Randomization (‒p)

Top<N> randomization

For each candidate trajectory, the score is computed and a random trajectory in the top 'N' matches is selected. The following uses one of the top 5 trajectories for each match:

wxgen sim -db db_agg.nc -n 1000 -t 365 -w 1,2.5,0.001 -p top5 -o top5.nc

This is the default randomization policy.

Weighted randomization

This picks a random trajectory, but with a weighted probability given by the score. That is, a segment with twice the score of another segment will have twice the probability of being selected. This gives more randomization than the top<N> approach, however it will allow poor matches to be chosen (however with low probability).

wxgen sim -db db_agg.nc -n 1000 -t 365 -w 1,2.5,0.001 -p random -o random.nc

Unweighted randomization

This strategy ignores the score and selects a random segment within the same climate regime. This can be achieved by not specifying any weights.

wxgen sim -db db_agg.nc -n 1000 -t 365 -o noweight.nc