Retrieval-Augmented Water Level Forecasting for Everglades

2025-08-06

AI Summary

To address the lack of effective adaptation mechanisms and poor generalizability across heterogeneous wetland environments in Everglades water-level forecasting, this paper proposes the first hydrology-oriented retrieval-augmented time-series foundation model. Without fine-tuning, our method retrieves analogous hydrological events from an external historical observation repository—using dual criteria of dynamic time warping similarity and mutual information—and injects matched segments as contextual prompts into a pre-trained foundation model, enabling cross-basin and cross-temporal adaptive prediction. Experiments on long-term, multi-site observational data from the Everglades demonstrate substantial improvements in short- to medium-term water-level forecasting accuracy (average MAE reduced by 23.6%) and strong generalization across diverse hydrological regimes. The code and datasets are publicly released.

Technology Category

• Machine Learning: Other Foundations of Machine Learning
• Cognitive Modeling & Cognitive Systems: Other Foundations of Cognitive Modeling & Systems
• Computer Vision: Other Foundations of Computer Vision

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Accurate water level forecasting is crucial for managing ecosystems such as the Everglades, a subtropical wetland vital for flood mitigation, drought management, water resource planning, and biodiversity conservation. While recent advances in deep learning, particularly time series foundation models, have demonstrated success in general-domain forecasting, their application in hydrology remains underexplored. Furthermore, they often struggle to generalize across diverse unseen datasets and domains, due to the lack of effective mechanisms for adaptation. To address this gap, we introduce Retrieval-Augmented Forecasting (RAF) into the hydrology domain, proposing a framework that retrieves historically analogous multivariate hydrological episodes to enrich the model input before forecasting. By maintaining an external archive of past observations, RAF identifies and incorporates relevant patterns from historical data, thereby enhancing contextual awareness and predictive accuracy without requiring the model for task-specific retraining or fine-tuning. Furthermore, we explore and compare both similarity-based and mutual information-based RAF methods. We conduct a comprehensive evaluation on real-world data from the Everglades, demonstrating that the RAF framework yields substantial improvements in water level forecasting accuracy. This study highlights the potential of RAF approaches in environmental hydrology and paves the way for broader adoption of adaptive AI methods by domain experts in ecosystem management. The code and data are available at https://github.com/rahuul2992000/WaterRAF.

Problem

• Improving water level forecasting accuracy in the Everglades
• Adapting deep learning models for hydrology applications
• Enhancing predictions using historical hydrological data retrieval

Innovation

• Retrieval-Augmented Forecasting for hydrological data
• External archive enhances historical pattern integration
• Similarity and mutual information methods compared