Planning migrant labor for green sugarcane harvest: A stochastic logistics model with dynamic yield prediction

The sugarcane harvest often requires migrant workers when there is insufficient supply of domestic workers or harvest machines. The recruitment of migrant workers is a long process, and the decision maker (DM) needs to determine the number of workers before the yield uncertainty is revealed. The prior distribution can be estimated from historical yields. During the sugarcane growing season, the DM can dynamically update the initial estimate of yield distribution. For this purpose, a stochastic logistics model with dynamic yield prediction is formulated. The model determines the flow of migrant workers from sources to sugarcane fronts in order to maximize the total expected contribution, subject to the worker availability constraint at each source. With respected to yield distribution, the expected contribution is defined as the expected revenue from green and burnt cane, minus the cost of labor, the cost of transporting workers from a source to a sugarcane front, and the cost of updating yield distribution. Under certain conditions, the optimization problem becomes convex programming, and a closedform optimal solution can be obtained. In a general case, three heuristic solutions are proposed, and their performances are evaluated in a numerical example: The optimality gap is less than 0.07 percent in a small example. The proposed method is applied to a case study of sugarcane harvest in 46 provinces in Thailand. The monetary value of the dynamic yield prediction is found to be 1.9 million THB, which can be used to justify the investment in yield prediction and precision agriculture technology to reduce the variability of the agricultural yield.

Automated summary

The sugarcane harvest often requires migrant workers, and decision makers can maximize contribution by dynamically updating yield distribution estimates. In a case study of 46 provinces in Thailand, the monetary value of dynamic yield prediction is found to be 1.9 million THB, supporting investment in yield prediction and precision agriculture technology to reduce yield variability.