One way to bring down costs a lot, is to combine on the same grid, regions where the wind over a period of weeks to years, are negatively correlated: that is to say, it helps if we can easily identify pairs of regions where low winds in one region tend to happen at the same time as high winds in the other.

At a scale of days, sufficiently distant regions are uncorrelated. At a scale of weeks to years, a set of correlations of reanalysis data (e.g. ECMWF or CFSR) suggests that there are pairs of regions which do exhibit negatively correlated wind, either seasonally or annually.

Is there a reliable way to identify such combinations of regions?

你可以在很多地方找到合适的代码，例如here.