TL;DR
A new real-time global CO2 forecast system, integrated within the ECMWF's numerical weather prediction framework, has been developed. This system shows promising skill in capturing CO2 variability across various timescales, particularly when accounting for daily land-surface flux variations, though biases in flux modeling lead to accumulating errors, especially during spring.
What Problem Does This Paper Address?
Accurately forecasting atmospheric CO2 concentrations is crucial for understanding and mitigating climate change. However, this is a complex challenge due to the intricate interplay between atmospheric transport and surface fluxes (sources and sinks of CO2). Previous efforts were hampered by the delayed availability of CO2 observations and uncertainties in modeling these fluxes and transport processes. This paper addresses the need for a real-time, operational CO2 forecast system that can provide timely information and support data assimilation efforts, bridging the gap between current capabilities and the demand for near-real-time CO2 monitoring.
Methods
The core of this work is the integration of a CO2 forecasting system within the European Centre for Medium-Range Weather Forecasts (ECMWF) Integrated Forecasting System (IFS). Key methodological aspects include:
- Online Land Surface Fluxes: A significant advancement is the online computation of terrestrial biogenic CO2 fluxes (Gross Primary Production and Ecosystem Respiration) within the IFS. This ensures consistency between flux calculations and meteorological forcing.
- Prescribed Fluxes: Other CO2 fluxes (anthropogenic, fires, ocean) are incorporated using prescribed inventories and models.
- State-of-the-Art Transport: The forecast leverages ECMWF's advanced numerical weather prediction (NWP) system for atmospheric transport, initialized daily with meteorological observations.
- Hindcasting for Evaluation: To assess the system's capability, extensive hindcasts (simulations of past years) were performed using the same configuration as the real-time forecast. This allows for a robust comparison with independent observational datasets.
Datasets
The study utilizes a variety of observational datasets for evaluating the CO2 forecast:
- NOAA/ESRL Surface Stations: Continuous measurements from background air sampling sites (e.g., Barrow, American Samoa, South Pole).
- ICOS Network: Continuous measurements from the Integrated Carbon Observation System network.
- TCCON Network: Total column FTIR measurements from the Total Carbon Column Observing Network.
- NOAA/ESRL Aircraft Profiles: Vertical profile data from the NOAA GMD Carbon Cycle Vertical Profile Network.
- HIPPO Flights: Data from the HIPPO (HIAPER Pole-to-Pole Observations) campaign for evaluating CO2 in the free troposphere.
These datasets represent a comprehensive suite of ground-based and remote sensing observations, providing coverage across different spatial and temporal scales.
Key Insights and Ideas
- The Power of Daily Flux Variability: The forecast's skill in simulating day-to-day CO2 variability, especially in the atmospheric boundary layer, is significantly enhanced when the daily variability of vegetation CO2 fluxes is modeled, rather than using smoothed monthly averages. This highlights the importance of capturing the dynamic nature of land-atmosphere exchange.
- Springtime Challenges: The largest biases in the atmospheric CO2 forecast occur in spring, coinciding with the onset of the growing season in the Northern Hemisphere. This suggests that accurately modeling the transition from respiration to photosynthesis is a critical area for improvement.
- Transport vs. Fluxes: While atmospheric transport plays a dominant role in CO2 variability, the study demonstrates that day-to-day variations in land-surface fluxes also significantly contribute to the synoptic skill of the forecast, particularly at lower atmospheric levels.
- Accumulating Errors: Biases in the modeled CO2 fluxes, even if small, lead to accumulating errors in the forecast over time. This underscores the need for accurate flux estimates and strategies to mitigate these accumulating biases.
Key Results
- Good Representation of Diurnal and Synoptic Cycles: The forecast generally represents the modulation of the CO2 diurnal cycle by near-surface winds and boundary layer height well. It also shows high skill in simulating day-to-day synoptic variability.
- Correlation Coefficients: Correlations between modeled and observed CO2 are predominantly higher than 0.5 in winter months, often reaching 0.65-0.95. Spring months show lower correlations, indicating challenges in this period.
- Seasonal Cycle Underestimation: The model tends to underestimate the amplitude of the seasonal cycle for both CO2 fluxes and atmospheric CO2 concentrations.
- Interannual Variability: The correlation between modeled and observed global annual atmospheric growth is 0.74, indicating a reasonable representation of interannual trends.
- Bias: The forecast exhibits biases, particularly in the Northern Hemisphere, with an underestimation of the seasonal cycle amplitude.
Limitations and Future Work
- Flux Biases: The primary limitation acknowledged is the presence of biases in the modeled CO2 fluxes, which lead to accumulating errors in the atmospheric CO2 forecast. These biases are particularly pronounced in spring.
- Future Re-initialization: To address accumulating errors, future versions of the forecast will be regularly re-initialized using atmospheric CO2 analyses derived from the assimilation of near-real-time satellite and in situ CO2 observations.
- Model Developments: Ongoing developments in ECMWF's IFS, including improvements in convection and vertical diffusion parameterizations, are expected to enhance the transport modeling and, consequently, the CO2 forecast.
- Satellite Data Assimilation: The assimilation of new satellite products (e.g., albedo, LAI) is anticipated to improve the modeling of CO2 fluxes.
- Vegetation Model Enhancements: Further improvements to the vegetation radiative transfer scheme are planned.
Summary updated