AUTHOR=Delgado-Bonal Alfonso , Marshak Alexander , Yang Yuekui 

TITLE=Size matters: the influence of pixel resolution on DSCOVR/EPIC reflectance and cloud metrics

JOURNAL=Frontiers in Remote Sensing

VOLUME=Volume 6 - 2025

YEAR=2025

URL=https://www.frontiersin.org/journals/remote-sensing/articles/10.3389/frsen.2025.1683919

DOI=10.3389/frsen.2025.1683919

ISSN=2673-6187

ABSTRACT=Satellite-derived reflectance and cloud retrievals are highly sensitive to spatial scale. Coarser pixels exaggerate cloud fraction, bias optical thickness and height estimates because unresolved subpixel variability violates plane-parallel assumptions. Here, we use DSCOVR/EPIC Level-1 reflectance (317–780 nm) and Level-2 cloud products (binary cloud mask, effective cloud height, ice/liquid optical thickness) to quantify these effects. Full-disk images were down-sampled to eight resolutions of 1,024, 512, 256, 128, 64, 32, 16, and 8 pixels across the disk at ∼12, 25, 50, 100, 200, 400, 800, and 1,600 km per pixel, respectively. Reflectances were aggregated by simple averaging: cloud masks by five subpixel thresholds (≥1, ≥25, ≥50, ≥75, and 100% cloudy), and cloud height and optical thickness by mean values when ≥50% of subpixels were valid. Global means of reflectance, cloud fraction, cloud height, and optical thickness were then calculated at each scale and threshold. While reflectance averages remained constant to within 1% across all scales, the cloud fraction rose steeply under permissive thresholds as resolution coarsened. Mean cloud height and optical thickness also increased, reflecting the dominance of taller, thicker clouds in coarse-pixel averages. These results quantify resolution-driven biases in EPIC cloud products and underscore the value of high-resolution observations and heterogeneity-aware retrieval methods for robust cloud characterization.