This publication contains spatial data, tabular data and scripts used to analyze the spatial patterns of refugia and associated plant communities following each of several fires in northern New Mexico. Four of the geotiff files were derived during the project (*Kernel.tif) using dNBR (delta Normalized Burn Ratio) or dNDVI (delta Normalized Difference Vegetation Index). The kernel raster data represent density of unburned/low severity grid cells in approximately 10-hectare neighborhoods following the Cerro Grande, Dome, La Mesa, and Las Conchas fire events in 2000, 1996, 1977, and 2011, respectively. The data were produced using a kernel smooth process, with output values range from 0 to 1, representing a gradient in neighborhood density of refugia. In addition, geotiff files of the dNBR for Las Conchas (this version is not available at mtbs.gov, but was provided for the study by S. Howard, USGS-EROS), the dNDVI for La Mesa and the La Mesa footprint (both developed for the Fire atlas for the Gila and Aldo Leopold Wilderness Areas project; https://doi.org/10.2737/RDS-2015-0023) are also included. Finally, the archive contains a digital elevation model (developed by USGS-EROS), cropped to the study area; the DEM was used to derive terrain metrics describing topographic heterogeneity at local and catchment scales. The text files contain R scripts and associated tabular data that can be used to repeat the analysis presented in the publication by performing the following functions: 1) generate the kernel rasters (kernel geotiffs described, above); 2) generate terrain metrics from DEM (geotiff included), 3) sample the kernel rasters, terrain metric outputs and 1 kilometer resolution bioclimatic data (downloaded from https://adaptwest.databasin.org/pages/adaptwest-climatena); 4) develop environmental models from the raster sample data (text file included); and 5) conduct a multivariate analysis of species and communities using species data recorded in the field (text file included).