Accessing data in a intermittently connected mobile ad hoc networks is a challenging problem that is caused by frequent network partitions due to node mobility and to impairments of wireless communications. This partitioning pattern is studied by examining the statistics of network partitions for a number of mobility models. This paper then establishes the relationship between the network partitioning pattern and the effectiveness of the data replication scheme, and then gives an upper bound of data availability, achieved by an ideal replication scheme under standard operational conditions. The data availability of a totally random scheme can achieve is also given. Based on these results, a novel replication scheme, RICMAN (replication in intermittently connected mobile ad hoc networks), which takes into account the fact that the network is often partitioned into smaller and uses only intermittent connectivity, thanks to mobile nodes traveling across partition, is proposed. In RICMAN, data items are replicated with the nodes using rather stable neighboring topology and with enough resources. A semi-probabilistic data disseminating protocol is employed to distribute the replicas and propagate the updates and can identify the potential mobile nodes traveling across partitions to maximize data delivery. To maintain replica consistency, a weak, eventual consistency model is utilized to ensure all updates eventually propagate to all replicas in a finite delay. Simulation results demonstrate that RICMAN scheme can achieve high data availability with low overhead. With optimized parameters, the data availability is just 10%～15% lower than the upper bound under certain network conditions.