| Sumario: | In this study, we aimed to map the frequency of recurrent desert locust invasions to predict future invasions and understand their correlation with climatic variables. Utilizing global locust data from the Desert Locust Hub (FAO) covering 1985 to 2020, we formulated an optimization problem to minimize grid size, ensuring each cell contained data for at least 35 years to calculate the invasion frequency using the ratio of locust presence per year. The computational implementation used Python, for numerical computations and geographical data handling. The model incorporated climatic predictors from the NASA POWER project, including relative humidity, temperature, evapotranspiration, wind speed, and precipitation. Using an Adaptive Neuro-Fuzzy Inference System (ANFIS) combined with a genetic algorithm (GA), we refined our predictive capabilities. The model underwent 3000 iterations, with an optimized crossover and mutation rate enhancing the genetic algorithm’s performance. The model demonstrated strong predictability and transferability with an R2of 0.87, an root mean square error (RMSE) of 0.44, and a low Mean Absolute Error (MAE) of 0.30 to 0.40. The model predicted that 40% of the study area is frequently invaded by desert locusts, highlighting the African Sahel, Arabian Peninsula, and parts of the Maghreb and Middle East as high-risk regions. These findings provide a robust framework for predicting desert locust invasions and developing effective management strategies.
|
|---|