach was established, which was used to verify and screen the yield separation results with five different methods. With the disaster index as the input factor, the climate yield model of peach was established by BP neural network based on genetic algorithm (GA-BP). The results showed that, from 1995 to 2018, the continuous rain during blossommature period and heavy precipitation during stone hardeningmature stage of Fenghua juicy peach decreased first and then increased. In recent years, the frequency of extreme precipitation and gale increased. Climate warming alleviated the chilling damage in flowering and fruit setting stage, but increased the risk of high temperature in mature stage. The simulation effect of trend yield by the third-order Fourier separation was good. The simulation accuracy of typical disaster years and low disaster years was as high as 88%, with a correlation coefficient of -0.8 and a comprehensive index of0.85. The GABP model based on the disaster index had the best simulation effect on the relative climate yield separated by the thirdorder Fourier method. The absolute error and root mean square error of the back-testing were 0.02 and 0.03, respectively, with the correlation coefficient being as high as 0.95. The absolute error and root mean square error of prediction test were 0.03 and 0.04, respectively, with the correlation coefficient being as high as 0.92. In conclusion, the thirdorder Fourier method was suitable for peach yield separation in Fenghua. The simulation accuracy and stability of characteristic crop yield based on disaster index were higher, with clear physical meaning of the model.

Key words: juicy peach, disaster index, yield separation, GA-BP method, yield simulation.