Estimates of Regional Carbon Dioxide Fluxes Using a Dense Network of Low-Cost In Situ Observations

Thumbnail Image
Publication or External Link
Martin, Cory
Zeng, Ning
Current inverse modeling-based estimates of carbon dioxide (CO2) fluxes in urban areas typically use a network of 10-20 observation sites featuring high-accuracy gas analyzers that can cost over $100,000 each. Recently, commercially available, low-cost sensors to measure both traditional meteorological quantities and trace gases such as CO2 have become a focus of atmospheric science research. These flux estimations are an ill-posed problem in the sense that, depending on resolution, the mathematical model may be optimizing fluxes for hundreds or even thousands of grid points, with only relatively few observations to use as constraint. Theoretically, by introducing many more observations into the system, the result will better represent the true state of the surface fluxes. This work comprises of three related studies that evaluate the viability of using a low-cost CO2 sensor combined with a mesoscale meteorology model with online tracers, and an advanced ensemble data assimilation technique, to estimate surface fluxes of CO2 in an urban region. First, the SenseAir K30 sensor is evaluated compared to a reference gas analyzer to determine the accuracy and precision of the observations from this sensor. Next, a simulation of atmospheric CO2 is evaluated against observations to understand the error in simulated mole fractions from variations in existing emissions inventories. Finally, a series of observing system simulation experiments (OSSEs) are conducted to understand the sensitivity of estimated CO2 fluxes to the ensemble data assimilation system configuration. From this work, it is found that the K30 sensor can be useful for urban ambient monitoring of CO2 after corrections for environmental factors such as temperature and pressure. Additionally, the modeled CO2 results show that the error in simulated mole fractions is likely larger from meteorological error than it is from uncertainty in emissions. Finally, the OSSEs find that this ensemble data assimilation system using a dense network of lower-accuracy observations can achieve comparable CO2 flux estimation results to that of using a sparse network of high-accuracy observations. However, the configuration of the system, particularly the inflation technique used, can significantly affect the quality of the analyzed fluxes.