Warming Temperatures Lead to Reduced Summer Carbon Sequestration in the U.S. Corn Belt

The study addresses how warming temperatures affect the seasonal dynamics of carbon exchange in highly productive croplands of the northern mid-latitudes, specifically the U.S. Corn Belt. Atmospheric CO2 has risen rapidly since the industrial revolution, while the terrestrial biosphere absorbs roughly a quarter of anthropogenic emissions, yet the magnitude, spatial distribution, and climatic controls of this sink remain uncertain. The amplitude of the northern hemisphere CO2 seasonal cycle has increased markedly since the 1960s, historically linked to warming-driven extensions and intensification of photosynthesis. However, recent evidence suggests weakened or reversed correlations between growing-season temperatures and CO2 drawdown at some sites, indicating possible shifts in carbon-climate relationships. Temperate regions, which include extensive croplands, have been less studied than tropical and boreal systems regarding their role in the CO2 seasonal cycle. This paper aims to quantify how interannual temperature variations influence regional CO2 exchange in the Corn Belt, distinguish responses of croplands versus natural ecosystems, and project impacts of future warming on the CO2 seasonal cycle and net uptake by 2050.

Prior studies have documented a ~50% increase in the amplitude of the northern hemisphere CO2 seasonal cycle since the 1960s, attributed to enhanced plant productivity and phenological shifts under warming. Evidence has emerged of attenuated or reversed temperature-CO2 drawdown relationships in recent decades at Mauna Loa and high northern latitudes, signaling changing carbon-climate feedbacks. Temperate forests show increased spring/autumn NPP and extended growing seasons, potentially enhancing carbon sequestration, yet the role of extensive croplands has been underappreciated. Top-down and bottom-up analyses suggest agricultural intensification at northern temperate latitudes has contributed 17–45% of the increased CO2 seasonal amplitude over the past five decades, with corn as a dominant driver due to concentrated production in the U.S. Corn Belt and northern China. Nonetheless, direct observational constraints remain limited temporally and spatially, leaving uncertainties around the strength of agricultural forcing and ecosystem responses to future warming.

Observations: A decadal record (2007–2019) of hourly atmospheric CO2 at 100 m was collected from the University of Minnesota tall tower Trace Gas Observatory (KCMP; 44.6888°N, 93.0728°W) using a tunable diode laser spectrometer (10 Hz, averaged to hourly; precision ~0.2 ppm). Wind speed and direction at 100 m were measured by sonic anemometry. Additional tall tower records from NOAA’s network were used: Park Falls, WI (LEF) and West Branch, IA (WBI), at 99–122 m height (2007–2018). A continental background site (Niwot Ridge, NWR) provided reference CO2. Data processing: Hourly CO2 series were de-spiked, gap-filled, and aggregated to daily values. Wavelet transform decomposition extracted a detrended seasonal cycle (3–18 month periods) and long-term growth (>18 months). Seasonal amplitude was defined as peak-to-trough of the detrended cycle. For KCMP, wind-sector sampling (2010–2018; wind speed >3 m s−1) produced sectoral monthly datasets: northwest (270°–360°, KCMPNW) and south/southeast (120°–210°, KCMPSSE) to assess transport and land-use influences. Footprint characterization: Prior STILT analyses showed ~80% of concentration sensitivity originates within ~300 km of KCMP across seasons. A 300 km radius was adopted as the intense concentration footprint for all towers; sensitivity to 150–450 km radii was tested. NEE inversions: Monthly net ecosystem exchange (NEE; 1° × 1°) within footprints was obtained from CarbonTracker CT2019 inversions (2007–2018). Annual NEE amplitude was calculated as cumulative dormant-season (Oct–Apr) minus growing-season (May–Sep) NEE. Land use: High-resolution NASS Cropland Data Layer (2008–2018) defined fcs, the fraction of land ecosystems in corn and soybean relative to total land ecosystems (croplands + natural ecosystems). Pasture, wheat, oats, and perennial forages were grouped with natural ecosystems for this analysis. Crop data: County-level corn and soybean harvested area and production (USDA Quick Stats, 2008–2018) were aggregated within/overlapping the 300 km footprints to compute site-year yields (t ha−1). Climate data: NCEP-NARR provided daily air temperature (2 m), precipitation, and shortwave radiation (2007–2019); PRISM supplied daily min/max temperatures. Future projections (2006–2050) from 10 CMIP5 GCMs under RCP4.5 and RCP8.5 were bias-corrected/downscaled to match 2008–2018 means/variances and used to estimate 2041–2050 minus 2010–2019 changes. Derived metrics and statistical analyses: The first time-derivative of monthly detrended seasonal CO2 (ΔCO2) was computed to better proxy regional net flux variations. For each month, multiple linear regression (MLR) related ΔCO2 or NEE anomalies to temperature, water availability (3-month cumulative precipitation anomaly including current month), and radiation, after linear detrending. Temperature sensitivity βT was the partial slope on temperature; uncertainties used bootstrap resampling (1000 iterations). A panel data model decomposed site-specific climate sensitivities into biome-specific (croplands vs natural ecosystems) using footprint fractions fcs and fNy=1−fcs, validated by reconstructing site sensitivities (NSE>0.9). Growing degree time (base 8 °C) accumulated at hourly resolution from PRISM min/max temperatures simulated corn leaf emergence date (CLED; threshold 450 units) to link phenology with June NEE anomalies. Panel regression linked corn and soybean yields to spring (May–June) temperature, summer (July–August) temperature, and growing season precipitation, including fixed effects and a linear trend term; quadratic temperature terms were omitted due to limited range and sample size.

