Suppose, I have 7 timeseries, that represent a changing value measured in a point with coordinates x,y. Also I have a group of 20 points distributed spatially within the coverage of these 7 points. Thus I want to get 20 time series, where every value is an interpolated value of initial 7 points on a corresponding moment. Timestep is day. I know that kriging is the best interpolation method in my case. Also I know that kriging interpolation of several points over a regular grid is easy to perform with scikit-learn or pykrige packages. But I want time series (time cycle? wouldnt it work too slowly?) and I want irregular points as target positions of interpolated values, not a regular grid.
So, what is the optimal decision here?
I've seen this theme but there is no time cycling.
On the scheme are shown: points with measured time-series (x,y,value) as o and points which are targets for interpolation as x. scheme of points