EarthPT: a time series foundation model for Earth Observation

TL;DR

EarthPT is a large, pretrained transformer model designed for Earth Observation data, capable of accurate future surface reflectance prediction and useful for downstream tasks like land use classification, with scalable potential due to abundant data.

Contribution

This paper introduces EarthPT, a 700-million parameter transformer model specifically trained for Earth Observation tasks, demonstrating its forecasting accuracy and semantic embedding utility.

Findings

EarthPT accurately predicts NDVI evolution with about 0.05 error over five months.

Embeddings from EarthPT encode meaningful information for downstream tasks.

EO data volume allows for theoretically unlimited scaling of EarthPT.

Abstract

We introduce EarthPT -- an Earth Observation (EO) pretrained transformer. EarthPT is a 700 million parameter decoding transformer foundation model trained in an autoregressive self-supervised manner and developed specifically with EO use-cases in mind. We demonstrate that EarthPT is an effective forecaster that can accurately predict future pixel-level surface reflectances across the 400-2300 nm range well into the future. For example, forecasts of the evolution of the Normalised Difference Vegetation Index (NDVI) have a typical error of approximately 0.05 (over a natural range of -1 -> 1) at the pixel level over a five month test set horizon, out-performing simple phase-folded models based on historical averaging. We also demonstrate that embeddings learnt by EarthPT hold semantically meaningful information and could be exploited for downstream tasks such as highly granular, dynamic land use classification. Excitingly, we note that the abundance of EO data provides us with -- in theory -- quadrillions of training tokens. Therefore, if we assume that EarthPT follows neural scaling laws akin to those derived for Large Language Models (LLMs), there is currently no data-imposed limit to scaling EarthPT and other similar `Large Observation Models.'