Erratic rainfall, within and between years, leads to high uncertainty in runoff reservoir operations in many Sub-Sahara African (SSA) countries. Severe food shortages attributed to drought affect millions of vulnerable households, particularly those whose livelihoods depend on agriculture. We construct and successfully simulate a stochastic dynamic model of a small runoff reservoir fashioned after the Tono Reservoir in Northern Ghana. Our model considers a reservoir authority who at the beginning of each season observes the stock of water in the reservoir and, in the wet season, the amount of rainfall, and must decide how much water to release from the reservoir for irrigation and how much acreage to irrigate in order to maximize returns to the agriculture over the three growing seasons. We drive optimal irrigation policies with and without access to emergency groundwater pump irrigation systems and with and without access to rainfall index insurance. The optimal irrigation policies including reservoirs operation policy, deficit irrigation management, and the available water resource allocation could be used to provide decision support for water resources management. Besides, the strategies obtained could help with the risk analysis of reservoirs operation stochastically.