Models transport Saharan dust too low in the atmosphere compared to observations

We investigate the dust forecasts from two operational global atmospheric models in comparison with in-situ and remote sensing measurements obtained during the AERosol properties – Dust (AER-D) field campaign. Airborne elastic backscatter lidar measurements were performed on-board the Facility for Airborne Atmospheric Measurements during August 2015 over the Eastern Atlantic, and they permitted to characterize the dust vertical distribution in detail, offering insights on transport from the Sahara. They were complemented with 15 airborne in-situ measurements of dust size-distribution and optical properties, and datasets from the CloudAerosol Transport System spaceborne lidar (CATS) and the Moderate Resolution Imaging Spectroradiometer (MODIS). We compare the airborne and spaceborne datasets to operational predictions obtained from the Met Office Unified Model (MetUM) and the Copernicus Atmosphere Monitoring Service (CAMS). The dust aerosol optical depth predictions from the models are generally in agreement with the observations, but display a low 20 bias. However, the predicted vertical distribution places the dust lower in the atmosphere than highlighted in our observations. This is particularly noticeable for the MetUM, which does not transport coarse dust high enough in the atmosphere, nor far enough away from source. We also found that both model forecasts underpredict coarse mode dust, and at times overpredict fine mode dust. An analysis of the processes driving dust uplift in the models suggests that errors in the large scale wind and dust size distribution at source could be the cause of differences 25 between model predictions and observations of the Saharan Air Layer. Mineral dust is an important component of the climate system, therefore it is important to assess how models reproduce its properties and transport mechanisms.


Airborne lidar
The Leosphere ALS450 elastic backscatter lidar (wavelength 355 nm) is deployed on the FAAM aircraft in a 180 nadir-viewing geometry. Marenco et al., 2011, andMarenco, 2013 describe the methodology for converting lidar beam returns at 355 nm wavelength into profiles of aerosol extinction coefficient. The system specifications are summarised in Marenco et al., 2014 and references therein, and a further description of the data processing methodology can be found in Marenco et al., 2016. During processing, the lidar data was integrated to 1 min temporal resolution, which corresponds to a 9 ± 2 km footprint at typical aircraft speeds. Smoothing to a 45 m 185 vertical resolution was also applied to reduce the effect of shot noise. The vertical profiles were processed using a double iteration. First we determined the lidar ratio (extinction-to-backscatter ratio), and subsequently we processed the full data set to determine the extinction coefficient and AOD (see Marenco et al., 2016 and references therein, where the same methodology is applied). The first iteration was conducted on a subset of the vertical profiles, where the signature of Rayleigh scattering above the dust layer could clearly be identified to 190 enable the lidar ratio to be determined. We obtained a campaign mean lidar ratio of 54 ± 8 sr, which is in reasonable agreement with other measurements of the lidar ratio for dust at 355 nm (Lopes et al., 2013). This value of the lidar ratio was subsequently used to process the full dataset in the second iteration. On average during this campaign, the uncertainty in the derived dust extinction coefficient was 8%; however with a significant variability of this figure in both the vertical and horizontal. The uncertainty is smaller than this near 195 the top of the profile (closer to the aircraft) and larger nearer the ground. The methodology described in Marenco et al., (2016) was used here.

In-situ aerosol measurements
A number of wing-mounted instruments permitted us to measure the aerosol size distribution between 0.1 and 100 μm. The Passive Cavity Aerosol Spectrometer Probe (PCASP;Liu et al., 1992;Osborne et al., 2008;200 Rosenberg et al., 2012) measured optical size from 0.1-2.5µm. The cloud droplet probe (CDP-100;Lance et al., 2010;Rosenberg et al., 2012) measured larger particles with diameters 5-40 µm (Knollenberg, 1981), and the two-dimensional stereo probe (2DS) measured large aerosol particles up to ~100 µm. Calibration of the PCASP was done before and after the campaign, whereas the CDP was also calibrated before most flights. The PCASP and CDP measurements (d <20 µm) and their calibration for the ICE-D campaign are discussed in more detail in 205 Ryder et al (2018), where the full size distribution measurements are described. The particle size spectra have been processed for an assumed refractive index for dust of 1.53−0.001i, thus correcting for the bin ranges calibrated using polystyrene latex spheres, and the first bin has been discarded due to its undefined lower edge.
The 2DS is a shadowing probe with 10 µm resolution, and it does not rely on refractive index to infer particle size. Profiles of in situ measurements were acquired on slant trajectories through the atmosphere (aircraft 210 profiles).

