FF-BERT: A BERT-based ensemble for automated classification of web-based text on flash flood events

<p>In the second half of the 20<sup>th</sup> century, hydrological regimes in central Western Europe were largely characterised by large-scale winter floods. This type of event was predominantly triggered by westerly atmospheric fluxes, bringing moist and mild air masses from the Atlantic Ocean to the European continent. Since the late 1990’s, major flooding events seem to have shifted in time and magnitude. Flash flood events, while being a well-known phenomenon in Mediterranean catchments, are increasingly also reported at higher latitudes. Unlike the large-scale winter flood events, flash floods are of very narrow spatial extension and triggered by rather short, but highly intense rainfall events.</p><p>Here, we focus on the specific case of rivers in Luxembourg that have experienced several flash flood events in recent years, while only small to moderate winter flood events have been reported since the late 1990’s. National hydro-meteorological monitoring and flood forecasting systems have been designed for large-scale floods and are not suited for simulating local flash flood events. Therefore, there is a need to increase our understanding of the hydro-meteorological processes underlying flash flood occurrences in our area of interest.</p><p>While increasing air temperature is known to allow a higher air moisture content that can lead to more intense rainfall events and possible flooding, we moreover hypothesize that the recent increase in flash flood occurrences in Luxembourg is reinforced by a change in atmospheric circulation patterns. To test this hypothesis, we analyse the prevailing atmospheric patterns on rainy days during summer and winter months over the period 1954 - 2019, with a particular focus on rainfall events that lead to moderate and extreme floods. In a next step, we intend to extend our findings for Luxembourg in a larger European context. This analysis should allow to better assess the current situation of hydrological extreme events in central Western Europe in order to take precaution measures and prepare for a diversifying hazard.</p>

More frequent high-intensity, short-duration rainfall events increase the risk of flash floods on steeply sloped watersheds. Where measured data are unavailable, numerical models emerge as valuable tools for predicting flash floods. Recent applications of various hydrological and hydrodynamic models to predict overland flow have highlighted the need for improved representations of the complex flow processes that are inherent in flash floods. This study aimed to identify an optimal modeling approach for characterizing leaf litter losses during flash floods. At a gauged watershed in the Hidegvíz Valley in Hungary, a physical-based model was calibrated using two distinct rainfall–runoff events. Two modeling methodologies were implemented, integrating canopy interception and leaf litter storage, to understand their contributions during flash flood events. The results from the model’s calibration demonstrated this approach’s effectiveness in determining the impact of leaf litter on steep-sloped watersheds. Soil parameters can estimate the behavior of leaf litter during flash flood events. In this study, hydraulic conductivity and initial water content emerged as critical factors for effective parametrization. The findings underscore the potential of a hydrodynamic model to explore the relationship between leaf litter and flash flood events, providing a framework for future studies in watershed management and risk-mitigation strategies.

Floods can be classified in four types based on characteristics of the flood event: flash floods of a few hours duration; single-event floods of longer duration; multiple-event floods; and seasonal floods. Flash floods most frequently occur on small headwater basins in association with short-duration convective, frontal, or orographic type storms with high-intensity rain cells. The economic, social, and environmental effects of flash floods are similar to impacts of “river floods” of other types; the major differences being that the effects of flash floods generally occur with little, if any, warning over a very short time period and are generally limited to relatively small areas. However, the effects of flash floods are often catastrophic and may, depending on development and land use in the affected basins, result in large loss of life because of the sudden onset of the flood events.

<strong class="journal-contentHeaderColor">Abstract.</strong> In recent years, flash floods repeatedly occurred in temperate regions of central western Europe. Unlike in Mediterranean catchments, this flooding behaviour is unusual. In the past, and especially in the 1990s, floods were characterized by predictable, slowly rising water levels during winter and driven by westerly atmospheric fluxes (Pfister et al., 2004). The intention of this study is to link the recent occurrence of flash floods in central western Europe to extreme precipitation and specific atmospheric conditions to identify the cause for this apparent shift. Therefore, we hypothesise that an increase in extreme precipitation events has subsequently led to an increase in the occurrence of flash flood events in central western Europe and all that being caused by a change in the occurrence of flash flood favouring atmospheric conditions. To test this hypothesis, we compiled data on flash floods in central western Europe and selected precipitation events above 40 mm h^&minus;1 from radar data (RADOLAN, DWD). Moreover, we identified proxy parameters representative for flash flood favouring atmospheric conditions from the ERA5 reanalysis dataset. High specific humidity in the lower troposphere (q &ge; 0.004 kg kg^&minus;1), sufficient latent instability (CAPE &ge; 100 J kg^&minus;1) and weak deep-layer wind shear (DLS &le; 10 m s^&minus;1) proved to be characteristic for long-lasting intense rainfall that can potentially trigger flash floods. These atmospheric parameters, as well as the flash flood and precipitation events were then analysed using linear models. Thereby we found significant increases in atmospheric moisture contents and increases in atmospheric instability. Parameters representing the motion and organisation of convective systems occurred slightly more often or remained unchanged in the time period from 1981&ndash;2020. Moreover, a trend in the occurrence of flash floods was confirmed. The number of precipitation events, their maximum 5-minute intensities as well as their hourly sums were however characterized by large inter-annual variations and no trends could be identified between 2002&ndash;2020. This study therefore shows that the link from atmospheric conditions via precipitation to flash floods cannot be traced down in an isolated way. The complexity of interactions is likely higher and future analyses should include other potentially relevant factors such as intra-annual precipitation patterns or catchment specific parameters.

