The capability of the Wireless Sensor Network (WSN) to perform the information processing and energy saving tasks and their effectiveness largely depends on ensuring stability of network properties. Unpredictable nature of WSN and at the same time strict application requirements call for the incorporation of the maintenance mechanisms. The aim of this thesis is to design a set of maintenance techniques for tackling the most common and important problems impacting WSN. This thesis defines four distinct maintenance problems and presents efficient and robust solutions. The considered problems include: Localized Energy Hole Profiling - mapping of an uneven energy distribution in the network, Topology Oriented Maintenance - finding and remedying topology irregularities, Distributed k-Connectivity Maintenance - assuring communication reliability by providing low resource cost localized k-connectivity with latency guarantees and Adaptive Spatial Sampling - dynamic adaptation of spatial sampling resolution in order to match the sampling to the dynamics of monitored phenomena. The presented in this thesis algorithms for tackling the above outlined problems constitute together Maintenance Framework for WSN. Localized Energy Hole Profiling: WSNs display non-uniform energy usage distribution, induced by non-uniform distribution of communication traffic or sensing activities, which manifests itself as energy holes. Energy holes are direct source of the network partitioning and sensing voids. The thesis presents new distributed energy profiling algorithms for generalized types of energy holes. The algorithms search for boundary nodes of energy hole and use them as a reference to calculate the energy needs of sensor nodes within the energy hole. These, when aggregated, create angular and radial energy profiles. Topology Oriented Maintenance: Sparse WSN networks even while connected, usually suffer from topology irregularities that negatively impact the network lifetime and responsiveness, i.e., sensor data delivery reliability and latency. The thesis directly targets the problem by proposing algorithms that use the discrepancy between Euclidean and hop distances, to provide in-network and localized strategy that efficiently (i) discovers generic topology irregularities, and (ii) identifies locations for minimal number of new augmented sensor deployments to remedy topology irregularities and sustain the desired operational requirements. Distributed k-Connectivity Maintenance: A common approach for providing reliability in WSN is to assure global k-connectivity. This property guarantees that the failure of up to k - 1 sensor nodes does not cause network partitioning. Thesis develops a technique that allows for localized, sustainable maintenance, capable of efficiently restoring the WSN desired k-connectivity. The approach is based on assuring that each sensor node in its routing tree has at least k direct or indirect neighbors, which are placed closer to the sink than sensor node itself. Adaptive Spatial Sampling: A prominent application of WSNs is the monitoring of physical phenomena. The monitored phenomena often tend to have unknown spatial distributions, that also change over time resulting in either under- or over-sampling of signals in space. The thesis outlines a Voronoi based adaptive spatial sampling algorithm, which aims at minimizing measured signal variation at neighboring Voronoi sensor nodes. This approach generates additional new sampling locations in the under-sampling regions to fulfill specified accuracy requirements. The effectiveness and efficiency of all outlined in this thesis algorithms were tested through a set of extensive simulations.