Precisely forecasting population dynamics of agricultural pests based on an interval type-2 fuzzy logic system: case study for oriental fruit flies and the tobacco cutworms

Traditional pest control approaches rely mostly on the experience of farmers, which may not be effective due to lack of scientific information regarding the environment where crops grow. Farmers can initiate a more effective integrated pest management program when precise and quantified results of forecasting pest population outbreaks are provided. Previous studies generally utilize long-term data to predict pest populations, but such a prediction approach might not be useful for farmers who grow fruit and vegetables with shorter life cycles. This paper therefore proposes an interval type-2 fuzzy logic system (IT2FLS) with short-term data to forecast the population dynamics of the oriental fruit fly (OFF, Bactrocera dorsalis (Hendel)) and the tobacco cutworm (TC, Spodoptera litura (Fabricius)). Two automatic monitoring systems are used to collect the data of the population dynamics of OFFs and TCs and the environmental parameters in farming areas. A univariate fuzzy time series forecasting model with difference-based intervals (UFTSFM_DI) and a bivariate fuzzy time series forecasting model with difference-based intervals (BFTSFM_DI) are developed, and integrated into the proposed IT2FLS. It is found that the BFTSFM_DI model yields better performances of forecasting OFF and TC populations when the atmospheric temperature data are employed. With the forecasting results, farmers will have a better understanding of the population dynamics of the OFF and TC in farming areas, so they can take proper measures, such as bagging their fruits and spraying pesticides, before pest outbreaks occur.

Automated summary

Traditional pest control methods may not be effective due to the lack of scientific information. This study proposes a fuzzy logic system with short-term data to forecast the population dynamics of the oriental fruit fly and the tobacco cutworm, and develops corresponding forecasting models.