HLS-GPT: A Generative Pretrained Transformer (GPT) for Continental-Scale NASA Harmonized Landsat and Sentinel-2 (HLS) Reflectance Reconstruction Across All Bands on Arbitrary Dates

2026-06-16

AI Summary

Existing methods for reconstructing remote sensing reflectance are limited in spectral coverage, spatiotemporal generalization, or context length. This work proposes a hierarchical Transformer-based generative pretraining model trained on over 250,000 pixels across the conterminous United States using 12-month time series from the NASA Harmonized Landsat and Sentinel-2 (HLS) dataset. Leveraging random temporal cropping and a 50% observation masking strategy, the model achieves, for the first time, high-fidelity, unified reconstruction of full-spectrum Landsat and Sentinel-2 reflectance—including Landsat’s missing red-edge bands—at any date and location across continental scales. Evaluated on 62,000 independent test pixels, the model attains a full-band RMSE below 0.026 under normal conditions and below 0.028 even with 50% masked inputs, substantially outperforming conventional approaches and the Prithvi model.

Abstract

Recent deep learning methods for Landsat and Sentinel-2 reflectance time series reconstruction remain limited by restricted spectral coverage, limited geographic scalability, or patch-based designs with short temporal contexts. We present HLS-GPT, a large-scale generative pretrained Transformer model for reconstructing NASA Harmonized Landsat Sentinel-2 30 m surface reflectance for all bands, any date, and any pixel location. HLS-GPT uses a hierarchical Transformer architecture to handle the different spectral band configurations of Landsat and Sentinel-2 and operates on single-pixel 12-month time series. To capture geographic and seasonal variability, the model was trained with nine years of HLS time series from more than 0.25 million training pixels across the conterminous United States. A random cropping and masking strategy extracts 12-month periods with varying start dates across epochs, masks 50% of valid observations, and trains the model to reconstruct the masked reflectance values from the remaining observations. Evaluation using more than 62,000 independent test pixels shows robust reconstruction under diverse land surface conditions, including complex crop phenology and sparse, irregular observations. Leave-one-observation-out evaluation achieved reconstruction RMSE below 0.026 for all HLS spectral bands, with relative RMSE below 35% for visible bands and below 13% for other bands. Red-edge band errors were comparable to red and near-infrared errors despite the absence of red-edge bands on Landsat. Sensitivity analyses that randomly masked 10% to 90% of test observations showed only modest degradation when 10% to 50% of observations were masked, with all-band RMSE below 0.028. Image reconstruction over nine independent 109 by 109 km CONUS HLS tiles further demonstrates that HLS-GPT outperforms two conventional methods and the NASA-IBM Prithvi model.

Problem

• reflectance reconstruction
• Landsat
• Sentinel-2
• time series
• continental-scale

Innovation

• HLS-GPT
• Generative Pretrained Transformer
• Reflectance Reconstruction
• Multi-sensor Fusion
• Time Series Imputation