Search

Sub-daily precipitation extremes are high-impact events that can result in flash floods, sewer system overload, or landslides. Several studies have reported an intensification of projected short-duration extreme rainfall in a warmer future climate. Traditionally, regional climate models (RCMs) are run at a coarse resolution using deep-convection parameterization for these extreme events. As computational resources are continuously ramping up, these models are run at convection-permitting resolution, thereby partly resolving the small-scale precipitation events explicitly. To date, a comprehensive evaluation of convection-permitting models is still missing. We propose an evaluation strategy for simulated sub-daily rainfall extremes that summarizes the overall RCM performance. More specifically, the following metrics are addressed: the seasonal/diurnal cycle, temperature and humidity dependency, temporal scaling and spatio-temporal clustering. The aim of this paper is: (i) to provide a statistical modeling framework for some of the metrics, based on extreme value analysis, (ii) to apply the evaluation metrics to a micro-ensemble of convection-permitting RCM simulations over Belgium, against high-frequency observations, and (iii) to investigate the added value of convection-permitting scales with respect to coarser 12-km resolution. We find that convection-permitting models improved precipitation extremes on shorter time scales (i.e, hourly or two-hourly), but not on 6h-24h time scales. Some metrics such as the diurnal cycle or the Clausius-Clapeyron rate are improved by convection-permitting models, whereas the seasonal cycle appears robust across spatial scales. On the other hand, the spatial dependence is poorly represented at both convection-permitting scales and coarser scales. Our framework provides perspectives for improving high-resolution atmospheric numerical modeling and datasets for hydrological applications.

The increase in extreme precipitation is likely to be one of the most significant impacts of climate change in cities due to increased pluvial flood risk. Hence, reliable information on changes in sub-daily extreme precipitation is needed for robust adaptation strategies. This study explores extreme precipitation over Denmark generated by the regional climate model (RCM) HIRHAM-ECEARTH at different spatial resolutions (8, 12, 25 and 50 km), three RCM from the RiskChange project at 8 km resolution and three RCMs from ENSEMBLES at 25 km resolution at temporal aggregations from 1 to 48 h. The performance of the RCM simulations in current climate as well as projected changes for 2081–2100 is evaluated for non-central moments of order 1–3 and for the 2- and 10-year events. The comparison of the RCM simulations and observations shows that the higher spatial resolution simulations (8 and 12 km) are more consistent across all temporal aggregations in the representation of high-order moments and extreme precipitation. The biases in the spatial pattern of extreme precipitation change across temporal and spatial resolution. The hourly extreme value distributions of the HIRHAM-ECEARTH simulations are more skewed than the observational dataset, which leads to an overestimation by the higher spatial resolution simulations. Nevertheless, in general, under current conditions RCM simulations at high spatial resolution represent extreme events and high-order moments better. The changes projected by the RCM simulations depend on the global climate model (GCM)–RCM combination, spatial resolution and temporal aggregation. The simulations disagree on the magnitude and spatial pattern of the changes. However, there is an agreement on higher changes for lower temporal aggregation and higher spatial resolution. Overall, the results from this study show the influence of the spatial resolution on the precipitation outputs from RCMs. The biases of the RCM simulations increase, and the projected changes decrease for decreasing spatial resolution of the simulations. This points towards the need for high spatial and temporal resolution RCMs to obtain reliable information on changes in sub-daily extreme precipitation.

Design rainfall time series are a critical input for urban flood modeling and require depth-duration-frequency curves along with a design temporal pattern. The association between rainfall depth and duration, which depicts the inherent structure of rainfall patterns, plays an important role in flood-causing potential; this association can be revealed through a bivariate rainfall frequency analysis to obtain the depth-duration-frequency curves. Unlike earlier approaches that use either parametric or nonparametric models, the present study proposes a new semiparametric model approach, which can evaluate all possible combinations of the parametric-nonparametric marginals without restrictions. A comparison between the copula-based bivariate frequency analysis and the generalized least-squares regression shows the former to be better in performance. The first three best-fit models-Gaussian kernel, triangle kernel, and Burr type XII combined with the generalized Pareto distribution-yield consistently similar results, indicating the robustness of the approach. Realistic design rainfall temporal pattern could also be derived from the observed set of temporal patterns using their skew, kurtosis, and bimodality measure to quantify the maximum flood-causing potential and also to generate the design rainfall time series. The proposed framework has been demonstrated for a severely flood-prone coastal megacity, Mumbai, in India. (C) 2015 American Society of Civil Engineers.

Pure radar rainfall, station rainfall and radar-station merging products are analysed regarding extreme rainfall frequencies with durations from 5 min to 6 h and return periods from 1 year to 30 years. Partial duration series of the extremes are derived from the data and probability distributions are fitted. The performance of the design rainfall estimates is assessed based on cross validations for observed station points, which are used as reference. For design rainfall estimation using the pure radar data, the pixel value at the station location is taken; for the merging products, spatial interpolation methods are applied. The results show, that pure radar data are not suitable for the estimation of extremes. They usually lead to an overestimation compared to the observations, which is opposite to the usual behaviour of the radar rainfall. The merging products between radar and station data on the other hand lead usually to an underestimation. They can only outperform the station observations for longer durations. The main problem for a good estimation of extremes seems to be the poor radar data quality.