4.1. Latitudinal Distribution
 The systematic examination of 5112 images from 90°N to 90°S between 180°E to ∼240°E revealed that 10% of the images exhibit TARs. Although this result confirms the prevalence of these aeolian bedforms on the surface of Mars, their latitudinal dependence (Figure 3) suggests these bedforms may not be as globally ubiquitous as previously suggested [Malin and Edgett, 2001].
 The frequency of TARs in 10° latitude bins reveals a latitude trend in both hemispheres (Figure 4). The average number of images in each 10° latitude bin is 284 (SD = 101), and the frequency of images with TARs ranges from 0–45% across all latitude bins. Almost all (92%) of the images with TARs are located between ±50°, with very few TARs in the middle to upper latitudes of either hemisphere, at the poles, or within the latitude bin from 0° to 10° north.
Figure 4. Percentage of examined MOC NA images with transverse aeolian ridges (TARs) from the Primary Mission of Mars Global Surveyor in 10° latitude bins from 90°N to 90°S latitude between 180°E and ∼240°E.
Download figure to PowerPoint
 Most (73%) of the images exhibiting TARs in the study region are located in the southern highlands, with the highest frequency (45%) between 30° and 40°S. Of the 379 images with TARs in the southern hemisphere, 69% occur between 10° and 40°S and 92% occur between 0° and 50°S. The frequency of images with TARs in the southern hemisphere decreases to ∼1−3% in latitude bins poleward of 50°S. Of the 141 images with TARs in the northern hemisphere, 70% occur between 10° and 30°N and 90% occur between 0° and 40°N. Most (∼60%) of the images with TARs in the northern hemisphere are associated with the Olympus Mons Aureole (10−40°N, ∼210−230°E), with the remainder located west of Amazonis Planitia from 0° to 40°N in rough, fractured terrain such as the HNu (undifferentiated) unit in the Highly Fractured Materials, the Ridged Plains Materials (Hr) and the Arcadia Formation. We found very few examples of TARs in the northern lowlands; the frequency of images with TARs in the northern hemisphere decreases to 0−3% in latitude bins greater than 40°N.
 The dearth of images with TARs in the northern lowlands poleward of 40°N is a good indication that the particles comprising TARs are not derived from sediments associated with the hypothesized northern ocean [Parker et al., 1989, 1993]. The small increase in the frequency of images with TARs (∼3%) between 80° and 90°N in the linear dune material (Adl) and between 70° and 80°S in the polar layered deposits (Apl) suggests that an adequate sediment supply exists at these latitudes, along with sufficient wind velocities required to form TARs. The dramatic decrease in the prevalence of TARs in the polar regions relative to the equatorial region and the midlatitudes, however, may be the result of very low seasonal temperatures or frost cover that somehow restrict the mobility of the particles that comprise TARs, thus inhibiting their formation. The latitudinal dependence of images with TARs in this study region reflects changes in regional elevation and roughness (discussed in section 4.5), with the majority of ridges located at elevations above that of the northern lowlands.
 The elevation values in our study region range from below −5 km to over 15 km relative to the MOLA datum [Smith et al., 1999]. To compensate for the relative paucity of images in elevation bins below −5 km and above 3 km (average = 6 images per bin; SD = 15), the data were grouped into the following elevation bins: −6 to −4 km; +2 to 16 km; and 1 km bins from −4 km through +2 km (Figure 5). The average number of images in each elevation bin is 682 (SD = 341). Although the study region includes elevations from below –5 to over 15 km, the vast majority (98%) of the images examined occur between −5 km and +3 km. The frequency of images with TARs between −3 and +2 km is moderately enhanced (12−26%), with the highest concentration between −1 and 0 km (26%). The frequency of images with TARs significantly decreases to <6% at all elevations below −4 km and above +2 km. Of the 520 images with TARs, 97% are located at elevations between −4 and +3 km and 74% are found at elevations between −3 and +2 km.
