North Atlantic hurricanes are one of the weather-related perils that most severely impact insured properties along the US East Coast, and thus represent a key exposure for most reinsurance and insurance-linked security (ILS) portfolios. The hurricane models traditionally used to quantify re/insurance risk account for the effect of fundamental climate circulation features, such as the El Niño Southern Oscillation (ENSO) and the Atlantic Multidecadal Oscillation (AMO). However, the longer-term impact of global warming on hurricane-exposed reinsurance portfolios is still largely unknown. Here, we leverage recent scientific insights and historical records to explore the potential link between global warming and hurricane insured losses. Historical records suggest that the annual frequency of North Atlantic hurricanes does have a material impact on industry losses (rank correlation coefficient ~0.4). However, both models and historical trends seem to indicate no change, or even a slight decrease, in North Atlantic hurricane frequencies in a warmer climate. We also find that a potential increase in the proportion of hurricanes that reach major intensities, expected to be of the order of 10-20% according to the 2021 IPCC report, might lead to a 5-10% increase in industry losses. A similar increase in industry losses might also result from the higher hurricane precipitation rates, which are projected to increase by 11- 28% according to the 2021 IPCC report.  This suggests that the impact of global warming on hurricane insured losses may be significant, albeit not as critical as other fundamental loss drivers, such as urban development, demographic growth, as well as economic and social inflation.

How to cite: Comola, F., Dawson, S., Stahel, M., Paul, H., Märtl, B., and Koller, P.: North Atlantic hurricane activity in a warmer climate: implications for property catastrophe reinsurance, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-11933, https://doi.org/10.5194/egusphere-egu23-11933, 2023.

As recently shown by the storm Vaia that hit Northeastern Italy in the fall of 2018, extreme wind represents a critical weather-related hazard in this region. Over the course of this century, changes in the frequency of extreme windstorms are expected. Obtaining an accurate understanding of wind speed distribution in present and future conditions is thus vital. Robust estimates of the probability of occurrence of extreme winds must be developed and employed to save lives and reduce economic losses. The objective of this study is to develop a novel non-asymptotic statistical method to estimate extreme wind return levels at multiple temporal scales (wind gusts, hourly, daily). Our approach is based on the identification of independent wind storms and estimation of ordinary wind speed events, the latter defined over the Veneto region using multi-year observations from 146 stations. To model the cumulative distribution function of ordinary wind speed events, different parametric distributions are compared, including mixture models specifications. By separately considering storm occurrence and conditional wind speed intensity, the proposed method could improve our understanding of wind speed extremes in the area and provides a tool for projecting future extreme wind speed return levels based on available model simulations.

How to cite: Fathollahzadeh attar, N., Canale, A., and Marra, F.: Extreme windstorm hazard in northern Italy using non-asymptotic statistics, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-731, https://doi.org/10.5194/egusphere-egu23-731, 2023.

On average, the largest registered hailstone per year in Europe has a diameter of about 11cm. Individual hail events can cause losses exceeding one billion Euros. Therefore, the insurance industry is interested in modelling local hail risk. In fact, questions can be as specific as ‘What is the probability that this specific solar panel or roof window gets destroyed by hail within the next year?’

Meteorological modelling on the other hand describes the probability of hail on synoptic scales. Moody’s RMS developed a hierarchy of statistical models downscaling the risk from the large synoptical scale down to individual objects at risk.

I will discuss how the models are designed and calibrated to characterize local risk as well as spatial correlation on various time scales. To make the final model applicable for the insurance industry a thorough validation analysis is performed and results of this validation are shown in this presentation.

How to cite: Grieser, J.: Modelling the Scales of Hail, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-8812, https://doi.org/10.5194/egusphere-egu23-8812, 2023.

Extreme precipitation events can lead to severe negative consequences on society, the economy, and the environment. To mitigate related risks, it is crucial to understand their natural causes. There is a vast number of methods in the literature analyzing their connection to large-scale drivers. Recently there has been much interest in using machine learning (ML) methods instead of traditional statistical models like regression. ML methods are based on algorithms adapting and learning from data. By contrast, regression models are based on theory and assumptions and benefit from domain knowledge for model specification. Because of its adaptability, ML is claimed to offer superior predictive performance than traditional statistical modeling and better manage a greater number of potential predictors. A few studies in climate research have compared the performance between these two approaches, but their conclusions are inconsistent, and some have limitations. 

