Observation and quantification of aerosol outflow from southern Africa using spaceborne lidar

FUNDING: None Biomass burning in Africa provides a prolific source of aerosols that are transported from the source region to distant areas, as far away as South America and Australia. Models have long predicted the primary outflow and transport routes. Over time, field studies have validated the basic production and dynamics that underlie these transport patterns. In more recent years, the advancement of spaceborne active remote-sensing techniques has allowed for more detailed verification of the models and, importantly, verification of the vertical distribution of the aerosols in the transport regions, particularly with respect to westerly transport over the Atlantic Ocean. The Cloud-Aerosol Transport System (CATS) lidar on the International Space Station has detection sensitivity that provides observations that support long-held theories of aerosol transport from the African subcontinent over the remote Indian Ocean and as far downstream as Australia.


Introduction
The African continent is a prolific source of aerosols flowing out over the Atlantic and Indian Oceans. Transport of Saharan dust off the continent and over the equatorial and North Atlantic Ocean is well documented. [1][2][3][4] It is now appreciated that dust from the African subcontinent, following that transport route, finds its way to the Caribbean and Amazon basin. [5][6][7][8][9][10] Similarly, evidence of sub-Saharan aerosol and trace gas transports comprising biomass burning smoke, dust and industrial emissions has been documented. These transports fall into three general categories: (1) out over the Atlantic Ocean (originating primarily in tropical Africa north of 20ºS) 11-18 ; (2) air mass recirculation from and over the southern portion of the subcontinent [19][20][21] ; and (3) westerly transport out over the Indian Ocean (south of 20ºS) 13,14,18,[22][23][24][25][26][27][28][29][30][31][32][33] . The transports and their emission sources mentioned above that contribute to the atmospheric aerosol loading over and off southern Africa exhibit a strong seasonality and tend to migrate from western tropical southern Africa in May to southeastern southern Africa and Mozambique in September and October.
The Southern African Fire-Atmosphere Research Initiative and Southern African Regional Science Initiative (SAFARI 92 and SAFARI 2000, respectively) were extensive field campaigns specifically designed to study the postulated aerosol and trace gas transports of combined emissions, in general, and biomass burning emissions, in particular, from the southern regions of the African continent. 31,34 While SAFARI 2000 focused on aerosol emissions and transports, it did so primarily over and very near the southern African subcontinent. Easterly and westerly atmospheric transports from southern Africa occur over expansive areas of the remote Atlantic and Indian Oceans where ground-based and sea surface measurements are sparse and airborne measurements are challenging to obtain. Understanding, following and documenting atmospheric features such as these, requires the use of atmospheric models and satellite data.
Near source regions, aerosol concentrations in outflows are dense and sufficiently optically thick to be rather easily detected by spaceborne passive sensors such as the Moderate Resolution Imaging Spectrometer (MODIS). [35][36][37] Although optically thick layers can be detected by MODIS or other passive sensors, over oceans, the aerosol optical depth cannot be accurately retrieved from MODIS for aerosol layers that have aerosol optical depths of less than 0.03. 36,37 For less optically thick outflows, active remote sensors, such as the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) 38 lidar, can be used to detect aerosol layers. However, CALIPSO also requires a minimum density of scatterers before an aerosol layer can be detected. A challenge to both passive and active sensors, as noted by Edwards et al. 39 is that high aerosol concentrations generally do not extend far from the source region. Far from the source region, aerosols are lofted and transported over the Indian Ocean. The aerosol plume tends to spread, somewhat in the horizontal but more in the vertical, thereby becoming too diffuse for spaceborne sensors to detect. This flow is in contrast to the easterly flow out over the Atlantic, which either occurs within the boundary layer, particularly over Namibia during offshore transport of surface dust 40 , or is bounded between 800 hPa and 500 hPa 41 .
Although transport models routinely predict aerosol plumes over Australia, measurements verifying the plume height and distribution are extremely limited. Some ground-based measurements from Australia have shown evidence of the outflow plume 26,28 but spaceborne measurements that can conclusively track the outflow from the source region to the Australian continent have been lacking. There were initial spaceborne lidar measurements made by the Laser In-space Technology Experiment (LITE) 42 that appear to capture a feature similar to those described in this paper during September 1994. As LITE was a technology demonstration onboard the Space Shuttle, those measurements were limited in coverage and, moreover, the 1064 nm data from LITE was never calibrated. In this paper, we present, for the first time, calibrated 1064 nm observations that support longheld (>25 years) postulated understandings of atmospheric transport modes from the biomass burning region of subequatorial Africa out over the Indian Ocean and towards Australia.

