Abstract:The rural population of a tropical semi-arid region such as the northeast of Brazil is directly dependent on rainfall for agricultural production, pasture for livestock, and the storage of drinking water for human consumption. The region is characterised by the frequent occurrence of dry years and consecutive dry days (CDD) during the rainy season, demonstrating the vulnerability of rainfed agriculture. The aim of this study therefore, was to identify homogeneous regions of the State of Ceará where there is greater or lesser uncertainty of the rainfall regime. Accordingly, the occurrence of dry and very dry years were investigated as uncertainties between years, together with the occurrence of two classes of CDD (10-20 and 21-30 days) during the rainy season. To define similar regions as to the uncertainty of the rainfall regime, the technique of Hierarchical Cluster Analysis (HCA) was used. The study considered time series of least 30 years, from 77 rain gauge stations around the state of Ceará. The use of HCA resulted in the formation of seven homogenous groups in both of the phenomena being investigated. Municipalities in the Central Backlands, the Inhamuns Backlands and the Jaguaribana Region displayed a higher frequency of both inter-annual (dry years) and intra-annual (CDD) uncertainty, demonstrating the greater vulnerability of rainfed agriculture. The occurrence of a CDD event is not dependent on the total annual rainfall, as several stations with an average rainfall higher than that of the state showed potential for the occurrence of CDD. The number of CDD events recorded in the 11-20 day class was higher than in the 21-30 day class.