Uncertainty and sensitivity analysis for natural risk and adaptation appraisal modeling with CLIMADA
- 1Institute for Environmental Decisions, ETH Zurich, Universitätstr. 22, 8092 Zurich, Switzerland
- 2Federal Office of Meteorology and Climatology MeteoSwiss, Operation Center 1, P.O. Box 257, 8058 Zurich-Airport, Switzerland
- 3Institute of Geography and Oeschger Centre for Climate Change Research, University of Bern, Switzerland
Modelling societal, ecological, and economic costs of natural hazards in the context of climate change is subject to both strong aleatoric and ethical uncertainty. Dealing with these is challenging on several levels – from the identification and the quantification of the sources of uncertainty to their proper inclusion in the modelling, and the communication of these in a tangible way to both experts and non-experts. One particularly useful approach is global uncertainty and sensitivity analysis, which can help to quantify the confidence in the output values and identify the main drivers of the uncertainty while considering potential correlations in the model. Here we present applications of global uncertainty analysis, robustness quantification, and sensitivity analysis in natural hazard modelling using the new uncertainty module of the CLIMADA (CLIMate ADAptation) platform.
CLIMADA is a fully open-source Python program that implements a probabilistic multi-hazard global natural catastrophe damage model, which also calculates averted damage (benefit) thanks to adaptation measures of any kind (from grey to green infrastructure, behavioral, etc.). With the new uncertainty module, one can directly and comprehensively inspect the uncertainty and sensitivity to input variables of various output metrics, such as the spatial distribution of risk exceedance probabilities, or the benefit-cost ratios of different adaptation measures. This global approach does reveal interesting parameter interplays and might provide valuable input for decision-makers. For instance, a study of the geospatial distribution of sensitivity indices for tropical cyclones damage indicated that the main driver of uncertainty in dense regions (e.g. cities) is the impact function (vulnerability), whereas in sparse regions it is the exposure (asset) layer.
CLIMADA: https://github.com/CLIMADA-project/climada_python
(1) Aznar-Siguan, G. et al., GEOSCI MODEL DEV. 12, 7 (2019) 3085–97
(2) Bresch, D. N. and Aznar-Siguan., G., GEOSCI MODEL DEV. (2020), 1–20.
How to cite: Kropf, C. M., Ciullo, A., Meiler, S., Otth, L., McCaughey, J. W., and Bresch, D. N.: Uncertainty and sensitivity analysis for natural risk and adaptation appraisal modeling with CLIMADA, EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-1773, https://doi.org/10.5194/egusphere-egu21-1773, 2021.
