ARTICLE | doi:10.20944/preprints202210.0167.v1
Subject: Engineering, Civil Engineering Keywords: compound flooding; D-vine copula; trivariate joint analysis; Bernstein estimator; Beta kernel estimator; parametric copulas; kernel density estimation; return periods
Online: 12 October 2022 (09:04:04 CEST)
Low-lying coastal communities are often threatened by compound flooding (CF), which can be determined through the joint occurrence of storm surges, rainfall and river discharge either successively or in close succession. The trivariate distribution can demonstrate the risk of the compound phenomenon more realistically rather than considering each contributing factor independently or in a pairwise dependency. Recently vine copula has been recognized as the highly flexible approach to constructing a higher dimensional joint density framework. In such construction, parametric class copula with parametric univariate marginal distributions is often involved. Such incorporation can lack flexibility due to parametric functions with prior distribution assumptions about their univariate marginal and/or copula joint density. This study introduces the vine copula approach in a nonparametric setting by introducing Bernstein and Beta kernel copula density in establishing trivariate flood dependence. The proposed model is applied to 46 years of flood characteristics collected on the west coast of Canada. The univariate flood marginal distribution is modelled using nonparametric kernel density estimation (KDE). The 2-D Bernstein estimator and Beta kernel copulas estimator are tested independently in capturing pairwise dependencies to establish D-vine structure in a stage-wise nesting approach in three alternative ways, each by permutating the location of the conditioning variable. The best-fitted vine structure is selected using goodness-of-fit (GOF) test statistics. The performance of the nonparametric vine approach is also compared with the vine constructed in the parametric and semiparametric fitting procedure. Investigation reveals that the D-vine constructed using Bernstein copula with normal KDE marginals nonparametrically performed well in capturing dependence of the compound events. Finally, the derived nonparametric model is used in the estimation of trivariate OR- and AND-joint return periods, further employed in estimating failure probability (FP) statistics. The trivariate return periods for the AND-joint case are higher than for the OR-joint case for the same flood combination. Also, the trivariate flood hazard results in a high-value FP than bivariate and univariate events. Ignoring the trivariate dependence could result in the underestimation of FP
ARTICLE | doi:10.20944/preprints202206.0259.v1
Subject: Engineering, Civil Engineering Keywords: compound flooding event; vine copula; trivariate joint analysis; joint return period; conditional return period; hydrologic risk
Online: 20 June 2022 (05:17:56 CEST)
The interaction between oceanographic, meteorological, and hydrological factors can result in an extreme flooding scenario in the low-lying coastal area, called compound flooding (CF) events. For instance, rainfall and storm surge (or high river discharge) can be driven by the same meteorological, tropical or extra-tropical cyclones, resulting in a CF phenomenon. The trivariate distributional framework can significantly explain compound events' statistical behaviour reducing the associated high-impact flood risk. Resolving heterogenous dependency of the multidimensional CF events by incorporating traditional 3-D symmetric or fully nested Archimedean copula is quite complex. The main challenge is to preserve all lower-level dependencies. An approach based on decomposing the full multivariate density into simple local building blocks via conditional independence called vine or pair-copulas is a much more comprehensive way of approximating the trivariate flood dependence structure. In this study, a parametric vine copula of a drawable (D-vine) structure is introduced in the trivariate modelling of flooding events with 46 years of observations of the west Coast of Canada. This trivariate framework searches dependency by combining the joint impact of annual maximum 24-hr rainfall and the highest storm surge and river discharge observed within the time ±1 day of the highest rainfall event. The D-vine structures are constructed in three alternative ways by permutation of the conditioning variables. The most appropriate D-vine structure is selected using the fitness test statistics and estimating trivariate joint and conditional joint return periods. The investigation confirms that the D-vine copula can effectively define the compound phenomenon's dependency. The failure probability (FP) method is also adopted in assessing the trivariate hydrologic risk. It is observed that hydrologic events defined in the trivariate case produce higher FP than in the bivariate (or univariate) case. It is also concluded that hydrologic risk increases (i) with an increase in the service design life of the hydraulic facilities and (ii) with a decrease in return periods.
ARTICLE | doi:10.20944/preprints202204.0086.v1
Subject: Engineering, Civil Engineering Keywords: Flood hazard; CaMa-Flood; Flood Map Viewer; Floodplain mapping; Flood risk; North American Regional Reanalysis; Property exposure
Online: 11 April 2022 (03:35:28 CEST)
Flood events and their associated damages have escalated significantly in the last few decades. To add to the gruesome situation, many reports and studies warn that flood risk would aggravate significantly in future periods due to significant alterations in the climate patterns and socio-economic dynamics. Floodplain mapping is looked upon as a viable option to tackle this global issue as it provides both quantitative and qualitative information on flood dynamics. Moreover, with the increasing availability of global data and enhancement in computational simulations, it has become easier to simlate flooding patterns at large scales. This study deter-mines the usability of publicly available datasets in capturing flood hazards over Canada. Run-off data set from the North American Regional Reanalysis (NARR), along with a few other rele-vant inputs are fed to CaMa-Flood, a robust global hydrodynamic model to generate flooding patterns for 1 in 100 and 1 in 200-yr return period events over Canada . The simulated maps are compared and validated with the existing maps of a few flood-prone regions in Canada, thereby establishing their performance over both regional and country-scale. Later, the simulated flood-plain maps are used in conjunction with property related information at 34 cities (within the top 100 populous cities in Canada) to determine the degree of exposure due to flooding in 1991, 2001, and 2011. The results indicate that around 80 percent of inundated spots belong to high and very-high hazard classes in a 200-yr event, which is roughly 4 percent more than simulated for 100-yr event. NARR derived floodplain maps perform very well while compared over the six flood-prone regions. While analyzing the exposure of properties to flooding, we notice an in-crease in the number during the last three decades, with the maximum rise observed in Toronto, followed by Montreal, and Edmonton. To disseminate the extensive flood-related information, a web-based public tool, Flood Map Viewer (http://www.floodmapviewer.com/) is developed. The development of the tool was motivated by the commitment of the Canadian government to provide $63 M over the next three years for the completion of flood maps for higher-risk areas. The study reaches out to demonstrate how publicly available datasets can be utilized with a lesser degree of uncertainty in representing flooding patterns over large regions. The flood re-lated information derived from the study can be used along with vulnerability for quantifying flood risk, which will help in developing appropriate pathways for resilience building for long-term sustainable benefits.