Winter climate preconditioning of summer vegetation extremes in the northern hemisphere DOI Creative Commons
Mohit Anand, Raed Hamed, Nora Linscheid

et al.

Environmental Research Letters, Journal Year: 2024, Volume and Issue: 19(9), P. 094045 - 094045

Published: Aug. 20, 2024

Abstract The impact of the spring climate on Northern Hemisphere’s summer vegetation activity and extremes has been extensively researched, but less attention devoted to whether how winter may additionally influence in summer. Here, we provide insights into temperature precipitation Hemisphere. To do this, identify positive negative leaf area index (LAI, a proxy for activity) assess effects those using logistic regression at regional scale. Over quarter regions Hemisphere show strong preconditioning LAI extremes, which is typically stronger croplands than forests. In with preconditioning, mediates link between through ecological memory seasonal legacy effects. Our findings suggest that extremely low both forests preconditioned by colder drier winters, while high associated warmer wetter winters. For croplands, winters are an increased likelihood mid-latitude reduced high-latitude regions. Consideration improve our understanding inter-annual variability support agricultural land management practitioners anticipating detrimental crop yields forest conditions.

Language: Английский

Using explainable artificial intelligence (XAI) methods to understand the nonlinear relationship between the Three Gorges Dam and downstream flood DOI Creative Commons

Xikun Wei,

Guojie Wang, Paula Farina Grosser

et al.

Journal of Hydrology Regional Studies, Journal Year: 2024, Volume and Issue: 53, P. 101776 - 101776

Published: April 11, 2024

In the Yangtze River basin of China. The emerging Explainable Artificial Intelligence (XAI) methods provide us an opportunity to understand nonlinear relationship that Deep Learning(DL) model learned inside. construction Three Gorges Dam (TGD) has successfully minimized likelihood flooding in basin. XAI can help know behind it. We apply Long Short Term Memory (LSTM) network, conjunction with two methods, SHapley Additive exPlanation (SHAP) and Expected Gradient (EG), do our work.In DL model, we use YiChang (YC) station runoff,Precipitation (Pre) vapour pressure deficit (VPD) data from middle lower river as input, while output generates runoff at DaTong (DT) station, enable calculate significance each input feature is for generating a model. this study, examine difference importance scores between Before (BTGD) period After (ATGD) period. BTGD period, YC was primary contributor DT station. However, ATGD largest contribution shifted reaches precipitation. Our results suggest show TGD downstream flood clearly effectively mitigate basins by regulating upper work shows potential explain hydrology field.

Language: Английский

Citations

6
Changes in the Climate System Dominate Inter‐Annual Variability in Flooding Across the Globe DOI Creative Commons
Hanbeen Kim, Gabriele Villarini, Conrad Wasko

et al.

Geophysical Research Letters, Journal Year: 2024, Volume and Issue: 51(6)

Published: March 21, 2024

Abstract Extreme flood events have regional differences in their generating mechanisms due to the complex interaction of different climate and catchment processes. This study aims examine capability drivers capture year‐to‐year variability global extremes. Here, we use a statistical attribution approach model seasonal annual maximum daily discharge for 7,886 stations worldwide, using season‐ basin‐averaged precipitation temperature as predictors. The results show robust performance our climate‐informed models describing inter‐annual discharges regardless geographical region, type, basin size, degree regulation, impervious area. developed enable assessment sensitivity changes, indicating potential reliably project changes magnitude

Language: Английский

Citations

5
Improving flood forecast accuracy based on explainable convolutional neural network by Grad-CAM method DOI Creative Commons

Xin Xiang,

Shenglian Guo, Zhen Cui

et al.

Journal of Hydrology, Journal Year: 2024, Volume and Issue: 642, P. 131867 - 131867

Published: Aug. 23, 2024

Language: Английский

Citations

4
An increase in the spatial extent of European floods over the last 70 years DOI Creative Commons
Beijing Fang, Emanuele Bevacqua, Oldřich Rakovec

et al.

Hydrology and earth system sciences, Journal Year: 2024, Volume and Issue: 28(16), P. 3755 - 3775

Published: Aug. 20, 2024

Abstract. Floods regularly cause substantial damage worldwide. Changing flood characteristics, e.g., due to climate change, pose challenges risk management. The spatial extent of floods is an important indicator potential impacts, as consequences widespread are particularly difficult mitigate. highly uneven station distribution in space and time, however, limits the ability quantify characteristics and, particular, changes extents over large regions. Here, we use observation-driven routed runoff simulations last 70 years Europe from a state-of-the-art hydrological model (the mesoscale Hydrologic Model – mHM) identify spatiotemporally connected events. Our identified spatiotemporal events compare well against independent impact database. We find that increase by 11.3 % on average across Europe. This occurs most Europe, except for parts eastern southwestern Over northern mainly driven overall magnitude caused increasing precipitation snowmelt. In contrast, trend central can be attributed heavy precipitation. Overall, our study illustrates opportunities combine long-term consistent regional with detection algorithm large-scale trends key their drivers. detected change should considered assessments it may challenge control water resource

Language: Английский

Citations

4
Fine-tuning long short-term memory models for seamless transition in hydrological modelling: From pre-training to post-application DOI Creative Commons
Xueying Chen, Yuhang Zhang, Aizhong Ye

et al.

Environmental Modelling & Software, Journal Year: 2025, Volume and Issue: unknown, P. 106350 - 106350

Published: Jan. 1, 2025

Language: Английский

Citations

0
Identification of time-varying parameters of a monthly Budyko function in US MOPEX catchments and its implications DOI
Weibo Liu, Pan Liu, Lei Cheng

et al.

