Spanning trees rooted at a sink node are often employed for data collection in sensor networks. The node?s double task of sensing and forwarding traffic from other nodes results in energy depletion which is dependent on the constructed spanning tree. The approach taken in this paper is to extend the lifetime of a sensor network by judiciously recalculating the data collection tree to adjust it pro-actively before nodes fail due to total energy depletion. The proposed scheme attempts to make energy-critical nodes leaf nodes, or, as a last resort, to balance the traffic load handled by them. Additionally, an exponentially decaying threshold is used for adaptively characterizing nodes as critical. Simulations demonstrate significantly improved network lifetime but also point to the need to carefully set the thresholds for characterizing nodes as critical.