We trained a recurrent neural network model to classify microblogging posts related to urban waterlogging and establish an online monitoring system of urban waterlogging caused by flood disasters. We manually curated more than 4400 waterlogging posts to train the RNN model so that it can precisely identify waterlogging-related posts of Sina Weibo to timely determine urban waterlogging.

We examined maximum rainfall values for different durations from 16 years of radar-based rainfall records for whole Germany. Unlike existing observations based on rain gauge data no clear linear relationship could be identified. However, by classifying all time series, we could identify three similar groups determined by the temporal structure of rainfall extremes observed in the study period. The study highlights the importance of using long data records and a dense measurement network.

To improve the quantitative degree of spatial analysis of vulnerability, we construct grid-scale drought vulnerability curves of European winter wheat based on model simulation, and we discuss their spatial differences through feature points and clustering features. These vulnerability curves show zonal differences, which can be divided into five loss types, and the vulnerability increases from south to north. The results can provide guidance for regionalized risk management.

A landslide-generated tsunami is a complex phenomenon that involves landslide dynamics, wave dynamics and their interaction. This phenomenon threatens numerous lives and infrastructures around the world. To assess this natural hazard, we developed an efficient numerical model able to simulate the landslide, the momentum transfer and the wave all at once. The good agreement between the numerical simulations and physical experiments validates our model and its novel momentum transfer approach.

In previous studies of landslide susceptibility mapping, one inventory is for one kind of landslide. However, this causes some problems for prevention and management. This study aims to map two kinds of landslides and use the results on the same map to explore the potential relationship. Through superimposition of two zoning maps, this provides a new way to evaluate the disaster chain and provides a valuable reference for land use planners.

Rockfalls are a major sources of danger, particularly along transportation routes. Thus, the assessment of their occurrence is a major challenge for risk management. One interesting factor involved in the occurrence of an event is the failure mechanism of rock bridges along the potential failure plane. This work proposes to study the phenomenology of this failure considering numerical modelling. The influence of rock bridge position in regard to the rockfall failure mode is highlighted.

We demonstrate a finding of a very sudden change in the nature of water level extremes in the Gulf of Riga which coincides with weakening of correlation with North Atlantic Oscillation. The shape of the distribution is variable with time; it abruptly changed for several years and was suddenly restored. If similar sudden changes happen in other places in the world, not taking into account the non-stationarity can lead to significant underestimation of future risks from extreme-water-level events.

This paper reviews new observations for the verification of high-impact weather and provides advice for their usage in objective verification. New observations include remote sensing datasets, products developed for nowcasting, datasets derived from telecommunication systems, data collected from citizens, reports of impacts and reports from insurance companies. This work has been performed in the framework of the Joint Working Group on Forecast Verification Research (JWGFVR) of the WMO.

We use high-resolution computer simulations of tropical cyclones to investigate extreme wind speeds over Bangladesh. We show that some northern provinces, up to 200 km inland, may experience conditions equal to or exceeding a very severe cyclonic storm event with a likelihood equal to coastal regions less than 50 km inland. We hope that these kilometre-scale hazard maps facilitate one part of the risk assessment chain to improve local ability to make effective risk management decisions.

In view of the shortage of data in alpine mountainous areas and the difficulty of a single drought index to reflect all the characteristics of drought, this paper constructs a comprehensive drought index (MAHDI) based on the SWAT model and the empirical Kendall distribution function, which connects multiple drought elements. The results show that MAHDI can simultaneously characterize meteorological, agricultural and hydrological drought and has strong applicability and comprehensiveness.