5.1. Topoclimatic Contrasts With Aspect
 Annual air temperatures on south-facing canyon slopes in the study area were 1.4–2.1°C warmer than north-facing slopes (Table 1), and differences in winter were 2–3°C warmer (Figure 8). Modern vertical temperature gradients in the southern Colorado Plateau are 5–6°C/1000 m [Stute et al., 1995], thus air temperature differences among aspects (Table 1) are roughly analogous to elevation differences of about 230–600 m. Paleoclimatic reconstructions for the southern Colorado Plateau and adjacent areas indicate a 2–4°C range of mean annual temperatures in the Holocene, and late Pleistocene temperatures 4–6°C cooler than present [Phillips et al., 1986; Cole, 1990; Davis and Shafer, 1992; Anderson, 1993; Stute et al., 1995; Zhu et al., 1998]. Temperature differences between north- and south-facing slopes thus approach long-term differences in Holocene mean temperatures, but both analogies are limited. For example, precipitation generally increases with elevation, and late Quaternary climate changes are likely associated with precipitation changes as well. In contrast, topoclimates on opposing aspects have small precipitation differences, and temperature contrasts change dramatically over the seasons. Numerical modeling of desert soils, however, implies that a temperature decrease of 4°C, with no precipitation increase, significantly increases infiltration depth, leaching, and weathering rates [McFadden and Tinsley, 1985].
 Ultimately, soil moisture is likely the most important aspect-influenced control on weathering and slope form in the study area. Differences in field-measured moisture potential are probably most related to differential evapotranspiration driven by insolation and temperature [Kirkby et al., 1990], as both latent heat transfer and evaporation are proportional to temperature and net radiation [Barry and Chorley, 1998]. Lower soil moisture is observed on south aspects during the warm season except during brief periods in the summer monsoon (Figure 7d). Evapotranspiration is higher on south-facing slopes, as indicated by more rapid drying after storm events and during the spring. In contrast, only the north-facing sensor on bedrock indicated a relatively high potential all year. Soil moisture potential is highest in winter, but the contrasts between aspects are smallest, even though insolation and temperature differences between north- and south-facing slopes reach their highest magnitude with low-angle winter sun. With lower temperatures, winter evapotranspiration is generally lower than precipitation, and soil moisture increases on all slopes, reducing differences between north and south aspects.
 Windward slopes experience greater precipitation and evapotranspiration [Barry and Chorley, 1998], but the net effect of wind on soil moisture in the study area is unknown. Prevailing winds indicated by Holocene eolian features in northeast Arizona [Stokes and Breed, 1993] and northwest New Mexico [Wells et al., 1990] are southwesterly. Wind direction during storms is much more variable. It is not likely that wind-generated precipitation contrasts between north and south slopes are large, as the study area canyons probably channel winds subparallel to both aspects.
 Only a relative assessment of the effects of soil moisture on slope vegetation can be made, as the permanent wilting point for plants is about −1500 kPa [Brady and Weil, 2000], well below the minimum potential resolved by the field instruments (−200 kPa). Calculations of evapotranspiration, potential evapotranspiration and water stress are also impossible without humidity data. Higher potential at north-facing sensors implies less water stress, particularly during the summer, which likely allows for almost three times greater vegetation cover on north aspects than on south-facing slopes [Burnett, 2004].
5.2. Slope Processes and Asymmetry
 Excluding cliffs, the mean angle of south-facing slopes is statistically greater than north-facing slopes, but differences are quite small. Slope gradients range broadly around the angle of repose (Figure 5) and are controlled over substantial areas by the mobility of the cover of weathered bedrock and colluvium. Many unweathered bedrock slopes lie near the angle of repose as well (Figure 10c) and were probably covered by a weathered mantle or colluvium in the recent past. Some unweathered bedrock slopes well exceed the angle of repose, however, and grade into cliffs, and gentle slopes exist on highly weathered sandstone and shale. If the small aspect differences in noncliff slope angles between north and south slopes are real, they probably relate to these weathering and bedrock strength contrasts. The difference in cliff area between north- and south-facing slopes, however, is very large (Table 5) and is the principal element of asymmetry within the canyons. Cliffs of 5–70 m height exist along the entire length of south facing-slopes except at one place in western basin 5. In contrast, north-facing slopes have no cliffs higher than 1–2 m, except in narrow canyon heads.