The Cloud-Aerosol Transport System
There have been only two in-space lidar sensors that have operated over multiple years to capture seasonal transport patterns: CALIPSO and the Cloud-Aerosol Transport System (CATS) onboard the International Space Station (ISS). The CATS sensor is a backscatter lidar instrument with depolarisation measurement. 43 A notable feature of CATS is the use of photon-counting detection, which permits high detection sensitivity. As a result, at least during night portions of each orbit, CATS has detection sensitivity (minimum detectable backscatter, at 1064 nm) as low as 5 x 10 -5 km -1 sr -1 , which is more sensitive than the CALIPSO minimum detection sensitivity (at 532 nm) of ~8 x 10 -4 km -1 sr -1 . 44 As noted in the previous section, although the LITE demonstration had detection sensitivity sufficient to detect diffuse aerosol layers, the limited lifetime (approximately 40 h total observation time) and the limited number of observations over the study region precludes an ability to track individual events as they vary with synoptic conditions. The more continuous and multi-year operation of CATS, coupled with the high detection sensitivity, can be used to demonstrate persistence of the aerosol outflow as well as tracking of the outflow from Africa towards Australia.
Operating from February 2015 until October 2017, CATS data contains observations of each of the outflow patterns identified in Garstang et al. 13 The most intriguing are observations of the westerly transport of aerosols out over the Indian Ocean and over Australia. The detection sensitivity of CATS at 1064 nm has enabled observations of the diffuse aerosol plumes transported off the African subcontinent over those regions, providing direct measurement of the transport predicted by Garstang et al. 13 Moreover, unlike other spaceborne lidar sensors, the CATS 1064 nm data is directly calibrated at 1064 nm 44,45 , thereby augmenting the available data record with additional wavelength information.
In addition to backscatter detection, the CATS 1064 nm channel provides a linear depolarisation measurement. The depolarisation measurement is exceptionally useful as an aid in determining cloud and aerosol type. 46,47 Relevant to African aerosol transport, where smoke and dust (and combinations of the two) are prevalent, the depolarisation ratio provides a critical determinant of aerosol type. Smoke tends to have a linear Aerosol outflow from southern Africa Page 2 of 6 depolarisation ratio of the order of 1-10%, whereas dust is in the range of 20-30%. Smoke combined with dust will lower the ratio somewhat, typically to the 10-25% range. The depolarisation measurement provides important substantiation that the elevated layers observed are, in fact, composed of smoke particles and, hence, are coming from the expected source region. The LITE demonstration did not have depolarisation measurement capability, thus making those prior measurements more challenging to relate to aerosol type.

CATS observations of westerly outflow
The CATS lidar onboard the ISS, with its unique precessing orbit, has captured multiple occurrences of westerly outflow from the African subcontinent towards Australia. The CATS data used herein are calibrated Level 1B data products, specifically 1064 nm attenuated total backscatter coefficients at a resolution of 350 m horizontal by 60 m vertical. 44 Three specific examples are described below.

Case 1: 7 September 2016
Multiple ISS passes on 7 September 2016 provide a unique Eulerian perspective with multiple snapshots of the resultant transport plume. As illustrated in Figure 1, data captured on subsequent orbits show evolution of an aerosol plume originating off the west coast of southern Africa (approximate latitude 25°S) and propagating across the Indian Ocean to south of Australia. Analysis of 5-day back trajectories, shown in Figure 2a, obtained from the Hysplit model 48 , indicates that parcels observed in the 9-km altitude range to the west of Australia on 7 September 2016 originated over the west coast of southern Africa (within the free troposphere) as well as from South America. Both these regions are large biomass burning source regions during the austral spring. The trajectory analysis illustrates how transport from the two continents merges in a transient westerly wave over southern Africa and exits the African subcontinent towards the southeast, south of the semipermanent Indian anticyclone (Figure 3). In this transport pathway, air parcels rise rapidly and within 2 days are separated from the surface layer to become a clearly defined lofted layer. This observation is consistent with the postulated expectations based on modelled outputs (e.g. Garstang et al. 13 , Tyson and D'Abreton 20 ) that the westerly plumes exiting the subcontinent in a westerly wave tend to ascend over the southern Indian Ocean, facilitating rapid transport toward Australasia.    30 found in the long-range transport of nitrogen dioxide plumes between southern Africa and Australia. Such transport is also consistent with previously described transports of water vapour 49 , trace gases 50 and aerosols 51 off the subcontinenttransports that have been demonstrated to impact atmospheric chemistry and composition as well as possibly the biogeochemical cycling of precipitation and the ocean surface along the path of transport 52 . Figure 4 shows a cross-section of the elevated plume as it approaches the west coast of Australia. As seen in Figure 4b, the elevated layer is distinct and extends from 3 km up to about 11 km, but with low backscatter of <5 x 10 -4 km -1 sr -1 . Although covering a large vertical extent, the median optical depth of the layer west of Australia is only of the order of 0.03-0.05 (±0.008). The median depolarisation ratio (integrated through the layer) is 0.05-0.08, indicating the elevated layer is composed primarily of smoke. The layer does start with higher optical depth (mean optical depth of 0.15±0.05) over the African subcontinent, which is attenuated as it is transported across the Indian Ocean through the loss of particles by wet and dry removal processes. 25

