AquaCast: Urban Water Dynamics Forecasting with Precipitation-Informed Multi-Input Transformer

2025-09-11

AI Summary

Urban hydrodynamic forecasting faces challenges in effectively fusing endogenous variables (e.g., water level, flow rate) with exogenous factors (e.g., historical and forecasted precipitation) and suffers from low modeling efficiency. To address this, we propose a Transformer-based multi-input multi-output deep learning model. Our key innovation lies in introducing an embedding layer that directly encodes external variables—bypassing separate prediction of exogenous inputs—and thereby captures cross-dimensional dependencies among heterogeneous time series. Leveraging self-attention mechanisms, the model jointly processes multi-source temporal data. Evaluated on the real-world LausanneCity dataset and three large-scale synthetic benchmarks, it consistently outperforms state-of-the-art methods in short-term urban flood risk prediction. The model demonstrates strong robustness to input perturbations and scalability to varying data scales and variable dimensions, enabling high-accuracy, operationally viable forecasting for smart city water management.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Planning, Routing, and Scheduling: Planning with Language Models
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

This work addresses the challenge of forecasting urban water dynamics by developing a multi-input, multi-output deep learning model that incorporates both endogenous variables (e.g., water height or discharge) and exogenous factors (e.g., precipitation history and forecast reports). Unlike conventional forecasting, the proposed model, AquaCast, captures both inter-variable and temporal dependencies across all inputs, while focusing forecast solely on endogenous variables. Exogenous inputs are fused via an embedding layer, eliminating the need to forecast them and enabling the model to attend to their short-term influences more effectively. We evaluate our approach on the LausanneCity dataset, which includes measurements from four urban drainage sensors, and demonstrate state-of-the-art performance when using only endogenous variables. Performance also improves with the inclusion of exogenous variables and forecast reports. To assess generalization and scalability, we additionally test the model on three large-scale synthesized datasets, generated from MeteoSwiss records, the Lorenz Attractors model, and the Random Fields model, each representing a different level of temporal complexity across 100 nodes. The results confirm that our model consistently outperforms existing baselines and maintains a robust and accurate forecast across both real and synthetic datasets.

Problem

• Forecasting urban water dynamics using multi-input deep learning
• Incorporating precipitation history and forecast reports
• Capturing inter-variable and temporal dependencies effectively

Innovation

• Multi-input transformer model with precipitation embedding
• Captures inter-variable and temporal dependencies simultaneously
• Eliminates exogenous forecasting via embedding layer fusion