- Spatial agricultural imprint: Within 300 km footprints, cropland fraction fcs averaged 0.43±0.01 (KCMP), 0.12±0.01 (LEF), and 0.56±0.01 (WBI). WBI exhibited the largest CO2 seasonal amplitude (28.1 ppm), followed by KCMP (27.9 ppm) and LEF (23.9 ppm), consistent with higher cropland fractions and stronger growing-season NEE. KCMP’s south/southeast sector (fcs≈0.52) showed stronger CO2 drawdown than its northwest sector (fcs≈0.23). - Trends: CO2 seasonal amplitude trends (2007–2019) were 0.18 ppm yr−1 at KCMP and 0.39 ppm yr−1 at WBI (not statistically significant), and 0.19 ppm yr−1 at LEF (significant at P<0.1). Annual NEE amplitude showed no significant trends at any site. - Decoupling at interannual scale: After detrending, crop yield anomalies did not significantly correlate with annual CO2 amplitude or NEE amplitude at any site, suggesting compensating sub-annual processes and transport influences. - Temperature sensitivities: Monthly βT for ΔCO2 and NEE showed strong seasonality. In July–August, βT was significantly positive at all sites (higher temperature reduced net CO2 uptake). In June, croplands exhibited significantly negative βT for NEE at KCMP and WBI (warmer conditions enhanced uptake), while natural ecosystems showed weak/variable spring-to-early-summer responses. - Biome-specific responses: Panel analysis revealed that croplands drive the negative June βT (enhanced early-season uptake under warmth), while both croplands and natural ecosystems have positive βT in July–August (reduced uptake with warming). Results were robust across footprint radii and biome definitions. - Mechanisms: Positive summer βT aligns with ecosystems operating near or beyond thermal optima for photosynthesis and with increased respiration and water stress (higher VPD and limited soil moisture). Dry summers amplified positive βT; however, even moderate-to-wet summers retained significantly positive βT, indicating adaptation to current thermal regimes. - Phenology linkage: June NEE anomalies correlated with anomalies in simulated corn leaf emergence date (CLED) at KCMP and WBI, independent of June temperature, indicating legacy effects of earlier warm springs and planting on early-season uptake. - Crop yield responses: Panel regression explained ~80% of corn yield variance across KCMP and WBI footprints. Corn yields decreased by −0.36 t ha−1 per °C increase in July–August (90% CI −0.54 to −0.19), about 3% per °C relative to a 23 °C baseline, consistent with prior literature. Spring temperature had no significant effect on final corn yield. Soybean yields were significantly positively related to growing-season precipitation, but not to spring or summer temperature over 2008–2018, possibly due to CO2 fertilization alleviating heat stress. - Future projections to 2050: Applying observed βT with CMIP5 ensemble mean warming yields attenuation of the CO2 seasonal amplitude by ~1.5 ppm (~5%, RCP4.5) to ~3 ppm (~10%, RCP8.5). Summer (July–August) net CO2 uptake reduction across the Corn Belt land ecosystems is projected at ~30 Tg C per °C (RCP4.5; 90% CI 10–60) to ~60 Tg C per °C (RCP8.5; 90% CI 20–117), about 10–20% of annual regional sequestration. Positive early-season effects (May–June) partially offset summer losses, yielding a small, statistically non-significant net reduction in growing-season uptake by 2050.

