In this study, we investigate the effect of climate change on tropical cyclone (TC) induced compound flooding and impacts for TC Idai, making landfall in Mozambique in 2019. TCs are one of the most damaging extreme events and are challenging to attribute using conventional, probabilistic methods. We develop a storyline attribution framework including a state-of-the-art modelling chain that combines hydrological, coastal, flood and impact models to simulate the changes in flooding and its impact under factual and counterfactual scenarios, with a plausible range of the climate trend removed from TC rainfall, maximum wind speed and sea level rise (SLR). For the case of TC Idai, we find that climate trends in SLR and wind-driven storm surge lead to the largest increase in flood damage of 6 %–27 %, compared to the counterfactual, while causing the smallest increase in flood volume and flood extent of 0.2 %–0.8 % and 0.1 %–0.6 %, respectively. Climate trends in rainfall lead to the largest increase in flood volume and flood extent of 5 %–19 % and 1 %–4 %, respectively, but account for the smallest increase in flood damage of 2 %–8 %. Changes in all drivers combined lead to about the same increase in flood volume and flood extent as the rain-only scenario (5 %–19 % and 1 %–5 %, respectively) but the largest increase in flood damage of 8 %–35 %. A non-linear relationship between flood hazard and flood damage results in a stronger climate footprint on TC impacts than hazards. Assessing the combination of all climate change-affected flood drivers is crucial for obtaining a comprehensive view on the effect of climate change. The attribution framework presented in this paper is applicable for TC-prone regions across the globe and can be applied in data-scarce, often highly impacted and vulnerable areas which are currently underrepresented in attribution studies.

This study focuses on exploring the potential of using Long Short-Term Memory networks (LSTMs) for low-flow forecasting for the Rhine River at Lobith on a daily scale with lead times up to 46 days ahead. A novel LSTM-based model architecture is designed to leverage both historical observation and forecasted meteorological data to carry out multi-step discharge time series forecasting. The feature and target selection for this deep learning (DL) model involves evaluating the use of different spatial resolutions for meteorological forcing (basin-averaged or subbasin-averaged), the impact of incorporating past discharge observations, and the use of different target variables (discharge Q or time-differenced discharge dQ). Then, the model is trained using the ERA5 dataset as meteorological forcing, and employed for operational forecast with ECMWF seasonal forecast (SEAS5) data. The forecast results are compared to a benchmark process-based model, wflow_sbm. This study also explores the flexibility of the DL model by fine-tuning the pretrained model with limited SEAS5 dataset. Key findings from feature and target selection include: (1) opting for subbasin-averaged meteorological variables significantly improves model performance compared to a basin-averaged approach. (2) Utilizing dQ as the target variable greatly boosts short-term forecast accuracy compared to using Q, with a mean absolute error (MAE) of 25 m³ s⁻¹ and mean absolute percentage error (MAPE) of 0.02 for the first lead time, ensuring reliability and accuracy at the onset of the forecast horizon. (3) While incorporating historical discharge improves the forecasting of Q, its impact on predicting dQ is less pronounced for short lead times. In the operational forecast with SEAS5, compared to the wflow_sbm model, the DL model exhibits skill in forecasting low flows as evidenced by Continuous Ranked Probability Skill Score (CRPSS) median values of all lead times above zero, and better accuracy in forecasting drought events within short lead times. The wflow_sbm model shows higher accuracy for longer lead times. In the exploration of fine-tuning approach, the fine-tuned model generates marginal short-term enhancements in forecasting low-flow events over a non-fine-tuned model. Overall, this study contributes to advancing the field of low-flow forecasting using deep learning approach.

State-of-the-art flood hazard maps in coastal cities are often obtained from simulating coastal or pluvial events separately. This method does not account for the seasonality of flood drivers and their mutual dependence. In this article, we include the impact of these two factors in a computationally efficient probabilistic framework for flood risk calculation, using Ho Chi Minh City (HCMC) as a case study. HCMC can be flooded subannually by high tide, rainfall, and storm surge events or a combination thereof during the monsoon or tropical cyclones. Using long gauge observations, we stochastically model 10,000 years of rainfall and sea level events based on their monthly distributions, dependence structure and cooccurrence rate. The impact from each stochastic event is then obtained from a damage function built from selected rainfall and sea level combinations, leading to an expected annual damage (EAD) of $1.02 B (95th annual damage percentile of $2.15 B). We find no dependence for most months and large differences in expected damage across months ($36–166 M) driven by the seasonality of rainfall and sea levels. Excluding monthly variability leads to a serious underestimation of the EAD by 72–83%. This is because high-probability flood events, which can happen multiple times during the year and are properly captured by our framework, contribute the most to the EAD. This application illustrates the potential of our framework and advocates for the inclusion of flood drivers' dynamics in coastal risk assessments.

Compound weather and climate events are combinations of climate drivers and/or hazards that contribute to societal or environmental risk. Studying compound events often requires a multidisciplinary approach combining domain knowledge of the underlying processes with, for example, statistical methods and climate model outputs. Recently, to aid the development of research on compound events, four compound event types were introduced, namely (a) preconditioned, (b) multivariate, (c) temporally compounding, and (d) spatially compounding events. However, guidelines on how to study these types of events are still lacking. Here, we consider four case studies, each associated with a specific event type and a research question, to illustrate how the key elements of compound events (e.g., analytical tools and relevant physical effects) can be identified. These case studies show that (a) impacts on crops from hot and dry summers can be exacerbated by preconditioning effects of dry and bright springs. (b) Assessing compound coastal flooding in Perth (Australia) requires considering the dynamics of a non-stationary multivariate process. For instance, future mean sea-level rise will lead to the emergence of concurrent coastal and fluvial extremes, enhancing compound flooding risk. (c) In Portugal, deep-landslides are often caused by temporal clusters of moderate precipitation events. Finally, (d) crop yield failures in France and Germany are strongly correlated, threatening European food security through spatially compounding effects. These analyses allow for identifying general recommendations for studying compound events. Overall, our insights can serve as a blueprint for compound event analysis across disciplines and sectors.

Current global riverine flood risk studies assume a constant mean sea level boundary. In reality high sea levels can propagate up a river, impede high river discharge, thus leading to elevated water levels. Riverine flood risk in deltas may therefore be underestimated. This paper presents the first global scale assessment of the joint influence of riverine and coastal drivers of flooding in deltas. We show that if storm surge is ignored, flood depths are significantly underestimated for 9.3% of the expected annual population exposed to riverine flooding. The assessment is based on extreme water levels at 3433 river mouth locations as modeled by a state-of-The-Art global river routing model, forced with a multi-model runoff ensemble and bounded by dynamic sea level conditions derived from a global tide and surge reanalysis. We first classified the drivers of riverine flooding at each location into four classes: surge-dominant, discharge-dominant, compound-dominant or insignificant. We then developed a model experiment to quantify the effect of surge on flood hazard and impacts. Drivers of riverine flooding are compound-dominant at 19.7% of the locations analyzed, discharge-dominant at 69.2%, and surge-dominant at 7.8%. Compared to locations with either surge-or discharge-dominant flood drivers, locations with compound-dominant flood drivers generally have larger surge extremes and are located in basins with faster discharge response and/or flat topography. Globally, surge exacerbates 1-in-10 years flood levels at 64.0% of the locations analyzed, with a mean increase of 11 cm. While this increase is generally larger at locations with compound-or surge-dominant flood drivers, flood levels also increase at locations with discharge-dominant flood drivers. This study underlines the importance of including dynamic downstream sea level boundaries in (global) riverine flood risk studies.