I'm looking for an effective way to measure changes in spatial location in a time series data set (whether an observation repeats in the same location between successive time-steps, or if it shifts and by how much).

What I have is a point file, showing the presence/absence of a species in four stages, in study region A. I have a similar dataset for four other regions (B-D). I am looking for an approach to measure how much movement there is in region A over the four time-steps, and if this is greater, lesser, or equal to comparable movements in regions B-D. Each region can be run as a separate analysis.

I am using ArcGIS (10.1 and 10.2), as well as GME. I don't find any out of the box solution to what I'm trying to show. The closest is to use a summary measure of point-distance. The closest theoretical approach I have found is Colwell's (1974) concept of temporal/spatial constancy, used in Ecology.

If anyone has suggestions or knows of a tool in ecological modeling for measuring constancy, this would be very helpful.


You could try a matrix regression to derive the temporal correlations (ie., Partial Mantel Test on two pairwise-distance matrices or Mantel Test on single, cross-distance matrix) or you could apply Dutilleul's (1993) modified t-test.

Sorry, I just can't think of an "out-of-box" solution available in ArcGIS or GME. You may want to take a look at the "Spatial Analysis in Macroecology" or "PASSaGE" GUI software which are both designed for pattern and spatial statistical analysis. There may be a suitable tool available in one of them. I know that Michael Rosenberg is consistently adding new functionality to PASSaGE and am fairly sure that there is a Mantel test available.

If you are expecting large deviations, you could also address change in autocorrelation structure using the family of point pattern analysis (PPA) statistics. In this case a appraoch may be the Ripley's-K (Ripley 1976), standardized as Besag's-L (Besag & Diggle 1977), where you can look at spatial structure by distance lag. Whereas, this is available in ArcGIS, there is only a univariate implementation and you are needing a bivariate cross K. Lynch & Moorcroft (2008) did propose a spatiotemporal extension of the Ripley's-K.

I would note that it would be critical to perform a Monte Carlo simulation thus, producing a simulation envelope, to ensure that the results are significant.


Besag, J. & Diggle, P.J. (1977). Simple Monte Carlo tests for spatial pattern, Applied Statistics, 26:327–333

Dutilleul, P. (1993). Modifying the t-test for assessing the correlation between two spatial processes. Biometrics 49:305-314.

Lynch, H.J., & P.R. Moorcroft (2008) A spatiotemporal Ripley’s K-function to analyze interactions between spruce budworm and fire in British Columbia, Canada. Canadian Journal of Forest Research, 38:3112–3119

Ripley, B.D. (1976). The second-order analysis of stationary point processes, Journal of Applied Probability, 13:255–266

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.