 Asymmetry in the study area canyons has similar aspect relations to that in the South Dakota badlands on weak rocks [Churchill, 1982], and is most likely the result of topoclimatically controlled weathering and erosion processes. Enhanced weathering and debris formation on north-facing slopes keeps slope angles lower by providing more erodible surface material, rapidly reducing ledges, and minimizing the area of cohesive bedrock that can hold steep slopes. Bedrock on little-weathered south-facing slopes, however, is strong enough to hold cliffs at least 70 m high, and noncliff slopes are also slightly steeper on average.
 In contrast, asymmetry documented elsewhere in the western U.S. is often characterized by steeper north-facing slopes [Melton, 1960; Dohrenwend, 1978; Branson and Shown, 1989; McMahon, 1998]. This pattern has been attributed largely to greater vegetation density on moister north slopes that (1) reduces runoff through increased rainfall interception, infiltration rates, and surface roughness and (2) increases erosion resistance through root strength [Melton, 1960; Selby, 1993; Schmidt et al., 2001]. As in Churchill's  badlands study, however, vegetation cover in our study area is low even on north-facing slopes [Burnett, 2004], with limited effect on erosion by surface runoff. Nevertheless, the greater vegetation cover on north-facing slopes may somewhat moderate slope angle asymmetry.
 By itself, bedding within 10° of horizontal has limited effect on slope stability [Selby, 1980, 1993], but given sapping processes, even slight inclination of a caprock unit may produce slope asymmetry with a steeper scarp slope facing away from the dip direction [Howard and Selby, 1994]. The slight southwesterly dip of the strata in our study area, however, would tend to favor cliffs on north-facing slopes, opposite of the pattern we document. In addition, asymmetry with steeper south faces exists where only the Morrison formation is present, and is also quite apparent in basins 1 and 2 where the Dakota caprock is completely absent. Caprock cliffs in the study area may be either Dakota or Morrison formation sandstones. Erosion of the weaker rocks below undermines cliffs [Howard and Selby, 1994], and cliff height in our study area is probably largely a function of weathering and erosion rates on the underlying slopes.
 An aspect of asymmetry that is apparent but unquantified is the greater development of stream drainages on north-facing slopes (Figure 2). This may result from the greater area for surface runoff generation on slopes as opposed to cliffs, as well as greater erodibility of more weathered north-facing slopes. Also, once a major cliff develops, focused headward erosion along a drainage is limited by the cliff face itself.
5.3. Morrison Formation Weathering
 Because slope processes in the study area change dramatically with weathering, we further consider weathering processes and effects. With progressive weathering of Morrison formation sandstones, Young's modulus and bulk density are reduced, and porosity increases. Strength reduction probably results from both loss of cohesion in the clay cements and reduced grain contact friction. Progressive reduction in bulk density requires mechanisms that expand the pore spaces (Figure 12) and (or) isovolumetrically remove mass without substantial compaction. There is no reduction in cement or matrix with weathering, but more weathered samples are less grain supported, indicating significant rock expansion [Burnett, 2004]. If mass is lost by leaching, it is likely small and subordinate to physical weathering by expansion. Frost action and the swelling of clays are two processes that can cause volumetric expansion.
 The specific processes involved in frost action are poorly understood for soft, porous rocks. Frost action in most bedrock requires temperatures of −5°C and −10°C and moisture conditions near saturation [Walder and Hallet, 1985; Matsuoka, 1990], but strain experiments [Matsuoka, 1988] in porous rocks indicated that most expansion occurs between 0 and −5°C. Winter surface temperatures in the study area often fell below freezing, and frost heave has been observed in the upper 1 cm of slope debris, but temperatures at 10 cm depth in weathered bedrock rarely cooled below 0°C and never reached −5°C, even with a 1–2°C correction for warmer than average temperatures in winter 2002–2003. The bedrock never reached saturation (>−10 kPa), but was relatively moist (>−100 kPa) on north faces all winter and on south faces March through May (Figure 7d). Soil moisture in colluvium also decreased with depth (Figure 7d), implying that frost action could not be responsible for producing >30 cm thick weathered mantles on bedrock. Temperature and moisture conditions are unfavorable for frost action below a few centimeters depth, but could accelerate breakdown of near-surface materials.