The results demonstrate that warming exerts contrasting seasonal controls on regional CO2 exchange in the Corn Belt: it enhances early-season uptake in croplands via accelerated phenology and canopy development but reduces peak-season uptake in both croplands and natural ecosystems due to surpassing thermal optima, increased respiration, and heightened atmospheric water demand. This reconciles the lack of interannual correlation between annual yield and CO2 amplitude by revealing compensatory sub-annual dynamics and transport effects. The strong positive summer temperature sensitivity indicates that current midlatitude ecosystems may already operate near thermal optima for CO2 uptake, leading to reduced summer sequestration under further warming. The agricultural imprint is clear, with higher cropland fractions linked to greater seasonal amplitude and stronger sensitivity. The linkage between July NEE anomalies and corn yields underscores the dominant role of corn reproductive-stage heat stress in driving both ecosystem carbon fluxes and agricultural outcomes. Projections using observed sensitivities suggest that continued warming will attenuate the CO2 seasonal cycle and reduce summer carbon uptake substantially, partially offset by stronger spring uptake, emphasizing that both magnitude and seasonal timing of warming will shape future temperate carbon sinks.

This study leverages a decadal tall-tower CO2 record, regional inversion products, and land-use stratification to quantify how temperature modulates regional carbon exchange in the U.S. Corn Belt. It shows a robust negative impact of summer warming on net CO2 uptake in both croplands and natural ecosystems, contrasted with a positive early-season effect in croplands. Agricultural land use strongly imprints the regional CO2 seasonal cycle and sensitivity, and corn yields decline with warmer summers. Projections indicate future warming will attenuate the CO2 seasonal amplitude and reduce summer sequestration by about 10–20% of annual uptake, though enhanced spring uptake may partly compensate. These findings challenge the paradigm that warming will uniformly favor northern mid-latitude carbon sinks and call for improved representation of cropland temperature sensitivities and seasonal processes in carbon cycle models. Future work should address extreme events, interactions with water availability and VPD, farmer adaptations (e.g., planting dates, cultivar selection), and slow biogeochemical feedbacks (CO2 fertilization, nutrient cycling, soil carbon turnover) to refine predictions of carbon-climate feedbacks.

- Projections assumed unchanged seasonal atmospheric transport and circulation; deviations could alter ΔCO2-NEE relationships. - Warming impacts were based on monthly mean temperatures, not explicitly capturing effects of extreme heat events, which are expected to intensify. - Slow processes (CO2 fertilization, soil carbon turnover, nutrient cycling) and potential nonlinearities over multi-decadal scales were not explicitly modeled and may confound temperature effects. - Farmer adaptations (earlier planting, longer-maturity cultivars) were not incorporated, potentially mitigating adverse summer heat impacts. - Biome categorization grouped pastures and some crops with natural ecosystems, potentially smoothing biome-specific sensitivities. - Reliance on inversion-based NEE and footprint assumptions (e.g., uniform weighting within 300 km) introduces uncertainties, though sensitivity tests suggested robustness. - Limited temporal span (2007–2019) and sample size for some monthly analyses constrain detection of trends and higher-order (nonlinear) responses.