Long-term column-averaged greenhouse gas observations using a COCCON spectrometer at the high-surface-albedo site in Gobabeb, Namibia

<p>In this study, we present column-averaged dry-air mole fractions of <span class="inline-formula">CO₂</span> (<span class="inline-formula">XCO₂</span>), <span class="inline-formula">CH₄</span> (<span class="inline-formula">XCH₄</span>) and <span class="inline-formula">CO</span> (<span class="inline-formula">XCO</span>) from a recently established measurement site in Gobabeb, Namibia. Gobabeb is a hyperarid desert site at the sharp transition zone between the sand desert and the gravel plains, offering unique characteristics with respect to surface albedo properties. Measurements started in January 2015 and are performed utilizing a ground-based Fourier transform infrared (FTIR) EM27/SUN spectrometer of the COllaborative Carbon Column Observing Network (COCCON). Gobabeb is the first measurement site observing <span class="inline-formula">XCO₂</span> and <span class="inline-formula">XCH₄</span> on the African mainland and improves the global coverage of ground-based remote-sensing sites. In order to achieve the high level of precision and accuracy necessary for meaningful greenhouse gas observations, we performed calibration measurements for 8 d between November 2015 and March 2016 with the COCCON reference EM27/SUN spectrometer operated at the Karlsruhe Institute of Technology. We derived scaling factors for <span class="inline-formula">XCO₂</span>, <span class="inline-formula">XCH₄</span> and <span class="inline-formula">XCO</span> with respect to the reference instrument that are close to 1.0. We compare the results obtained in Gobabeb to measurements from the Total Carbon Column Observing Network (TCCON) sites at Réunion Island and Lauder. We choose these TCCON sites because, while 4000 km apart, the instruments at Gobabeb and Réunion Island operate at roughly the same latitude. The Lauder station is the southernmost TCCON station and functions as a background site without a pronounced <span class="inline-formula">XCO₂</span> seasonal cycle. We find a good agreement for the absolute X<span class="inline-formula">_gas</span> values, apart from an expected <span class="inline-formula">XCH₄</span> offset between Gobabeb and Lauder due to significantly different tropopause height, as well as representative intraday variability between TCCON and COCCON. Together with the absence of long-term drifts, this highlights the quality of the COCCON measurements. In the southern hemispheric summer, we observe lower <span class="inline-formula">XCO₂</span> values at Gobabeb compared to the TCCON stations, likely due to the influence of the African biosphere. We performed coincident measurements with the Greenhouse Gases Observing Satellite (GOSAT), where GOSAT observed three nearby specific observation points, over the sand desert south of the station, directly over Gobabeb and over the gravel plains to the north. GOSAT H-gain <span class="inline-formula">XCO₂</span> and <span class="inline-formula">XCH₄</span> agree with the EM27/SUN measurements within the 1<span class="inline-formula"><i>σ</i></span> uncertainty limit. The number of coincident soundings is limited, but we confirm a bias of 1.2–2.6 ppm between GOSAT M-gain and H-gain <span class="inline-formula">XCO₂</span> retrievals depending on the target point. This is in agreement with results reported by a previous study and the GOSAT validation team. We also report a bias of 5.9–9.8 ppb between GOSAT M-gain and H-gain <span class="inline-formula">XCH₄</span> measurements which is within the range given<span id="page5888"/> by the GOSAT validation team. Finally, we use the COCCON measurements to evaluate inversion-optimized CAMS model data. For <span class="inline-formula">XCO₂</span>, we find high biases of 0.9 <span class="inline-formula">±</span> 0.5 ppm for the Orbiting Carbon Observatory-2 (OCO-2) assimilated product and 1.1 <span class="inline-formula">±</span> 0.6 ppm for the in situ-driven product with <span class="inline-formula"><i>R</i>²</span> <span class="inline-formula">&gt;</span> 0.9 in both cases. These biases are comparable to reported offsets between the model and TCCON data. The OCO-2 assimilated model product is able to reproduce the drawdown of <span class="inline-formula">XCO₂</span> observed by the COCCON instrument at the beginning of 2017, as opposed to the in situ-optimized product. Also, for <span class="inline-formula">XCH₄</span>, the observed biases are in line with prior model comparisons with TCCON.</p>