 In contrast, Schumm and Chorley  inferred that frost action was the most important weathering process causing disintegration of sandstone talus on the Colorado Plateau. They measured mass loss of boulders exposed to natural weather augmented by simulated rainfall over ∼1.5 year in Denver, Colorado. The most mass loss per unit precipitation occurred with freeze-thaw cycles, but with reduced evaporation in winter, hydration would also be enhanced. Also, since precipitation was artificially doubled, the rocks were more likely near saturation during freeze-thaw cycles. The most rapid weathering occurred when 80 mm of simulated rainfall was applied during midwinter, and when the rocks were buried in intermittently melting snow. These are not common conditions in our study area, and in particular, saturated conditions are unlikely on moderate to steep bedrock slopes and cliffs. Frost action is probably most effective in talus weathering in our study area, consistent with Schumm and Chorley's  observations and experimental design.
 Swelling of smectite clays within the sandstones has been suggested as the primary weathering mechanism in the Morrison formation [Tillery, 2003] and in other clay-rich sandstones [Vicente, 1983]. X-ray diffraction analysis of Morrison sandstones reveals that the major clays are smectite and kaolinite [Tillery, 2003]. Swelling of smectites is directly related to the mass water content in soils [Fu et al., 1990; Cygan, 2002]. Since volumetric and mass water contents in sandy materials change only slightly below −100 kPa [Brady and Weil, 2000], we assume that only fluctuations in moisture potential above this limit are important for smectite hydration. The period over which soil moisture potential is above −100 kPa or the number of wetting cycles recorded provide estimates of weathering potential, and both measures suggest that potential clay expansion on north-facing slopes is twice that of south-facing slopes [Burnett, 2004]. The moisture on north-facing slopes is also more likely to reach greater depths and produce thick weathered mantles.
 Clay expansion is also likely to be the dominant weathering process in shale units of the Morrison, Dakota, and Mancos formations. Weathered surfaces on these units exhibit “popcorn” morphology attributed to cycles of clay hydration [e.g., Churchill, 1982]. Shattered pieces of well-cemented Dakota sandstone comprise most of the coarse material in colluvium, showing that they weather much more slowly than Morrison formation sandstones.
5.4. Slope Processes and Implications for Landscape Evolution
 Repeat ground-based LiDAR measurements in the study area have shown that prolonged, intense precipitation causes significant erosion of the weathered sandstone mantle and some debris slopes [Wawrzyniec et al., 2007]. The importance of heavy rainfall in stripping weathered material is supported by soil and dendrogeomorphic data in the study area that indicate rapid erosion in major storms following prolonged droughts [McAuliffe et al., 2006]. Once bedrock is exposed (e.g., Figure 3), surface runoff increases dramatically. These observations imply that two major positive feedbacks exist between weathering, erosion, infiltration and runoff on the Morrison formation (Figure 13). In feedback 1, increased soil moisture enhances production of a weathered mantle. This weathered bedrock increases soil moisture infiltration and retention that accelerates hydration weathering, and maintains or expands the mantle [cf. Wahrhaftig, 1965]. Feedback 1 is directly linked to evapotranspiration differences driven by topoclimatic contrasts in insolation and temperature and is largely responsible for the differences in weathering and rock strength observed on north- and south-facing slopes. Feedback 1 enhances weathering on moist north-facing slopes and increases erodibility, but reduces surface runoff, the primary agent of slope erosion.
Figure 13. Feedbacks between the weathered mantle and the (a) soil moisture and (b) runoff systems. Arrow links indicate a positive relationship between the two elements joined, and a links that end in dots indicate a negative relationship between the elements.
