I have two dimensional spatial (x,y - coordinates of meteo stations) data for small region (so I could neglect the shape of earth globe), for each (x,y) I have one observation of wind direction and speed I need to design metric for clustering problem of such data. It's simple to do when we consider only directions or only the speed of wind but how to combine these quantities ? I try to analyse this problem using R.
Use a flexible clustering algorithm, such as e.g. OPTICS. Where you can put in arbitrary dissimilarity functions. Then define a dissimilarity that uses you value attributes, but filters on the geographic distance, i.e. the dissimilarity is maximal when the objects are geographically distant. Try with different distances to find clusters of different geographical size.