Satellite data
Two sources of satellite data are used here, the Cloud-Aerosol Transport System (CATS) and the Moderate Resolution Imaging Spectroradiometer (MODIS). CATS is a multi-wavelength lidar instrument (wavelengths 532 https://doi.org/10.5194/acp-2020-57 Preprint. Discussion started: 16 April 2020 c Author(s) 2020. CC BY 4.0 License.
The MetUM extinction coefficient only includes dust, which could potentially make the results lower compared 270 to total aerosol extinction which also includes other aerosol types. However, data from the CATS lidar, as well as the in-situ measurements including filter samples discussed in Ryder et al (2018), confirm that the aerosol sampled during AER-D/ICE-D was predominantly dust, with a contribution from marine aerosol in the MBL. For this reason, for this study we neglect the conceptual difference between the dust-only extinction of the MetUM, and total-aerosol properties in CAMS and the observations. https://doi.org/10.5194/acp-2020-57 Preprint. Discussion started: 16 April 2020 c Author(s) 2020. CC BY 4.0 License.

Individual case studies
Case study 1: 7 th August 2015, B920 (Fig 2 -7). This flight took place near Praia and was co-located with an 285 overpass of the CATS spaceborne lidar. There were two high level sections during the flight that have been looked at, R1 and R6 (see Table 1 for run times and locations). Fig 2 displays the airborne, spaceborne and model data for R6, which coincided with a CATS overpass. A deep dust layer was observed between ~ 2 and 5 km, with marine aerosol mixed with dust in the boundary layer, and a broken cloud field at the top of the boundary layer. Both the extent and amount of aerosol observed agree well between the airborne and the spaceborne lidars 290 (Fig 2 a and d). The aerosol type classification from CATS (not shown here) also agrees well with the in-situ measurements, which found a marine aerosol layer below the dust layer. The dust layer was well mixed, with moderate extinction coefficients (100 -180 Mm -1 ) and AOD's between 0.28 and 0.44 observed by the airborne lidar.  Table 1. For this run, the MetUM mean extinction was 55± 38 Mm -1 , ECMWF forecast 58± 41 Mm -1 and the aircraft lidar measured a mean extinction value of 56± 40 Mm -1 . surface. It is moreover dominated by the smaller size bin (d1, 0.2 -4.0 μm diameter), in particular for the aerosol below 1km primarily. The concentration predicted by CAMS for this case is less than half of that in the MetUM, and however the magnitude of the predicted extinction is similar, although with differences in the dust layering. typically makes up about a third of the total dust concentration measured, and d2 is around two thirds. In contrast, the measurements show very little dust in the CAMS d1 and d2 size bins, and the concentration in the d3 size bin https://doi.org/10.5194/acp-2020-57 Preprint. Discussion started: 16 April 2020 c Author(s) 2020. CC BY 4.0 License.

In
is very close to the total measured. Comparing the model data (lines with markers on) to the measurements (lines 320 of the same colour with no markers) in Fig 4 and 5 we can see that both models struggle to accurately capture dust concentration for each size bin. This adds to the difficulty in attributing dust to the right altitude. For example, in P2 (Fig5a) the MetUM has more d1 dust than d2, while the aircraft measurements show the opposite.
For the same profile (Fig 5b), CAMS has more d2 dust than d3, however the measurements show that there is very little d1 or d2 dust, and the predicted CAMS d3 is about a factor of 3 too low.

