translate model output back to geographical units

[JannisHoch]

to assess the accuracy of the predictions, it is necessary to convert back the boolean information to the corresponding geographical unit (now: water province).

[JannisHoch]

@wande001, question to be asked: at the moment, we throw all areas on one big pile (multiplied by the number of simulation years this yields a big number of values) and then try to find 'one model to rule them all'.

would it not make more sense to fit a model per polygon? this model could then be based more on the local specifics rather on correlations for the entire continent. moreover, we would immediately see for which polygon the model results in good performance in for which ones not.

[JannisHoch]

answer to comment above: no, we will stick to establishing relations based on all water provinces and not per water province individually. mostly because the model will never project conflicts in areas where no conflicts occured thus far, even though environmental/climate conditions may change.

[JannisHoch]

in 79fd7d6 splitting/adding geometry information per data point was achieved.

now needed to correctly combine the predicted conflicts with the geometry information used as input to the predict function. the data can then be plotted (somehow).