Extreme events tools
Extreme values statistics:
Following tools are available through R using:
install.packages('...')for CRAN packages (Renext, SpatialExtremes, …),devtools::install_github('IRSN/...')(Windows: assuming devtools & Rtools are available on your system)install.packages('https://github.com/IRSN/.../releases/download/....tgz|zip|tar.gz', repos=NULL, type='binary')(precompiled for Linux (x86), Windows, or MacOSX)
| Name | Short description | Location | Supports unsteady data | Processing historical data | Sampling type | Distribution laws | Covariables | Censored data | Data types | Confidence interval type | Other tool features |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Renext | Statistical tool for estimating extreme values in a frequentist framework with many methods | CRAN + IRSN GitHub | ✘ | ✔ | POT block maxima r-LOS | GPD, expo, GEV, gumbel, mixed expo, log normal, Weibull, gamma (any with given parameters) | ✘ | ✔ | events, yearly, struct. | Delta | Various anova statistical tests |
| NSGEV | Statistical tool for estimating extreme values in an unsteady frequentist framework based on annual maxima | IRSN GitHub | ✔ | ✘ | maximum block | GEV, gumbel | x(t) | ✘ | yearly | Delta Likelihood Profile Bootstrap | Various anova statistical tests |
| potomax | Statistical tool for estimating extreme values with a new sampling method (MA censoring threshold) | IRSN GitHub | ✘ | ✔ | POT block maximum | GPD, expo, GEV, gumbel | ✘ | ✔ | events, yearly, struct. | Delta Likelihood Profile | Validation method to check the convergence of results (argument check=TRUE) |
| SpatialExtremes | Fork of the CRAN package by Mathieu Ribatet by Yves Deville: spatgev class to take into account temporal covariates | Y. Deville GitHub | ✔ (using it with NSGEV) | ✘ | maximum block | GEV, gumbel | Delta Likelihood Profile Bootstrap | Package to extrapolate temperatures in places where there is data (spatial covariates) | |||
| bever | Statistical tool for estimating extreme values in a Bayesian framework (with focus on Markov chain Monte-Carlo) | IRSN GitHub | ✘ | ✔ | POT block maximum | GPD, expo, GEV, gumbel | ✘ | ✘ | events, yearly | Credibility intervals with HPD and equal tails methods | Calculation of the predictive distribution |
| beverstan | Rstan statistical tool, dependent on NSGEV and bever, for the estimation of extreme values in a Bayesian framework | IRSN GitHub | ✔ | ✔ | maximum block | GEV, gumbel | x(t) | ✔ | yearly | Credibility intervals with HPD and equal tails methods | |
| dailymet | Tool for analyzing daily weather series (mainly temperatures) using POT methods | IRSN GitHub | ✔ | ✘ | POT | GPD, exhibition | x(t), x | ✘ | daily | Functions to describe seasonality and annual trend | |
| SeaLev | Estimation of extreme sea levels by combining tide and surge | ✘ | ✔ (with Renext) | POT or maximum block (premiums) | Renext laws | ✘ | ✘ | tidely | Delta | Theoretical tide distribution |
Extreme events quantization
Sample/cluster extreme events (like flooding maps) to keep representativeness of rare probability / high impact:
