Inventory routing for humanitarian water distribution in drought-affected regions

Emergency Water Transport (EWT) is the reactive response most commonly adopted by developing countries to cope with droughts. As droughts become more frequent, severe, and prolonged due to climate change, this operation becomes an important factor in public policies to address the problem of water shortage. However, when applied on a large scale to geographically dispersed rural populations, EWT requires the use of a significant fleet of vehicles, leading to high operating costs. Thus, it is crucial to propose solutions that guarantee the financial and technical viability of EWT. Therefore, this work proposes an algorithm to solve the water distribution problem in drought scenarios, which is modeled as an Inventory Routing Problem (IRP), to optimize the weekly vehicle routes and the inventory at each demand point for a period of one month. We explore the use of a hybrid metaheuristic of Iterated Local Search with Randomized Variable Neighborhood Descent (ILS-RVND) to search for efficient solutions. The algorithm was applied to 798 classical IRP benchmark instances and to a real case of water distribution in the semi-arid region of Brazil. Comparing our results with those from the DIMACS Challenge revealed improvements in 12 previously best-known solutions. The results for the real case are compared with the actual routes adopted in Brazilian water distribution, with solutions using routes obtained by solving the Capacitated Vehicle Routing Problem (CVRP). The proposed application reduces travel costs by up to 45.9% and improves equity in water distribution, offering a scalable solution for humanitarian logistics.