The first example of terrestrial dust devil tracks is presented in this paper. Tracks found in Ténéré Desert, Niger are formed by transient events not related to regional winds. Compared to the Martian tracks, Ténéré tracks are generally longer and show higher average density. We interpreted these differences as due to different intensities of the dust devil vortices combined with different surface properties. We also suggest that grain size distribution and sorting of surface material is crucial to allow track formation. Major surface changes of Ténéré tracks have been observed in selected areas over a time span of 2 years, confirming the very low preservation potential of the tracks. However, no clear evidence for seasonal variations has been found on the available dataset. The Ténéré Desert represents a unique site to study the formation and evolution of these peculiar features and to compare their behavior on other planets.
 Dust devils are warm-cored and low-pressure vortices of air and dust that originate from unstable air at the near-surface boundary layer caused by surface heating [Ives, 1947; Sinclair, 1966; Rennó et al., 2000]. When dust devils occur, they mobilize dust and fine-grained unconsolidated deposits from the surface. Terrestrial dust devils have radii of 1–150 m and can reach heights of several kilometers, transporting mobilized particles over great distances [Sinclair, 1969]. On Earth, they are mainly formed in hot arid or semiarid regions.
 Dust devils have been observed on Mars by both orbiters and landers since the time of the Viking missions [Thomas and Geriash, 1985; Metzger et al., 1999]. A large number of dark, filamentary tracks observed on the Martian surface have been interpreted as being caused by dust devil passages [Edgett and Malin, 2000; Balme et al., 2003], and active devils associated to these tracks have been observed [Edgett and Malin, 2000]. Dust devil tracks are produced by erosional process and they are not related to regional wind directions. Conversely, wind streaks represent depositional events controlled by winds. Dust devils are, generally, not affected by local topography [Balme et al., 2003]. Also, they show higher sinuosity than wind streaks due to the random movement of the vortex.
 The common finding of dust devil tracks on the surface of Mars led us to look for similar features on Earth. We discovered surface structures with a striking similarity with Martian dust devils tracks in Western Ténéré Desert, Niger (Figure 1).
 The peculiar geological, morphological and climatic conditions of Ténéré Desert along with the availability of high resolution (15 m/pixel) satellite images allowed the observation of these unusual features and the monitoring of their temporal variation. This study has been performed on a multitemporal Advanced Spaceborne Thermal Emission Spectrometer (ASTER) dataset consisting of about 30 radiometrically corrected frames (Figure 1), georectified and co-registered in a GIS archive. The frames were acquired from July 19th, 2000 to July 16th, 2002. The data coverage is not uniform in time and space but important temporal track variations have been observed in selected areas as summarized in Table 1.
Table 1. Tracks Temporal Variation
Date of Frames Acquired
Amount of Tracks and Direction
May 26th, 2001 frames record the largest occurrence of tracks.
Tracks seen in Area 4 in December 1st, 2000 are not the same of September 28th, 2000.
 The study area (Figure 1) is an extremely flat and gently southward sloping plain located at the western border of Ténéré Desert, Niger. The Air Mountains, a series of crystalline and volcanic massifs ranging in age from Precambrian to Quaternary [Choubert et al., 1987], limit the study area to the West and reach the height of about 2000 m asl. Another remarkable range characterizes the study area, the Adrar Madet, a narrow massif (∼3 km wide and 20 km long) of Cretaceous sedimentary rocks [Choubert et al., 1987] elongated in NNW-SSE direction. The Adrar Madet is about 200 m higher than the surrounding plains. Today, dune fields and sand sheets are mainly present in the study area [Hugot, 1962; Warren, 1972; Mainguet and Callot, 1978]. Widely spaced longitudinal (seif) dunes surround the Adrar Madet (Figure 2). Transverse dune fields form the Grand Erg du Fachi-Bilma to the S-SE, and the Erg Brusset to the NW (Figure 1). The seif dunes can be up to 7 m high and are generally higher than the transverse dunes [Warren, 1972]. Sand sheets are concentrated on the western side of the Adrar Madet massif (Figure 2). The unconsolidated sand sheets are of few centimeters up to decimeters thick. Sand sheets cover brown paleosols formed during the Holocene humid climatic periods [Felix-Henningsen, 2000].
