This tool can be used to perform nieghbourhood-based (i.e. using roving search windows applied to each grid cell) correlation analysis on two input rasters (input1 and input2). The tool outputs a correlation value raster (output1) and a significance (p-value) raster (output2). Additionally, the user must specify the size of the search window (filter) and the correlation statistic (stat). Options for the correlation statistic include pearson, kendall, and spearman. Notice that Pearson's r is the most computationally efficient of the three correlation metrics but is unsuitable when the input distributions are non-linearly associated, in which case, either Spearman's Rho or Kendall's tau-b correlations are more suited. Both Spearman and Kendall correlations evaluate monotonic associations without assuming linearity in the relation. Kendall's tau-b is by far the most computationally expensive of the three statistics and may not be suitable to larger sized search windows.

See Also

image_correlation, image_regression

Function Signature

def image_correlation_neighbourhood_analysis(self, raster1: Raster, raster2: Raster, filter_size: int = 11, correlation_stat: str = "pearson") -> Tuple[Raster, Raster]: ...

Project Links

WbW Homepage User Manual Support WbW