Figure 5. Percentage of examined MOC NA images with transverse aeolian ridges (TARs) (bars) and the total number of examined MOC NA images (line) from the Primary Mission of Mars Global Surveyor in elevation bins from 90°N to 90°S between 180°E and ∼240°E.
Download figure to PowerPoint
 The thin atmospheric conditions on Mars should cause wind-blown material to be more mobile at lower elevations and under presumably higher atmospheric pressures. This study, however, shows that the processes responsible for forming TARs occurred over a wide range of elevations on Mars, and most commonly at elevations above that of the northern lowlands. The abundance of TARs at low elevations above the northern lowlands (e.g., −3 km to +2 km) suggests that atmospheric pressure is not a limiting factor on TAR development over most or all of the study area. Of the 520 images with TARs, ∼1% occur at elevations greater than 3 km, supporting the observation that aeolian processes do occur at high elevations and under presumably lower than current average atmospheric pressure [Edgett and Malin, 2000a; Malin and Edgett, 2001]. Elevations over 3 km may represent an effective upper limit to where the aeolian processes responsible for TARs can commonly occur. Alternatively, the atmospheric pressure at the time these high-elevation TARs formed may have been very different from present-day conditions due to changes in obliquity and global climate [e.g., Head et al., 2003]. Additional examination of images at higher elevations is necessary to further explore the prevalence of TARs at elevations greater than 3 km to avoid sampling effects that may have occurred within the study region.
4.3. Thermal Inertia
 The map of global thermal inertia (TI) [Mellon et al., 2000; Putzig et al., 2003] provides insight into the physical characteristics of the Martian surface, as it relates to factors such as grain size, degree of induration, abundance of rock, and exposed bedrock. Generally, low TI values are associated with fine-grained and loosely packed material, whereas high TI values are related to rocks, indurated surfaces, and exposed bedrock.
 The majority of the study region contains TI values that range from low to moderate (∼40–340 J m−2 s−1/2 K−1), with the highest values near the north pole (∼500 J m−2 s−1/2 K−1). The lowest TI signature is between ∼35°N to ∼20°S, a region covered by broad dust deposits [Mellon et al., 2002]. Higher TI signatures occur from ∼35°N to 70°N and 20°S to 60°S, a region interpreted to consist of coarser soil, surface rocks, and some degree of indurated soils [Mellon et al., 2002]. The images with TARs in the northern hemisphere (0°–40°N) correspond only to very low values of TI (∼40 J m−2 s−1/2 K−1), whereas the TARs in the southern hemisphere are associated with TI values that range from ∼40–255 J m−2 s−1/2 K−1 (Figure 3). The increase in TI from ∼180 to 330 J m−2 s−1/2 K−1 between ∼50° to 60°S correlates to the sudden decrease in the frequency of TARs in the southern hemisphere poleward of 50°S.
 Dust cover likely inhibits movement of particles by the wind, suggesting that TARs in low TI regions are older than the large-scale dust deposits. Although the dust layer may only be centimeters to decimeters thick, it is responsible for the observed low TI signatures but not sufficiently thick to mute the underlying topography, including the observed aeolian ridges. The age of the large-scale dust deposits is estimated to be 105–106 years on the basis of the estimated thickness and rate of accumulation [Christensen, 1986]. If the TARs in this region are older and therefore underlie the thin dust deposits that influence the regional TI signature, there is no temporal or material connection between the dust and the TARs. It is also possible that the lack of a clear relationship between the distribution of images with TARs and global TI signatures in the study region may be influenced by scale differences. The TI map was derived from surface temperature observations measured by the Thermal Emission Spectrometer (TES) at a resolution of ∼3 × 6 km [Mellon et al., 2002], which is much larger than the scale of the aeolian bedforms under consideration.
