In this work, we propose a scheme, named LECRAD, for the re-construction of temporarily lost connectivity in the presence of dumb nodes in stationary wireless sensor networks. A sensor node is termed as “dumb”, when it continues its physical sensing, but fails to communicate due to shrinkage in communication range, typically attributed to adverse environmental effects such as rainfall, fog and high temperature. On the resumption of favorable environmental conditions, the node starts to behave normally. So, dumb behavior is dynamic in nature. Such behavior of a node leads to network partitioning and node isolation, which results in disruption of connected topology. Therefore, the proposed scheme in this paper reconstructs the lost connectivities between nodes by activating intermediate sleep nodes or by adjusting the communication range of the sensor nodes, while there is no neighbor node within the reduced communication range. In the proposed scheme, a learning automata-based approach is used for activation of intermediate sleep nodes or adjustment of communication range of isolated or intermediate nodes to decrease the message overhead and energy consumption of the network. Simulation results show that the proposed scheme, LECRAD, exhibits better energy efficiency and message overhead than that of the recently proposed topology management protocols, LETC and A1, if they are applied in such scenario.