325
Temperature and specific humidity profiles from the aircraft in-situ instruments were also compared with data from the MetUM and ECMWF. An example is shown for this flight for P2 (Fig 6) and P7 (Fig 7). The temperature profiles are in good agreement for all profiles looked at, with no systematic bias for either model.
Both models also generally get the humidity profiles about right, capturing the main features. Generally, the models predict a correct vertical structure of the atmosphere in terms of thermodynamic profiles; however the 330 predicted dust vertical distribution seems to depart excessively from the thermodynamic structure.
Case study 2: 12 th August 2015, B923 and B924 (Fig 8-11). Flights B923 and B924 both took place on the 12 th August flying between Praia and Fuerteventura to sample the outflow from a dust uplift event that had happened on the 10 th August in Northern Mali. These flights were able to reach the main dust plume, which means that highest AOD's and extinction coefficients of the campaign were measured on this day (Marenco et al, 2018). The 335 two flights sampled the same plume at different times during the day, and only B923 is shown here as results for flight B924 are similar. The AOD measured by the airborne lidar reached 2, with an aerosol extinction coefficient of 100 -1300 Mm -1 near the Western African coast. As in the previous case study, both models captured the spatial distribution of the dust AOD well (Fig 8 d-f), however they also under-predicted the AOD.
For this section of flight B923, both models showed a dust layer up to ~ 5 km, with an enhanced extinction 340 coefficient at 13-17°W, between the surface and 1 km, where the extinction coefficient increases from an average in-layer value of 100-150 Mm -1 to 500-700 Mm -1 (Fig 8a-b and Fig 9). This spatial distribution along the flight track is similar to the observed one (Fig 8c and 9); however, the maximum dust extinction is observed at ~ 1 km altitude, whereas the models predict it closer to the surface, and the dust maximum extinction coefficient along the flight track was under-predicted in the MetUM and CAMS (Table 1) In P1 the measurements show very large amounts of dust, up to 3000 μg/m -3 concentration (Fig 10), with both models predicting an order of magnitude less. Interestingly in this aircraft profile, which is closer to the area affected by the intense dust, both models have the greatest proportion of dust in the largest size bins, in 355 agreement with the in situ measurements. https://doi.org/10.5194/acp-2020-57 Preprint. Discussion started: 16 April 2020 c Author(s) 2020. CC BY 4.0 License.
for this case study. Note that the CAMS d2 dust mass concentration of R1 (Fig 14b) and P4 (Fig 15b) is virtually identical to the d3 mass concentration, with the two lines over-lapping.

395
For all the case studies the MetUM and ECMWF global dust forecasts capture the spatial distribution of dust AOD reasonably well in comparison with observations. The model predictions show some positioning errors compared to MODIS AOD, and this can affect the local comparisons made at the aircraft location. The models tend to underpredict the AOD, but as shown by case study 4 the opposite can also occur.
The model prediction of the vertical distribution of the dust extinction coefficient is not always consistent with 400 observations. As a general rule, both models tend to have the dust too low in the atmosphere compared with the observations, with ECMWF generally doing a better job at capturing elevated dust layers. In the next section we will use data from the CATS spaceborne lidar, in comparison with predictions from the MetUM, to investigate what could be causing the observed discrepancies in the dust vertical distribution.
We noted very large differences between the measured and modelled dust concentration, associated however with 405 a modelled extinction closer to the observations, which may appear surprising because concentration is the modelled variable, from which optical properties are computed. We need to bear in mind, however, that AOD is the mostly used metric used to compare aerosol model predictions and observations: AERONET AOD is often used in model verification, and both the MetUM and the CAMS model use MODIS AOD in data assimilation. It is not so surprising, therefore, that modelled optical properties are pulled towards the observations, even when the 410 microphysical properties from which they are computed are out of scale (in this case, an underestimated dust concentration). Finer particles make a greater contribution to the aerosol extinction coefficient per unit mass than coarser ones, and the mismatch between the representation in concentration and in optical properties can be compensated in the models through the size-distribution. For all the aircraft profiles studied here, the models have too much of the dust concentration in the smaller size bins, meaning that an underpredicted dust concentration 415 can yield an aerosol extinction coefficient of the right order of magnitude.
For the flights which sampled dust nearer the source regions (case studies 2 and 4) the models had a greater proportion of dust concentration in larger size bins than for the other flights. This seems to indicate that the models may represent the dust size distribution better nearer the source. The observations from the AER-D and ICE-D campaigns suggest that, as the dust travels away, the observed size distribution changes little, with large 420 particles transported in significant quantities as far as Cape Verde (Liu et al., 2018;Ryder et al., 2018). In contrast, the models appear to lose particles from the larger size bins rapidly with increasing dust mass age, due to the gravitational sedimentation processes.

425
We compared almost every CATS overpass covering North Africa and the Eastern Atlantic during AER-D and ICE-D with the MetUM. CATS and model data were compared for overpasses between 6 and 25 August 2015, in the study region off the Western African coast between 40°N and 10°S latitude and 40°W and 40°E longitude, for a total of 45 overpasses. The four most significant cases are discussed here. For each overpass, the CATS aerosol extinction coefficient was compared with the MetUM dust extinction coefficient, and the modelled 430 contribution to extinction of each of the two size bins was also analysed.
In fig 16, a CATS overpass at 00 UTC on the 7 th August over the African continent is shown, with significant amounts of dust between 1 and 7 km. The MetUM predicts the dust in more or less the right places across the CATS track, but underpredicts the magnitude of the extinction coefficient. As for the case studies in the section 4.1, most of the predicted dust is also lower in altitude than in the observations (below 5km) than observed and 435 extends to the surface (although the model does predict some dust reaching as high as 7 km). The smaller size bin contributes most to the modelled extinction coefficient.
In fig 17,  African coast and then moves over the ocean. As in the previous example, the model predicts a deep dust layer extending up to 6km. The model does predict some dust over the ocean, but beyond 15°W the aerosol extinction is again underestimated, with no dust in division d2 over ocean.