In recent years, flash floods repeatedly occurred in temperate regions of central western Europe. Unlike in Mediterranean catchments, this flooding behaviour is unusual. In the past, and especially in the 1990s, floods were characterized by predictable, slowly rising water levels during winter and driven by westerly atmospheric fluxes (Pfister et al., 2004). The intention of this study is to link the recent occurrence of flash floods in central western Europe to extreme precipitation and specific atmospheric conditions to identify the cause for this apparent shift. Therefore, we hypothesise that an increase in extreme precipitation events has subsequently led to an increase in the occurrence of flash flood events in central western Europe and all that being caused by a change in the occurrence of flash flood favouring atmospheric conditions. To test this hypothesis, we compiled data on flash floods in central western Europe and selected precipitation events above 40 mm h−1 from radar data (RADOLAN, DWD). Moreover, we identified proxy parameters representative for flash flood favouring atmospheric conditions from the ERA5 reanalysis dataset. High specific humidity in the lower troposphere (q ≥ 0.004 kg kg−1), sufficient latent instability (CAPE ≥ 100 J kg−1) and weak deep-layer wind shear (DLS ≤ 10 m s−1) proved to be characteristic for long-lasting intense rainfall that can potentially trigger flash floods. These atmospheric parameters, as well as the flash flood and precipitation events were then analysed using linear models. Thereby we found significant increases in atmospheric moisture contents and increases in atmospheric instability. Parameters representing the motion and organisation of convective systems occurred slightly more often or remained unchanged in the time period from 1981–2020. Moreover, a trend in the occurrence of flash floods was confirmed. The number of precipitation events, their maximum 5-minute intensities as well as their hourly sums were however characterized by large inter-annual variations and no trends could be identified between 2002–2020. This study therefore shows that the link from atmospheric conditions via precipitation to flash floods cannot be traced down in an isolated way. The complexity of interactions is likely higher and future analyses should include other potentially relevant factors such as intra-annual precipitation patterns or catchment specific parameters.

In recent years, flash floods repeatedly occurred in temperate regions of central western Europe. Unlike in Mediterranean catchments, this flooding behaviour is unusual. In the past, and especially in the 1990s, floods were characterized by predictable, slowly rising water levels during winter and driven by westerly atmospheric fluxes (Pfister et al., 2004). The intention of this study is to link the recent occurrence of flash floods in central western Europe to extreme precipitation and specific atmospheric conditions to identify the cause for this apparent shift. Therefore, we hypothesise that an increase in extreme precipitation events has subsequently led to an increase in the occurrence of flash flood events in central western Europe and all that being caused by a change in the occurrence of flash flood favouring atmospheric conditions. To test this hypothesis, we compiled data on flash floods in central western Europe and selected precipitation events above 40 mm h−1 from radar data (RADOLAN, DWD). Moreover, we identified proxy parameters representative for flash flood favouring atmospheric conditions from the ERA5 reanalysis dataset. High specific humidity in the lower troposphere (q ≥ 0.004 kg kg−1), sufficient latent instability (CAPE ≥ 100 J kg−1) and weak deep-layer wind shear (DLS ≤ 10 m s−1) proved to be characteristic for long-lasting intense rainfall that can potentially trigger flash floods. These atmospheric parameters, as well as the flash flood and precipitation events were then analysed using linear models. Thereby we found significant increases in atmospheric moisture contents and increases in atmospheric instability. Parameters representing the motion and organisation of convective systems occurred slightly more often or remained unchanged in the time period from 1981–2020. Moreover, a trend in the occurrence of flash floods was confirmed. The number of precipitation events, their maximum 5-minute intensities as well as their hourly sums were however characterized by large inter-annual variations and no trends could be identified between 2002–2020. This study therefore shows that the link from atmospheric conditions via precipitation to flash floods cannot be traced down in an isolated way. The complexity of interactions is likely higher and future analyses should include other potentially relevant factors such as intra-annual precipitation patterns or catchment specific parameters.

