Morphometric analysis of slipface processes of an aeolian dune: Implications for grain‐flow dynamics

Grain flows are an integral part of sand dune migration; they are a direct response to the local wind regime and reflect complex interactions between localized over‐steepening of a dune slipface and complex turbulent airflow on the lee slope. Grain flows are primarily responsible for delivering sediment to the base of a dune, thus driving slipface advancement; yet, there are few constraints on their morphological and spatial characteristics or the amount of sediment that is redistributed by these flows. Using a combination of high‐resolution terrestrial laser scanning and video recordings, four distinct grain‐flow types are identified based on morphology and area on a dune slipface. Grain‐flow morphologies range from small, superficial flows to larger flows that affect greater portions of the slipface, moving significant amounts of sediment. Detailed field observations are presented of the dynamics of lee slopes, including measurements of the initiation location, thickness, magnitude and frequency statistics of grain flows, as well as volume estimates of redistributed sediment for each grain flow observed. High‐resolution laser scans enable accurate quantification of bulk sediment transfer from individual grain flows and can be used to study grain flows in a variety of environments. A categorization of grain‐flow morphologies is presented that links styles of flows with wind strength and direction, turbulent airflow, sediment deposition and environment.


INTRODUCTION
Grain flows, or avalanching, on aeolian dunes is an essential process in dune migration. The migration of dunes is largely accomplished via a cycle involving the occurrence of a grain flow that redistributes sediment downslope on the lee slope, advancing the slipface locally. Grainflow activity is followed by the restoration of the slope from grain-fall deposition and grain saltation over the brink whereupon another grain flow may take place (e.g. Bagnold, 1941;Allen, 1970;Hunter, 1977Hunter, , 1985Anderson, 1988;McDonald & Anderson, 1995;Kok et al., 2012). Grain flows have been the focus of many studies including in situ observations of active dunes, interpretations of flow signatures and paleoenvironments in aeolianites, wind tunnel experiments, and numerical modelling approaches. Despite these studies, there is still a lack of knowledge regarding grain-flow formation, initiation, magnitude, flow morphologies, and behaviour under a variety of wind conditions and environments (e.g. Walker, 1999Walker, , 2000Cupp et al., 2005;Breton et al., 2008;Sutton et al., 2013a;Nield et al., 2017). There is a need for additional field observations, wind tunnel experimentation, and numerical modelling over a wide range of aeolian settings and dune morphologies, as well as a classification system of grain flows to begin constraining spatial, morphological and dynamic characteristics of grain flows. This study presents the foundation of a classification system which can be used by future studies to describe grainflow attributes observed in the field or during wind tunnel experiments and can also be used to compare grain-flow behaviour between a variety of environments.

Grain-flow formation
The frequency of grain flows is thought to be dependent on factors such as the depositional rate (correlated to wind velocity) and sediment characteristics or environmental conditions which may affect grain movement or angle of repose (e.g. Allen, 1968Allen, , 1970Sweet & Kocurek,1990;McDonald & Anderson, 1992, 1995Nickling et al., 2002;Cupp et al., 2005;Breton et al., 2008;Sutton, 2012;Sutton et al., 2013a,b;Pelletier et al., 2015;Nield et al., 2017). However, the mechanism of formation and initiation of grain flows is more complex. Grain flows have been observed to be triggered by a localized over-steepening of the slipface slope near the dune brink where saltating grains are transported from the stoss slope onto the lee slope tens of centimetres below the brink (e.g. Allen, 1968Allen, , 1970Bor owka, 1979;Hunter, 1985;Anderson, 1988;Nickling et al., 2002;Walker & Nickling, 2002;Cupp et al., 2005;Kok et al., 2012;Sutton et al., 2013a,b). This process continues to form a bulge of sediment until it reaches the angle of initial yield (or exceeds the angle of repose of the sediment) and then fails. Failure occurs near the lower inflection point of the sediment bulge from which a tongue of sediment subsequently flows downslope (e.g. McDonald & Anderson, 1992, 1995Tischler et al., 2001;Cupp et al., 2005;Dasgupta & Manna, 2011;Sutton, 2012;Sutton et al., 2013a,b;Nield et al., 2017). As sediment is removed, a scarp forms and simultaneously migrates upslope, spreading laterally (e.g. Hunter, 1977;Fryberger & Schenk, 1981). The resulting morphology is that of an alcove or zone of depletion just below the dune brink and a tongue of sediment or zone of accumulation directly downslope of the alcove (Sutton, 2012;Sutton et al., 2013a,b).
An alternative grain-flow mechanism was described by Fryberger & Schenk (1981) from wind tunnel experiments. These authors characterized two mechanisms of grain flow including scarp recession [originally described by Hunter (1977) and related to the formation of an alcove detailed above] and slump degeneration. Fryberger & Schenk (1981) observed that slump degeneration was the most common mechanism during their experiments. Each slump began as a series of tensional features near the top of the artificial slipface. The flow progressed with compressional features (folds) forming in places where the sediment slowed and overrode other parts of the flow. The resulting 'slump sheet' produced minor deformational structures in cross-section suggesting cohesive sand, although the cause of cohesion in the dry sand was never discovered. Sutton et al. (2013b) applied the term slump degeneration to a small non-cohesive slab (100 to 400 cm 2 ) which disintegrated after flowing a few centimetres downslope. Breton et al. (2008) characterized two types of grain flows based on flow triggers during field observation. The first was named primary flow and was observed to initiate by sediment accumulation on the slipface near the brink, consistent with previous descriptions of localized slope over-steepening with the formation of an alcove and sediment tongue (Allen, 1968(Allen, , 1970Hunter, 1977;Bor owka, 1979;Fryberger & Schenk, 1981;Hunter, 1985;Anderson, 1988;McDonald & Anderson, 1992, 1995Tischler et al., 2001;Nickling et al., 2002;Walker & Nickling, 2002;Cupp et al., 2005;Dasgupta & Manna, 2011;Kok et al., 2012;Sutton, 2012;Sutton et al., 2013a,b;Nield et al., 2017). The second, referred to as a secondary flow, was thought to be initiated by disturbances on the slipface due to primary flows and were generally smaller and occurred soon after primary flows. The secondary flows reported in that study are reminiscent of grain flows that occur mid slope in 'lock up zones' on taller dunes (McDonald & Anderson, 1992, 1995, 1996 where grain flows are unsuccessful in transporting sediment to the base of the slipface. Additionally, the upper-mid slope is also known to be the location where finer, suspended particles from the sediment cloud launched from the dune brink accumulate, creating a potential secondary location for localized over-steepening and grain flow (McDonald & Anderson, 1992, 1995, 1996Nickling et al., 2002;Nield et al., 2017). However, the possible influence of a 'lock up zone' or suspended sediment settling mid slope and the occurrence of smaller grain flows was not made in Breton et al. (2008), potentially suggesting that destabilization from larger grain flows may be an additional grain-flow trigger. There also remains the possibility that the slipface slope is more easily destabilized by primary flows due to the accumulation of sediment from a 'lock up zone' or mid-slope particle settling.

