Precip distribution in Toposub

[joelfiddes]

Not really a bug as behaviour is I think as expected. Map below is monthly precip total over davos with range of c. 90-130mm. I have applied an elevation lapse rate. The distribution of precip is correct (mainly from NW). SE is drier. Tops are wetter valleys are drier. What I think is up though is that there is way too much slope detail. Aspect seems to be comming through here. I would expect to see mainly an elevation gradient with also the NW to SE Precip gradient. I think this is because elev and aspect are weighted the same in toposub. I guess for a variable like precip we just want elevation and x y as dimensions of variability. I think in my Toposub elevation was weighted much higher so if I have few samples the main variability is elevation. I had a weighting factor for each dimension before it went into Kmeans.

[krisaalstad]

This can be a nice idea in general to use different clustering approaches for different variables. It totally makes sense NOT to weight aspect for precip, but weight it for radiation etc...