We used five predictor variables - Geopotential height at 500hPA, Convective available energy (CAPE), Total column water (TCW), Sea Surface Temperature (SST), and Surface Temperature (SAT) using ERA5, the latest reanalysis dataset from ECMWF, and data produced by the Danish Meteorological Institute. All the predictors were not used directly as inputs but were preprocessed before modeling. We trained models using logistic regression (LR) and three commonly used supervised machine learning algorithms - random forests (RF), neural networks (NNET), and support vector machines (SVM) to predict whether an extreme event occurred over Copenhagen. In the LR framework, the predictor variables were modeled using restricted cubic splines to address potential nonlinearity. The training data are highly unbalanced, so using a traditional performance metric such as accuracy (ACC) could be misleading. In light of this, we use performance metrics specialized for unbalanced datasets: the ROC (receiver operating characteristic) curve as the primary measure and the area under the precision-recall curve, the Brier score, and ACC together with the true positive rate and the false positive rate at the optimal threshold as secondary measures.

During the variable selection process, it was found that SST has the weakest relationship with extreme events, and its inclusion did not increase the model performance. Furthermore, the results showed that the LR performs similarly to more complex ML algorithms. SVM had the worst performance in all cases. While most of the top-ranked impacting predictors were nearly comparable amongst models, especially CAPE and TCW, we found discrepancies; SAT contributed to RF and NNET but not to LR.

How to cite: Antoniadou, N., Sørup, H. J. D., Pedersen, J. W., Bülow Gregersen, I., Schmith, T., and Arnbjerg-Nielsen, K.: Comparison of data-driven methods for linking extreme precipitation events to large-scale drivers: A case study from Copenhagen, Denmark, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-5074, https://doi.org/10.5194/egusphere-egu23-5074, 2023.

Extreme precipitation is among the most devastating atmospheric phenomena, causing severe damage worldwide, and is likely to intensify in strength and occurrence in a warming climate. Quantifying the frequency of occurrence of short-duration extreme rainfall exceeding certain amounts is important for hydrologists and urban planners. In Germany, the official design storms are determined from long time series of stations measurements (KOSTRA DWD2010R). Stations, however, only represent a limited region and cannot provide information for ungauged areas. Weather radar networks represent an alternative to overcome these issues, but their time series currently span periods of up to a few decades. Therefore, estimating design precipitation from radar observations with traditional methods is prone to large uncertainties because only few extreme events are included in the statistical analyses. 
Recently, non-asymptotic approaches, such as the simplified metastatistical extreme value, proved promising in estimating design precipitation from short records of sub-daily data, such as the ones from weather radar archives. 
We will present the results obtained applying the simplified metastatistical extreme value approach to the 21-year time series of the German weather radar network, and compare them to traditional estimates from the KOSTRA DWD2010R and to estimates from station observations. Using the simplified metastatistical approach, we find a clear reduction in uncertainty of hourly to daily design precipitation with return periods of 5, 20 and 100 years. Last, we show that the simplified metastatistical approach is less sensitive to particularly extreme events (such as the July 2021 event in Western Germany) than traditional methods.

How to cite: Lengfeld, K. and Marra, F.: Improving estimations of design storms from weather radar observations using a simplified metastatistical extreme value approach, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-7371, https://doi.org/10.5194/egusphere-egu23-7371, 2023.

The statistics of extreme precipitation over a location of interest are crucial for designing hydraulic structures and mitigating extreme events impact. These statistics emerge from (i) the presence of different storm types, (ii) the different intensity of storms of a given type, (iii) the spatial variability of storms during their life-cycle, combined with (iv) the advection of storms across the domain. Explicit separation of these components could help us establish links between atmospheric dynamics (i.e., the occurrence and frequency of different types of storms) and thermodynamics (i.e., the properties of different storm types) on one side, and the emerging statistics of extremes. Here, we make a first step in this direction by focusing on a semi-arid region in the southeastern Mediterranean in which precipitation is almost solely related to convective processes, minimizing the effect of point (i).

