In this paper, we consider a sensor network consisting of a set of sensors deployed in a two-dimensional region. This network of sensors is required to sense a random scalar field and transport these values to a collector node where this field is reconstructed using this data. The number of times this task can be repeated is termed as the functional lifetime of the sensor network. The maximum mean square error of a snapshot of the field generated at the collector node is the maximum distortion. We consider the problem of sensor placement and routing measurements so as to maximize the functional lifetime and minimize maximum distortion. A centralized gradient descent approach is developed where both quantities are improved at each iteration starting from an initial sensor placement. This scheme requires solving convex programs at each iteration. The centralized algorithm can be used to improve upon any sensor placement strategy. We then consider situations where the initial deployment of sensors cannot be controlled. Post-deployment strategies for each sensor, that require them to run convex programs using only local information, are developed that can be employed to move to a new position so as to increase functional lifetime and decrease maximum distortion. [PDF]