Intercomparison of MAX-DOAS, FTIR and direct sun DOAS HCHO retrievals in Xianghe (China)

MAX-DOAS, direct sun DOAS and FTIR measurements are increasingly used as fiducial reference measurements (FRM) for the validation of HCHO satellite observations. Understanding their strengths and limitations, and assessing their consistency is therefore crucial to produce robust and consolidated validation results. So far, only a few studies have explored the complementarity between MAX-DOAS and FTIR HCHO measurements.

In the present study, we take benefit of MAX-DOAS and FTIR instruments being simultaneously operated at the Xianghe station (39.75° N, 116.96° E, approximately 55 km southeast of Beijing) to compare HCHO vertical columns (VCDs) retrieved from both instruments during one full year in a site under the influence of strong VOC emissions from biogenic and anthropogenic origins. In addition to its standard MAX-DOAS geometry, the IAP/BIRA instrument also provides regular direct sun measurements suitable for comparison with FTIR solar absorption data. The two direct sun measurements, in the UV and IR, show an excellent agreement and form a reliable HCHO VCD reference.

We investigate results obtained using MMF and MAPA MAX-DOAS algorithms and their combined use within the FRM4DOAS centralized processing facility, assessing the agreement reached with respect to the direct sun reference data and investigate the reasons for the observed differences. A good correlation (~0.96) but with a systematic under-estimation of about -20% is found for all the MAX-DOAS approaches. We explore whether this discrepancy can be understood by the known lack of sensitivity of MAX-DOAS measurements in the free-troposphere, above 4 km of altitude. Changing the MMF a-priori profiles to monthly averaged profiles coming from CAMS or TM5 models shows a reduction of the bias of 10 to 15% with respect to the direct sun reference data. By further considering the MAX-DOAS and FTIR respective vertical sensitivities through application of their averaging kernels, we reduce the remaining bias to -2% for the different MAX-DOAS datasets.