450
Two things stand out from the above examples: (1) over the African continent, where the dust is uplifted, the model generally agrees better with the observations than over the ocean further away from the source region, and (2) the smaller dust particles (division d1) in the model reach the same altitude as the dust layer observed by CATS, but the coarser particles (division d2) appear to be distributed much lower in the atmosphere (e.g Fig 16, 18 and 19). As already mentioned, we looked at similar plots for 45 overpasses in total, and the comparison gave 455 similar results.
In the MetUM there is a size dependence in the dust uplift scheme, where finer particles are lofted more easily.
However, previous studies suggest that the MetUM division d2 dust would be expected to reach higher altitudes away from source regions than it does. The behaviour downstream from the source seems to indicate that as the

Effect of large scale wind and boundary layer height
In this section we investigate potential drivers for the observed discrepancies in the vertical distribution of dust in the MetUM and ECMWF CAMS. This is a difficult task as there are many competing factors that influence how dust is lifted into the atmosphere and subsequently transported, and these vary considerably between models. In 465 the MetUM the three processes which are most likely to have an impact on the vertical distribution of dust are the convection scheme, boundary layer (BL) height at source, and the largescale wind. Looking at the largescale wind field and BL height should show whether the modelled dust layer height is controlled by the largescale wind or by boundary layer mixing processes at the source. If examination of these processes cannot explain why the dust is too low in altitude, then the most likely cause is to be researched in the convection scheme. There is, 470 however, no direct measure of convection in the model output fields from the MetUM, and therefore any influence can only be inferred from the data that is available to us.
Backtrajectories from HYSPLIT and NAME and SEVIRI dust RGB imagery were used to determine the central trajectory of the dust sampled during each case study from source (Fig 1). The dust concentration for each size bin, the large-scale wind (w) and the BL height were extracted from the model output along the track, and plotted 475 as a cross-section every 6 hours from the time of uplift to the time of sampling by the aircraft. . This suggests that problems with the BL height in the MetUM may not be the cause for the dust layer being represented too low in the atmosphere away from the source region.
From the data presented here it is not possible to determine how well the models represent largescale wind in the dust source regions. Previous studies which have looked at this issue more comprehensively do however suggest 485 that there is an underprediction of wind fields in the models, which is also linked to coarse resolution modelling (eg. Chouza et al., 2016). Evan et al, (2016) showed that desert dust emission is to first order a function of wind speed, and it is against this quantity that models parametrise the dust source. Therefore, it seems reasonable that improving the wind speed in the models is a key part of getting the amount of dust uplift right.