We use very-high-resolution (60 m x 60 m, 1 min) weather radar observations to track convective cells during 11 storms that occurred over 2 years (>1200 cells). We mimic rain gauge observation of the tracked cells by sampling the rainfall fields at random locations and we alter advection by applying synthetic velocities to the Lagrangian fields of the cells. This allows us to isolate the impacts of (ii) storm intensity, and (iv) advection. Then, we generate sets of synthetic cells with analogous properties (peak intensity, area, velocity) and different profiles to examine the impact of (iii) spatial variability.

We find that the spatial sampling of convective cells occurred during the 11 storms explains most of the variability of extreme precipitation in the region. The extremes emerging from this sampling are well described by Weibull tails. Return levels estimated from the 11 storms using a non-asymptotic extreme value method are comparable to the ones derived from 25 years of rain gauge observations (error in the 100-year return levels <15%). We discuss the sensitivity of extreme return levels to changes in properties and velocities of the convective cells.

How to cite: Marra, F., Dallan, E., Morin, E., and Armon, M.: Emergence of extreme precipitation statistics from the properties of convective cells, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-5278, https://doi.org/10.5194/egusphere-egu23-5278, 2023.

Design extreme rainfall maps are essential for the construction of many water systems and works, and are typically achieved by regionalizing extreme rainfall statistics from ground-based observations. Different methods are used for such task, where the most popular are kriging and index-based regionalization. In a previous study conducted in Germany, Shehu et al. (2022) revealed that kriging with external drift performed better than index-based regionalization in terms of accuracy (smaller error obtained from cross-validation), however it is still unclear which of the method is superior in terms of precision (wideness of prediction intervals). As the risk may be underestimated due to different sources of uncertainty, a more certain method (in terms of narrower prediction intervals) is preferable (while maintaining a good accuracy). Therefore, the objective of this study is to investigate the propagation of different uncertainty sources for both kriging and index-based regionalization and compare these two in terms of precision and accuracy.

To conduct this study, around 1200 ground-based observations at fine temporal scales (5min) from the German Weather Service (DWD) for whole Germany are employed. For each ground-based observation the annual maximum volumes at different durations (from 5mins up to 7days) are extracted, and local IDF curves are estimated according to Koutsoyiannis et al. (1998). For spatial uncertainty evaluation in the kriging system sequential Gaussian simulation (sGs) together will local sample bootstrapping are employed as shown in Shehu and Haberlandt (2022). On the other hand, the uncertainty in index-based regionalization is evaluated based on a combination of regional sample bootstrapping and spatial simulations of the index. The precision of IDF curves from both methods in terms of 95% confidence interval width is compared on a cross-validation procedure at the locations with more than 40 years of observation.

 The results of this study reveal how the uncertainty of annual rainfall extremes propagates from local estimation to the regionalization of IDF curves based on kriging and index-based regionalization. The comparison of the uncertainty in terms of precision sheds light on which method can produce narrower prediction intervals and hence is more precise in regionalizing IDF curves. Additionally, the accuracy of both methods is advised in order to discuss the advantages and disadvantages of each method for generating spatial IDF curves.

References: 

Koutsoyiannis, D., Kozonis, D. and Manetas, A.: A mathematical framework for studying rainfall intensity-duration-frequency relationships, J. Hydrol., 206(1–2), 118–135, doi:10.1016/S0022-1694(98)00097-3, 1998.

Shehu, B., Willems, W., Stockel, H., Thiele, L., and Haberlandt, U.: Regionalisation of Rainfall Depth-Duration-Frequency curves in Germany, Hydrol. Earth Syst. Sci. Discuss. [preprint], https://doi.org/10.5194/hess-2022-118, in review, 2022.

Shehu, B. and Haberlandt, U.: Uncertainty estimation of regionalised depth–duration–frequency curves in Germany, Hydrol. Earth Syst. Sci. Discuss. [preprint], https://doi.org/10.5194/hess-2022-254, in review, 2022.