Grain-flow initiation
More recently, the exploration of complex airflow patterns in the lee side of aeolian dunes using high resolution computational modelling has suggested a possible role for turbulent winds disturbing areas of the slipface and increasing the potential to trigger grain flow. For example, secondary lee slope flow patterns generated by flow separation at the dune crest and the subsequent reattachment flow have been observed to produce complex eddies and vortices (Parsons et al., 2004a,b;Jackson et al., 2013a,b;Pelletier et al., 2015;Smith et al., 2017a). While the role of these complex airflow patterns in sediment transport on the lee side of a dune is not well-understood, they may have a significant localized effect on surface shear stress, resulting in sediment redistribution (e.g. Wiggs, 2001;Nickling et al., 2002;Walker & Nickling, 2002;Cupp et al., 2005). Pelletier et al. (2015) demonstrated that the angle of initial yield can decrease with increasing wind velocities, leading to grain flows at lower critical angles than otherwise possible, and another study noted that oblique incident airflow can result in deflected lee-flow, which promotes lateral transport of sediment on a slipface (Walker, 1999). Research is ongoing in the investigation on how lee slope airflow patterns potentially affect surface shear stress and sediment transport, and may be useful in distinguishing various grain-flow formation mechanisms.

Grain-flow magnitude
Grain-flow magnitude is under debate in the aeolian community. There is a dearth of information related to grain-flow surface area as well as flow thickness. Previous studies have discounted a connection between sediment flux (wind velocity) and grain-flow magnitude where only a relationship between grain-flow frequency and sediment flux was observed (McDonald & Anderson, 1995;Cupp et al., 2005;Breton et al., 2008;Sutton, 2012;Sutton et al., 2013a,b;Pelletier et al., 2015). In contrast, Nield et al. (2017) argued that there was a correlation between grain-flow frequency as well as magnitude with greater wind velocity, where weaker winds produced smaller grain flows near the dune brink and stronger winds triggered avalanches further downslope resulting in more sediment being mobilized upslope. Additional field observations are needed to resolve the inconsistency of the relationship between wind velocity and grain-flow magnitude and frequency.
Related to the topic of grain-flow magnitude and the formation mechanism of large grain flows is the hypothesis that grain-flow thickness directly correlates with dune height. This relationship has been used to estimate palaeo dune heights from aeolianite deposits (e.g. Kocurek & Dott, 1981;Sweet et al., 1988;Kocurek, 1991) but it has not been tested extensively with modern day field observations of grain-flow thicknesses involving a variety of dune heights and morphologies. Field observations have been made of thick, large areal extent grain flows (e.g. McDonald & Anderson, 1995;Nickling et al., 2002;Breton et al., 2008) but what triggers these flows and an exact definition of what qualifies as 'large ' have not yet been determined.
There is a possibility that the formation of large (or thicker) grain flows is not entirely controlled by dune height and may be influenced by environmental conditions or involve an alternative formative mechanism (e.g. McDonald & Anderson, 1995). For example, the expansion of sediment suspension and saltation trajectories on the slipface during high velocity winds was observed to transform the localized bulge of sediment near the dune brink into a large wedge of sediment, and accumulate more sediment mid slope resulting in much larger grain flows on dunes (e.g. Nickling et al., 2002;Cupp et al., 2005;Nield et al., 2017). It has also been suggested that larger grain flows on dunes may increase in thickness and areal extent if sediment is entrained from the disruption of unstable grains deposited in the 'lock up zone' mid slope (e.g. McDonald & Anderson, 1992, 1995, 1996Nickling et al., 2002). Both examples of high-magnitude grain flows occur on large dunes but the specific relationship between height and flow thickness has not been constrained.

Grain-flow classification
Descriptions of variations in grain-flow planform morphology are extremely scarce in the literature; however, they may be the key to consolidating knowledge of grain-flow dynamics across a wide range of aeolian settings. With the growing literature and interest by researchers in slipface processes, there is now a need to provide a structure to support discrete observations of grain flows between different aeolian environments. Grain flows probably operate on a dynamic continuum, meaning that similar formation mechanisms and morphologies exist over a wide range of aeolian environments. Differing grain-flow morphologies and magnitudes may indicate failure mechanisms due to the influence of the local wind regime, turbulent wind patterns on the lee slope, grain characteristics, or other environmental factors such as presence of salts or liquid/frozen water. Therefore, a classification of grain-flow types based on surface area and planform shape is useful for comparative studies across a range of aeolian environments.
The classification for grain flows introduced in this paper aims to quantify their morphometric characteristics. It builds upon previous work but is based on new observations of formation patterns and morphology. It is not intended to be exhaustive and is flexible to future expansion with the potential for additional grain-flow morphologies from different wind regimes and climatic environments.
This study constrains grain-flow thickness using sub-millimetre resolution three-dimensional laser scans along with simultaneous video documentation. This allows better estimates of the volume of redistributed sediment for each grain-flow morphology. In addition, there is a need to undertake further work to distinguish the controls on the specific styles of flows so that classification can be usefully linked to controlling parameters from environmental factors, such as climate and wind regime, to more detailed secondary airflow patterns as well as influences from physical grain characteristics and dune morphology. Further work in these areas would benefit the interpretation of aeolian deposits (e.g. McKee et al., 1971;McKee & Fig. 1. Maspalomas dune field, Gran Canaria, Spain, study site location (located at 27°44 0 41Á2″N, 15°34 0 23Á8″W). Displayed here are satellite images of Gran Canaria island (upper left), where the white box represents the extent of the dune field magnified in the middle image. The red dot represents the location of the study site within the dune field and is magnified in the upper right-hand box where the black circle shows the location of the slipface for this study. Green lines indicate dune crests and orange outlines indicate interdune areas. The white arrow represents wind direction on the day of observation. Bigarella, 1972;Morris et al., 1972;Bigarella, 1975;Hunter, 1977;Bor owka, 1979;Fryberger & Schenk, 1981;Sweet & Kocurek, 1990;Bourke, 2005;Anderson & Walker, 2006) and further elucidate paleoenvironments of aeolianites (e.g. Kocurek & Dott, 1981;Sweet et al., 1988;Kocurek, 1991;Eastwood et al., 2012).