Journal of Hydrology Regional Studies, Journal Year: 2025, Volume and Issue: 59, P. 102348 - 102348

Published: April 11, 2025

Language: Английский

Citations

0
Basin-informed flood frequency analysis using deep learning exhibits consistent projected regional patterns over CONUS DOI Creative Commons
Rehenuma Lazin, Giuliana Pallotta, C. Bonfils

et al.

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: April 13, 2025

Climate change poses a significant threat to flood-prone areas by altering precipitation patterns and the water cycle. Here, we analyzed impact of climate on future flood trends. We trained Long Short-Term Memory (LSTM) model estimate long term discharge at 638 river sites over contiguous United States (CONUS) based inputs from gridMET meteorological datasets, downscaled bias-corrected Coupled Model Intercomparison Project 5 (CMIP5) projections. Our results indicate that LSTM can replicate observed with reliable accuracy. The projected changes in magnitude for 10-year 100-year return periods reveal consistent geographical robust across models, increasing trends approximately + 10 40% East West coastal regions decreasing about - 30% Southwestern areas. exhibiting an trend are likely driven increase total seasonal extreme timing amount peak flow. In contrast, result reduction snowpack. To support adaptation planning, developed interactive map providing historical 10- floods selected basins CONUS.

Language: Английский

Citations

0
The Value of Large‐Scale Climatic Indices for Monthly Forecasting Severity of Widespread Flooding Using Dilated Convolutional Neural Networks DOI Creative Commons
Larisa Tarasova, Bodo Ahrens, Amelie Hoff

et al.

Earth s Future, Journal Year: 2024, Volume and Issue: 12(2)

Published: Feb. 1, 2024

Abstract Spatially co‐occurring floods pose a threat to the resilience and recovery of communities. Their timely forecasting plays crucial role for increasing flood preparedness limiting associated losses. In this study we investigated potential dilated Convolutional Neural Network (dCNN) model conditioned on large‐scale climatic indices antecedent precipitation forecast monthly severity widespread flooding (i.e., spatially floods) in Germany with 1 month lead time. The was estimated from 63 years daily streamflow series as sum concurrent exceedances at‐site 2‐year return periods within given across 172 mesoscale catchments (median area 516 km 2 ). trained individually whole country three diverse hydroclimatic regions provide insights heterogeneity performance drivers. Our results showed considerable using dCNN especially length training increases. However, event‐based evaluation skill indicates large underestimation rainfall‐generated during dry conditions despite overall lower these events compared rain‐on‐snow floods. Feature attribution wavelet coherence analyses both indicated difference major drivers regions. While North‐Eastern region is strongly affected by Baltic Sea, North‐Western more global patterns El‐Niño activity. Southern addition detected effect Mediterranean while less important region.

Language: Английский

Citations

3
Extreme and compound ocean events are key drivers of projected low pelagic fish biomass DOI Creative Commons
Natacha Le Grix, William W. L. Cheung, Gabriel Reygondeau

et al.

Global Change Biology, Journal Year: 2023, Volume and Issue: 29(23), P. 6478 - 6492

Published: Oct. 10, 2023

Ocean extreme events, such as marine heatwaves, can have harmful impacts on ecosystems. Understanding the risks posed by events is key to develop strategies predict and mitigate their effects. However, underlying ocean conditions driving severe ecosystems are complex often unknown arise not only from hazards but also interactions between hazards, exposure vulnerability. Marine may be impacted in single drivers rather compounding effects of moderate anomalies. Here, we employ an ensemble climate-impact modeling approach that combines a global fish model with output large simulation Earth system model, identify ecosystem associated most total biomass 326 pelagic species. We show low net primary productivity influential driver extremely over 68% area considered especially subtropics mid-latitudes, followed high temperature oxygen eastern equatorial Pacific latitudes. Severe loss generally driven anomalies at least one driver, except tropics, where combination sufficient drive impacts. Single never biomass. Compound either or necessary condition for 78% ocean, compound variable 61% ocean. Overall, our results highlight crucial role

Language: Английский

Citations

8
Uncovering the Dynamic Drivers of Floods Through Interpretable Deep Learning DOI Creative Commons

Yuanhao Xu,

Kairong Lin,

Caihong Hu

et al.

Earth s Future, Journal Year: 2024, Volume and Issue: 12(10)

Published: Oct. 1, 2024

Abstract The formation of floods, as a complex physical process, exhibits dynamic changes in its driving factors over time and space under climate change. Due to the black‐box nature deep learning, use alone does not enhance understanding hydrological processes. challenge lies employing learning uncover new knowledge on flood mechanism. This study proposes an interpretable framework for modeling that employs interpretability techniques elucidate inner workings peak‐sensitive Informer, revealing response floods 482 watersheds across United States. Accurate simulation is prerequisite provide reliable information. reveals comparing Informer with Transformer LSTM, former showed superior performance peak (Nash‐Sutcliffe Efficiency 0.6 70% watersheds). By interpreting Informer's decision‐making three primary flood‐inducing patterns were identified: Precipitation, excess soil water, snowmelt. controlling effect dominant regional, their impact steps shows significant differences, challenging traditional variables closer timing event occurrence have greater impact. Over 40% exhibited shifts between 1981 2020, precipitation‐dominated undergoing more changes, corroborating change responses. Additionally, unveils interplay among variables. These findings suggest through reverse deduction, transforms data‐driven models from merely fitting nonlinear relationships effective tools enhancing characteristics.

Language: Английский

Citations

2