Exploring the Relationships Between XCO2 Concentrations and Climate Variables in Brazil
Year: 2026
Type: article
Abstract: Climate conditions influence ecosystem carbon exchange and, consequently, atmospheric CO2 through variations in temperature, precipitation, and soil moisture. Brazil provides an important case study because of its extensive forest cover, land-use change, and pronounced climatic variability. This study combines OCO-2-constrained GEOS assimilated estimates of the column-averaged dry-air mole fraction of CO2 (XCO2; OCO-2 GEOS Level 3 monthly product, version 10r, February 2015–January 2022) with ERA5-Land air temperature and top-layer soil moisture and CHIRPS precipitation. All inferential analyses were performed on the native 0.5° × 0.625° grid of the XCO2 product (2264 grid cells), to which the climate fields were aggregated by area-weighted averaging. Monthly anomalies were obtained by jointly removing a monthly climatology and a linear trend at each grid cell, and statistical significance was assessed with an autocorrelation-adjusted effective sample size and Benjamini–Hochberg false-discovery-rate (FDR) control applied to pre-defined families of tests spanning all grid cells and, for the lagged analysis, all seven candidate lags. After FDR control, 861 of the 2264 grid cells (38.0%) retained a significant contemporaneous correlation between XCO2 and precipitation, compared with 207 (9.1%) for soil moisture and 0 for temperature. The precipitation association was positive, concentrated in central Brazil, predominantly contemporaneous, retained after conditioning on temperature and soil moisture, and robust to alternative trend definitions, rank-based correlation, and removal of the zonal-mean XCO2 background. Temperature was associated with XCO2 only when leading it by 1–4 months (21.3% of cells), a result that depended on the multiple-testing procedure, and no significant association was found with XCO2 leading climate. Without detrending, nominally significant correlations were widespread, but none survived autocorrelation adjustment and FDR control, and the sequential deseasonalize-then-detrend procedure of earlier analyses introduced an artifact that masked these associations. Out of sample, the three variables predicted XCO2 anomalies poorly (median leave-one-year-out cross-validated R2 = −0.023). These results reveal a spatially coherent short-term precipitation–XCO2 association in central Brazil whose physical origin, local fluxes or atmospheric transport, cannot be established from concentration data alone.
Source: Atmosphere
Institutions: EIA University, Universidade Federal de Santa Catarina
Language: English
Subfield: Global and Planetary Change
Field: Environmental Science
Domain: Physical Sciences
SDG: Climate action
Open Access status: gold
Estimated APC paid: $2,896
OpenAlex ID: W7221805337