Arid regions are prone to unprecedented extreme rainfall events that often result in severe flash floods. Using near-real-time precipitation data in hydrological modelling can aid in flood preparedness. This study analyzed rainfall data obtained from Integrated Multi-satellite Retrievals for Global Precipitation Measurement (IMERG V. 06) since 2001 to highlight recent trends of extreme rainfall indices for three selected watersheds in the UAE. Additionally, to validate the trends, the present study incorporated CHIRPS (Climate Hazards Group InfraRed Precipitation with Station data) into the analysis. Furthermore, for the first time, this study assessed the performance of the three products of IMERG in modelling flash flood events in the selected watersheds of UAE. A physical-based, fully distributed model was used to simulate the heaviest storm event. Also, a sensitivity analysis of the model’s output to variations in the input parameters was conducted using the one-factor-at-a-time method. The result of the trend analysis indicated that IMERG and CHIRPS show similar trends in both datasets, indicating agreement and reliability in their observations. However, there are a few instances where IMERG and CHIRPS show slight discrepancies in the nature of the trend. In general, the trend analysis results indicated an increasing trend of total precipitation (mm) and consecutive wet days, which suggests a rise in the risk of flash floods. The simulation of the flash flood event showed that the IMERG final product outperformed the other two products, closely matching the model output of the gauge rainfall data with mean absolute error (MAE) of 1.5, 2.37, and 0.5 for Wadi Ham, Wadi Taween, and Wadi Maidaq, respectively. The model’s performance was positively correlated with the size of the watershed. The sensitivity analysis results demonstrated that the model’s output was most sensitive to infiltration parameters. The study’s outcomes provide a good opportunity to improve near-real-time impact evaluation of flash flood events in the watersheds of the UAE.

Abstract. Recent flash flood impact studies highlight that road networks are often disrupted due to adverse weather and flash flood events. Road users are thus particularly exposed to road flooding during their daily mobility. Previous exposure studies, however, do not take into consideration population mobility. Recent advances in transportation research provide an appropriate framework for simulating individual travel-activity patterns using an activity-based approach. These activity-based mobility models enable the prediction of the sequence of activities performed by individuals and locating them with a high spatial–temporal resolution. This paper describes the development of the MobRISK microsimulation system: a model for assessing the exposure of road users to extreme hydrometeorological events. MobRISK aims at providing an accurate spatiotemporal exposure assessment by integrating travel-activity behaviors and mobility adaptation with respect to weather disruptions. The model is applied in a flash-flood-prone area in southern France to assess motorists' exposure to the September 2002 flash flood event. The results show that risk of flooding mainly occurs in principal road links with considerable traffic load. However, a lag time between the timing of the road submersion and persons crossing these roads contributes to reducing the potential vehicle-related fatal accidents. It is also found that sociodemographic variables have a significant effect on individual exposure. Thus, the proposed model demonstrates the benefits of considering spatiotemporal dynamics of population exposure to flash floods and presents an important improvement in exposure assessment methods. Such improved characterization of road user exposures can present valuable information for flood risk management services.

Flash flood events are expected to become increasingly common with the global increases in weather extremes. They are a significant natural hazard that affects karst landscapes, which host large resources of drinking water worldwide. The role played by underground flood events in the karst aquifer recharge is complex due to the heterogeneity of the basement which remains poorly understood. We present the analysis of 20 incave flash flood events affecting the Rochefort karst system (Belgium) using continuous gravity measurements at one single station and water level sensors installed in caves. Underground flood events typically produce a peak in the gravity signal, due to an increase in the associated mass change. After the flood, the gravity values drop but remain slightly increased compared to before the flood event. Via forward gravity modeling, we demonstrate that this remaining anomaly can be reasonably explained by the infiltration of local rainfall within the karst system rather than by allogenic recharge of the aquifer. Flash floods are mainly restricted to connected voids. This allows us to utilize them as proxies to investigate the distribution of cavities in the karst system. Forward modeling of the gravitational attraction induced by the mapped caves being flooded yields a gravity signal much smaller than the observed one. We conclude that at least 50% more cavities than those previously mapped are required to match the measured anomalies. This presents opportunities for implementing similar approaches in other diverse porous media, using gravity monitoring of hydrological processes (e.g., infiltration fronts, hydrothermalism, or tide effects in coastal aquifers) as proxies to characterize underground properties.