Download figure to PowerPoint
 A second positive feedback (Figure 13) helps prevent development of a weathered mantle on unweathered bedrock. High runoff generation on unweathered slopes strips away weathered material below, maintaining or expanding the area of unweathered bedrock (Figure 3). Feedback 2 has the greatest effect where a drier topoclimate hinders weathering, and is negated where a thick weathered mantle limits runoff. The two feedbacks push slope evolution toward two end-member geomorphic expressions for this sandstone bedrock: (1) a transport-limited slope mantled by 10 to 30 cm of disintegrated bedrock and (2) an unweathered, sediment-limited bedrock slope. The observed behavior of this overall slope system is similar to that proposed by Gilbert [1877, p. 97], where presence of a soil mantle promotes water retention and further weathering (although weathering is thought to slow with increasing thickness); but where the mantle is stripped, weathering is negated. The system may thus be approximated by a “humped” curve of soil production rate as a function of soil depth, where the maximum production rate (by weathering) exists at some nonzero soil depth, and declines with either increasing or decreasing depth [Carson and Kirkby, 1972; Ahnert, 1976]. This system contrasts with models of exponentially declining soil production rate with soil depth proposed for more continuously soil-mantled landscapes, where the production rate is maximum at zero soil depth [e.g., Heimsath et al., 1997].
 The prevalence of south-facing cliffs implies that steepening of drier bedrock slopes ultimately proceeds to a mass failure-dominated system. Slab failures and topples typically result from slope erosion removing support from cliff bases, where overburden stress is highest [Selby, 1982]. Runoff from cliffs may also promote slab failures by enhancing hydration weathering and erosion of the slopes near the base of a cliff, causing undercutting. Erosion processes therefore differ markedly with aspect. Neither feedback directly implies that long-term erosion rates are different with aspect, but if cliff retreat rate is controlled mainly by the rate of undermining by slope erosion below, then south-facing canyon walls may erode more slowly.
 The potential for gully incision and erosion is also enhanced by bedrock weathering, but buildup of coarse debris in channels inhibits flow and sediment transport, and armoring can lead to enhanced lateral erosion and channel migration [Bryan, 1940; Mills, 1981; Twidale and Campbell, 1986]. East-facing gully sideslopes are gentler and more often debris covered than opposing west-facing sideslopes (Table 3, Figure 10). Material from the east sideslopes armors that channel wall, forcing undercutting of the less protected west-facing wall, resulting in a steeper sideslopes with more exposed bedrock, and eastward gully migration. This also undercuts any colluvial mantle on the adjacent east-facing slope, but the west-facing wall is too steep to maintain a debris cover, so that colluvium adds to the armor. Small debris flows are also common, and their deposits may have aided the armoring process in some gullies. These self-reinforcing processes produce short, steep bedrock slopes that are most common in the west-facing facet data.
 Cliff formation and destruction processes are clearly key in interpreting long-term landscape evolution in these canyons. Cliffs appear to begin as either steep unweathered Morrison slickrock slopes (Figure 3b) or as small resistant ledges in the Dakota formation. Where headward gully erosion or slopewash exceeds weathering on Morrison formation slopes, progressively stronger bedrock is exposed (Figure 14a, line a). Under low soil moisture, the upper slopes remain unweathered and become steeper, as they erode more slowly than the slopes below that receive runoff (Figure 14a, line b). The unweathered slopes expand laterally and vertically until failure occurs along tectonic or slope parallel unloading fractures (Figure 14a, line c). Where weathering exceeds erosion, steep slickrock slopes required for cliff initiation in the Morrison formation cannot form. Cliff initiation also occurs where Dakota or upper Morrison formation sandstones behave as erosion resistant caprocks. Slopes below the caprock continue to erode, leaving a cliff below the resistant unit until overburden stresses cause the weaker lower unit to fail [Koons, 1955]. Both of these processes are diminished on north-facing slopes that generally have enhanced moisture and weathering.
Figure 14. Cliff initiation from a slope. (a) Headward gully erosion produces a steep, unweathered upper slope (dotted lines a and b), which is ultimately undercut and fails along a nearly vertical fracture producing a cliff (line c). This process will continue and extend the existing cliff. (b) Where unweathered bedrock slopes are not extensive and not long-lived, slope parallel retreat occurs with only small ledges forming where resistant units (shaded area), typically Dakota Sandstones, outcrop. (c) Cliff reduction can occur where high weathering and sediment production rates on the cliff and upper slope reduce retreat rates of the lower slopes. (d) Cliff expansion is also aided where near-vertical unloading joints or tectonic fractures (dashed lines) facilitate slope weathering and provide a vertical erosional contact.
Download figure to PowerPoint
 Conversion of slopes to cliffs is rarely discussed in landscape evolution models, but is commonly apparent on south-facing slopes in these canyons. The slope-cliff transition is wavy and rarely follows lithologic contacts; continuous rock units pass laterally from cliff to slope and back to cliff (Figure 4). Depressions in the slope-cliff transition occur where erosion is enhanced by tributaries that drain over the canyon rim and by runoff from steep bedrock slopes (Figure 4). The contact is also depressed near the eastern mesa points, where two major cliffs intersect. The slope-cliff transition is thus controlled by weathering and erosion rather than by stratigraphy
 Once initiated, cliffs may expand downward by slope erosion combined with relative stability of the near-vertical face. Slab failure along stress fractures parallel to the cliff free face also facilitates rapid conversion of bedrock slopes to cliffs. Cliff parallel fractures were observed mostly in slopes below higher cliffs or those aligned with the main tectonic fracture set. The fractures also extend into bedrock below cliffs (Figures 3c and 14d) and aid infiltration, weathering and erosion of cliff-fronting slopes.
 On north-facing slopes, cliff reduction processes have clearly dominated. Enhanced weathering and debris accumulation on slopes below cliffs leads to a reduction in cliff height (Figure 14c), similar to classic slope decline and replacement models [Davis, 1899; Penck, 1924]. A persistent weathered mantle or debris cover on moister north-facing slopes retards slope retreat and cliff expansion, instead promoting upward extension of the slope. The debris mantle shed from retreating Dakota sandstone caprock is unlikely to be the primary factor controlling long-term slope erosion, as present debris covers are thin mobile layers, and relict talus slopes or “flatirons” [Howard and Selby, 1994] are not present. Cliffs in the study area are not observed to buried by basal debris. Thus, valley asymmetry is largely a reflection of the net relative effectiveness of initiation, growth and reduction processes of cliffs, largely controlled by bedrock weathering and erosional processes on subjacent north- versus south-facing slopes. Modern topoclimatic contrasts are sufficient to place north- and south-facing slopes in different modes of development. Under modern and Holocene climates, cliff expansion has clearly been dominant on south-facing slopes, but north-facing slopes show more diversity, and some evidence for cliff initiation, growth and decline can be locally observed on slopes of all aspects at present.
 Although strongly suggestive, modern monitoring data cannot directly indicate whether topoclimatic differences were effective over timescales sufficient for slope evolution, under very different, pre-Holocene climates. Estimated slope retreat rates in the study area based on dendrogeomorphic analyses [McAuliffe et al., 2006] and cosmogenic nuclide accumulation in Jurassic sandstone concretions [McFadden et al., 2005] are similar at a few mm per year, despite 10–102 and 103 year measurement timescales, respectively. Perhaps coincidentally, retreat rates over 107 year based on assumed post-Laramide initial escarpment positions are also a few mm per year [Schmidt, 1989]. These data suggest that major changes in slope morphology could occur over 104–105 years in the middle to late Quaternary.
 Although Pleistocene climates were not uniformly wetter than present, episodes of greater effective moisture may have favored cliff reduction and debris slope processes, with very little cliff area on north aspects. It is possible that at times in the Quaternary, cliffs were mostly erased from these canyons, with slopes covered by continuous debris blankets, as inferred on other Colorado Plateau escarpments [Howard and Selby, 1994], but no field evidence exists to test this. Modern debris slopes appear to be largely relict late Pleistocene features that are degrading, particularly on south-facing slopes where backwasting of the escarpment appears minimal at present. While some smaller south-facing cliffs may have formed entirely within the Holocene, many are too high for this to be the case given estimated rates of slope retreat and cliff formation. For example, cliffs in the eastern study area are up to 70 m tall, which requires an unlikely combination of vertical slope lowering of 7 mm a−1 (where a is years) and a virtually noneroding, downward-extending cliff, if restricted to the Holocene. Even faster slope lowering is necessary if some cliff retreat is allowed. Thus, some net growth likely occurred on those cliffs within the late Pleistocene.