The Mineral Aerosol Profiling from Infrared Radiances (MAPIR) algorithm: version 4.1 description and evaluation

<p>The Mineral Aerosol Profiling from Infrared Radiances (MAPIR) algorithm retrieves vertical dust concentration profiles from cloud-free Infrared Atmospheric Sounding Interferometer (IASI) thermal infrared (TIR) radiances using Rodgers' optimal estimation method (OEM). We describe the new version <span class="inline-formula">4.1</span> and evaluation results. Main differences with respect to previous versions are the Levenberg–Marquardt modification of the OEM, the use of the logarithm of the concentration in the retrieval and the use of Radiative Transfer for TOVS (RTTOV) for in-line radiative transfer calculations. The dust aerosol concentrations are retrieved in seven <span class="inline-formula">1</span>&thinsp;<span class="inline-formula">km</span> thick layers centered at <span class="inline-formula">0.5</span> to <span class="inline-formula">6.5</span>&thinsp;<span class="inline-formula">km</span>. A global data set of the daily dust distribution was generated with MAPIR v<span class="inline-formula">4.1</span> covering September <span class="inline-formula">2007</span> to June <span class="inline-formula">2018</span>, with further extensions planned every 6 months. The post-retrieval quality filters reject about <span class="inline-formula">16</span>&thinsp;% of the retrievals, a huge improvement with respect to the previous versions in which up to <span class="inline-formula">40</span>&thinsp;% of the retrievals were of bad quality. The median difference between the observed and fitted spectra of the good-quality retrievals is <span class="inline-formula">0.32</span>&thinsp;<span class="inline-formula">K</span>, with lower values over oceans. The information content of the retrieved profiles shows a dependence on the total aerosol load due to the assumption of a lognormal state vector. The median degrees of freedom in dusty scenes (min <span class="inline-formula">10</span>&thinsp;<span class="inline-formula">µm</span> AOD of <span class="inline-formula">0.5</span>) is <span class="inline-formula">1.4</span>. An evaluation of the aerosol optical depth (AOD) obtained from the integrated MAPIR v<span class="inline-formula">4.1</span> profiles was performed against <span class="inline-formula">72</span> AErosol RObotic NETwork (AERONET) stations. The MAPIR AOD correlates well with the ground-based data, with a mean correlation coefficient of <span class="inline-formula">0.66</span> and values as high as <span class="inline-formula">0.88</span>. Overall, there is a mean AOD (<span class="inline-formula">550</span>&thinsp;<span class="inline-formula">nm</span>) positive bias of only <span class="inline-formula">0.04</span> with respect to AERONET, which is an extremely good result. The previous versions of MAPIR were known to largely overestimate AOD (about <span class="inline-formula">0.28</span> for v<span class="inline-formula">3</span>). A second evaluation exercise was performed comparing the mean aerosol layer altitude from MAPIR with the mean dust altitude from Cloud–Aerosol LIdar with Orthogonal Polarization (CALIOP). A small underestimation was found, with a mean difference of about <span class="inline-formula">350</span>&thinsp;<span class="inline-formula">m</span> (standard deviation of about <span class="inline-formula">1</span>&thinsp;<span class="inline-formula">km</span>) with respect to the CALIOP cumulative extinction altitude, which is again considered very good as the vertical resolution of MAPIR is <span class="inline-formula">1</span>&thinsp;<span class="inline-formula">km</span>. In the comparisons against AERONET and CALIOP, a dependence of MAPIR on the quality of the temperature profiles used in the retrieval is observed. Finally, a qualitative comparison of dust aerosol concentration profiles was done against lidar measurements from two ground-based stations (M'Bour and Al Dhaid) and from the Cloud–Aerosol Transport System (CATS) instrument on board the International Space Station (ISS). MAPIR v<span class="inline-formula">4.1</span> showed the ability to detect dust plumes at the same time and with a similar extent as the lidar instruments. This new MAPIR version shows a great improvement of the accuracy of the aerosol profile retrievals with respect to previous versions, especially so for the integrated AOD. It now offers a unique <span class="inline-formula">3</span>-D dust data set, which can be used to gain more insight into the transport and emission processes of mineral dust aerosols.</p>