Purpose: This paper aims to address a location-distribution-routing problem for distributing relief commodities during a disaster under uncertainty by creating a multi-stage model that can consider information updates during the disaster. This model aims to create a relief network that chooses distribution centers with the highest value while maximizing equity and minimizing response time. Design/methodology/approach: A hybrid algorithm of adaptive large neighborhood search (ALNS) and multi-dimensional local search (MDLS) is introduced to solve the problem. Its results are compared to ALNS and an augmented epsilon constraint (AUGMECON) method. Findings: The results show that the hybrid algorithm can obtain high-quality solutions within reasonable computation time compared to the exact solution. However, while it yields better solutions compared to ALNS, the solution is obtained in a little longer amount of time. Research limitations/implications: In this paper, the uncertain nature of some key features of the relief operations problem is not discussed. Moreover, some assumptions assumed to simplify the proposed model should be verified in future studies. Practical implications: In order to verify the effectiveness of the designed model, a case study of the Sarpol Zahab earthquake in 2017 is illustrated and based on the results and the sensitivity analyses, some managerial insights are listed to help disaster managers make better decisions during disasters. Originality/value: A novel robust multi-stage linear programming model is designed to address the location-distribution-routing problem during a disaster and to solve this model an efficient hybrid meta-heuristic model is developed.