GEO-CAPE Regional Chemical OSSE
The GEO-CAPE Regional/Urban Observation System Simulation Experiment (OSSE) Working Group was initiated in 2013 to assess the value of geostationary observations of ozone (O3), nitrogen dioxide (NO2) and formaldehyde (HCHO) over the continental US (CONUS) in addressing the scientific and applications objectives of GEO-CAPE. The main components of the Regional/Urban OSSE (nature run, observation simulator, data assimilation system) are illustrated in Figure 1; we follow the recommendations of Timmermans et al. (2015), which provided a framework for the use of OSSEs for assessing the impact of satellite trace gas retrievals on air quality forecasts, including requirements for the individual components.
Development of the OSSE framework was based on three overarching goals: (1) the nature run must provide a reasonable representation of the real atmosphere; (2) the observation simulator must be able to produce synthetic “measurements” that account for the spectral resolution, signal to noise ratio, and averaging kernel (AK, sensitivity of measurement to true state) of the instrument being assessed; and (3) the model used within the data assimilation system should be different than the model used to generate the nature atmosphere. The first two years of the Regional/Urban OSSE activities focused on completion of ultraviolet (UV) visible (VIS) and thermal infrared (TIR) radiative transfer (RT) modeling, generation of multi-spectral retrievals for a subset of CONUS profiles, and AK regression to extend the training set to all of North America for the O3 OSSE studies.
The Regional/Urban OSSE Working Group activities resulted in several innovations. A hyperspectral surface reflectivity/emissivity database was created that combined GOME, MODIS and ASTER measurements with dual regression fitting for the spectral gap between the near-infrared and the thermal infrared. A multiple linear regression method was developed to provide O3 retrievals over the entire CONUS region from selected full optimal estimation retrievals. Since NO2 and HCHO have significantly more spatial and temporal variability than O3, the fast Two Stream Exact Single Scattering (2S-ESS) Radiatie Transfer (RT) model was developed to avoid regression but instead perform full optimal estimation retrievals for every cloud free grid point of the nature run. This approach resulted in a 200-fold speed increase compared to the full multiple scattering LIDORT RT model with negligible loss of accuracy. A new aerosol single scattering property database was created for six aerosol types (black carbon, dust, nitrate, insoluble and soluble organic carbon and sulfate) that spanned the entire ultraviolet to thermal infrared wavelength range and accounted for hygroscopic effects.
The results were presented at the Second Atmopsheric Composition OSSE Workshop, which was hosted by the European Center for Medium Range Weather Forecasting in Reading, UK from 9–11 November, 2016. The GEO-CAPE Regional O3 OSSE demonstrates systematic and significant increase in lower to mid tropospheric correlations and reductions in root mean square (rms) errors and biases when hourly geostationary UV/VIS, and UV/VIS/TIR ozone retrievals are assimilated, compared to UV-only measurements. Results show improvements in lower tropospheric correlations and rms errors for all experiments, but the UV and UV/VIS experiments introduce higher biases. Comparisons with the nature run at US Environmental Protection Agency surface monitoring sites show that the overall positive impacts obtained with UV/VIS/TIR retrieval assimilation are due to reductions in nighttime biases, which highlights the importance of the TIR measurements in the multi-spectral retrievals. A manuscript describing the O3 OSSE is currently in prepartion (Pierce et al., 2018).
We have also completed the NO2 OSSE and are nearing completion of the HCHO OSSE. Using the 2S-ESS RT model, we were able to generate synthetic radiances for all cloud free gridpoints at hourly intervals for the full nature run timeperiod (July 2011) utilizing the Supercomputer for Satellite Simulations and Data Assimilation Studies (Boukabara et al, 2016) at the University of Wisconsin-Madison Space Science and Engineering Center. Since NO2 and HCHO are short lived species, assimilation of NO2 or HCHO column retrievals does not lead to systematic changes in the concentrations of these species. Instead, the NO2 and HCHO column retrievals must be used to constrain the emissions of these species. As part of the Regional/Urban OSSE working group activities during 2015–2018 we developed an offline approach to use satellite based trace gas retrievals to constrain area and point source emissions. The approach involves calculating the sensitivity of the trace gas column to changes in emissions following Lamsal et al. (2011) and then using this sensitivity, combined with the monthly mean trace gas analysis increment, to adjust the emissions. The results of the NO2 OSSE were presented at the Joint Committee on Earth Observation Satellites Atmospheric Composition-Virtual Constellation and GEO-CAPE Meeting, which was hosted by the NOAA Center for Weather and Climate Prediction in College Park, MD from May 2–4, 2018. The GEOCAPE NO2 OSSE demonstrates significant adjustments in a priori NOx emissions using hourly TEMPO-like NO2 retrievals compared to daily OMI NO2 emission adjustments. However, the NO2 OSSE results show low surface ozone sensitivity to changes in NOx emissions, possibly due to high urban NOx levels leading to VOC sensitive ozone production. The O3, NO2, and HCHO assimilation experiments were conducted using the NOAA gridpoint statistical interpolation (GSI; Wu et al., 2002; Kleist et al., 2009), which is a physical space-based 3-dimensional variational analysis. The observation operator for the O3, NO2, and HCHO profile retrievals was developed for GSI based on the approach used by Verma et al. (2009) for assimilation of ozone profiles from the Tropospheric Emission Spectrometer. This observation operator accounts for the AK and a priori used in the retrieval.