Field location and sediment characteristics
The study site is in the central region of the Maspalomas dune field, which has a total area of about 4 km 2 and is situated on the southern end of Gran Canaria island, Spain ( Fig. 1; Her nandez et al., 2017). All observations collected took place ca 500 m inland from the ocean at 27°44 0 41Á2″N, 15°34 0 23Á8″W on a large transverse dune. The dune field is comprised of several large transverse dunes, decreasing in size and transitioning to a mixture of small transverse and barchanoid dune shapes to the east. This dune field was chosen because of the arid climate, where the annual precipitation is <100 mm year À1 (Marzol, 1987), resulting in sparse vegetation, and regular aeolian activity, thus making it a suitable field site for dune morphodynamic studies. Observations were collected on 8 December 2014, during the winter season when the eastern and north-eastern winds are the strongest and have the most influence on dune migration (Hernandez et al., 2002;Jackson et al., 2013b;Hernandez et al., 2014;Smith et al., 2017b).
Meteorological data was retrieved from a weather station located approximately 2Á5 km south-west of the study site (27°44 0 09Á0″N, 15°35 0 45Á2″W). Between 10:04 am and 12:50 pm (Western European Time Zone) on 8 December 2014, during the time of field observation, the Maspalomas weather station recorded an average temperature of 21°C, humidity levels around 52%, and wind from ca 55°(north-east) with an average wind velocity at 1 m height of 9Á51 m sec À1 (AE0Á09 m sec À1 ) and a maximum gust speed of 23Á14 m sec À1 .
Sediments in the Maspalomas dune field are composed of a mixture of calcium carbonate/biogenic marine materials (Alcantara-Carrio, 1998;Hernandez et al., 2002) and phonolitic rocks weathered from the island volcanic deposits (Martinez, 1986) with an average grain size of 220 lm (Alcantara-Carrio, 1998). Although a compositional analysis was not conducted on the sediment at the study site, components of both carbonate/biogenic marine materials and volcanic deposits were recognized in the sediment samples. The darker, volcanically-derived sediments were easily distinguished from the carbonate marine materials due to aeolian sorting effects, where the volcanic sediments concentrated on ripples and were frequently observed coalescing along the edges of an active grain flow. Six sediment samples were collected at the study site, including the surface of the lower stoss slope, mid stoss, crest, mid lee, lower lee slope and interdune, and measured by laser diffraction. An average grain size of 244 lm was measured for the entire sand dune (excluding the interdune) and is consistent with the grain-size analysis reported by Alcantara-Carrio (1998).
Larger grains were present on the stoss slope, with a 50th percentile of 304 lm on the lower stoss slope and 218 lm mid stoss. The range of grain sizes for the lower and mid stoss slope was between 86 to 756 lm and 76 to 516 lm, respectively. The 50th percentile for grains collected from the dune crest was 254 lm with a range of 98 to 586 lm. The 50th percentiles for the mid lee slope and lower lee slope were 214 lm and 229 lm, respectively, and had identical grain-size ranges between 76 lm and 516 lm. Lastly, the 50th percentile for interdune sediment was 194 lm, ranging between 14 lm and 516 lm grain sizes. These grain sizes do not include any salt component because the laser diffraction measurements were conducted in water, thereby dissolving any salt component. Water and salt weight percentages of each sediment sample were calculated by weighing samples before and after overnight drying took place in an oven and prior to laser diffraction measurements. These percentages were calculated to investigate any potential influences of moisture on grain-flow formation.
Particle-size distribution was investigated prior to laser diffraction measurements to identify a potential salt component in the sediment samples. Sample splitting was conducted followed by standard sieve analysis using a series of nested sieves with quarter phi set. Salt was easily identified in all of the sediment samples after sieving took place because salt entirely comprised grain sizes <100 lm. The weight percentages for salt presented here are likely to be an underestimate depending on how much salt was dislodged from sand grains during the sieving process. There was no detectable amount of moisture in the sediments collected on the dune but there was a small amount of salt, ranging from 0Á7% weight near the crest to 3Á4% weight near the base of the slipface. The interdune sediment sample contained about 2Á6% water weight with 16Á2% weight salt. Knowledge of water and salt components in the sediment may be useful for future studies conducting similar grain-flow observations under differing environmental conditions and a better understanding of the relationship between moisture and grain-flow styles may be formulated.

Slipface observations
Data collection included a series of groundbased laser scans, video recordings and images taken on a portion of a slipface of a large, transverse dune ( Fig. 1; located at 27°44 0 41Á2″N, 15°34 0 23Á8″W). Twenty five consecutive laser scans of the slipface began at 10:00 am and continued until 12:50 pm using a FARO Focus 3D X 330 terrestrial laser scanner with a laser beam wavelength of 1550 nm (FARO, Lake Mary, FL, USA). Simultaneous video recording occurred between 9:55 am and 11:20 am. Observations were collected during this time of day when winds were the strongest and most consistent, and there were no periods of calm or inactivity in sediment transport.
The FARO Focus 3D X 330 is designed to capture fast and accurate measurements of complex objects or buildings, making it ideal for recording detailed measurements of grain flows in the field. The scanner is calibrated yearly to guarantee accuracy and precision, minimizing data error from the electronic, distance, intensity, temperature and angular sensors and inclinometer. The FARO Focus 3D X 330 has a scan range distance between 0Á6 mm and 330 m with a ranging error of AE2 mm at 10 m and a ranging noise error at this distance of 0Á3 mm with 90% reflectance and 0Á4 mm with 10% reflectance. These error and noise estimates increase at greater distances. The measurements collected in this study were about 1Á5 m from the lee slope, well within the 10 m distance for the above-mentioned error estimates, and were collected at full resolution (0Á61 mm m À1 , or about 160 000 points per degree), producing a highdensity point cloud totalling approximately 28Á6 million data points.
The study site dune had a height of approximately 9Á62 m and a lee slope of 31Á04°, as measured by the ground-based, high-resolution laser scanner. A 5 m span of the slipface, directly in front of the TLS (terrestrial laser scanner) is where the highest resolution observations were recorded is referred to here as the 'main slipface' (Fig. 2). The main slipface was intersected perpendicularly by a larger dune form ( Fig. 1), resulting in a smaller portion of the slipface being exposed. This exposed portion had a slipface height of 2Á2 m and a slope of 32Á15°. Laser scan data were used primarily to quantify change on the slipface slope and observations were set to record a 180°view approximately 1Á5 m from the base to the slipface, thereby capturing the entire 2Á2 m high main slipface section as well as portions of the larger 9Á62 m high slipface. Three target boards were placed around the study site and real time kinematic (RTK) Global Positioning System (GPS) points were taken of each target location to allow for accurate spatial comparisons between each laser scan. Laser scans were taken at 8 min intervals from 10:00 am to 12:00 pm at full resolution. Between 12:00 pm and 1:00 pm the scan frequency was increased to 4 min intervals, whilst retaining the resolution. Video footage and images of the lee slopes were taken concurrently with the laser scans between 9:55 am to 10:21 am, 10:40 am to 11:00 am and 11:14 am to 11:20 am using an iPad TM (Apple, Cupertino, CA, USA) with 1080p HD video recording at 30 frames per second and 8-megapixel photographs.
Post-processing of the laser scans involved filtering the data using FARO SCENE software (FARO Technologies UK Ltd, Warwickshire, UK). The filters applied to each individual scan included a dark scan points filter, a distance filter and a stray filter. Each of these filters minimized noise and eliminated erroneous data points that can be common in laser scan data. The dark scan points filter is defined by a reflectance threshold where dark points (having weak signals and low accuracy) below the defined threshold value are removed. The distance filter can be specified for any distance and points greater than that distance are deleted. This filter was useful for minimizing noise from particles, such as airborne sediment. Lastly, the stray filter removed edge effects and other outliers from the dataset.

