From data perspectives only, three kind of spatial pattern can be detected.
(1) Value questions: where is the hotspot. This can be detected by visualization;
(2) change questions: where variable change rapidly. This can be detected by gradient plot;
(3) Association questions: "whether areas of high values overlap significantly for 2 variables".

Currently, I am working to firstly find the locations where two variables exceed their p quantile value (e.g. p=70%,..,90%), then calculate the location similarity as indicators of how the two hotspots overlap on a certain quantile value. Finally plot the similarity vs quantile value to find interesting indications.

For example, in the plot for Manitowoc4 path, blacks points means the overlap level of BAT hotspot and total hotspot. yellow points means the overlap level of BAT coldspot and total coldspot. Red points means the overlap level of BAT hotspot and total coldspot, and green points means the overlap level of BAT coldspot and total hotspot. In the plot, yellow points is very high, indicates the distribution of low BAT value is always consistent with the distribution of low total value. however, the black points increase as quantile increase, which indicates only very high value of BAT is correlated with very high value of total.