Climate attribution for extreme events. A statistically based approach.

Weather data observed all over the world show an increase in the frequency of extreme events, leading to higher economical losses and numbers of victims. It is thus crucial to investigate causes of such trends, and future evolutions thereby. A method for climate attribution is to assess as to whether anthropogenic climate change affected either the probability of occurrence/or the magnitude of extreme events. Results thereby are fundamental to link extreme event impacts and global warming, and could potentially be used to assess responsible subjects, under the perspective e.g. of setup of compensation policies. Attribution studies so far have shown large uncertainty, especially concerning weather events affected by other drivers than temperature (e.g. precipitation, drought, snow-fall events). As a first step to perform an attribution study one has to identify/quantify a trend of one or more variable(s)/indexes affecting the event under analysis.

Here, we applied statistical methods to identify potential trends, and to back-attribute them to global warming. Using synthetically generated a-priori known, trend-affected series of meteorological variables (P,T, etc..), we (try and) back-trace the presence/magnitude of trends, and try and verify measurable changes of the statistical (extreme values) distribution thereby, for the purpose of robust  climate attribution.  

The goal is to quantify how much results of an attribution study depend upon data type (ground based, climate models, etc..), and accuracy thereby, and upon (robustness of) the trend detection method applied in the analysis.

For an application to a real world case study, we selected variables of interest (precipitation/snowfall extremes) in the Alps of Italy, and we tested the results of the methodology, by assessing trend presence/magnitude of extreme events distributions’ parameters, and robustness thereby.