Spatio-Temporal Transformers for Long-Term NDVI Forecasting

TL;DR

This paper introduces STT-LTF, a novel spatio-temporal transformer model that effectively predicts long-term NDVI changes in heterogeneous Mediterranean landscapes using multi-scale spatial and temporal data, outperforming existing methods.

Contribution

The paper presents a unified transformer framework that integrates spatial and temporal modeling for long-term satellite image forecasting, utilizing self-supervised learning on unlabeled data.

Findings

Achieves MAE of 0.0328 and R^2 of 0.8412 for next-year NDVI prediction.

Outperforms traditional statistical, CNN, LSTM, and transformer models.

Handles irregular sampling and variable horizons effectively.

Abstract

Long-term satellite image time series (SITS) analysis in heterogeneous landscapes faces significant challenges, particularly in Mediterranean regions where complex spatial patterns, seasonal variations, and multi-decade environmental changes interact across different scales. This paper presents the Spatio-Temporal Transformer for Long Term Forecasting (STT-LTF ), an extended framework that advances beyond purely temporal analysis to integrate spatial context modeling with temporal sequence prediction. STT-LTF processes multi-scale spatial patches alongside temporal sequences (up to 20 years) through a unified transformer architecture, capturing both local neighborhood relationships and regional climate influences. The framework employs comprehensive self-supervised learning with spatial masking, temporal masking, and horizon sampling strategies, enabling robust model training from 40 years of unlabeled Landsat imagery. Unlike autoregressive approaches, STT-LTF directly predicts arbitrary future time points without error accumulation, incorporating spatial patch embeddings, cyclical temporal encoding, and geographic coordinates to learn complex dependencies across heterogeneous Mediterranean ecosystems. Experimental evaluation on Landsat data (1984-2024) demonstrates that STT-LTF achieves a Mean Absolute Error (MAE) of 0.0328 and R^2 of 0.8412 for next-year predictions, outperforming traditional statistical methods, CNN-based approaches, LSTM networks, and standard transformers. The framework's ability to handle irregular temporal sampling and variable prediction horizons makes it particularly suitable for analysis of heterogeneous landscapes experiencing rapid ecological transitions.