In the semiarid region of Northeast Brazil rivers are intermittent. The annual cycle is divided into two parts: in the first half of the year, the occurrence of significant inflows is possible; in the second half of the year, there is no river flow. As such, in the second half of the year, available surface water comes solely and exclusively from reservoirs. Knowing the volume of stored water in July is the same as knowing the surface water availability for this period. Each July the regional water users meet with the state water agencies to decide the water macroallocation in a participatory process. Use licenses with specific quantities are decided at such a meeting based on the reservoir storage levels. Macroallocation uses such information to make a deterministic decision for a 6-month period (July-December). Irrigation and urban water use are the primary competing uses, and the peak irrigation demand occurs in the second half of the year, while the urban demand is relatively constant over the year. The water allocation problem would be solved by the above deterministic approach if there was no need for multi-annual supply regularization. The recognition that in many years reservoir inflows can be close to zero leads to the following question: how much water should I keep in the reservoir at the end of the year to meet future year demands? The attempt to answer this question in the absence of an ability to forecast the future inflows, leads to a strong restriction to the volume allocable over the six-month time horizon. Historically the answer to this question was to keep at least the volume of water sufficient to meet the non-rationed human demands in the following year in the reservoir, assuming that the reservoir inflows in the following year will be zero. The operation plan horizon would be then increased from 6 months to 18 months under a deterministic formulation. This allocation pattern is associated with an optimization aimed at Maximizing the minimum expected benefit (MAXMIN). The worst natural conditions observed were those of no annual inflows. Thus, the lowest benefits would be associated with this nature conditions. Thereby, it would only be necessary to maximize the benefits under such nature conditions to achieve the MAXMIN results. This maximization would require the use of water for irrigation as soon as possible (because of its demand curve linearity), and make human water supply rationing as low and continued as possible over the horizon under analysis (because of the strong decrease of marginal benefits to higher demand satisfaction levels). This approach minimizes an aspect of the climate risk, however it may lead to a high regret level, because both human supply rationing and irrigation in July-December period could prove fully unnecessary after the reservoir recharge in the period of January-June of the following year. Long-term benefits are lower under MAXMIN than under other possible approaches. Adopting a methodology that would maximize the benefits over a long-term analysis, or that would otherwise reduce regret at every application is desirable. Such approaches would be more subject to risk than MAXMIM. The maximum desired risk could be related to those objectives in such a way that the higher the risk the greater the water availability. Scenarios that are planned for this analysis may use the historical series or the climate forecast for inflows. The objective of forecasts is to reduce the error of use of historical series (climate) when estimating the expected value and reducing the dispersion (variance) of inflows adopted for the construction of scenarios. A forecast is considered good when it manages to reduce the climate error. This paper presents a methodology to integrate seasonal climate forecasting in water macroallocation in Ceará State, Brazil. The key steps adopted are reservoir inflow forecasts. These are used with a reservoir operation model using network flow whose results can be shared in the participatory allocation process. By making the risks explicit, transactions or negotiations across user groups are facilitated. These transactions reflect the risk sensitivity of each user and their willingness to pay or act to reduce their risk. The individual risk sensitivity is hard to assess by an outside group except through an analysis of transactions. However, the transactions can facilitate movement towards a macro allocation process that is better informed and does not require the elicitation of preferences and utility by state planners to perform a top down allocation or impact analysis. Mechanisms for the implementation of this process in a regulated setting are discussed and illustrated with simulations.