Following the empirical evidence of a clear correlation between rainfall events and cholera resurgence that was observed in particular during the recent outbreak in Haiti, a spatially explicit model of epidemic cholera is re-examined. Specifically, we test a multivariate Poisson rainfall generator, with parameters varying in space and time, as a driver of enhanced disease transmission. The relevance of the issue relates to the key insight that predictive mathematical models may provide into the course of an ongoing cholera epidemic aiding emergency management (say, in allocating life-saving supplies or health care staff) or in evaluating alternative management strategies. Our model consists of a set of dynamical equations (SIRB-like i.e. subdivided into the compartments of Susceptible, Infected and Recovered individuals, and including a balance of Bacterial concentrations in the water reservoir) describing a connected network of human communities where the infection results from the exposure to excess concentrations of pathogens in the water. These, in turn, are driven by rainfall washout of open-air defecation sites or cesspool overflows, hydrologic transport through waterways and by mobility of susceptible and infected individuals. We perform an a posteriori analysis (from the beginning of the epidemic in October 2010 until December 2011) to test the model reliability in predicting cholera cases and in testing control measures, involving vaccination and sanitation campaigns, for the ongoing epidemic. Even though predicting reliably the timing of the epidemic resurgence proves difficult due to rainfall inter-annual variability, we find that the model can reasonably quantify the total number of reported infection cases in the selected time-span. We then run a multi-seasonal prediction of the course of the epidemic until December 2015, to investigate conditions for further resurgences and endemicity of cholera in the region with a view to policies which may bring to the eradication of the disease in Haiti. The projections, although strongly depending on still uncertain epidemiological processes, show an endemic, seasonal pattern establishing in the region, which can be better forestalled by an improvement of the sanitation system only, rather than by vaccination alone. We thus conclude that hydrologic drivers and water resources management prove central to prediction, emergency management and long-term control of epidemic cholera.