Influence of Vegetation Cover on Atmospheric CO2 Mixing Ratios in the São Paulo Metropolitan Area

Urban vegetation plays a key role in modulating atmospheric carbon dioxide (CO₂) in megacities. However, studies that explicitly quantify the effect of urban vegetation on CO₂ remain scarce. This study investigates how vegetation cover affects CO₂ mixing ratios in the Metropolitan Area of São Paulo (MASP) during 2020–2022, using four monitoring sites with contrasting vegetation fractions: Pico do Jaraguá (PJ; 59.75%), IAG (36.38%), ICESP (22.42%), and UNICID (10.42%). Vegetation cover was derived from Sentinel-2 Level-2A imagery using NDVI-based pixel classification, while CO₂ observations were obtained from the METROCLIMA network and analyzed together with concurrent meteorological variables (temperature, humidity, and wind). The analysis comprised characterizing temporal variability and quantifying vegetation effects using regression models and probability distribution functions (PDFs). Clear seasonal and diurnal patterns were observed, with lower CO₂ concentrations during summer and afternoon hours (420-414 ppm), and higher values during winter and nighttime periods (447-425 ppm). The greener and less urban site, PJ, exhibited the lowest and most stable CO₂ levels, whereas the highly urban UNICID site showed the highest average mixing ratios. Elevated CO₂ values at IAG (428.30 ppm in summer and 435.49 ppm in winter), despite substantial vegetation cover, suggest the influence of local emissions and boundary-layer dynamics, while relatively low CO₂ values at ICESP (422.10 ppm in summer and 428.03 ppm in winter) likely reflect the elevated measurement height (~100 m a.g.l.), which favors regional-scale mixing and reduces sensitivity to local emission sources. NDVI revealed a bimodal phenological cycle (April–May and October–November), which was mirrored by CO₂ variability at PJ. Among 132 fitted PDFs, the Normal Inverse Gaussian distribution best captured CO₂ variability, with greener sites showing flatter and more symmetric distributions and urban sites showing increased skewness and peakedness. Regression results indicate a significant vegetation signal: a 0.1 increase in NDVI was associated with CO₂ reductions of ~3.92 ppm (PJ), 2.81 ppm (IAG), and 7.66 ppm (ICESP; likely conditional on site features), while no significant effect was detected at UNICID. Overall, urban vegetation influences both mean CO₂ levels and their distributional characteristics, supporting the role of green infrastructure and phenology in urban carbon management.

Keywords: carbon dioxide, vegetation cover, linear regression, Metropolitan Area of São Paulo.