search

ISI-MIP / attrici

Produce counterfactual climates for ISIMIP.
GNU General Public License v3.0
9 stars 2 forks source link

Most extreme events are often not adjusted #73

Closed matthiasmengel closed 3 years ago

matthiasmengel commented 4 years ago

For example for precipiation at lat = 69.75, lon = -139.75 there is one extreme rainfall event around the year 1920. This event fails to be estimated in quantile mapping as scipy.stats.distribution most probably cannot map it. As we map inf and nan values from quantile mapping to the original values, this event stays unadjusted.

This is a problem as the most extreme events are of key interest, and just leaving them as they are takes them out for attribution assessments. image

SimonTreu commented 4 years ago

The Gamma-distribution might not be optimal to model high precipitation amounts. Some approaches use a different distribution for precipitation amounts that exceed a certain threshold.