Wireless sensor networks (WSN) consist of large numbers of sensor nodes embedded in physical space for continuously sensing/monitoring a specified physical phenomenon. Typically, WSN requires spatially dense sensor deployment to satisfactorily respond to the requirements of the application. The spatial proximity of nodes and the temporal behaviour of the sensing operation causes the observations sensed to be associated spatio-temporally. However, due to unfavourable environmental conditions and low energy resources available in WSN, there are instances when the nodes lie with in the same spatial space and measure the same attribute yet the observations sensed by them are not related. No mechanism is present that identifies the degree of association or disassociation between them.In this paper we propose a methodology that uses concepts from computational geometry to identify and isolate relevant spatially related subsets from a given universal set of nodes in WSN, and validate the existence of spatio-temporal associations among the observations sensed by them. The node identification process is achieved by employing classical object intersection solutions from computational geometry and the validation process is achieved by using similarity verification methods. By means of simulations on real time data we demonstrate the procedure for identification and validation by using Jaccard's coefficient, Cosine similarity and Pearson product correlation coefficient.