Conclusions
The vertical distribution, particle size distribution, and mass concentration are the key properties that are predicted in a dust transport model. On the other hand, the main observable quantity on a global scale is https://doi.org/10.5194/acp-2020-57 Preprint. Discussion started: 16 April 2020 c Author(s) 2020. CC BY 4.0 License. aerosol optical depth, from AERONET (Holben et al., 1998), MODIS (Hsu et al, 2004, 2006, 2013Levy et al, 2013;Sayer et al, 2013Sayer et al, , 2014, and potentially other sources such as PMAP (Lang et al, 2017), VIIRS 495 (Hsu et al, 2019, and several others. Aerosol optical depth is at the same time an optical property and a vertically integrated quantity, meaning that a same observable AOD can be retrieved e.g. with differing combinations of concentration and particle size distribution, or with a differing vertical distribution. It is good practice to pull the model towards the observations, and this can be achieved through tuning and through data assimilation: this means that we can expect a good model to yield a sensible prediction of the 500 AOD. This is however insufficient to state that the underlying microphysical properties, from which AOD is derived, are correctly balanced. The vertical distribution and particle size distribution heavily affect how dust is transported and how quickly it is deposited. Wind speed and direction are altitude dependent, meaning that transport is heavily dependent on the altitude of a layer. Residence time and transport range are affected by both the particle 505 size distribution (coarse particles tend to be deposited more quickly) and vertical distribution (turbulent mixing in the boundary layer speeds up deposition, compared to the free troposphere). The representation of these properties in a model can affect the predicted AOD gradient across the Atlantic, for example. All this means that in the case of a model constrained by AOD observations only, other processes may need to compensate for a potential imbalance in the microphysical representation, such as e.g. the intensity of 510 sources and sinks. The microphysical properties and the three-dimensional spatial distribution of dust are thus deeply interconnected.
We have used a combination of remote sensing and in-situ measurements to characterize the vertical distribution and transport of Saharan dust over the Eastern Atlantic and West Africa during August 2015, and to evaluate the dust forecasts from two operational global atmospheric models (MetUM and ECMWF 515 CAMS). The dust AOD predictions at short forecast lead times from both models were in agreement with the aircraft, satellite, and AERONET observations, but with a low bias (note that both models assimilate AOD). On the other hand, we found that the vertical distribution of aerosol extinction coefficient and dust concentration could benefit from improvements. Our results show that the predicted vertical distribution places the dust low in the atmosphere, when compared to observations. Agreement between measured and 520 modelled profiles was better near source, with differences increasing downstream. These results are in agreement with previous studies (e.g Kim et al., 2014, Ansmann et al., 2017. This issue was particularly noticeable in the MetUM, where the coarser dust was not transported high enough in the atmosphere, or far enough away from source, compared with the observations. This suggests that the model could be settling the coarse mode dust too quickly, and similar findings have also been 525 observed in previous studies (e.g Kim et al., 2014;Mona et al., 2014;Binietoglou et al., 2015). We also found that both models underpredict the coarse mode, and overpredict the fine mode. The discrepancy between the magnitude of the measured and modelled extinction coefficient is much less than for the concentration profiles. This is likely to be due to the microphysical representation, since small particles are more optically efficient. Due to MODIS AOD data assimilation, and model tuning against AERONET 530 https://doi.org/10.5194/acp-2020-57 Preprint. Discussion started: 16 April 2020 c Author(s) 2020. CC BY 4.0 License. observations, the large under prediction of coarse mode dust in the models is compensated with a relatively small effect on the forecast average extinction coefficient and aerosol optical depth, even with the discrepancies in size distribution and dust concentration.
The overestimation of dust concentration in the finer ECMWF CAMS bins, and the underestimate of coarser dust is something that ECMWF are aiming to address in the future. In order to do this an updated 535 dust emission scheme based on Remy et al (2019) using the Kok et al., (2012) estimates of size distribution at emission would be used. It is expected that this would increase the total dust concentration and shift it to the larger sizes, thus keeping total extinction similar to its present values, but more accurately representing the dust size distribution. After these changes have been implemented, a further study like the present one can help quantify the improvement introduced.

540
We have also investigated the processes driving dust uplift in the models, and our analysis suggests that uncertainties in the large-scale wind and the emitted size distribution are likely causes of differences between observations of the Saharan Air Layer (SAL) and MetUM predictions. The crude representation of the dust size-distribution in the MetUM 2-bin dust scheme is another important factor. The MetUM operational dust forecast is intended to be used primarily for AOD forecasts and extinction for visibility 545 purposes, and although improvements of the microphysical properties would be desirable, the current implementation is satisfactory to an extent, and has the advantage of being computationally cheap. We also note that the dust scheme used in the Met Office climate model differs, using 6 size bins rather than 2, with the 6-bin version yet to be evaluated as in this article.
The scheme used to represent dust microphysical properties in models deserves attention as a key element 550 to pursue accurate mineral dust predictions. Simple schemes (such as for instance the 2-bin dust sizedistribution in the operational version of the MetUM) have the obvious advantage of being viable in terms of computing resources required, but on the other hand there is the consequence of giving a less accurate representation of the microphysical properties. This could be addressed by increasing the number of variables used to represent the size distribution, for example by using a scheme with 2 or more modes, each 555 defined by 2 variables, such as in the GLOMAP-mode aerosol scheme in UKCA (Mulcahy et al, 2020), although the ability of this scheme to represent the coarse and giant modes correctly still needs to be proven. Whatever approach is chosen, it needs to allow for coarse and giant particles to be represented, a capability currently missing in many models (Huneeus et al, 2011). It is to be noted that there are plans in place to move to GLOMAP dust within the operational Global MetUM in the near future, and also ongoing 560 experimentation with this scheme in the ECMWF IFS within CAMS. Moreover, there are plans to modify the latter scheme by adding a third (super coarse) mode: these are changes in the right direction.
As the size-distribution affects gravitational settling, it indirectly affects the three-dimensional distribution.