Threshold computation for daily data
Hi @nperez,
Doing some literature review about methods to compute daily climatologies, I found these two papers discussing the implications of computing daily climatologies/thresholds under scarce historical samples. This has relevant implications when computing indices (results shown in both references). Usually the 30 years of data available are not enough to characterise the climatology of an specific day of the year, providing noise climatologies and unrealistic thresholds. To correct this, several methodologies have been proposed, in practice they apply an smoothenning to the daily time series while preserving as much as possible the internal variability.
- This first work uses a bootstrap resampling procedure. This method has been implemented in the climdex package for temperature thresholds calculations. source
- The second work uses a loess filter before computing the climatologies. source
I was wondering if this problem is tackled in csindicators, looking at the code I was not able to detect a similar procedure.
If it is not implemented, we could discuss how much of an effort it will take and if you find that it is something that the package should incorporate.
Best,