CORTEXA
← Browse
openalexWater2026-07-25

Physics-Informed CNN-LSTM for Street-Scale Urban Flood Prediction: Reconciling Aggregate Accuracy and Street-Level Plausibility

Luc D’Costa, Yidi Wang, Jonathan L. Goodall, Rohan Chandra

Deep learning surrogate models trained with mean-squared-error loss produce statistically accurate but physically unconstrained flood predictions: water may flow uphill, appear spontaneously, or smooth over street-level corridors. In this work, a physics-informed training framework is developed for CNN-LSTM models that predict urban flood depths at 15 min intervals over a 128×128 spatial grid. Three differentiable penalty terms are embedded directly into the loss function: (i) a gravity loss that penalizes depth increases against the water-surface-elevation gradient, (ii) a continuity loss enforcing local mass conservation with rainfall-adaptive thresholds, and (iii) a topography-aware false-alarm penalty modulated by the topographic wetness index (TWI). The framework is evaluated on the Norfolk, Virginia, flood dataset spanning two major storm events (August 2017 and September 2022) comprising 300 samples, with all variants trained on identical splits and robustness assessed over repeated random splits and leave-one-storm-out tests. A road-proximal evaluation restricted to a TWI-derived street mask quantifies street-level skill. The physics-constrained model achieves near-zero gravity violations (∼10−6) and the highest street-channel recall (0.77 ± 0.09 versus 0.44 ± 0.10 for the unconstrained baseline), the capability most relevant to downstream traffic routing, and its recall advantage more than doubles on a held-out storm, while a uniform false-alarm variant attains 16% lower mean absolute error but suppresses street recall to 0.25. The proposed TWI-modulated penalty reconciles this trade-off: it improves upon the uniform variant on every metric measured, recovering 60% higher street recall at the lowest MAE among all constrained variants and the best street-level F1 score. These results expose a fundamental tension between aggregate pixel-level error metrics and application-specific physical plausibility, and demonstrate that terrain-aware loss modulation offers a principled resolution.

Related papers

openalexWater2026-07-26

A Machine Learning Approach for Water Quality Assessment in the Lower Rio Grande Valley Watershed

Saika Nowshin Nowrin, Chu‐Lin Cheng, Jungseok Ho, Jinwoo An, Fatemeh Nazari

Water quality analysis plays an essential role in maintaining the health and sustainability of river ecosystems, especially in semi-arid regions like the Arroyo Colorado Watershed in South Texas. Since the river is a vital source of water supply for local communities, agriculture…

crossrefWater2026-07-23

Modelling Shallow Groundwater Level Fluctuations in Very Flat Landscapes Based on Satellite Data and Machine Learning

Javier Houspanossian, Francisco Diez, Raul Rivas, Esteban Jobbagy, Mauro Holzman, Gabriëlle J. M. De Lannoy

Groundwater level fluctuations play a critical role in shaping hydrological extremes in flat sedimentary landscapes, where shallow water table depth (WTD) and strong surface-subsurface connectivity modulate the impacts of floods and droughts. The Western Pampean Plain (Argentina)…

openalexWater2026-07-23

Interannual Responses of Common Reed (Phragmites australis) to Fluctuating Water Flows Entering the Ili River Delta, Kazakhstan

Sabir T. Nurtazin, Steven G. Pueppke, Ruslan Salmurzauli, Niels Thevs, Altynbek Mirzakul, Azim Baibagyssov, et al.

Kazakhstan’s Ili River delta nourishes a unique wetland ecosystem in arid Central Asia. The delta is dominated by common reed [Phragmites australis (Cav.) Trin. ex Steud.], an ecologically and economically significant species that is sensitive to water levels. We used machine lea…

crossrefWater2026-07-14

Integrated Satellite-Derived Bathymetry and Morphodynamic Assessment for Regulated River Monitoring Using Machine Learning and Sentinel-2 Data

Ahmed Nour-Eldeen, Rofyda Abdelrehem, Alban Kuriqi, Ismail Abd-Elaty, Hickmat Hossen

This study presents an integrated, data-driven framework for satellite-derived bathymetry and morphodynamic assessment in large, regulated rivers, providing a spatial database to support reach-scale hydromorphological monitoring and river management. Satellite-derived bathymetry…

crossrefWater2026-07-02

Uncovering the Drivers of Greenhouse Gas Emissions from Hydropower Reservoirs in China Based on Machine Learning

Haixia Li, Qiang Liu, Xiaolin Tang, Lian Ai, Hongqiao Chen, Jie Xiong, et al.

China is expanding hydropower capacity as a key climate change mitigation strategy, yet greenhouse gas (GHG) emissions from reservoirs can substantially offset this benefit. The influence of specific environmental drivers on these emissions remains poorly understood, and previous…

crossrefWater2026-04-14

Benchmarking Machine Learning and Deep Learning Models for Groundwater Level Prediction in Karst Aquifers: The Dominant Role of Hydrogeological Complexity

Qingmin Zhu, Yinxia Zhu, Jie Niu, Jinqiang Huang, Fen Huang, Xiangyang Zhou, et al.

Karst aquifers present unique challenges for groundwater level prediction due to their dual-porosity structures and highly nonlinear hydrological responses. This study systematically evaluates nine machine learning and deep learning models (RF, XGBoost, LSTM, CNN, Transformer, N-…