Case 2: 11-18 October 2015
In contrast to the Eulerian view of Case 1, over the 8-day period of 11-18 October 2015, CATS captured a Lagrangian view of the evolution and transport of multiple plumes. Data captured during this period, displayed in Figure 5, show evolution of multiple dust or aerosol plumes that originate over southern Africa (approximate latitude 25°S) and propagate across the Indian Ocean to the south of Australia. Back trajectory analysis (Figure 2b) again indicates that 5-6 days are required for a plume to transit to Australia. Back trajectory analysis suggests that near the west coast of Australia the layer should be in the 3-km altitude range, and CATS profiles show the layer extending from an altitude of about 1 km up to about 8 km.
The median optical depth of the layer west of Australia is of the order of 0.01-0.03 (±0.09). The median depolarisation ratio (integrated through the layer) is 0.05-0.10, again indicating primarily smoke.

Case 3: 13-19 September 2016
Similar to Case 2, this example presents a Lagrangian view of a plume transiting off southern Africa towards Australia. Figure 6 shows data captured during the period 13-19 September 2016, highlighting the evolution of a plume that transits directly over Australia and then continues on to the south. Similar to the other two cases, back trajectory analysis (Figure 2c) again indicates that 5-6 days are required for the plume to transit to Australia. Back trajectory analysis suggests the layer should be in the 5-6 km altitude range, and CATS profiles show the layer extending from an altitude as low as 1 km up to as high as 10 km. Although covering a large vertical extent, the median optical depth of the layer over Australia is low, only of the order of 0.008-0.01 (±0.003) and again becomes progressively lower the farther east the plume travels. The median depolarisation ratio (integrated through the layer) is 0.02-0.08, once again indicating that the layer is composed primarily of smoke. Garstang et al. 13 showed that during the dry season in southern Africa (April through October), the dominating synoptic weather pattern is an anticyclonic circulation that results in horizontal recirculation at spatial scales as high as thousands of kilometres. Aerosols exit this anticyclonic flow in the southernmost part of Africa to the east into the Indian Ocean via westerly wave and trough disturbances. These westerly disturbances peak in the spring months (September-November) and in very dry seasons such as those observed during the SAFARI project in 1992, can direct as much as 90% of aerosol transport into the Indian Ocean. 13

Conclusions
Biomass burning in Africa has long been recognised as a significantly important source of aerosols and trace gas. Focused field studies, such as SAFARI 92 and SAFARI 2000, validated the primary source and outflow patterns for smoke and trace gases from the African subcontinent. The primary outflow and transport routes from Africa to the Atlantic Ocean and over to South America, and from Africa to the Indian Ocean and Australia, have long been predicted via models.
While measurements of a number of these transports have been captured at least spatially and temporally, it was not until the advent of spaceborne active remote sensing by lidar that characterisation of these transports in the vertical became possible. Even with spaceborne sensors, detection of aerosol plumes can only be accomplished if the aerosol concentration is sufficient to meet minimum detection thresholds. An aerosol layer that is dense and easily detectable near the source region eventually spreads and disperses beyond the minimum detectable limit for the sensor.
Hence, discerning information on aerosol and trace gas transports has been heavily reliant upon modelled information which is itself suspect in such a data-limited part of the world as the African continent and remote Indian Ocean. The CATS lidar on the ISS had detection sensitivity sufficient to identify diffuse elevated aerosol plumes originating on the African subcontinent and transported towards Australia. As a result, CATS has provided an opportunity to test and validate long-held assumptions regarding the nature of aerosol and trace gas transports in these remote regions of the world.

Data availability
The CATS data used in this paper are archived in NASA's Atmospheric Science Data Center (ASDC) Distributed Active Archive Center (DAAC), and are accessible via the CATS website (https://cats.gsfc.nasa.gov).