How to cite: Shehu, B. and Haberlandt, U.: Comparing uncertainty propagation of different methods for regionalized IDF curves in Germany, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-11147, https://doi.org/10.5194/egusphere-egu23-11147, 2023.

Intensity-Duration-Frequency curves are useful in water resources engineering for the planning and design of hydrological structures such as sewer lines, culverts, drains, dams, dykes.  They provide the mathematical link between the rainfall intensity, I, over a given duration, D, that is expected to be exceeded on average, once every T years (frequency). As opposed to the common use of only extreme data to build IDF curves, here, we use all the non-zero rainfall intensities, thereby making efficient use of the available information. As a parametric model, we use the Extended Generalized Pareto Distribution (EGPD) of Naveau et al. (2016)  for the non-zero intensities. We consider three commonly used approaches for building IDF curves. The first approach is based on the scale-invariance property of rainfall, the second relies on the general IDF formulation of Koutsoyiannis et al. (1998) and the last approach is purely data-driven (Overeem et al., 2008), where the linkage of parameter and duration is empirically determined from data. Using these three approaches, and some extensions around them, we build a total of 10 models for the IDF curves. We then compare them based on their in-sample performance, parsimony in parameterization, as well as their robustness and reliability in a split-sampling cross-validation framework. We consider a total of 81 stations at 10 min resolution in Switzerland.  As a result of the marked seasonality of rainfall in the study area, we adopted a seasonal-based analysis.  The results reveal that the model based on the data-driven approach is the best model. It is able to correctly model the observed intensities across duration while being reliable and robust. It is also able to reproduce the space and time variability of extreme rainfall across Switzerland. While our study focused on Switzerland, the results can be generalized everywhere, especially for locations with high-resolution data availability. To our knowledge, our work is the first to consider using the EGPD in IDF curve modeling.

References

Koutsoyiannis, D., Kozonis, D., & Manetas, A. (1998, April). A mathematic cal framework for studying rainfall intensity-duration-frequency relationships. Journal of Hydrology, 206 (1-2), 118–135. Retrieved 2021-09-06, from https://linkinghub.elsevier.com/retrieve/pii/S0022169498000973 doi: 10.1016/S0022-1694(98)00097-3

Naveau, P., Huser, R., Ribereau, P., and Hannart, A.: Modeling jointly low, moderate, and heavy rainfall intensities without a threshold selection, Water Resour. Res., 52, 2753–2769, https://doi.org/10.1002/2015WR018552, 2016.

Overeem, A., Buishand, A., & Holleman, I. (2008, January). Rainfall depth duration-frequency curves and their uncertainties. Journal of Hydrology, 348 (1-2), 124–134. Retrieved 2021-11-30, from https://linkinghubelsevier.com/retrieve/pii/S0022169407005513 doi:10.1016/j.jhydrol.2007.09.044

How to cite: Haruna, A., Blanchet, J., and Favre, A.-C.: Modeling Intensity-Duration-Frequency curves for the whole range of precipitation, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-2313, https://doi.org/10.5194/egusphere-egu23-2313, 2023.

As Catastrophe models tend to focus more on Fluvial Flood Risk, Pluvial Flood Risk can be sometimes underestimated or neglected. However, when high intensity and short duration events such as hurricanes Ida and Ian occur in urban and semi urban areas, failure to account for Pluvial Flood Risk is consequential. To this end, the importance of accurately estimating Pluvial Flood Risk has strengthened.

In this study, we investigate the sensitivity of Pluvial Flood Risk to design rainfall characteristics. In particular, we explored the trade-off between flood extent and flood depths for different design rainfall durations, as well as the resulting economic losses at different spatial scales covering local, catchment and county levels. The study focuses on urban and semi-urban areas, where besides rainfall duration and intensity, drainage characteristics are expected to play a significant role in Pluvial Flood Risk.

How to cite: Nicotina, L., Moges, E., Sharifian, M., Jankowfsky, S., Li, S., and Hilberts, A.: Design Rainfall controls on Pluvial Flood Risk at different spatial and temporal scales – a U.S. case study, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-13498, https://doi.org/10.5194/egusphere-egu23-13498, 2023.