Contents lists available at ScienceDirect Computers and Electronics in Agriculture journal homepage: www.elsevier.com/locate/compag Original papers Decision support system for designing sustainable multi-stakeholder networks of grain storage facilities in developing countries E. Essien a, ⁎ , K.A. Dzisi b , Ahmad Addo b a University of Ghana, Legon, Ghana b Kwame Nkrumah University of Science and Technology, Kumasi, Ghana ARTICLE INFO Keywords: Goal programming Grain storage facilities Warehouses Facility location Developing countries ABSTRACT Governments in developing countries often face the daunting task of designing a network of grain storage fa- cilities to simultaneously benefit the various stakeholders (farmers, market traders, exporters, etc.) in the grain supply chain irrespective of the conflicting objectives of these stakeholders. Existing decision support systems either require data that are unavailable in most developing countries or have objectives irrelevant in the context of developing countries. This paper therefore develops a decision support system that integrates transportation, pseudo p-median, forecasting and goal programming models to optimally design networks of grain storage facilities to reduce the transportation cost of respective stakeholders. The effectiveness of the proposed decision support system has been demonstrated by comparing phantom networks developed with the decision support system to the Government of Ghana’s network of 48 grain storage facilities. 1. Introduction The alarmingly high levels (30–50%) of postharvest losses in de- veloping countries compel governments and donor agencies to provide grain storage facilities (McNeil, 2013). These facilities essentially have systems to aid in the efficient processing and storage of agricultural commodities to reduce commodity deterioration. There have however been concerns about the sustainability of these facilities in some de- veloping countries for several reasons. For instance, the climate de- pendent nature of agriculture in these countries makes the facilities highly susceptible to climatic shocks. Thus a minor change in the cli- matic pattern could render the facilities dormant for an entire year. Furthermore, unlike the farmer in the developed country who sizes a storage facility based on the production capacity he/she can guarantee (because of mechanization), grain storage facilities in developing countries are mostly sized based on the aggregated capacity of the cluster of farmers it is supposed to service. There is therefore an in- herent uncertainty in the production capacity of any cluster as most of the smallholder farmers do not have stable production levels. Thus, storage capacities of facilities determined using the aggregated capacity of the cluster may result in high levels of unused capacity. Coulter et al., for instance report of high levels of unused storage capacity in Ghana (Coulter et al., 2000). The resulting insufficient revenue generation makes the storage facilities economically unsustainable. A possible solution to improving the sustainability of the storage facilities is to site them strategically to allow for its concurrent use by other stakeholders. The unused capacity resulting from uncertainties in agricultural production can be ameliorated by incentivizing other sta- keholders (market traders, exporters/importers, etc.) whose capacities are relatively easy to forecast to concurrently use the facilities. This will effectively eliminate any dormancy within the storage facilities as well as improve their sustainability. In a country like Ghana where there is reported unused capacity for grain storage but high demand for storage space for imported commodities such as sugar, rice, and fertilizer, one could site the storage facility to simultaneously suit farmers, market traders, importers and exporters (Coulter et al., 2000). The quantum of possible solutions to the problem of siting facilities to suit all stakeholders precludes the use of human intuition for solving such problems. Several mathematical models have therefore been de- veloped to solve such location problems in agriculture. These models as applied in agriculture are usually multi-faceted. They solve for the optimal locations to site the facilities, determine the catchment areas to be served by respective facilities and the corresponding routing deci- sions (Lucas and Chhajed, 2004). The plurality of features significantly increase the size of most real world location problems found in agri- culture. They are therefore mostly solved with commercial solvers or specialized heuristic algorithms. Another important feature of the lo- cation models found in agriculture is their ability to deal with varia- bility. The models account for variability in production which may occur as a result of climatic changes within and across years. https://doi.org/10.1016/j.compag.2018.02.019 Received 16 May 2017; Received in revised form 14 February 2018; Accepted 16 February 2018 ⁎ Corresponding author. E-mail address: ing.essien@gmail.com (E. Essien). Computers and Electronics in Agriculture 147 (2018) 126–130 0168-1699/ © 2018 Elsevier B.V. All rights reserved. T