Table 3 Overview of strongly conditioned (storyline) event attribution methods
From: The concept of spectrally nudged storylines for extreme event attribution
Strongly Conditioned (Storyline) Event Attribution Methods | ||
|---|---|---|
Attribution Methods | How does it work? | Advantages/Disadvantages |
Statistical | The dynamic and thermodynamic components of a spatial timeseries (e.g., daily maps of surface temperature) are isolated statistically | No climate model simulations are needed The dynamic component is available for separate analysis, e.g., in terms of its long-term trends The thermodynamic component is estimated as a residual and may contain unidentified dynamic components A target variable must be chosen, restricting the method to univariate extremes Full 3-D fields are not available, limiting the use for impact studies |
Analogues | The most similar circulation patterns (analogues) to a particular event are selected to represent the dynamical conditions | Observed changes in the frequency of the analogue states provide an estimate of the observed changes in the likelihood of the dynamical conditions conducive to the event, providing a bridge to probabilistic attribution Good analogues may not always be found |
Dynamical Adjustment | Dynamically induced variations are statistically removed to isolate the anthropogenic climate change signal | The nature of the dynamical trend is not immediately interpretable |
Modelled | A physical model is run in factual and counterfactual mode, with the circulation constrained in some way | Not restricted to the historical record so can consider unseen events The role of changing circulation is not directly estimated The model may have biases |
Pseudo Global Warming | Differences between global simulations for various climates are added to reanalysis data to constrain the boundary conditions for regional extreme event simulations | Very similar in spirit to spectrally nudged storylines, but cannot be applied globally and suffers from the limitations of limited area models driven only from the boundary conditions |
Short-term Forecasts | Historical extreme event forecasts with weather models are re-forecast under counterfactual conditions | Can use existing operational forecast systems Can represent any extreme that can be forecast Target events need to be identified Only useful for short-term events Interpretation of simulations is challenging because of statistical non-stationarity |
Spectrally Nudged Storylines | Extreme events are simulated for different climates using similar large-scale circulation | Output is statistically homogeneous in space and time, so analysis is easy Events do not need to be predefined; different users can use different event definitions from the same simulations Set-up is easy to implement Small-scale events are not nudged and may differ between storylines The nudging needs to be adjusted |
