Stratified random sampling for Landscape Analysis

Random sampling is common in ecology. We often stratify the sample to ensure an equal (or minimum) amount of plots fall within some category (e.g. soils, elevation, cover type, etc.) that is meaningful to the central question of the study. Ecologists often stratify their study area based on landscape facets (Assal et al. 2014). Furthermore, it can be advantageous to use this approach when working with remotely sensed covariates, as segregating the landscape into sub-regions of similar biophysical characteristics can isolate spectral gradients (Homer et al. 2004)

This can be accomplished very efficiently using a raster-based approach in R. In this example, I will randomly sample 5 cells in three strata, then obtain the coordinates of the cell centroid and convert the points to a SpatialPointsDataFrame using the raster library. The code below is very basic, but it can be modified to work with any discrete raster that represents the complex landscape facets of interest.

Plot of a continuous raster layer.

Plot of reclassified raster into three strata (white, yellow, green).

Plot of reclassified raster with sample points.

The sample points can now be projected and/or exported to another format (e.g. shapefile) for further analysis.

Literature Cited

Assal, T.J., Sibold, J., and R. Reich. 2014. Modeling a Historical Mountain Pine Beetle Outbreak Using Landsat MSS and Multiple Lines of Evidence. Remote Sensing of Environment 155:275-288. 

Homer, C., Huang, C., Yang, L., Wylie, B., & Coan, M. 2004. Development of a 2001 Na- tional Land-Cover Database for the United States. Photogrammetric Engineering & Remote Sensing 70:829-840.

Leave a Reply

Your email address will not be published. Required fields are marked *