Rethinking deep learning: linear regression remains a key benchmark in predicting terrestrial water storage

2025/10/12 by Wanshu Nie, Nie, Wanshu, Sujay V. Kumar +15
Environmental Science · #Atmospheric and Oceanic Physics (physics.ao-ph) #FOS: Computer and information sciences #FOS: Physical sciences #Flood Risk Assessment and Management #Geophysics (physics.geo-ph) #Hydrology and Watershed Management Studies #Machine Learning (cs.LG) #Plant Water Relations and Carbon Dynamics

paper · pdf · doi:10.48550/arxiv.2510.10799

openalex publication_date 2025/10/12 · openalex created_date 2025/10/15 · openalex updated_date 2026/07/28

Abstract

Recent advances in machine learning such as Long Short-Term Memory (LSTM) models and Transformers have been widely adopted in hydrological applications, demonstrating impressive performance amongst deep learning models and outperforming physical models in various tasks. However, their superiority in predicting land surface states such as terrestrial water storage (TWS) that are dominated by many factors such as natural variability and human driven modifications remains unclear. Here, using the open-access, globally representative HydroGlobe dataset - comprising a baseline version derived solely from a land surface model simulation and an advanced version incorporating multi-source remote sensing data assimilation - we show that linear regression is a robust benchmark, outperforming the more complex LSTM and Temporal Fusion Transformer for TWS prediction. Our findings highlight the importance of including traditional statistical models as benchmarks when developing and evaluating deep learning models. Additionally, we emphasize the critical need to establish globally representative benchmark datasets that capture the combined impact of natural variability and human interventions.

Citations

• Prithvi-EO-2.0: A Versatile Multi-Temporal Foundation Model for Earth Observation Applications
• SatVision-TOA: A Geospatial Foundation Model for Coarse-Resolution All-Sky Remote Sensing Imagery
• An Abrupt Decline in Global Terrestrial Water Storage and Its Relationship with Sea Level Change
• Deep Time Series Models: A Comprehensive Survey and Benchmark
• Are Language Models Actually Useful for Time Series Forecasting?
• PhilEO Bench: Evaluating Geo-Spatial Foundation Models
• Foundation Models for Generalist Geospatial Artificial Intelligence
• Probing the limit of hydrologic predictability with the Transformer network
• Temporal Fusion Transformers for Streamflow Prediction: Value of Combining Attention with Recurrence
• Why do tree-based models still outperform deep learning on tabular data?
• Are Transformers Effective for Time Series Forecasting?
• Forecast Evaluation for Data Scientists: Common Pitfalls and Best Practices
• On the Opportunities and Risks of Foundation Models
• Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecasting
• Scaling Vision Transformers
• Do We Really Need Deep Learning Models for Time Series Forecasting?
• Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecasting
• Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecasting
• Scaling Laws for Neural Language Models
• Temporal Fusion Transformers for Interpretable Multi-horizon Time Series Forecasting
• Temporal Fusion Transformers for interpretable multi-horizon time series forecasting
• BART: Denoising Sequence-to-Sequence Pre-training for Natural Language\n Generation, Translation, and Comprehension
• Optuna: A Next-generation Hyperparameter Optimization Framework
• Towards Learning Universal, Regional, and Local Hydrological Behaviors\n via Machine-Learning Applied to Large-Sample Datasets
• The Modern-Era Retrospective Analysis for Research and Applications, Version 2 (MERRA-2)
• Attention Is All You Need
• Global river hydrography and network routing: baseline data and new approaches to study the world's large river systems
• Long Short-Term Memory
• LTSM-Bundle: A Toolbox and Benchmark on Large Language Models for Time Series Forecasting
• BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

Related

• Reanalysis and Forecasting of Total Water Storage and Hydrological States by Combining Machine Learning With CLM Model Simulations and GRACE Data Assimilation
• Time Distributed Deep Learning Models for Purely Exogenous Forecasting: Application to Water Table Depth Prediction using Weather Image Time Series
• HydroStartML: A combined machine learning and physics-based approach to reduce hydrological model spin-up time
• Tackling water table depth modeling via machine learning: From proxy observations to verifiability
• SEN2DWATER: A Novel Multispectral and Multitemporal Dataset and Deep Learning Benchmark for Water Resources Analysis