Supplementary Materials for: Physically Interpretable Stacking Ensemble for GOCI-II TSM Retrieval Beyond Sensor Saturation over Korean Coastal Waters

This record contains the supplementary materials (twelve figures, two tables, and a combined Word document with full captions) accompanying the manuscript "Physically Interpretable Stacking Ensemble for GOCI-II TSM Retrieval Beyond Sensor Saturation over Korean Coastal Waters," submitted to Remote Sensing (MDPI). Abstract: Total Suspended Matter (TSM) is a key coastal water-quality indicator, yet the operational GOCI-II retrieval algorithm is constrained by an empirical ceiling of approximately 28.13 g/m3 because its high-turbidity module is disabled, precluding accurate monitoring of turbid nearshore environments critical for ecosystems and coastal management. Here, we developed and evaluated PISE (Physically Interpretable Stacking Ensemble), a model that fuses three complementary base learners via a Ridge meta-learner on five physically interpretable GOCI-II Rrs(555, 620, 709 nm) features, trained on 212 matchup samples (2022-2025). On an independent held-out test set, PISE achieves R2 = 0.793 and RMSE = 3.670 g/m3, improving on the operational GOCI-II product by 8.9 percentage points in R2 and 16.3% in RMSE, with 90% conformal prediction intervals achieving 90.6% empirical coverage. Applied to a full day of GOCI-II imagery (4 May 2025), PISE provides the first quantitative satellite-derived characterization of diurnal TSM dynamics across the flood-ebb tidal cycle over Korean coastal waters, extending operational TSM retrieval above the 28.13 g/m3 saturation ceiling up to its trained limit of 50 g/m3. Contents:PISE_GOCI_II_TSM_Supplementary_Material.docx — combined document with all figures and tables below, fully captionedTableS1_FeatureDefinitions.csv — the five physically interpretable PISE spectral features and their selection rationaleTableS2_EvaluationMetrics.csv — definitions of the six evaluation metrics (R2, RMSE, MAE, Bias, SMAPE, LogRMSE)FigS1 — PISE vs. GOCI-II ATBD scatter by turbidity regimeFigS2 — Cap sensitivity summary (R2, RMSE, SMAPE vs. Winsorization cap)FigS3 — Permutation feature importance for PISEFigS4 — Spatial block cross-validation resultsFigS5 — Relative error (%) by turbidity regimeFigS6 — Ten-panel hourly PISE TSM maps, 4 May 2025FigS7 — Ten-panel hourly GOCI-II ATBD TSM maps, 4 May 2025FigS8 — Ten-panel PISE 90% conformal prediction interval width mapsFigS9 — Ten-panel PISE inter-model spread mapsFigS10 — Residual diagnostics for PISEFigS11 — Cap sensitivity scatter plots across all six cap valuesFigS12 — Nested leave-year-out cross-validation results Funding: This research was supported by the Korea Institute of Marine Science and Technology Promotion (KIMST), funded by the Ministry of Oceans and Fisheries (RS-2022-KS221660). Contact: Jong-Kuk Choi (corresponding author), [e-mail], Korea Institute of Ocean Science and Technology (KIOST).