 The climate of the study area is mainly controlled by the annual movements of the inter-tropical convergence zone (ITCZ), characterized by the mixing of the dry tropical air and wet equatorial air. The mean annual rainfall from 1978 to 1990, has been 100–170 mm, with a minimum of <10 mm [Gasse, 2002]. Scattered rainfalls mainly occur in July and August due to monsoon circulation. During the dry and hot season (approximately March–May), the region is under the influence of the Harmattan circulation characterized by NE trade winds. The Adrar Madet massif is an obstacle for these regional NE-SW winds [Hugot, 1962; Mainguet and Callot, 1978], shielding the area immediately to the SW from the wind [Mainguet and Callot, 1978]. For this reason, arcuate seif dunes form on the NE side of the Adrar massif (Figure 2).
3. Description of Dust Devil Tracks
 The track mapping has been performed on ASTER Green Band images (0.52–0.60 μm) because the highest albedo contrast between the tracks and the surroundings is in the Green Band, decreasing towards Red and Near Infrared wavelengths. Similar behavior has been observed for some Thermal Emission Imaging System (THEMIS) images of Mars.
 Tracks in Ténéré Desert have been found in several frames acquired in different seasons (Table 1). The following track description is based on the May 26th, 2001 scenes where the largest number of tracks (more than a thousand) have been observed (Figures 2a and 2b) whereas in other images we found only a few tens of tracks when present.
Figure 2a shows the density of tracks mapped in May 26th, 2001. The tracks are concentrated on various terrain types: transverse dune fields, sand sheets and interdune seif zones. Tracks are not present SE and SW of the Adrar Madet where seif dunes are closely spaced (<1 km). The average density is ∼2 track per km2. The highest concentration of tracks (∼4 track per km2) have been found in smooth interdune and sand sheets areas and in the transverse dune fields of the easternmost portion of Erg Brusset (Figures 2a and 2b).
 The tracks can be recognized as low albedo filaments with variable width and length (Figures 3a–3e). They generally show a low degree of sinuosity but sometimes, curling tracks are observed (Figure 3a). Ténéré tracks show relatively small variations both in thickness and length, over a large population. The width of mapped tracks is on average a few tens of meters, but widths of several hundreds of meters have been measured for a limited number of tracks. The average length is about 3 km, reaching a maximum of 8.5 km. The length to width ratio is generally high (>20:1) but can be about 10:1 for a small population. The tracks often cross-cut each other (Figures 3a–3e). However, an age relationship cannot be inferred at the given image resolution. Undulated terrain or dune-covered surfaces are characterized by shorter and more curvilinear tracks (Figures 3b and 3c). Large, flat featureless sand sheets generally contain more linear and longer tracks (Figures 3d and 3e).
 Although most of the tracks maintain their original direction when crossing linear dunes (Figures 3b and 3c), we have noted that local relief formed by seif dunes may affect track directions in places (Figure 3e).
 The tracks mapped on the May 26th, 2001 scenes show a preferred NW trend, although a significant percentage of the track population spans from NNW to NNE (Figure 4). Tracks observed on December 1st 2000 and November 15th 2000 show an EW trend (Table 1). The small number of tracks found in the September 28th 2000 are characterized by NNE direction (Table 1). Tracks seen in the same area at different times are not the same. We did not observe any active dust devils on the ASTER images.
4. Discussion and Conclusion
 Since dust devils are formed by thermodynamic air instability caused by surface heating, they are not controlled by wind direction. If the Ténéré tracks were formed by dust devil passages they should not be aligned to local wind direction. Regional wind direction in the study area comes mainly from the NE (Figure 4) during the acquisition time of the May 26th, 2001 scenes as suggested by wind streaks observed to the West of Adrar Madet (Figure 2b). This confirms that the observed features cannot be formed by the regional winds.
