Processed Landsat–MODIS inputs for spatiotemporal fusion evaluation under sparse agricultural observations

This repository provides the processed remote-sensing inputs used in the study “Reliability assessment of minimal-input Landsat–MODIS spatiotemporal fusion under sparse observations in a winter wheat–summer maize system.” The archived data include three groups of imagery: (1) Landsat reference images for eight retained observation dates in 2024: 12 February, 7 March, 15 March, 31 March, 24 April, 26 May, 30 August, and 28 December; (2) actual MODIS inputs, based on MOD09GA Version 6.1 surface reflectance and spatially aligned to the Landsat grid for use in the fusion models; and (3) Landsat-derived same-source coarse inputs, generated by aggregating the corresponding Landsat images onto a 500-m projected grid using the arithmetic mean of valid pixels and subsequently aligning them to the 30-m Landsat grid using nearest-neighbor resampling. Six common optical bands are included: Blue, Green, Red, near-infrared (NIR), shortwave-infrared 1 (SWIR1), and shortwave-infrared 2 (SWIR2). Reflectance values are stored using the scaled-reflectance convention adopted in the study. These data were used to evaluate STARFM, FSDAF, Fit-FC, and FastVSDF across 28 forward prediction pairs under actual MODIS and same-source input settings. The repository contains the processed reference and model-input imagery used in the experiments. Third-party model source code is not redistributed. Additional derived model outputs and evaluation results are not included in the present archive. The original Landsat and MODIS products are publicly available through Google Earth Engine.