Updates to impact yearset module and review of tutorial by ValentinGebhart · Pull Request #1075

Updates to impact yearset module and review of tutorial by ValentinGebhart · Pull Request #1075 · CLIMADA-project/climada_python

This is an example for which the earlier sampling without replacement was not reasonable. One of the impact events has a significantly larger frequency. Once it was drawn in the sampling, the other events were drawn until no events were left. With replacement, the high-frequency event would be drawn more often.

import numpy as np
import climada.util.yearsets as yearsets
from climada.engine import Impact

imp = Impact()
imp.at_event = np.arange(10, 110, 10)
imp.frequency = np.concatenate([np.array([1.]), np.ones(9) * 0.01]) # highly unbalanced frequencies, first event 100 times more likely

# the number of years to sample impacts for (length(yimp.at_event) = sampled_years)
sampled_years = 10

# sample number of events per sampled year
lam = np.sum(imp.frequency)
events_per_year = yearsets.sample_from_poisson(sampled_years, lam, seed=123)
sampling_vect = yearsets.sample_events(events_per_year, imp.frequency, seed=123)
sampling_vect

Result when setting the optional parameter with_replacement=False, similar to the old version

2025-07-22 16:59:12,925 - climada.util.yearsets - WARNING - The frequencies of the different events are not equal. This can lead to distorted sampling if the frequencies vary significantly. To avoid this, please set with_replacement=True to sample with replacement instead.
[array([0]),
 array([], dtype=int64),
 array([7, 1]),
 array([6, 2]),
 array([8]),
 array([5]),
 array([], dtype=int64),
 array([9]),
 array([], dtype=int64),
 array([], dtype=int64)]

Result in the new version

[array([0]),
 array([], dtype=int64),
 array([0, 0]),
 array([0, 0]),
 array([0]),
 array([1]),
 array([], dtype=int64),
 array([0]),
 array([], dtype=int64),
 array([], dtype=int64)]