4.4. Regional Geology
 The study region contains a total of 53 geologic units as defined by Scott and Tanaka , Greeley and Guest , and Tanaka and Scott . The 520 images with TARs occur in 38 different geologic units. The 38 units containing TARs are grouped into the following ten formations: Channel System Materials (Achp, Hch, Hchp, Hcht); Olympus Mons Formation (Aoa4, Aoa1, Aoa3, Aos); Ridged Plains Materials (Hr); Plateau Sequence (Npl1, Npl2, Nplh, Nplr, Hpl3); Tharsis Formation (Ht1, AHt3, At4, Ht2); Highly Deformed Terrain Materials (v, HNu, Nb, Nf); Medusae Fossae Formation (Aml, Amm, Amu); Impact Crater Materials (cb, cs, s); Arcadia Formation (Aa1, Aa3, Aa4); and Surficial Materials (Ad, Adl, Ae, Am, Api, Apl, As). The total number of images from each group ranges from 74 in the Channel System Materials to 1344 in the Plateau Sequence, (average = 461; SD = 495). The number of images with TARs from each group ranges from 20 in the Impact Crater Materials to 225 in the Plateau Sequence (average = 52; SD = 62). The cumulative area of the units within each formation containing TARs ranges from 156,600 km2 to 5,515,100 km2 (average = 1,709,000 km2; SD = 1,767,300 km2).
 The percent of images with TARs in each formation ranges from 3–39% (average = 17%; SD = 11%, Figure 6). The greatest frequency of images with TARs occurs in the Channel System Materials (39%) and the Olympus Mons Formation (33%), likely indicating the importance of numerous topographic traps within these terrains. The frequency of images with TARs for the remaining formations are <20%, with the lowest values occurring in the Arcadia Formation and Surficial Materials (4% and 3%, respectively). Image density (total number of images in each formation/cumulative area of units in each formation) is a parameter that provides a relative sense of the area captured by the images from each formation (see section 3 for further discussion). The image density for each geomorphic group ranges from 0.00007 for the Olympus Mons Formation to 0.0012 for the Impact Crater Materials (average = 0.00041; SD = 0.00038).
Figure 6. Percentage of examined MOC NA images with transverse aeolian ridges (TARs) in each geologic formation versus image density (total number of images per formation/cumulative area of units within each formation). The following units are associated with each formation: Channel System Materials (Achp, Hch, Hchp, Hcht); Olympus Mons Formation (Aoa4, Aoa1, Aoa3, Aos); Ridged Plains Materials (Hr); Plateau Sequence (Npl1, Npl2, Nplh, Nplr, Hpl3); Tharsis Formation (Ht1, AHt3, At4, Ht2); Highly Deformed Terrain Materials (v, HNu, Nb, Nf); Medusae Fossae Formation (Aml, Amm, Amu); Impact Crater Materials (cb, cs, s); Arcadia Formation (Aa1, Aa3, Aa4); and Surficial Materials (Ad, Adl, Ae, Am, Api, Apl, As).
Download figure to PowerPoint
 The following 15 units from the study region did not contain any images with TARs: Vastitas Borealis Formation (Hvk, Hvm, Hvr); Alba Patera Formation (Aam, Hal); Dorsa Argentea Formation (Hdu); Elysium Formation (d); Aa2 and Aa5 in the Arcadia Formation; Npld and Nple in the Plateau Sequence; Aop in the Olympus Mons Formation; Hf and m in the Highly Deformed Terrain Materials; and Ach in the Channel System Material. The total number of images from these units and formations range from 1 in the Dorsa Argentea Formation and Elysium Formation to 181 in the Vastitas Borealis Formation (average = 51; SD = 62). The cumulative area of these units ranges from 7,400 km2 to 820,200 km2 (average = 236,100 km2; SD = 292,900 km2). The image density for units or formations without TARs is 0.0001 to 0.0003 (average = 0.00021; SD = 0.00008). Although the total number of images examined from these units and formations are far fewer relative to the formations with TARs, their surface areas are much smaller so their image densities are roughly comparable (Figure 6).