Grain-flow morphologies and volume estimates
Four types of grain flows were identified in this study; they were classified using their surface area of grain flow and planform shape. All calculations of area and volume estimates were conducted using ground-based laser scan data registered with RTK GPS points. Volume estimates were determined using the polygonal area of redistributed sediment on the slipface in Esri ArcMap and measurements of grainflow thicknesses were collected using SCENE software.
Despite having high resolution scans of grain flows, deriving volume estimates of sediment from some individual flows proved difficult and unreliable, primarily due to the amount of noise from airborne sediment and secondarily due to the highly dynamic nature of the slipface. Noise from airborne sediment increased upslope near the brink, complicating the delineation of surface points and laser points bouncing off the sediment cloud. Removing the sediment cloud left behind small 'shadow zones' on the upper slopes of the slipface, where the laser could not penetrate to the surface. These occluded survey locations precluded an accurate area measurement of grain-flow areas using software techniques but their extent was otherwise easily discerned by an observer.
In cases where noise from airborne sediment was more subdued, another difficulty arose due to the time lapse between scans. Ideally, a volume estimate of redistributed sediment from a single grain flow could be derived if a previous scan recorded an undisturbed slipface surface prior to a grain-flow event. However, it was rare that each grain flow that was scanned had a previous scan of the same area undisturbed prior to the flow. Smaller grain flows occurred every 30 sec across the entire slipface and larger grain flows occurred every 2 to 3 min. Laser scans in this study recorded a 180°view every 4 to 8 min making it difficult to capture the ideal conditions for a reliable volume estimate of each individual grain flow.
Therefore, a modified formula of a wedge was used to approximate the shape of the grain flows and estimate the volume of sediment displaced for each grain flow. A wedge is used to approximate the thickening of grain flows downslope from their initiation point and is represented by the following formula: where b is twice the thickness of the grain flow at the base of the grain flow, h is the height of the grain flow normal to the dune substrate, a is the width of the base of the grain flow, and c is the width of the grain flow near the initiation point (Fig. 2). Height is calculated by taking the sine of slipface slope angle (32Á12°), multiplied by the length of the grain flow on the slipface surface. The volume of the wedge is halved to approximate a single side of the wedge, representing the grain-flow level with the slipface surface. For each grain flow, flow thicknesses were measured in three locations along the edge of the flow, including   immediately below the initiation point, midway down the flow, and at the base (source sink) of the grain flow (Table 1) to ensure that there was a gradual thickening of sediment, as assumed by the wedge formula. These measurements do not account for any potential concavity of the underside of the grain flow and only include areas of redistributed sediment, excluding other features associated with the flow such as alcoves and incised 'channels' upslope of the grain-flow deposit. An attempt was made to test the accuracy of the wedge formula by restoring a small portion of the slipface with a recent grain flow to pre-grain-flow conditions from a single laser scan. The grain flow selected from the scan data (#15, Table 1) had lower amounts of noise to reduce error in volume calculation. To restore the slipface, the data points containing the grain flow were removed, leaving the surrounding, undisturbed slipface on either side of the flow. The remaining data points were used to interpolate the surface prior to the flow event. Volume was then calculated by subtracting the interpolated slipface surface from the surface with the grain flow.
The calculated volume using an interpolated surface of grain flow #15 was approximately 4348 cm 3 , where the wedge method estimated a volume of 5398 cm 3 (Table 1). It is likely that the volume estimates using an interpolated surface are underestimates, especially higher upslope where the grain flow cut into the slipface below the alcove. Any redistributed sediment deposited below the interpolated surface was calculated as negative volume. The volume calculations using the wedge formula include all redistributed sediment from a grain flow, including sediment deposited where the grain flow cut into the slope. Considering the uncertainty introduced in restoring the slipface via interpolation and how redistributed sediment volume was calculated, these two estimates appear to agree well.
Locations of grain-flow initiation were primarily isolated through GPS-registered laser scans. In cases where high levels of noise were present, the grain-flow initiation point was approximated using still frame images from video recordings using the GPS-registered target boards to constrain location (Table 1). Airborne sediment ejected from the dune brink introduced significant amounts of noise at the top of the slipface in the laser scans and entirely obscured approximately 10 to 20 cm of the lee slope just below the brink. Therefore, grain flows that initiated near the brink were measured in ArcMap using images captured from video. A few grain flows reported in Table 1 do not include an initiation point because the laser scan captured the grain flow after the alcove had already formed, making it difficult to estimate the exact initiation point, and there was no concurrent video record to indicate the exact location of initiation.
Lastly, overall horizontal slipface advancement was approximated using two laser scans 92 min apart. The horizontal change (slipface Fig. 3. Example of an 'hourglass' grain flow in the Maspalomas dune field in Gran Canaria, Spain. 'Hourglass' flows initiate tens of centimetres from the dune crest (left), eventually developing an alcove with a broadened sediment fan below (right). Time progression between the left and right images is approximately 13 sec. The checkered A4 paper (210 9 297 mm) on the target board at the crest of the dune was used for GPS registration and can be used for background scale of the slipface. A foreground scale is provided in the lower right-hand corner. Slipface height is approximately 2Á2 m. advancement) was determined by taking the difference of the two laser scans in ArcMap GIS, imported with 1 cm gridding. This provided a visual representation and quantified slipface advancement during observation.

Grain-flow morphology
Grain flows displayed a variety of morphologies during observation in the Maspalomas dune field (Figs 3, 4 and 5). Previous, grain-flow classifications have broadly separated 'primary' and 'secondary' flows (Breton et al., 2008). The distinction focused on potential triggers of grainflow activity but did not distinguish between differences in morphology and respective sediment loads. Initiation of primary flows is attributed to the accumulation and destabilization of a sediment bulge centimetres below the dune brink on the slipface. Secondary flows are generally smaller and thought to be initiated in response to disturbances from primary flow activity.
The basic distinction between 'primary' and 'secondary' flows (Breton et al., 2008) is retained in this paper but is extended by identifying subsets of flow features with morphological descriptions and quantifying sediment redistribution volumes for each. New terminology ('hourglass', 'slab', 'lobe' and 'funnel' flows) is introduced to further distinguish different flow types based on morphology. A total of 71 grain-flow events were captured during laser scanning and approximately 30 of these were also recorded on video. For each grain-flow type, the location of flow Fig. 4. An example of a 'slab' flow in the Maspalomas dune field, Gran Canaria, Spain. These flows begin with the appearance of horizontal tensional cracks in close proximity and parallel to the dune brink (emphasized by arrows on the left). Sediment is mobilized shortly after tensional cracks appear (right). Time progression between the left and right images is approximately 5 sec. The checkered A4 paper (210 9 297 mm) on the target board at the crest of the dune was used for GPS registration and can be used for background scale of the slipface. A foreground scale is provided in the lower right-hand corner. Slipface height is approximately 2Á2 m. These smaller grain flows occurred shortly before or after larger grain flows. These flows commonly occurred in clusters (multiple flow initiations seconds apart) but also occurred one at a time. In a few cases, 'lobes' and 'funnels' occurred together. The 'lobes' on the left formed seconds apart. Similarly, the two 'funnel' flows on the left also formed seconds apart. The checkered A4 paper (210 9 297 mm) on the target board at the crest of the dune was used for GPS registration and can be used for background scale of the slipface. A foreground scale is provided in the lower right-hand corner. Slipface height is approximately 2Á2 m. initiation, areal extent, grain-flow thickness, and volume estimates of redistributed sediment are quantified and discussed.
Hourglass grain flows Sediment flows observed in wind tunnel experiments and field observations display a localized over-steepening of the slipface from the formation of a sediment 'bulge' a short distance from the dune brink due to an accumulation of sediment by grain fall, saltation and reptation (e.g. Bagnold, 1941;Allen, 1968Allen, , 1970Hunter, 1985;Anderson, 1988;McDonald & Anderson, 1995;Nickling et al., 2002;Cupp et al., 2005;Sutton, 2012;Sutton et al., 2013a,b). The subsequent initiation of the grain flow results in the formation of an alcove via scarp recession just below the crest that spreads upward and expands laterally (Lindsay, 1973;Hunter, 1977;Sutton, 2012;Sutton et al., 2013b). The alcove grows as sediment flows downslope, and forms a bottleneck at the point of steepest gradient on the slipface (Anderson, 1988), then the sediment accumulates as a fan at the base of the lee slope. In this study, these types of previously observed sediment flows are referred to as 'hourglass' grain flows based on their morphology (Figs 3 and 6); they are considered a type of primary grain flow because they are a direct response to localized slope steepening, independent of other grainflow activity.
The 'hourglass' morphology occurred frequently at the Maspalomas study site and tended to move moderate amounts of sediment, relative to 'lobes', 'funnels' and 'slab' grain flows. The 28 'hourglass' flows recorded affected areas between 4352 cm 2 and 23 624 cm 2 on the slipface, averaging 10 979 cm 2 ( Table 1). The point of failure of the grain flow ranged between 27 cm and 36 cm downslope from the brink, consistent with previous studies (e.g. Anderson, 1988;Nickling et al., 2002;Breton et al., 2008). Grainflow thicknesses ranged from ca 0Á1 to 8Á0 cm with an average thickness of 1Á43 cm. The average estimated volume of redistributed sediment from these 'hourglass' flows per flow event was approximately 22 494 cm 3 , ranging between 582 cm 3 (a grain flow in progress) to 85 891 cm 3 .

Slab grain flows
Larger grain flows have been mentioned in previous work as 'large area extent' grain flows (e.g. McDonald & Anderson, 1995;Nickling et al., 2002;Breton et al., 2008) but these grain flows have not been rigorously studied. In the absence of a clear description of trigger mechanisms or spatial or temporal quantifiers (for example, estimates of area dimensions, formation frequency and flow duration), 'slab' flows might be one of these 'large areal extent' flows or a new type of large primary grain flow. In this study, 'slab' flows, affected large areas of the slipface, ranging from 210 779 to 685 563 cm 2 , and each mobilized a large volume of sediment that successfully reached the base of the slipface in one episode. The initiation point of these grain flows was first visible in the form of uneven horizontal tensional cracks appearing at variable distances parallel to the brink (Figs 4 and 6). The tensional cracks commonly appeared within 2 to 12 cm of the brink, with an average initiation location of about 5Á83 cm from the brink. 'Slab' flows were relatively rare compared to the other grain flows with only seven events captured during observation, most of which occurred on the longer slipface slopes of the dune (Table 1). Grain-flow thicknesses increased downslope and ranged from 0Á07 to 5Á52 cm with an average thickness of 1Á04 cm. The volume of sediment moved by these grain flows per flow event was estimated to range between 103 724 cm 3 and 559 334 cm 3 with an average volume of 182 111 cm 3 .
Lobe and funnel grain flows Secondary grain flows are significantly smaller than primary flows and have been observed to occur shortly after larger grain-flow events (Breton et al., 2008). Smaller grain flows were also observed to form seconds before large grain-flow events during the Maspalomas observations and were often obscured by the overriding primary grain flow. The smaller Maspalomas grain flows initiated in localized areas of slope destabilization, specifically where alcove walls failed or in areas where previous grain flows 'locked up' mid slope (McDonald & Anderson, 1995). Two types of distinct secondary grain flows were observed (Figs 5 and 6), referred to here as 'lobes' and 'funnels'. Nearly all observed 'lobe' and 'funnel' grain flows formed in the mid-slope region of the slipface. The 'lobe' and 'funnel' grain flows that occurred higher upslope, near the brink, were frequently triggered by the failure of an alcove wall from an 'hourglass' grain flow. 'Lobes' and 'funnels' typically formed independently of one another but, in some cases, there were clusters of activity where two or more flows initiated at the same time or seconds apart from one another at various locations of the slipface. In other cases, a single 'lobe' or 'funnel' flow formed without any further activity.
'Lobes' were typically thin, broad, flattened features that rarely transported sediment to the base of the slipface. A total of 25 'lobe' morphologies were identified in the laser scan data (Table 1). Grain-flow thicknesses of 'lobes' averaged about 0Á77 cm and thinned upslope to the point of initiation, and sometimes thinned downslope as well. These flows moved small amounts of sediment, averaging about 1944 cm 3 per flow event and ranging between approximately 124 cm 3 and 5091 cm 3 .
'Funnel' grain-flow morphologies distinctly manifested as long, linear features that transported small amounts of sediment downslope through a shallow, narrow trough (Figs 5 and 6). In contrast to 'lobes', 'funnels' often successfully delivered sediment directly to the base of the slipface where it accumulated in a small depositional fan. These grain flows were quickly infilled, making it difficult to measure and identify them in the laser scan data, and therefore only nine 'funnel' morphologies were identified in the laser scans with confidence (Table 1). From the few measurements collected, 'funnel' flow troughs averaged 1Á89 cm in depth, and had widths ranging from about 2 to 5 cm. The average sediment thickness, deposited at the base of the slipface from 'funnel' flows, was measured to be approximately 1Á19 cm and tapered in thickness upslope. The estimated volume of sediment transported downslope by 'funnel' flows averaged about 1312 cm 3 with a range of about 60 to 6386 cm 3 .

Frequency and magnitude
In addition to measurements from laser scans, video coverage of the main slipface (the 2Á2 m high exposed slipface portion of the larger 9Á62 m dune; Figs 1 and 2) between 10:00 am and 11:00 am provided the means to investigate frequency of grain-flow occurrence. This information was used to compare with the results of Breton et al. (2008). Table 2 lists the frequency of grain flows along with the grain-flow morphology corresponding to the grain flows listed in Table 1. Breton et al. (2008) reported 21 grain flows over 1 h of observation along a 14 m stretch of slipface of a 5 m high linear dune with wind speeds between 7Á38 m sec À1 and 7Á49 m sec À1 at a height of 0Á5 m. The grain flows had an average area of 32 800 cm 2 and occurred in 'temporal clusters' with an average reoccurrence of 2Á7 min.
The main slipface of the Maspalomas study site exhibited similar clusters of activity separated by brief periods of inactivity which was accentuated directly after large grain flows occurred ( Table 2). The Maspalomas grain-flow activity differs slightly from Breton et al. (2008) in terms of the time in between flows, duration of grain-flow activity, and area estimates (Tables 1 and 2). Maspalomas grain flows occurred more frequently (on average every minute) over a smaller area of slipface. The grain flows were also of shorter duration, averaging about 20 sec. Including all grain flows from Table 1, the spatial area was generally smaller. 'Hourglass' grain flows averaged 11 000 cm 2 , and 'funnels' and 'lobes' averaged between 200 cm 2 and 2900 cm 2 , respectively. 'Slab' flows (averaging 425 000 cm 2 ) were the outlier regarding surface area measurements, far exceeding any grain flow reported in Breton et al. (2008).

Slipface advancement
For the calculation of slipface advancement, laser scan data were restricted to a smaller portion of the slipface, overlapping with the main slipface, to investigate the relationship between grain-flow activity and slipface advancement. Based on laser scan measurements and video recordings, within 92 min of observation, approximately 277 832 cm 3 of sediment was redistributed by 22 different grain flows on the main slipface (Table 1), equivalent to nearly half a metric ton of sediment. 'Hourglass' flows were responsible for redistributing about 51Á6% of that sediment. The single recorded 'slab' flow for this portion of the slipface moved approximately 103 724 cm 3 of the sediment, (about 43% of the total). Secondary flows redistributed minor amounts of sediment comprising the remaining 5Á4% of the volume total. Figure 7 is a difference map of the two laser scans illustrating slipface advancement within the 90 min observation for an 8 m span of the slipface, overlapping with Figs 2, 3, 4 and 5. The difference map shows that some portions of the slipface advanced about 17 cm within this time. Using ArcMap and restricting the slipface area to a span of 5 m to correlate with the area of the main slipface, the volume difference between these scans was also calculated and resulted in a total of about 1 185 000 cm 3 of sediment.

Grain-flow formation
The in situ field observations and high resolution-measurements reported here provide new insights into morphodynamics of different grain flows and enable genetic insights into various grain-flow forms. 'Hourglass' morphologies are the best documented while smaller 'lobe' and 'funnel' morphologies are poorly known. The formational cause of particular grain-flow morphologies and the genetic relationships between various flow types are unknown. A few alternative formation mechanisms have been proposed to explain the occurrence of spatially-small and large flows that do not result in an hourglass morphology. The various grain-flow types are discussed below.

Hourglass flows
The most widely discussed grain-flow phenomenon involves the formation and failure of a bulge or larger wedge of sediment tens of centimetres downslope from the dune brink due to the accumulation of saltating sediment and grain fall. While the laser scan data in this study were not capable of observing the presence or growth of a sediment bulge/wedge due to noise near the dune brink, the location of failure downslope from the brink in relation to this pattern of sediment accumulation is consistent with previous studies (e.g. Allen, 1968Allen, , 1970Bor owka, 1979;Hunter, 1985;Anderson, 1988; Nickling et al., Table 2. Duration of grain flows observed on the main slipface shown in Figs 3, 4 and 5 with video coverage and scan data. Grain-flow failure numbers correlate with the failure numbers shown in Table 1. Grain flows 1 and 4 of the main slipface were only partially recorded and have not been included in this table. The start time for grain flow 6 is uncertain because video recording began at 9:55:38 am, seconds after its formation. Grain flows 6, 8 and 9 are from a separate video recording and do not have a time continuum with the remaining grain flows in this table which were all recorded in the same video. 2002; Walker & Nickling, 2002;Cupp et al., 2005;Kok et al., 2012;Sutton et al., 2013a,b).

Funnel and lobe flows
Factors invoked to explain other grain-flow morphodynamics include disturbance from large grain flows, mid-slope accumulation of sediment, and lee side airflow patterns. The formation of 'funnel' and 'lobe' flows may be influenced by one or more of these factors. Breton et al. (2008) observed that smaller 'secondary' flows formed as a direct response to disturbance from larger 'primary' flows, and they occurred soon after the formation of the 'primary' flow. Some of the 'lobe' flows in this study also occurred as a direct result of disturbance from larger flows. Typically, the collapse of an alcove wall from an 'hourglass' flow produced a smaller 'lobe' flow. In a few instances, the tongue of the 'hourglass' flow disturbed nearby sediment as it moved downslope, triggering simultaneous smaller 'lobe' flows. None of the 'funnel' flows were associated with 'hourglass' flows but they often occurred with 'lobes' Fig. 7. Colourized difference map of the Maspalomas slipface depicting change in metres on the slipface, where cool colours represent advancement of the dune toward the laser scanner. Topographic change on the lee slope was created by subtracting two 3D laser scans taken 1 h, 32 min apart with 0Á001 m gridding. During this time, the maximum slipface advancement was approximately 17 cm. Outlines of a few individual avalanche events can still be seen. Noisy data near the dune crest is the result of airborne sediment. The white box in the lower righthand corner contains a representation of the laser scanner, which is added automatically to each of the scans in the processing software 'SCENE' to show its location. The diagonal line radiating out from the scanner location is an artefact due to sensor vibrations probably caused by the windy conditions. The vertical black lines indicate the bounds of a 4 9 5 m area of the main slipface where volume calculations were derived from this difference map. This area also overlaps with the portion of the slipface shown in Figs 3, 4 and 5.
on the mid to lower lee slope preceding a larger flow. The occurrence of smaller flows prior to and independent of larger flows differs from the observations of Breton et al. (2008) and requires a different trigger mechanism.
The accumulation of sediment mid slope due to a 'lock up zone' (McDonald & Anderson, 1992, 1995, 1996 or the increase of aeolian transport during stronger winds (McDonald & Anderson, 1992, 1995, 1996Nickling et al., 2002;Nield et al., 2017) are both potentially valid alternative formation mechanisms of 'lobe' and 'funnel' flows that form independent of 'hourglass' flows. It is possible that sediment from previous grain flows that did not reach the base of the slipface accumulated mid slope in this 'lock up region', facilitating the buildup and destabilization of sediment. In addition, observations took place during strong wind conditions (9Á5 m sec À1 ) which probably increased the downslope distance of where sediment settled, augmenting the accumulation of sediment in this mid-slope region.
Secondary lee slope airflow patterns generated by flow separation at the dune crest and subsequent reattachment flow generate complex eddies and vortices which affect surface shear stress and sediment redistribution (Wiggs, 2001;Walker & Nickling, 2002;Parsons et al., 2004a,b;Cupp et al., 2005;Jackson et al., 2013a,b;Pelletier et al., 2015;Smith et al., 2017a). As more airflow modelling is conducted, the effect of lee side airflow patterns on the initiation of 'lobe' and 'funnel' flows may become better understood, and a direct link to turbulent airflow and localized slope destabilization may be discovered (e.g. Walker, 1999;Walker & Nickling, 2002). 'Lobe' and 'funnel' flows may also be sensitive to the magnitude and direction of wind, where increasingly stronger winds can decrease the angle of initial yield leading to an increase of failures (Pelletier et al., 2015), and oblique incident airflow may promote lateral redistribution of sediment on the slipface (Walker, 1999).

Slab flows
Large flows appear to be initiated by grain fall and grain saltation, similar to 'hourglass' flows; however, differences in scale and morphology suggest that an additional process influences the initiation of these spatially-large flows. Potential influences include moisture content (e.g. McKee et al., 1971;McKee & Bigarella, 1972;Morris et al., 1972;Bigarella, 1975), complex airflow cells on the lee side of the dune, or an alternative failure process.
The presence of moisture is an obvious explanation for the formation of 'slab' flows, especially for coastal dunes which may have varying degrees of internal cohesion. Heavy rainfall occurred three days before with some light rainfall the night before observations in this study. Despite the absence of detectable moisture in the sediment samples, internal layers of damp and dry sand could potentially influence the style of grain flow. Layers of relatively dry sand would provide a failure plane for large slabs of sediment such as those observed in this study. Similar cohesion is also thought to be present in polar aeolian deposits with alternate layers of snow, ice and sediment (e.g. Morris et al., 1972) but few field observations have been done in these environments and there are no known descriptions or observations of 'slab' flows on polar dunes.
Another explanation for the initiation of spatially-large grain flows may involve complex airflow cells which carry and deposit sediment in the lee of the dune (McDonald & Anderson, 1995;Nickling et al., 2002). It is possible that sediment transported in flow cells may provide additional sediment, thereby increasing the magnitude of grain flows . Cupp et al. (2005) showed from wind tunnel experiments that sediment deposition on the lee slope involves two transport zones. In the upper part of the lee slope, fallout of saltating grains results in reptation down the slope. Below this fallout area, there is a return cell that results in very minor upslope sediment transport. It has been hypothesized that the interaction between these two transport zones may produce larger grain flows Cupp et al., 2005). However, this explanation does not match well with the morphology and the initiation location of the 'slab' flows observed in this study.
Lastly, Fryberger & Schenk (1981) characterized two differing mechanisms of grain-flow, or failure processes, in a laboratory setting. These included scarp recession (detailing the 'hourglass' flow morphology) and slump degeneration. The description of the slump degeneration agrees well with the observations of 'slab' flows in this study. The progression of the observed slump degeneration began as a series of tensional features near the top of the artificial slipface until sediment flowed downslope forming compressional features (folds) while preserving cohesion under dry conditions (Fryberger & Schenk, 1981). The 'slab' flows observed in Maspalomas formed identically to this description of slump degeneration. The onset was manifest as multiple tensional cracks centimetres from the dune brink and compressional folds formed as the 'slab' flow began to travel downslope (Table 1; Fig. 4). Sediment in the Maspalomas 'slab' flows appeared to be semi-cohesive despite the absence of detectable moisture in the sediment samples. The cause of cohesion in dry sediment was never discovered by Fryberger & Schenk (1981) but it cannot be discounted as a possible flow mechanism.

Sediment redistribution
Each of the grain-flow morphologies observed in Maspalomas has differing capacities of sediment transport which affects slipface advancement and dune migration. Details of these different types of grain flows will lead to a better understanding of slipface processes as well as an improved interpretation of dune stratigraphy and aeolian environment.
Grain-flow morphology and magnitude 'Slab' flows displaced the greatest amount, nearly an order of magnitude greater volume than a single 'hourglass' flow and approximately two orders of magnitude greater than the average 'lobe' and 'funnel' flow. It therefore appears that 'slab' flows had the greatest influence on slipface advancement and sediment redistribution. However, 'slab' flows were comparatively rare and it is likely that these grain flows formed under special conditions, potentially dependent on moisture which may have added cohesion to the sediment and may not be representative of typical slipface processes.
Due to the higher frequency of occurrence during observation in Maspalomas, 'hourglass' flows were probably responsible for moving the most sediment downslope. The regularity and frequency in which these grain flows occurred suggest that 'hourglass' flows play an important role in slipface advancement. 'Lobe' and 'funnel' flows appear to play a secondary role despite sediment from these flows not reaching the base of the slipface. These grain flows commonly occurred mid slope in areas on the slipface previously suggested to be 'lock-up zones' (McDonald & Anderson, 1992, 1995, 1996 or zones of accumulated sediment from suspension settling (McDonald & Anderson, 1992, 1995, 1996Nickling et al., 2002;Nield et al., 2017). These smaller grain flows were the most frequent but were rapidly obscured by settling airborne sediment making it difficult to capture all the activity that occurred during observation. The sediment that accumulates in the mid slope from grain fall or smaller grain flows may supply additional sediment to other grain flows, potentially creating larger magnitude flows (McDonald & Anderson, 1992, 1995, 1996Nickling et al., 2002), thus playing a minor role in slipface advancement.

Overall slipface advancement
There is a large discrepancy between the volume change of sediment on the slipface calculated from taking the difference between laser scans in ArcMap and measurements of individual grain flows alone from Table 1. Aggregated volume measurements of individual grain flows provide a total volume estimate of redistributed sediment on the main slipface of approximately 277 832 cm 3 . In contrast, the difference of the two laser scans spanning 92 min indicated a total volume change of approximately 1 185 000 cm 3 (1Á185 m 3 ) for a 5 m span of the slipface.
This discrepancy is probably due to the fact that scans were acquired every 4 to 8 min and did not capture all grain flows which typically occurred at 30 sec to 1 min intervals (Table 2). In addition, the discrepancy in volume estimates between individual grain-flow measurements and the volume difference of laser scans may reflect how grain fall influences slipface advancement. Grain fall was responsible for triggering individual grain flows and rebuilding the slipface after an event but may have also had a significant role in slipface advancement. The precise influence that grain fall had on slipface advancement is difficult to quantify because it is a gradual process and varies depending on aeolian conditions.

Correlation of dune height and flow thickness
The thickness of individual grain-flow crossstrata preserved in aeolianite deposits has been previously used as a method to estimate palaeodune height (e.g. Kocurek & Dott, 1981;Sweet et al., 1988;Kocurek, 1991). This relationship was originally suggested by Hunter (1977) and relies on the assumption that grain-flow thickness is consistent on a variety of dune forms of the same height and increases in thickness with increasing dune height. While this hypothesis appears logical, there may be multiple influences not accounted for that affect the thickness of a grain flow. For example, Nield et al. (2017) observed that with increasing wind velocities, the length and thickness of grain flows also increased. Additionally, the entrainment of sediment as grain flows travel down the slipface may vary. Deposits of grain-flow thickness may increase due to the addition of grains from the mid slope where sediment can become destabilized in a 'lock up zone' or secondary location of localized over-steepening (McDonald & Anderson, 1992, 1995, 1996Nickling et al., 2002;Nield et al., 2017).
The grain flows observed in Maspalomas had a variety of flow thicknesses and did not have a strong correlation with dune height (related to grain-flow length; Fig. 8). The most variable grain-flow thicknesses were measured from 'hourglass' flows, where the thicker flows possibly entrained additional sediment from the mid slope. 'Funnel' flows had the strongest correlation between grain-flow length and thickness but also had the fewest number of observations. 'Lobe' flow thickness had a very weak correlation to grain-flow length but both the 'funnel' and 'lobe' flows generally did not extend the entire length of the slipface and would be a poor indicator of potential dune height. 'Slab' flows had the most consistent base thickness, ranging between 1 cm and 2 cm with one outlier of 5Á5 cm thickness (Fig. 8). These flows were relatively thin compared to other grain-flow morphologies but may also be affected by entrainment of additional sediment. Until more observations and measurements are collected, it is unclear how reliable grain-flow thickness is as a link to dune height, and should be treated with caution when interpreting palaeoenvironments.

