Data Collection And Forecasting Model Establishment Of Cruciferous Insect Pests In Shandong Based On Big Data

Cruciferous vegetables are stable food for mankind.The main pests of cruciferous vegetables such as the cabbage worm Pieris rapae,the diamondback moth Plutella xylostella and the beet armyworm Spodoptera exigua cause the decrease of yield and lower quality of vegetable crops.Since the growth cycle of cruciferous vegetables is short,the P.rapae,P.xylostella and S.exigua are polyphargous herbivores,so the prediction of its occurrence and occurrence degree is lack of effective means,especially in the short-term prediction.It is impossible to guide the scientific and effective prevention and control program without accurate prediction.The excessive use of chemical pesticides,resulting in the growth of insecticide resistance,insect pest resurgence and pesticide residues,and other severe problems,and the ecological security,environmental safety and agricultural products and food security are difficult to be guaranteed.Here,5 representative areas of cruciferous vegetables in Shandong Province,Dingtao,Feicheng,Lanling,Shouguang and Jiaozhou(linear distance greater than 100 km)were selected as the investigation sites.Two algorithms,decision tree and random forest were used to analyze the correlation between the occurrence degree of P.rapae,P.xylostella and S.exigua and 16 kinds of meteorological data,such as the average temperature and the daily maximum temperature.The monitoring and forcasting model was established for providing theoretical basis and technical support to these insect pests control.The results are as follows:1.According to the average number of insect pest larvae per hundred plants of cruciferous vegetables accross the spatial scale of Shandong province from 2005 to 2015 years,it could be seen that the occurrence degrees of P.rapae,P.xylostella and S.exigua in Shouguang were the highest in the 5 investigation areas,followed by the amount of P.rapae in Lanling,P.xylostella in Dingtao,and S.exigua in Jiaozhou.The P.xylostella is the dominant species of the 3 cruciferous vegetable species with the highest amount in Dingtao,Feicheng,Shouguang and Jiaozhou.The number of the beet armyworm in the 5 observation areas is the lowest among the 3 cruciferous vegetable species.From time scales,the incidence of P.xylostella in Dingtao,S.exigua in Feicheng,P.rapae in Shouguang and the above 3 insect pests in Jiaozhou increased year by year.2.Based on the decision tree classification algorithm,16 kinds of meteorological data such as the average temperature and daily maximum temperature and the occurrence degree of the 3 insect pest lavae in 5 investagition areas were adopted to build the forcasting models.According to the information gain rate,the results show that:(1)The average temperatures in Dingtao and Lanling had the highest correlation with the occurrence degree of P.rapae(the information gain rates were 0.69,0.69 respectively).The daily minimum temperature in Feicheng and Shouguang had the highest correlation(0.63,0.66).The daily precipitation and the average temperature in Jiaozhou had highest correlation(0.45,0.37).(2)The average temperatures in Feicheng and Shouguang had the highest correlation with the occurrence degree of P.xylostella(the information gain rates were 0.67 and 0.64 respectively).The daily minimum temperature in Dingtao had the highest correlation(0.58).The minimum relative humidity in Lanling had the highest correlation(0.44).The daily precipitation in Jiaozhou had the highest correlation(0.72).(3)The daily minimum temperature in Dingtao,Feicheng and Lanling had the highest correlation with the occurrence degree of S.exigua(0.72,0.69 and 0.57 respectively).The daily maximum temperature in Shouguang had the highest correlation(0.54).The average relative humidity in Jiaozhou had the highest correlation(0.67).3.Based on the random forest algorithm,16 kinds of meteorological data such as the average temperature and daily maximum temperature and the occurrence degree of the 3 insect pest lavae in 5 investagition areas were adopted to build the forcasting models.According to the mean decrease Gini,the results showed that(1)The average temperature was the most important correlation factor with the occurrence degree of P.rapae in Dingtao,Lanling,Shouguang and Jiaozhou,and also the main correlation factor in Feicheng.(2)The average temperature was the most important correlation factor with the occurrence degree of P.xylostella in Dingtao and Lanling,the daily maximum temperature was the most important correlation factor in Feicheng and Jiaozhou,and the daily minimum temperature is the most important correlation factor in Shouguang.(3)The average temperature is the most important correlation factor with the occurrence degree of S.exigua in Dingtao and Lanling,the daily maximum temperature was the most important correlation factor in Shouguang and Jiaozhou,and the average water vapor pressure is the most important correlation factor in Feicheng.4.The comprehensive correlation factor analysis of the results of the prediction model constructed by the decision tree and random forest algorithms showed that temperature(including average temperature,daily maximum temperature and daily minimum temperature),humidity and precipitation(including average relative humidity,minimum relative humidity,daily precipitation and average water vapor pressure)and atmospheric pressure(including average station pressure,daily maximum station pressure and daily minimum station pressure)were the main reasons for the change of the ocurrence degrees of the 3 insect pests in the cruciferous vegetable fields of Shandong province.The confidence degrees of the decision tree models established in this study is above 90% and the operation was stable.The random forest models were verified by the practice of 3 insect pest species from 2016 to 2017,and the predicted anastomosis is up to 80%.These 2 models could forcast the short-term occurrence degree of the 3 pests of cruciferous vegetables,so as to guide scientific green prevention and control services.