 The conditions that facilitate the formation of aeolian bedforms are sufficient sediment supply, adequate wind velocity and low surface roughness. The prevalence of TARs in geologic units at elevations above the relatively smooth northern lowlands correlate to materials displaying considerable topographic relief and many depressions that can serve as sediment traps. The presence of local and regional topography affects wind velocity and direction, which can enhance erosion and sediment transport [Greeley et al., 2002]. In a modeling study of aeolian bedforms in troughs on Mars, Bourke et al. [2004b] show that wind velocities increase by ∼30% on the downwind side of trough margins, indicating that inclined locations are more likely to exceed the threshold wind speed required for sediment transport. Most images with TARs are found in low-standing areas such as Newton Crater, Gorgonum Chaos and Atlantis Chaos as well as the floors of extensional grabens and troughs including the Sirenum Fossae and Memnonia Fossae systems in Terra Sirenum and the Olympus Mons Aureole. Since topographic traps appear to enhance the development of TARs along the floors of these low-lying depressions, geologic units such as the Channel System Materials (Achp, Hch, Hchp, Hcht) and the Olympus Mons Formation (Aoa4, Aoa1, Aoa3, Aos) that are characterized by troughs, valleys, grabens, faults and scarps typically contain the highest frequency of TARs. Although the presence of TARs in channels and topographic depressions might suggest a local sediment source, MOC images do not provide conclusive evidence to support or disprove this hypothesis. Bourke et al. [2004b] suggest that the trough floors may be a sediment source for transverse ridges in linear depressions, and other researchers also propose a local source [Fenton et al., 2003]. The TARs associated with the Olympus Mons Aureole may indicate a local volcanic sediment source, but the pervasive dust cover precludes testing this hypothesis using TES and THEMIS data. The relative paucity of TARs near Elysium suggests that volcanic material may not be a source for TARs in this area, or the volcanic activity associated with Elysium Mons is sufficiently older than Tharsis so that sand-sized particles are not available in this region.
4.5. Local Slope and Surface Roughness
 Since TARs tend to occur on flat-lying surfaces on the floors of steep-sided troughs and craters, we analyzed slopes at 1.85 km/pixel to represent the effects of small-scale, local topography that likely has the greatest influence on the development of these bedforms. Slope values were calculated at 1.85 km/pixel for all images from 60°N to 60°S between 180°E to ∼240°E, resulting in slope values for 483 images with TARs and 2510 images without TARs. The images with and without TARs have average slopes of 1.92° (SD = 0.12°) and 1.36° (SD = 0.04°), respectively. The difference between average slope values for images with and without TARs is statistically significant (p-value <0.001). The slope values correspond to the latitude and longitude point at the center of each MOC image, which does not necessarily represent the exact location of the TARs in the image. The mean slope values for images without TARs represent a myriad of terrains from the smooth northern lowlands to the cratered highlands. The slightly higher average slope values of MOC images with TARs is possibly influenced by the high frequency of TARs at the bottom of steep-sided depressions in the Olympus Mons Aureole and the lack of TARs in the lowlands. In order to better understand the relationship between slopes, roughness and the distribution of TARs, we qualitatively assessed the distribution of TARs in relation to surface roughness and quantitatively analyzed the frequency of TARs in specific geologic units of known roughness (as described by Kreslavsky and Head [1999, 2000]).
 Kreslavsky and Head  mapped the kilometer-scale surface roughness from MOLA data using long, intermediate and small-scale baseline lengths of 19.2 km, 2.4 km and 0.6 km, respectively. They found latitudinal trends of roughness in both hemispheres and distinctive roughness characteristics associated with several geologic units. In general, the southern hemisphere of Mars is rough at all scales, but the equatorial region has a ∼3 times larger small-scale (0.6-km baseline length) roughness than high southern latitudes. In the study region, this latitudinal roughness trend is delineated by a smoothed terrain boundary around 50°S, which directly corresponds to a significant decrease in the occurrence of TARs poleward of this boundary. The difference in the characteristic vertical scale along this boundary is estimated to be on the order of several meters on the basis of model data [Kreslavsky and Head, 2000]. Similar to the southern hemisphere, the northern hemisphere exhibits a relative decrease (∼2−3 times) in small-scale roughness around 47°, although this boundary is not well defined in the study region. The study region in the northern hemisphere has smooth surfaces associated with Arcadia and Amazonis Planitia. From the equator to roughly 40°N, this smooth region with few TARs is bound by regions of high surface roughness that correlate to the Olympus Mons Aureole (between 210° and 240°E) and the knobby and ridged terrain in eastern Elysium. These rough areas in the northern hemisphere directly correlate to high occurrences of TARs.
