Quantifying Spatial Cross-Validation Inflation in Satellite-Ground Air Quality Coupling: Mesoscale NDVI/LST Regimes and Land Use Regression Across an Urban Megacity

Coupling satellite remote sensing with ground telemetry enhances urban exposure modeling, yet conventional random cross-validation substantially inflates predictive performance. We quantify spatial cross-validation inflation and evaluate mesoscale biophysical drivers across 13 regulatory air quality stations in the Kolkata-Howrah conurbation during 2024-2025. Landsat-8/9 OLI/TIRS surface reflectance and thermal imagery (30 m resolution) were integrated with ground-level particulate matter (PM2.5 and PM10) telemetry across 500 m to 5000 m buffer radii. Standard random 10-fold cross-validation yielded an inflated R2 of 0.84 for PM2.5, whereas rigorous spatial block cross-validation and Leave-One-Station-Out (LOSO) holdout validation revealed true spatial generalizability at R2 = 0.58 (an inflation factor of 1.45x). Normalized Difference Vegetation Index (NDVI) and Land Surface Temperature (LST) exhibit distinct non-linear cooling and scavenging thresholds, with high vegetative density (NDVI > 0.45) mitigating ambient PM2.5 concentrations by up to 24.3% relative to barren or built-up impervious surfaces (NDBI > 0.15). Multi-buffer spatial regression demonstrates that environmental covariates operate across multi-scale zones of influence, with thermal and built-up drivers dominating at localized scales (500-1000 m) and green-space buffering governing mesoscale transport (2000-5000 m). These findings demonstrate that spatial autocorrelation must be explicitly addressed during satellite-ground calibration to avoid spurious exposure estimates in epidemiological health studies and statutory environmental decision-making.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-20
DOI
https://doi.org/10.21203/rs.3.rs-11793536/v1
Primary Topic
Air Quality and Health Impacts
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Quantifying Spatial Cross-Validation Inflation in Satellite-Ground Air Quality Coupling: Mesoscale NDVI/LST Regimes and Land Use Regression Across an Urban Megacity

Saptarshi Sadhu
Research Square
Air Quality and Health Impacts
preprint

Quantifying Spatial Cross-Validation Inflation in Satellite-Ground Air Quality Coupling: Mesoscale NDVI/LST Regimes and Land Use Regression Across an Urban Megacity

Saptarshi Sadhu
preprint en

Abstract

Coupling satellite remote sensing with ground telemetry enhances urban exposure modeling, yet conventional random cross-validation substantially inflates predictive performance. We quantify spatial cross-validation inflation and evaluate mesoscale biophysical drivers across 13 regulatory air quality stations in the Kolkata-Howrah conurbation during 2024-2025. Landsat-8/9 OLI/TIRS surface reflectance and thermal imagery (30 m resolution) were integrated with ground-level particulate matter (PM2.5 and PM10) telemetry across 500 m to 5000 m buffer radii. Standard random 10-fold cross-validation yielded an inflated R2 of 0.84 for PM2.5, whereas rigorous spatial block cross-validation and Leave-One-Station-Out (LOSO) holdout validation revealed true spatial generalizability at R2 = 0.58 (an inflation factor of 1.45x). Normalized Difference Vegetation Index (NDVI) and Land Surface Temperature (LST) exhibit distinct non-linear cooling and scavenging thresholds, with high vegetative density (NDVI > 0.45) mitigating ambient PM2.5 concentrations by up to 24.3% relative to barren or built-up impervious surfaces (NDBI > 0.15). Multi-buffer spatial regression demonstrates that environmental covariates operate across multi-scale zones of influence, with thermal and built-up drivers dominating at localized scales (500-1000 m) and green-space buffering governing mesoscale transport (2000-5000 m). These findings demonstrate that spatial autocorrelation must be explicitly addressed during satellite-ground calibration to avoid spurious exposure estimates in epidemiological health studies and statutory environmental decision-making.

Research Square
Adamas University (IN)
Air Quality and Health Impacts
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon