Geomorphological evidence for a dry dust avalanche origin of slope streaks on Mars

Abstract

Mars has several different types of slope feature that resemble aqueous flows. However, current cold, dry conditions are inimical to liquid water, resulting in uncertainty about its role in modern surface processes. Dark slope streaks were among the first distinctive young slope features to be identified on Mars and the first with activity seen in orbital images. They form markings on steep slopes that can persist for decades, and the role of water in their formation remains under debate. Here I analyze the geomorphic features of new slope streaks using high-resolution orbital images. Comparison of before-and-after images reveals how the streak formation process affects the surface and provides information about the cause. These observations demonstrate that slope streaks erode and deposit material in some instances. They also reveal that streaks can jump slopes and may be erosive very near their termini. These observations support a formation model where dark slope streaks form as ground-hugging, low-density avalanches of dry surface dust. Such streaks need not be treated as Special Regions for planetary protection.

Dark slope streaks (Fig. 1a) are typically wedge-shaped features on steep slopes, initiating at a point and broadening downslope¹. Sullivan et al.² provided a thorough summary of early slope streak observations, primarily using the Mars Orbiter Camera (MOC³) and incorporating results from earlier studies^1,4. The streaks show no relief at MOC scale and have a wide range of aspect ratios. The apex or initiation point is often located at an outcrop or protrusion. The albedo within a streak is typically uniform, and textural features on the slope are continuous inside and outside the streaks. They may divert around obstacles but are also capable of overrunning low-relief features, sometimes leaving undisturbed zones in the wake of obstacles². Bright streaks are less common than dark, and both may occur on the same slopes². New streaks observed by MOC were the darkest on the slope, suggesting that streaks form with maximum contrast and then fade over time². Slope streaks form in dusty regions with high albedo and low thermal inertia and require steep local slopes^1,2,5. Dark slope streaks are only dark relative to their surroundings. In an absolute sense they are brighter than low-albedo regions of Mars^6,7; hereinafter, the terms bright and dark are relative to adjacent surfaces. They are distinct from Recurring Slope Lineae (RSL), which form in low-albedo regions and grow incrementally^7,8.

Figure 1. Formation of ridges corresponding to lobe boundaries within a new slope streak.

Downhill is to the bottom. b and d show insets before streak formation and c and e show the same regions afterwards. Note ridges in e, but streak-border relief is minimal and small topographic features within the streak are preserved. All figures are map-projected with north up and stretched to show local contrast, and have illumination from the left. See Table S3 for image information for all figures.

When first observed, it was proposed that dark slope streaks were streamers of debris fallen from small outcrops⁹. However, subsequent work favored either wet or dry flows. Wet debris flows or seeps were proposed^1,10,11, as were mass movements triggered by small amounts of liquid water⁵. Dry models include dust avalanches², or removal of dust from above shifting talus⁴.

Subsequent observations provide additional information. Streaks form in locations where dust preferentially accumulates¹² and appear to be indifferent to the local geology provided that the surface is blanketed by dust¹³. One large streak documented in detail had distinct mounds in the interior and a scarp along the edge¹⁴, suggesting substantial mass movement, similar to triangular avalanche scars¹⁵. Additional high-resolution observations revealed indications of relief along streak edges and showed observations indicating triggering of some streaks by rockfalls or impacts, as well as longitudinal ridges within some bright streaks¹⁶. Topographic relief was also reported along the edge of a newly formed streak¹⁷, although pre-streak MOC images lacked the resolution to determine whether relief at streak edges was pre-existing. These observations were all interpreted as consistent with dark streaks forming as dust avalanches ranging from superficial to meters-thick mass movements. However, the topographic effects of streaks have been questioned¹³, and streaks do flow for significant distances on low slopes, suggestive of a fluid process^13,18,19. However, steep slopes >20° are required for streak initiation¹⁸. Spectral evidence for wet origins has also been proposed²⁰, and correlations with atmospheric water vapor and surface hydrogen, chlorine, and iron have been taken as evidence for wet flows¹⁹.

Change-detection studies have also contributed to the understanding of slope streaks. After the initial discovery of new streaks^2,21, a survey of MOC images found that new streaks formed at a rate of 7% per existing streak, per Mars year, with little indication of fading over 28 years²². This suggested that formation and erasure are not in equilibrium. Follow-on work found a lower formation rate of 3% per streak per Mars year and reported indirect evidence of fading, but still observed an imbalance⁶, while a balanced cycle was shown at one site, with a streak lifetime on the order of forty years²³. An early study at three sites found that dark slope streaks form sporadically throughout the Mars year²⁴, but a more comprehensive survey showed seasonal variations that correlated with high surface temperature and wind speed²⁵. Bright streaks might form via modification of dark streaks⁶.

This work presents a survey of dark slope streak formation using monitoring images from the High Resolution Imaging Science Experiment (HiRISE²⁶); see supporting material and Methods for details. Before-and-after images enable an improved assessment of the effects of the streaks. Additionally, observations of slope streak morphology are described for some streaks without constrained formation times.

Observations of new streaks

Some new streaks have identifiable triggering mechanisms, such as rockfalls¹⁶, dust devil tracks^6,27 or new impact craters^16,28 at the source. However, in >98% of the new streaks observed, no specific cause is apparent. Most streaks initiate from point sources. Some appear to have multiple point sources but this is attributed to near-synchronous formation of adjacent streaks.

Some evidence from the most closely monitored sites indicates bursts of streak formation in brief intervals and a formation rate that varies inter-annually. For instance, thirty-eight new streaks formed at site 3 between mid-Mars Year (MY) 28 and late MY 31, but only five more formed by early MY 34. At site 5, forty-one streaks formed in mid/late MY 31, but only six more formed over the next two Mars years. Further evidence for brief intervals of intense streak formation comes from sites with numerous streaks of identical brightness, suggesting that they formed approximately simultaneously (Fig. E1).

Only 4% of large new slope streaks have confirmed topographic effects. Small topographic features are unchanged within many large streaks, and can clearly redirect the flows, although they are sometimes overtopped (Fig. 2, E2). Previous examples of possible streak-edge scarps^16,17 may simply reflect this topographic control. The streaks also do not remove the entire erodible layer on their slopes (Fig. E3). However, definitive topographic changes occur in some cases (Fig. 1, E4, E5). Fig. 1 shows the formation of longitudinal ridges within a streak. The ridges parallel the streak edges and align with boundaries between terminal lobes, suggesting that the flow broke into distinct lobes. However, small-scale topography within the upper part of the streak was largely preserved, and some pre-existing features remain in the lower section. The behavior of some new streaks sheds light on the long-term effects of repeated flows. Fig. 2 shows the terminus region of a new streak which had no resolved topographic effects. The streak was channeled between ridges and fits within existing troughs incised between them. This suggests that previous streaks at this site were erosive very near their termini (see also Fig. E6).

Figure 2: Evidence for erosion near slope streak termini.

a: Flat-floored channels with sharp, narrow walls (arrows show example) are incised into the troughs only near the uphill parts of the ridges. b: A new streak is confined within the channels and closely fits their width, indicating that it is similar in size and extent to the flows that eroded the channels. This suggests that the flows were erosive near their termini. Downhill is to the bottom.

Several examples demonstrate that slope streak flows can skip over terrain, leaving some of the surface undisturbed before continuing downslope (Fig. 3). These surfaces are oriented downhill (facing away from the direction of streak movement). Some are at rock outcrops but others appear dust-covered.

Figure 3: Slope streak topographic interactions.

Slope streaks appear to skip over downhill-facing slopes in some cases (arrows show examples), indicating that the flow lost contact with the ground. The slopes are dust-covered, not bedrock, and should have been disturbed by the passage of a flow. In a, most of the crater wall is excluded but the flow edges are aligned, indicating that it was not deflected. The toe of the lowermost streak in b is entirely detached. Downhill is to the upper left in a, and to the lower left in b.

In comparing images for new streaks, the temporal evolution of hundreds of existing streaks was also tested. In no case did a streak extend incrementally, as would be expected if growth was slow and images were sometimes acquired partway through formation. All observations are consistent with dark slope streaks forming in brief events^29,30 and never extending further. Some do overprint older flows, which could appear as incremental growth in lower-resolution data, but in all cases this is a new event that rapidly modifies its entire area including part or all of an existing streak, rather than slow growth (Fig. 4ab). As previously suggested², existing streaks appears to be unfavorable for propagation of new streaks, although it is not precluded (Fig. 4cd). Multitemporal observations also revealed cases where slope streaks fade on timescales much shorter than the multi-decade intervals reported elsewhere (Fig. E1).

Figure 4: Interactions between new and existing slope streaks.

a/b: A younger streak (arrow) overprints an older example with little apparent effect. Note sheltered zones downhill from boulders. Downhill is to the bottom. c/d: A new streak overprinted the head of an older streak but halted after a short distance. However, narrow tendrils (arrows) continued to flow just to the outside of the old streak, indicating that conditions are more favorable for propagation there; wet debris flows would develop levees and favor propagation at the center. Downhill is to the left.

Formation by Dust Avalanches

Because topographic changes are visible to HiRISE, slope streaks mobilize decimeters of surface material in some cases, and subresolution topographic changes could be common. Additionally, subsurface flow would not be deflected by small obstacles nor leave untouched zones in their wakes, and could not skip over downhill-facing slopes. This rules out formation models based only on percolation or subsurface water. Antarctic water tracks due to seeping water, an analog for wet slope streaks¹³, do not leave such undisturbed zones downhill from boulders³¹, since percolation can continue through the subsurface.

The observation that slope streaks can pass over obstacles suggests that they can lose contact with the surface. (Passing over downhill-facing slopes is the reverse of what would be expected for seepage.) This suggests a ground-hugging flow, fast enough to lose contact with steep slopes and run up small obstacles. The flow must have little momentum (low density) to be redirected by minor topography that it does not overtop. This is most consistent with the dust avalanche model², which envisioned incorporation of air into a thin dust flow. The flowing material is a ground-hugging body of dust, and could traverse low slopes if fluidized by ingested atmosphere, somewhat resembling a pyroclastic flow. This hypothesis explains large avalanche scars¹⁵ and aspects of slope streak morphology that resemble them¹⁶ as large end-members, transitioning from a superficial ground-hugging flow to a massive avalanche as a thicker body of surface dust is incorporated. The lowest-density end-member would more closely resemble a cloud of saltating particles, which can traverse level ground; there is evidence that dust aggregates saltate on Mars³², and saltation can be self-sustaining as grain impacts dislodge more particles³³.

Dust avalanches could lose 89–99% of their mass into suspension². This is consistent with the observation that some slope streaks can be erosional near their termini and usually lack substantial terminal deposits. The flow would continually be replenished by dislodging and incorporating the surface layer of dust, which can explain why old slope streaks are unfavorable for propagation—the most mobile dust has been removed. Dust aggregates occur but are very weak³⁴, so the particles of the flow may disintegrate as it proceeds, helping to convert the flow into suspendable material. Wet flows, by contrast, should not lose material into suspension and should consistently have terminal deposits.

Dust avalanches may be initiated via impact cratering, rockfalls, earthquakes, or dust devils, but the likely cause for streaks with no obvious trigger is high winds. This explains seasonality²⁵ and the occasional occurrence of bursts of formation, and clusters with similar albedo (age) are readily explained via stochastic high-wind events, which also reset dust surface textures (Fig. E1). This is consistent with the multi-year aeolian behavior observed by Viking Lander 1, where surface dust was episodically modified³⁵ and thin failures occurred in dust drifts³⁶. Avalanches could even begin as a burst of saltating dust aggregates causing runaway failure. This trigger could be supplemented, and dust lifting/fluidization enhanced, by electrostatic effects on charged dust³⁷ and thermal creep and insolation processes^38–41, which should be strongest in fine-grained material. Additionally, winds or temperature changes could affect soil hydration and cohesion⁴².

Additional support comes from similarities to spire streaks (Fig. E7). These are spindle-shaped wind streaks that initiate at point sources and propagate downwind⁴³. The planform shapes resemble slope streaks but they can occur on low slopes and propagate uphill. They appear sharp-edged at low resolution, although the edges may be less distinct at meter scales; like slope streaks, they can leave undisturbed dust wedges in the lee of obstacles. They are interpreted to form via erosion by high winds⁴³, although the darkened area retains a low thermal inertia, which suggests a change in photometric properties (as proposed for slope streaks¹²) rather than complete removal of the dust. When present in the vicinity of slope streaks, the latter have stronger contrast, suggesting greater disturbance of the surface. Slope streaks and spire streaks thus have many similarities; they may differ in that on flat ground the dust disturbance and erosion is controlled by saltation and aeolian turbulence, while on steeper slopes thicker failures are possible and result in denser bodies of flowing dust, controlled by gravity and with more sharply defined boundaries.

Implications of a dry origin

New observations support the dust avalanche hypothesis for dark slope streaks and illuminate its operation. This indicates that slope streaks need not be treated as Special Regions for planetary protection⁴⁴. Additionally, although Recurring Slope Lineae appear to be granular flows²⁹, the distinct behavior of slope streaks suggests that dust behaves differently from other granular materials on Mars.

Methods

The primary data set for this study was images from HiRISE²⁶. HiRISE images are normally 5–6 km wide and 10–20 km long, and have a pixel scale of 25–60 cm depending on latitude and binning. The survey used the Reduced Data Record (RDR) observations, which are map-projected at a uniform scale of 25 or 50 cm/pix and stretched for reasonable contrast across the image. HiRISE has a central color swath that includes blue-green and near-infrared filters, in addition to the broad red-filter coverage, but this survey focused on the red data. Images are consistently acquired near 3 PM local time (15:00) due to the Mars Reconnaissance Orbiter (MRO) orbit⁴⁵, but the lighting conditions vary seasonally. We selected 136 possible or definite slope streak sites with HiRISE images spanning a long temporal baseline (at least 4000 MRO orbits), acquired through MRO orbit 53000. The sites examined are listed in Table S1.

New slope streaks were identified using before-and-after HiRISE images, using methods similar to previous studies of gullies⁴⁶. For thirteen sites with good temporal coverage, images were blink-compared at a reduced resolution of 1 m/pix, focusing on prominent slopes with existing streaks on the slope or nearby; it is not practical to search every slope, particularly at full resolution. For locations with many images, it is not practical or useful to compare all possible pairwise combinations. Instead, comparisons were selected in order to span the full time interval of HiRISE data and with comparison pairs that were similar in lighting and geometry; this choice is dependent on the particular set of data available for each site. For each new streak, data collected include the overall downhill orientation of the slope near the streak initiation point, the source-to-toe length of the streak (not accounting for curvature or surface slope), and any evidence for the triggering process or topographic effects of the streak. These data, as well as formation timing constraints and streak coordinates, are given in Table S2. Similar to other surveys, the new streaks in the population analyzed here range from 10 m to >1 km long (any smaller streaks were not tabulated) and have a preference for south-facing slopes. Higher-resolution comparisons were also conducted for subsections of larger streaks (source-to-terminus distance >500 m) to look for topographic effects. It should be noted that a lack of observed topographic effects does not mean that there were none, as topographic changes could be unresolved or difficult to recognize due to different lighting.

For an additional 123 sites with steep slopes in dusty regions, images were inspected but not blink-compared, and only large new streaks (length >500 m) were documented, in order to increase the sample of large streaks where topographic changes are most likely to be detectable. This data set included hundreds of new streaks (including more than a hundred large streaks) and was sufficient to characterize the morphological effects of streak formation.

Some limitations of the data should be noted. First, the consistent 3 PM time of day means that west-facing slopes are always more directly illuminated than east-facing, which may lead to detection biases. This is mitigated for slope streaks because all of the study sites are at latitude <±38° (most are <±25°; all of the 123 additional sites used were chosen within this latitude range), so the Sun is typically high enough that in most cases dark streaks can be identified on all slope aspects. Dark slope streaks can be seen in shadow and new streaks are the darkest at a site and have the strongest contrast², so they are unlikely to be missed for this reason. Second, it is not possible to conduct temporally uniform monitoring of all locations on Mars with HiRISE. Instead, the monitoring sites have variable numbers of images with variable time intervals, and in many cases were selected for repeat imaging because they had sufficient numbers of slope streaks that new formation events appeared likely. Finally, it is likely that some small new slope streaks that are observable by HiRISE have been missed; only detections considered definite are included here, and in addition to resolution limitations, narrow dark streaks can be difficult to distinguish from shadowing associated with subtle linear topographic features. However, the largest streaks should be comprehensively detected, and offer the most information about the morphologic effects of streak formation.

The HiRISE image survey recorded formation of 787 dark streaks. No new bright streaks were observed herein. In addition to the detections in the survey, observations of other slope streaks are also described when they provide additional evidence regarding the formation and effects of the streaks. These include both streaks from the survey locations and elsewhere.

Data Availability

All of the data used in this study are available via the NASA Planetary Data System and/or the HiRISE team website, http://hirise.lpl.arizona.edu. Table S1 provides the geographic locations of all images used.