The GMES atmospheric services include global and European air quality monitoring and forecasting which require near real time delivery of atmospheric CO abundances. To achieve this, a numerically efficient retrieval approach for operational data processing is needed to derive CO column densities from shortwave infrared measurements in the 2.3μm band of the Sentinel 5 missions and its Precursor mission. The expected performance of both spectrometers will allow for clear-sky CO column retrievals over land with a precision of ≤10% and an overall accuracy of ≤15% even for background CO abundance and low surface reflection in the shortwave infrared spectral range. In this context, we present a new algorithm approach of the retrieval of CO from shortwave infrared measurements in clear sky and partially cloudy atmospheres over land and ocean surfaces. The algorithm employs simplified radiative transfer, where the model atmosphere is separated in a clear sky part, and a part which is bounded below by an elevated Lambertian reflector to account for atmospheric scattering by clouds and aerosols. Within the inversion scheme, Tikhonov regularization is used to determine, for each individual measurement, not only the vertically integrated CO column density and its retrieval error, but also the column averaging kernel. For the retrieval, a prior estimate of methane abundance is used to characterize the light path by retrieving effective cloud parameters from the shortwave infrared band itself. A performance analysis shows that, for a single cloud layer in the middle and lower troposphere, the bias on the CO retrieval due to the Lambertian cloud model is less than 2–3%. The effect of boundary layer aerosols can also be treated with similar accuracy. In contrast, the presence of elevated dust plumes above bright surfaces or a single layer cirrus cloud causes significant errors and, in these cases, a reasonably low retrieval bias can only be achieved for an optical depth in the shortwave infrared spectral range lower than 0.4. Another relevant error source for the CO retrieval algorithm is given by the prior uncertainty of methane. It is found that a 5% uncertainty in the methane column density causes biases of 3–9% on the retrieved CO column, depending on cloud fraction.
Read full abstract