Beyond same-source benchmarks: product performance and failure modes of minimal-input Landsat–MODIS fusion under sparse agricultural observations

This repository provides the study-specific data and supporting materials for the manuscript “Beyond same-source benchmarks: operational reliability and failure modes of minimal-input Landsat–MODIS fusion under sparse agricultural observations.” The study evaluates four representative minimal-input spatiotemporal fusion models, STARFM, FSDAF, Fit-FC, and FastVSDF, using eight operationally usable Landsat–MODIS observations acquired in 2024. The eight dates form 28 chronologically valid forward prediction pairs. Each prediction was evaluated under two coarse-resolution input settings: actual Terra MODIS MOD09GA observations and Landsat-derived same-source coarse-resolution inputs. The deposited data support the comparison of benchmark and operational multisensor conditions, the evaluation of radiometric accuracy, pixelwise correspondence, local structural similarity, and output completeness, and the diagnosis of incomplete STARFM predictions. The repository includes the processed Landsat and coarse-resolution input data used in the experiments, together with the associated metadata and supporting materials provided in the deposited files. Original Landsat and MODIS products are publicly available through Google Earth Engine. Third-party fusion implementations are not redistributed and should be obtained from the original software sources cited in the manuscript.