Spatial or temporal data mining tasks are performed in the context of the relevant space, defined by a spatial neighborhood, and the relevant time period, defined by a specific time interval. Furthermore, when mining large spatio-temporal datasets, interesting patterns typically emerge where the dataset is most dynamic. This dissertation is motivated by the need to mine large sensor datasets where a phenomenon is measured at a spatial location over a period of time. In particular, the focus is two-fold where the first is to find spatio-temporal intervals and neighborhoods for the purpose of providing naturally occurring regions in the data for specific time intervals where knowledge discovery tasks can be performed. The second focus is to find subspaces in the form of spatial locations and time periods where the spatio-temporal pattern of the phenomenon being measured is most dynamic. Two approaches to finding spatio-temporal intervals and neighborhoods are presented. "A"gglomerative "S"patio- "T"emporal "I"ntervals and "N"eighborhoods (ASTIN) first applies an agglomerative approach where spatio-temporal intervals are delineated for the time series across all sensors in a sensor network based on between-sensor relationships. Then for each temporal interval, the spatial pattern of the data is found using a graph-based approach. The second approach "M"ultiresolution "S"patio- "T"emporal "I"ntervals and Neighborhoods first finds the spatial neighborhoods of the entire dataset then delineates multiresolution spatiotemporal intervals where the resolution of the interval is based on the amount of spatial change occurring between time steps. The results of ASTIN and MrSTIN are then mined using methods to analyze the connectivity of the spatio-temporal neighborhoods. We then present an approach for the discovery of "Dyna"mic "S"patio- "T"emporal "S"ubspaces in Large Sensor Datasets (DynaSTS). This approach begins by measuring the change in local spatial autocorrelation to track spatial change over time. This provides a global dynamic subspace in the form of spatial locations and time periods. We then drill down into these global dynamic spatial nodes and time periods to find local changes which may not be as widespread. Finally an approach is presented to mine the trajectories and extents of the dynamic spatio-temporal subspaces. These methods are tested on real life datasets including (a) sea surface temperature data from the Tropical Atmospheric Ocean Project (TAO) array in the Equatorial Pacific Ocean and (b) NEXRAD precipitation data from the Hydro-NEXRAD system. The results are validated by ground truths from real life phenomena. We also quantify the results of our approach by performing a hypothesis testing to establish the statistical significance using Monte Carlo simulations. We compare our approach with existing approaches using validation metrics namely spatial autocorrelation and between interval dissimilarity. The results of these experiments show that both ASTIN and MrSTIN indeed identify highly refined spatio-temporal intervals and neighborhoods. Using the DynaSTS approach, we also found promising results in discovering trajectories and extents of highly dynamic subspaces in these datasets depicting several real environmental phenomenon that we validated from various sources as actual events of interest. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]