 The latitudinal trend in roughness is due to the presence of a high-latitude debris mantle that is between 1 and 10 meters thick [Kreslavsky and Head, 2000; Mustard et al., 2001]. Mustard et al.  suggest the deposit is a layer of cemented dust that was emplaced in the last 0.15 Myr and is undergoing desiccation, disaggregation and removal. The emplacement of this mantling layer is thought to be related to a series of climate-driven cycles resulting from changes in obliquity [Kreslavsky and Head, 2000; Mustard et al., 2001], the most recent of which occurred in the last ∼2.1 to 0.4 Myr [Head et al., 2003]. Unlike the thin dust cover draped over TARs in regions of low TI (section 4.3), a mantling deposit up to 10 m in thickness could completely bury TARs that are thought to be on the order of ∼2 to ∼6 m in height [Williams and Zimbelman, 2003; Wilson et al., 2003; Bourke et al., 2004a].
 Kreslavsky and Head  identified characteristic roughness values of several geologic units including Aoa4, Aoa3 and Aoa1 of the Olympus Mons Formation, Npl1 from the Plateau Sequence in the southern highlands, the Medusae Fossae Formation and Aa3 from the Arcadia Formation in the northern hemisphere. The upper member of the Olympus Mons Aureole, Aoa4, is the roughest terrain on Mars (at all scales with a maximum at ∼2.5-km baseline length). This unit forms broad, circular, flat lobes consisting of numerous faults, scarps, and deep troughs and grabens [Scott and Tanaka, 1986]. The Aoa3 member of the Olympus Mons Aureole has a similar characteristic spatial scale of topography relative to the Aoa4 member, but is slightly less rough. The Aoa1 member is essentially smoother than both Aoa4 and Aoa3, with a shorter characteristic spatial scale of topography. The Npl1 unit from the highlands is characterized by craters with smooth floors and steep walls, resulting in rough, spotty terrain at all scales. The surface of the Medusae Fossae Formation is relatively rougher than the northern lowlands and smoother than the equatorial highlands. Lastly, the Aa3 member of the Arcadia Formation is very smooth. On the basis of the relative surface roughness of the Olympus Mons Aureole (very rough), the Plateau Sequence (rough), Medusae Fossae Formation (intermediate roughness) and Arcadia Formation (smooth), we compared the frequency of TARs from each unit in the aforementioned formations (Figure 7). There is a positive linear correlation (R2 = 0.87) between the frequency of images with TARs and relative surface roughness associated with specific geologic units. As surface roughness decreases from Aoa4 to Aoa3 to Aoa1, the percentage of images with TARs systematically decreases (46%, 33% and 29%, respectively). The frequency of images with TARs from presumably rough units in the Plateau Sequence (Hpl3, Nplh, Nplr, Npl1 and Npl2) ranges from 15−25%. The frequency of TARs from units in the Medusae Fossae Formation (Amm, Amu and Aml) identified as “intermediate” roughness range from 12−13%. The smooth units in the Arcadia Formation (Aa3, Aa1, and Aa4) have the lowest frequency of TARs (3−4%). Therefore the frequency of TARs in geologic units and formations appears to be directly related to (small-scale) surface roughness, and the availability of sediment traps associated with this roughness.
Figure 7. Percentage of examined MOC NA images with transverse aeolian ridges (TARs) versus image density (total number of images per formation/cumulative area of units within each formation) for select geologic formations across a range of surface roughness values.
Download figure to PowerPoint