Comparison of grain-flow behaviour with previous studies
Grain flows observed in Maspalomas and those reported by Breton et al. (2008) differ slightly in behaviour but are consistent when physical differences in dune morphologies and wind regime are considered. The linear dune observed by Breton et al. (2008) was 5 m high, while the main Maspalomas slipface was 2Á2 m high. The shorter height of the Maspalomas slipface would shorten the duration of grain flows.
The Maspalomas dune had a higher frequency of grain-flow activity (Table 2), occurring on average every minute whereas Breton et al. (2008) observed an average recurrence time of 2Á7 min. Wind measurements in Breton et al., were collected at a height of 0Á5 m and ranged between 7Á38 m sec À1 and 7Á49 m sec À1 . Adjusting the average Maspalomas wind speed of 9Á5 m sec À1 from a height of 1Á0 m to a height of 0Á5 m using the wind profile power law results in a velocity of 8Á6 m sec À1 . The higher frequency of grain-flow activity at Maspalomas is probably due to the greater wind speeds at the study site. Greater wind velocities trigger grain flows more frequently due to the larger amounts of sediment delivered to the slipface. Ignoring 'slab' flows, which do not appear to have occurred during the Breton et al. (2008) study, the spatial area of the Maspalomas grain flows was on average about 2 m 2 smaller. This difference may also be due to the dune height where taller dunes produce greater planform areas of grain flows as they flow downslope. Nield et al. (2017) collected data on a 5Á12 m high dune using a Terrestrial Laser Scanner with wind measurements from anemometers at a height of 0Á5 m over two days of observation. Differing grain-flow behaviour was observed and related to wind velocity, where lower wind velocities (<6 m sec À1 ) resulted in small, discrete failures that occurred near the dune brink  Table 1. Similar relationships were present when plotting grain-flow thicknesses from the upper and mid slope. 'Hourglass' grain flows (blue dots) have highly variable grain-flow thicknesses where thicknesses <4 cm are less common and do not correlate with grain-flow length. 'Funnel' (red triangles) and 'lobe' (green diamonds) sediment flows have less variability in grain-flow thickness where most thicknesses are <2 cm. There exists a very weak correlation with grain-flow length and thickness for both types of flow but is stronger for 'funnel' flows. 'Slab' flows (orange squares) are generally thin compared with 'hourglass', 'funnel' and 'lobe' flows and do not have an apparent relationship with grainflow length. with limited grain-flow lengths and thicknesses (Nield et al., 2017). Moderate winds (ca 6Á5 m sec À1 ) produced grain-flow activity further downslope of the dune brink with the ability to transport sediment to the bottom of the slipface. Strong winds (>8Á4 m sec À1 ) enabled 'multiple families of failures', where larger grain flows initiated near the top of the slipface and smaller grain flows occurred partway down the slope (Nield et al., 2017). Under strong winds, there was frequent grain-flow activity and the grain flows that occurred had longer and thicker tongues of sediment that successfully transported sediment to the bottom of the slipface (Nield et al., 2017). Data also show that there was a downslope shift in the location from the dune brink of larger grain-flow initiation points (location of failure) from 30 cm during wind velocities around 6 m sec À1 to 40 cm for velocities >6 m sec À1 with some grain flows initiating 1 m below the dune crest (Nield et al., 2017).
While the Maspalomas observations did not include a study of grain-flow behaviour under a variety of wind conditions, the data agree well with the behaviour observed by Nield et al. (2017) during stronger wind velocities. The wind velocity at the study site, adjusted to a height of 0Á5 m, was approximately 8Á6 m sec À1 and exhibited a high frequency of grain-flow activity, where some grain flows occurred simultaneously or shortly after one another, or in 'multiple families of failures'. No downslope shift was detected for the initiation point of 'hourglass' grain flows, where the location of failure remained around 30 cm below the dune brink but many 'lobes' and 'funnels' initiated further downslope. Additional observations need to be made to determine if the downslope shift in grain-flow failure is directly linked to wind velocity alone or if slipface length and dune morphology may have an effect. More specifically, an investigation of the conditions that produce sediment bulges (e.g. Allen, 1968Allen, , 1970Hunter, 1985;Anderson, 1988;Walker & Nickling, 2002;Kok et al., 2012;Sutton, 2012;Sutton et al., 2013a,b) near the dune brink as opposed to larger sediment wedges (e.g. Nickling et al., 2002;Cupp et al., 2005;Nield et al., 2017) may be helpful in elucidating this difference in observation.

CONCLUSION
Using a combination of high resolution, groundbased continuous laser scans and simultaneous video recordings, various grain-flow morphologies are presented for the first time. This study includes precise measurements of grain-flow thickness, as well as area and volume estimates of redistributed sediment for each grain-flow morphology observed (Table 1). Typical grain-flow thicknesses ranged from about 0Á05 to 8Á0 cm, where the thicker portions of grain flows were located where sediment accumulated at the bottom of the slipface. On average, grain flows were approximately 1 cm thick. 'Slab' flows were responsible for redistributing significant amounts of sediment, averaging about 182 111 cm 3 per event but were not as common as 'hourglass' grain-flow morphologies and 'funnel' or 'lobe' flows. 'Hourglass' grain flows redistributed an average of 22 494 cm 3 per flow event while 'funnels' and 'lobes' redistributed an average of 1312 cm 3 and 1944 cm 3 , respectively.
The high-resolution measurements of grainflow characteristics presented herein will be vital in verifying and improving laboratory and modelling efforts to identify grain-flow triggers for morphologically diverse grain-flow events and sediment volume flux in aeolian environments. With the growing literature on aeolian slipface processes, there is a need for a classification of lee slope grain flows based on formation patterns (for example, flow area, frequency and duration) and morphology. The morphology classification presented in this study provides a foundation for future study and a structure to support discrete observations between different aeolian environments. For example, additional research is needed to investigate how grain-flow style is affected in coastal and polar environments in contrast to arid environments where cohesion from salts, liquid water and ice may affect grainflow behaviour. This classification will aid in distinguishing the influences on specific styles of flows which can be later linked to controlling parameters such as wind strength and direction, turbulent airflow, sediment deposition, grain texture and environmental influences. This will ultimately lead to a better interpretation of aeolian strata and palaeoenvironments of dunes preserved in the rock record (for example, estimates of dune height; McKee et al., 1971;McKee & Bigarella, 1972;Morris et al., 1972;Bigarella, 1975;Hunter, 1977;Kocurek & Dott, 1981;Sweet et al., 1988;Sweet & Kocurek, 1990;Cooke et al., 1993;Bourke, 2005;Anderson & Walker, 2006;Eastwood et al., 2012), as well as a better understanding of the intricacies of slipface processes and overall dune migration.