Climate-Informed Predictive Optimal Control of Desert Locust Dynamics Under Machine-Learning Climate Forcing
by Dejen K. Mamo, Mathew N. Kinyanjui & Nourridine Siewe.
Read the paper
Desert locusts swarm when the weather cooperates: rain greens the desert and warmth speeds their growth. Can better forecasts make control smarter? Instead of describing the seasons with a smooth sine wave, we trained two machine-learning models on decades of records from Amibara, Ethiopia — one for temperature, one for rainfall — and fed their forecasts into a population model tracking eggs, hoppers, bands, adults and swarms alongside the vegetation they strip. We then computed the treatment schedule that holds locusts lowest and vegetation highest for the least effort. Treating young and adult locusts together beat either alone, cutting the management burden by over 95%, and that ranking held when we deliberately corrupted the forecasts.
Image Description: Machine-learning forecasts of temperature and rainfall drive a stage- and phase-structured desert locust population model; the optimal control problem then schedules juvenile and adult treatment to jointly minimise locust abundance and vegetation loss.
