Abstract:Spatiotemporal forecasting (STF) extends traditional time series forecasting or spatial interpolation problem to space and time dimensions. Here, we review the statistical, physical and artificial intelligence (AI) methods, data and model uncertainties, predictability and future directions for STF problems. Statistical STF methods have limitations in high-level feature extractions and long-term memory modeling. Physical models are computationally intensive and are imperfect in model structure and parameterization. AI models lack the interpretability and require elaborate training but can model complex nonlinear and non-Gaussian problems. Integrating datadriven and physical model-driven methods could facilitate the improvement of interpretability and forecasting accuracy. The predictive uncertainty comes from data and models, which could be measured by probability distribution and Bayesian inference, respectively. The predictive uncertainty is generally missing in AI models and could be resolved by incorporating Bayesian frameworks. The predictability of dynamic earth systems is spatiotemporally heterogeneous and is generally examined by diagnostic and prognostic approaches. Diagnostic methods analyze the predictability empirically from a theoretical perspective, while prognostic methods investigate the predictability through real experiments. Unraveling the predictability in space and time and the predictability sources will greatly improve earth system understanding and operational forecasting development. Current STF systems are largely not user-friendly to provide probabilistic and understandable forecasting services in near real-time. Intelligent STF systems should automatically prepare various data sources, train the models in a self-adaptative way and provide timely predictive information services for users to make decisions. This review provides state-of-the-art advances in forecasting sciences and highlights new directions for new-generation STF systems.

Towards Learning in Grey Spatiotemporal Systems: A Prophet to Non-consecutive Spatiotemporal Dynamics

Interpretable and Generalizable Spatiotemporal Predictive Learning with Disentangled Consistency

STONet: A Neural-Operator-Driven Spatio-temporal Network

Multi-Factor Spatio-Temporal Prediction based on Graph Decomposition Learning

Towards Dynamic Spatial-Temporal Graph Learning: A Decoupled Perspective

Combining machine learning with knowledge-based modeling for scalable forecasting and subgrid-scale closure of large, complex, spatiotemporal systems

STUaNet: Understanding uncertainty in spatiotemporal collective human mobility

A Multi-graph Fusion Based Spatiotemporal Dynamic Learning Framework.

Towards Unifying Diffusion Models for Probabilistic Spatio-Temporal Graph Learning

LSTGCN: Inductive Spatial Temporal Imputation Using Long Short Term Dependencies

SAMSGL: Series-aligned multi-scale graph learning for spatiotemporal forecasting

DGSTD: Learning on Dynamic Graph with Spatio-Temporal Disentanglement

SAMSGL: Series-Aligned Multi-Scale Graph Learning for Spatio-Temporal Forecasting

Machine Learning for Spatiotemporal Sequence Forecasting: A Survey

Less is More: a New Machine Learning Methodology for Spatiotemporal Systems

Learning Time-aware Graph Structures for Spatially Correlated Time Series Forecasting

Spatiotemporal Dependence Learning with Meteorological Context for Transportation Demand Prediction

Spatiotemporal Forecasting in Earth System Science: Methods, Uncertainties, Predictability and Future Directions

ComS2T: A complementary spatiotemporal learning system for data-adaptive model evolution

Predicting Collective Human Mobility via Countering Spatiotemporal Heterogeneity

Fine-grained Graph Convolutional Network with Learning-Based Bi-Relational Graph for Spatiotemporal Forecasting