Search

Due to global climate change, flooding is predicted to become more frequent in the coming decades. Recent literature has highlighted the importance of river morphodynamics in controlling flood hazards at the local scale. Abrupt and short-term geomorphic changes can occur after major storms. However, our ability to foresee where and when substantial changes will happen is still limited, hindering our understanding of their ramifications on future flood hazards. This study sought to understand the implications of major storm events for future flood hazards. For this purpose, we developed self-organizing maps (SOMs) to predict post-storm changes in stage‐discharge relationships, based on storm characteristics and watershed properties at 3,101 stream gages across the continental United States (CONUS). We tested and verified a machine learning (ML) model and its feasibility for (1) mapping the variability of geomorphic impacts of extreme storm events and (2) representing the effects of these changes on stage‐discharge relationships at gaged sites as a proxy for changes in flood hazard. The established model allows us to select rivers with stage-discharge relationships that are more prone to change after severe storms, for which flood frequency analysis should be revised on a regular basis so that hazard assessment can be up to date with the changing conditions. Results from the model show that, even though post-storm changes in channel conveyance are widespread, the impacts on flood hazard vary across CONUS. The influence of channel conveyance variability on flood risk depends on various parameters characterizing a particular landscape or storm. The proposed framework can serve as a basis for incorporating channel conveyance adjustments into flood hazard assessment. [ABSTRACT FROM AUTHOR]

The exposure of High Mountain Asia (HMA) to disaster risks is heightened by extreme weather conditions and the impacts of climate change. Obtaining knowledge about the long-term response of the landscape to hydroclimatic variations in HMA is paramount, as millions of people are affected by these changes every year. During monsoons, substantial human suffering, and damage to crops and infrastructure in populated communities result from the flooding and debris flow caused by the increase in precipitation extremes each year. Although a few initiatives have undertaken the estimation of flood risk locally, the use of traditional techniques in ungauged basins is, unfortunately, not always possible because of the lack of extensive data required. To address this problem, we present in this study a geomorphologically guided machine learning (ML) approach for mapping flood impacts across HMA. We defined socioeconomic flood impact using the Lifeyears Index (LYI), a systematic index that measures the economic cost and loss of life caused by flooding. This index quantifies the importance of the destruction to infrastructure, capital, and housing in an overall assessment. We trained the proposed model with over 6000 flood events, from 1980 to 2020, and their computed five-year and ten-year LYIs. We used as predictors, (1) the five-year rainfall concentrations (which correlate the magnitude of precipitation events with the time of occurrence) of events retrieved from ERA5 daily data; (2) a geomorphic classifier (flood geomorphic potential) based on hydraulic scaling functions automatically derived from an 8 and 30-meter digital elevation model (DEM) for the region and (3) population. This model proved capable of identifying the hotspots of flood susceptibility on a national scale and showing its variability from 1980 to 2022. The study also highlights the severity of the impacts of hydroclimatic extremes in the entire HMA region. The framework is generic and can be used to derive a wide variety of flood vulnerability and subsequent risk maps in data-scarce regions. [ABSTRACT FROM AUTHOR]

Wildfire is a critical ecological disturbance in terrestrial ecosystems. Australia, in particular, has experienced increasingly large and severe wildfires over the past two decades while globally fire risk is expected to increase significantly due to the projected increase in fire weather severity and drought condition. Therefore, understanding and predicting fire severity is critical for evaluating current and future impacts of wildfires on ecosystems. Here, we firstly introduce a vegetation16 type specific fire severity classification applied on satellite imagery, which is further used to predict fire severity using antecedent drought conditions, fire weather, and topography of the fire season. Based on a 'leave-one-out' cross-validation experiment, we demonstrate high accuracy for both the fire severity classification and the regression using a suite of performance metrics: determination coefficient (R²), mean absolute error (MPE) and root mean square error (RMSE), which are 0.89, 0.05, and 0.07, respectively. Our results also show that the fire severity prediction results using the vegetation-type specific thresholds could better capture the spatial patterns of fire severity, and has the potential to be applicable for seasonal fire severity forecasting due to the availability of seasonal forecasts of the predictor variables. [ABSTRACT FROM AUTHOR]