Editing IPCC:AR6/SRCCL/Chapter-2 (section)

==== 2.4.1.2 Effects of past climate change on dust emissions and feedbacks ====

<div id="section-2-4-1-2-effects-of-past-climate-change-on-dust-emissions-and-feedbacks-block-1"></div>

A limited number of model-based studies found that dust emissions have increased significantly since the late 19th century: by 25% from the preindustrial period to the present day (e.g., from 729 Tg yr <sup>–1</sup> to 912 Tg yr <sup>–1</sup> ) with about 50% of the increase driven by climate change and about 40% driven by land use cover change, such as conversion of natural land to agriculture ( ''low confidence'' ) (Stanelle et al. 2014 <sup>[[#fn:r821|821]]</sup> ). These changes resulted in a clear sky radiative forcing at the top of the atmosphere of –0.14 W m <sup>–2</sup> (Stanelle et al. 2014 <sup>[[#fn:r822|822]]</sup> ). The authors found that, in North Africa, most dust is of natural origin, with a recent 15% increase in dust emissions attributed to climate change. In North America two-thirds of dust emissions take place on agricultural lands and both climate change and land-use change jointly drive the increase; between the pre-industrial period and the present day, the overall effect of changes in dust was –0.14 W m <sup>–2</sup> cooling of clear sky net radiative forcing on top of the atmosphere, with –0.05 W m <sup>–2</sup> from land use and –0.083 W m <sup>–2</sup> from changes in climate.

The comparison of observations for vertically integrated mass of atmospheric dust per unit area (i.e., dust mass path (DMP)) obtained from the remotely sensed data and the DMP from CMIP5 models reveal that the model-simulate range of DMP was much lower than the estimates (Evan et al. 2014 <sup>[[#fn:r823|823]]</sup> ). ESMs typically do not reproduce inter- annual and longer timescales variability seen in observations (Evan et al. 2016 <sup>[[#fn:r824|824]]</sup> ). Analyses of the CMIP5 models (Evan 2018 <sup>[[#fn:r825|825]]</sup> ; Evan et al. 2014 <sup>[[#fn:r826|826]]</sup> ) reveal that all climate models systematically underestimate dust emissions, the amount of dust in the atmosphere and its inter- annual variability ( ''medium confidence'' ).

One commonly suggested reason for the lack of dust variability in climate models is the models’ inability to simulate the effects of land surface changes on dust emission (Stanelle et al. 2014 <sup>[[#fn:r827|827]]</sup> ). Models that account for changes in land surface show more agreement with the satellite observations both in terms of aerosol optical depth and DMP (Kok et al. 2014 <sup>[[#fn:r828|828]]</sup> ). New prognostic dust emissions models are now able to account for both changes in surface winds and vegetation characteristics (e.g., leaf area index and stem area index) and soil water, ice and snow cover (Evans et al. 2016 <sup>[[#fn:r829|829]]</sup> ). As a result, new modelling studies (e.g., Evans et al. 2016 <sup>[[#fn:r830|830]]</sup> ) indicate that, in regions where soil and vegetation respond strongly to ENSO events, such as in Australia, inclusion of dynamic vegetation characteristics into dust emission parameterisations improves comparisons between the modelled and observed relationship with long-term climate variability (e.g., ENSO) and dust levels (Evans et al. 2016 <sup>[[#fn:r831|831]]</sup> ). Thus, there has been progress in incorporating the effects of vegetation, soil moisture, surface wind and vegetation on dust emission source functions, but the number of studies demonstrating such improvement remains small ( ''limited evidence, medium agreement'' ).

<div id="section-2-4-1-3-future-changes-of-dust-emissions"></div>

<span id="future-changes-of-dust-emissions"></span>