A fuzzy spatial and temporal adaptive reflectance fusion model with local information

The 16-day revisit cycle of Landsat satellites constrains global phenology research particularly in cloudy regions, where only 2–3 cloud-free images are available per year while coarse resolution sensors such as MODIS struggle to quantify biophysical processes in heterogeneous landscapes. To address these limitations, this study proposes a novel fuzzy local information spatiotemporal adaptive reflectance fusion model (FLISTARFM) for blending Landsat and MODIS surface reflectance data. The algorithm integrates high-frequency temporal information from daily 250-meter resolution MODIS data and high spatial resolution information from 16-day revisit 30-meter resolution Landsat ETM+ data to generate synthetic “daily” surface reflectance products at ETM+ spatial resolution. The FLISTARFM improves upon traditional STARFM by replacing its spectral similarity measurement with a similarity metric based on Fuzzy Local Information C-Means (FLICM) clustering. The FLICM incorporates local spatial and grayscale information via a fuzzy factor Gki, which effectively suppresses noise interference (e.g., atmospheric haze or cloud shadows) and preserves boundary details in heterogeneous areas addressing two major limitations of conventional FCM clustering and STARFM. The algorithm redefines spectral distance using membership matrices derived from fine-resolution clustering, decomposes the temporal distance of coarse pixels into class-specific components, and computes weights via normalized integration of spectral, temporal, and spatial distances.