 The striking morphological similarity between terrestrial and Martian tracks (Figure 3f) is evident at various scales. Martian dust devils tracks observed on Mars Orbiter Camera (MOC) and THEMIS images show a wider range of geometries and dimensions [Edgett and Malin, 2000; Balme et al., 2003]. Martian tracks in Hellas and Argyre Basins have average width of ∼20 m and average length ∼0.5 km [Balme et al., 2003]. Compared to Martian tracks, the Ténéré ones have similar widths but greater lengths. The average density of tracks in Ténéré (∼2 tracks per km2) is higher than those averaged for Argyre Planitia (0.81 tracks per km2) and the Hellas Basin of Mars (0.47 tracks per km2) [Balme et al., 2003]. However, densities of 50–100 tracks per km2 have been measured on Mars as well [Balme et al., 2003], outlining the large heterogeneity of track population.
 If the tracks reflect the dimension and intensity of the dust devil vortices, the difference in track lengths between Earth and Mars may reflect a different intensity of the vortices themselves. It seems that the vortices formed in Ténéré have higher energy which dissipates over a larger distance.
 A previously proposed hypothesis [Rossi, 2002] to explain the dark signature of the tracks considered the possible exposure of brown paleosols which underlie the sand sheets. This process requires a removal of tens of centimeter of loose sand for large areas (several hundreds of meter). Calculation done by Balme et al.  demonstrate that a dust devil may remove a layer of 2–40 μm of loose sediment from the surface. This thickness is not enough to expose the brown paleosols in the Ténéré study area.
 The role of the substrate in track formation is still not well understood and may play an important role in this process. The work of Balme et al. , suggests that the presence of a thin surface layer of dust is necessary to preserve the track morphology. Our observations of Ténéré Desert tracks, suggest that the grain size and sorting of the available loose sediments may affect the track formation mechanism. We found that the highest concentration of tracks is on flat sand sheet areas and low relief dunes. These eolian deposits of the Ténéré Desert are characterized by bi-modal sands with modes of very fine sand (63–125 μm) and coarse sand (500–1000 μm) [Warren, 1972; P. Felix-Henningsen, personal communication, 2002]. During a dust devil passage, the very fine sands can be easily removed from the surface and entrained in the vortex plume whereas the exposed surface results enriched in coarse sands. The exposure of coarse sand increases the surface roughness that can cause the strong albedo contrast between the track and the surroundings. Seif dunes consisting of well sorted fine sands (125–250 μm) [Warren, 1972], should not allow the formation of track during a dust devil passage. In fact, we did not observe any tracks on seif dunes.
 Another key aspect of the observed tracks is the high temporal variability. Selected areas with a good multitemporal data coverage during the 2 year campaign have been compared in order to study the surface variations with time (Table 1). We discovered that tracks shortly disappear, in few weeks or months later, at most. Also, newly formed tracks do not generally maintain the same direction of previous ones.
 The tracks seen in May 26th, 2001 images were not present in frames acquired in 19th May, 2001, just a week before. Images taken two months later (19th July, 2001) do not show any tracks (Table 1). Thus, all the tracks have been formed in a just few days and then erased within a few weeks. We have found a few tracks during winter (Sept-Nov-Dec. 2000) but not in late summer 2001 and 2002. Based on our dataset, we did not find any evidence for the season-related dust devil activity as observed on Mars [Balme et al., 2003]. Also, we do not know if an important track occurrence such as that on May 26th, 2001, have been observed subsequently, due to the low preservation potential of the tracks or the gaps in coverage of the ASTER satellite. Study of a larger multitemporal dataset is probably needed in order to search for large occurrences of tracks and their temporal variability.
 The strong morphological similarity with Martian dust devil tracks and the observed characteristics of Ténéré tracks suggest that these features were formed by transient events not related to regional wind direction. For these reasons, dust devil vortex passages are the best candidates for explaining the observed tracks. The observation of these features on Mars has been critical to interpret the origin of the enigmatic tracks found in Ténéré Desert. This is one of the rare examples of features recognized on other planets that help in the interpretation of terrestrial geological features and not vice-versa.
 The authors wish to thank Prof. Peter Felix-Hennigsen for providing useful information on Ténéré sand sheets, Gian Gabriele Ori for fruitful discussion and the Japan ASTER User Service for kind advises in image processing. We are grateful to Victor R. Baker and an anonymous reviewer for their useful comments. We also warmly thank Martha S. Gilmore for revising the final manuscript. Some of the ASTER images were obtained under a NASA approved data acquisition proposal. This work has been supported by an Italian Space Agency grant.