Using statistical functions and hydro-hydraulic models to develop human vulnerability curves for flash floods: The flash flood of the Taitou catchment (China) in 2016

The Thomas Fire began on December 4, 2017 and burned 281,893 acres over a 40‐day period in Ventura and Santa Barbara Counties, making it one of California's most destructive wildfires to date. A major rainstorm then caused a flash flood event, which led to the containment of the fire. Both airborne ash from the fire and the runoff from the flash flood entered into the Santa Barbara Basin (SBB). Here, we present the results from aerosol, river, and seawater studies of black carbon and metal delivery to the SBB associated with the fire and subsequent flash flood. On day 11 of the Thomas Fire, aerosols sampled under the smoke plume were associated with high levels of PM2.5, levoglucosan, and black carbon (average: 49 μg/m3, 1.05 μg/m3, and 14.93 μg/m3, respectively) and aerosol metal concentrations were consistent with a forest fire signature. Metal concentrations in SBB surface seawater were higher closer to the coastal perimeter of the fire (including 2.22 nM Fe) than further off the coast, suggesting a dependence on continental proximity rather than fire inputs. On days 37–40 of the fire, before, during, and after the flash flood in the Ventura River, dissolved organic carbon, dissolved black carbon, and dissolved metal concentrations were positively correlated with discharge allowing us to estimate the input of fire products into the coastal ocean. We estimated rapid aerosol delivery during the fire event to be the larger share of fire‐derived metal transport compared to runoff from the Ventura River during the flood event.

Flash flooding affects a growing number of people and causes billions of dollars in losses each year with the impact often falling disproportionally on underdeveloped regions. To inform post-flood mitigation efforts, it is crucial to determine flash flooding extents, especially for extreme events. Unfortunately, flood hazard mapping has often been limited by a lack of observations with both high spatial and temporal resolution. The CubeSat constellation operated by Planet Labs can fill this key gap in Earth observations by providing 3 m near-daily multispectral imagery at the global scale. In this study, we demonstrate the imaging capabilities of CubeSats for mapping flash flood hazards in arid regions. We selected a severe storm on 13–14 August 2021 that swept through the town of Gila Bend, Arizona, causing severe flood damages, two deaths, and the Declaration of a State of Emergency. We found the spatial extent of flooding can be mapped from CubeSat imagery through comparisons of the near-infrared surface reflectance prior to and after the flash flood event (ΔNIR). The unprecedented spatiotemporal resolution of CubeSat imagery allowed the detection of ponded (ΔNIR ≤ −0.05) and flood-affected (ΔNIR ≥ +0.02) areas that compared remarkably well with the 100-year flood event extent obtained by an independent hydraulic modeling study. Our findings demonstrate that CubeSat imagery provides valuable spatial details on flood hazards and can support post-flood activities such as damage assessments and emergency relief.

Simulating flash flood hydrographs and behavior metrics across China: Implications for flash flood management

Comparative hazard analysis of processes leading to remarkable flash floods (France, 1930–1999)

Sentiment analysis of news media articles is essential for understanding the dynamics of conflict and cooperation in transboundary rivers. However, it is not known which machine learning model(s) can best meet the requirement of sentiment analysis for transboundary rivers. This study presents a comparative examination of ten machine learning models commonly used in the field of text sentiment analysis, including K-Nearest Neighbors, Naive Bayes, Support Vector Machine, Decision Tree, Random Forest, Gradient Boosting Decision Tree, Extreme Gradient Boosting, Multilayer Perceptron, Long Short-Term Memory and Bidirectional Encoder Representations from Transformers, for five-class sentiment classification of 9382 news articles (1977–2022) attending to transboundary water conflict and cooperation. By evaluating their performance in terms of accuracy, precision, recall and F1-score, the Bidirectional Encoder Representations from Transformers (BERT) model demonstrated good overall performance and prediction capabilities for news articles with conflictive sentiments. By comparing with the AFINN sentiment dictionary, BERT showed superior performance in the prediction and identification of conflictive sentiment labels. And by validating against historical water events in the three river basins, BERT performed best in the Indus River basin. The findings of this study hold significant implications for government agencies in transboundary rivers, allowing them to promptly assess and respond to public sentiment, thereby preventing water conflict and promoting water cooperation.