Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
 Understanding interactions among tectonics, topography, climate, and erosion is fundamental to studies of mountainous landscapes. Here, we combine topographic analyses with modeled distributions of precipitation, insolation, and flexural isostasy to present a conceptual model of topographic evolution in the Teton Range, Wyoming, and test whether efficient glacial relief production has contributed to summit elevations. The conceptual model reveals a high degree of complexity inherent in even a relatively small, glaciated, mountain range. Back tilting has caused topographic asymmetry, with the greatest relief characterizing eastern catchments in the center of the range. Two high summits, Grand Teton and Mount Moran, rise hundreds of meters above the surrounding landscape; their elevations set by the threshold hillslope angle and the spacing between valleys hosting large, erosionally efficient glaciers. Only basins >20 km2 held glaciers capable of eroding sufficiently rapidly to incise deeply and maintain shallow downvalley gradients on the eastern range flank. Glacial erosion here was promoted by (1) prevailing westerly winds transporting snow to high-relief eastern basins, leading to cross-range precipitation asymmetry; (2) the wind-blown redistribution of snow from open western headwaters into sheltered eastern cirques, with the associated erosion-driven migration of the drainage divide increasing eastern accumulation areas; and (3) tall, steep hillslopes providing shading, snow influx from avalanching, and insulating debris cover from rockfalls to valley floor glaciers. In comparison, the topographic enhancement of glacial erosion was less pronounced in western, and smaller eastern, basins. Despite dramatic relief production, insufficient rock mass is removed from the Teton Range to isostatically raise summit elevations.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 An important caveat to the glacial buzz saw hypothesis is that, in some ranges, elevated peaks may rise high above the surrounding topography, forming notable isolated summits. Such features, termed “topographic lightning rods” by Brozović et al.  and “teflon peaks” by Anderson , are commonly formed in the hard crystalline core of mountain ranges undergoing relatively rapid rock uplift, and are characterized by steep hillslopes that allay glacier buildup, mitigating erosion. These peaks have the potential to disrupt atmospheric flow, and bolster snow influx and the efficacy of glacial erosion in neighboring valleys [Brozović et al., 1997; Anderson, 2005]. For example, in the Olympic Mountains of Washington state, the largest valleys with the greatest volumes of eroded material surround the flanks of Mount Olympus [Montgomery and Greenberg, 2000], while glaciated valleys with shallow longitudinal profile gradients and extensive hillslopes more than 3 km tall surround Nanga Parbat, Pakistan [Brocklehurst and Whipple, 2007].
 In the Teton Range, maximum topography generally parallels along-strike trends in last glacial maximum (LGM) and long-term glacier ELAs. We have previously suggested that this indicates a glacial buzz saw limit to maximum elevations for much of the range (Figure 2) [Foster et al., 2008]. However, two high peaks, Grand Teton and Mount Moran rise conspicuously hundreds of meters above the surrounding landscape [Foster et al., 2008]. The dramatic relief associated with these peaks makes for a suitable setting in which to test suggestions that differential erosion between valley floors and peaks can isostatically raise summit elevations [e.g., Molnar and England, 1990]. Published distributions of fault slip rate [Machette et al., 2001; O'Connell et al., 2003] are taken to be indicative of the pattern of long-term rock uplift rate. We compare these with metrics of range- and basin-scale topography, reconstructed LGM ice extents, and modeled modern patterns of precipitation and solar radiation to examine how topography influences patterns of glacier accumulation, ablation and erosion. These insights are used to construct a conceptual model of topographic development in the Tetons over the last several glaciations and highlight a number of feedbacks between climate and topography that we see as fundamental to carving the present landscape at both the range and catchment scale. Finally, we implement a simple model of flexural isostasy in order to assess whether the superelevation of Grand Teton and Mount Moran, and the relatively rapid rates of fault slip and rock uplift, can be attributed, at least in part, to deep valley incision, which we suggest results from glacial-topographic feedbacks bolstering the efficacy of glacial erosion.
2. Field Area
 The Teton Range trends ∼010°, is approximately 70 km long by 20 km wide, and lies at the eastern terminus of the Snake River Plain (SRP), such that it forms the first major orographic barrier to the transport of moist Pacific westerlies up the low-topography corridor of the SRP [Love et al., 2003; Meyer et al., 2004]. On the east side, the Tetons rise abruptly from the western edge of Jackson Hole, reaching elevations >2 km above the neighboring valley floor (Figures 1 and 2). Well-constrained slip rates along the east dipping normal Teton Fault, bounding the eastern side of the range, are among the fastest recorded in the conterminous United States [Machette et al., 2001; O'Connell et al., 2003]. Back tilting results in a topographic asymmetry, with the drainage divide offset west of the range crest (the zone of the highest peaks) by as much as 8 km (Figure 3), resulting in adjacent basins with vastly different topographies on opposite sides of the range.
 Glaciers have had a fundamental role in carving the present Teton landscape. LGM (Pinedale) valley glaciers were up to 15 km long and ∼500 m thick, as shown by moraines and trimlines [Pierce and Good, 1992; Love et al., 2003]. Cirques and tributaries in many larger basins preferentially form on southern valley sides, likely a result of the northerly aspect, with topographic shading enhancing ice buildup and frost cracking, and facilitating lateral erosion at the cirque headwall [e.g., Evans, 1977; Naylor and Gabet, 2007]. Although multiple glacial episodes likely affected the Tetons over the course of the Quaternary Period, evidence for only the two most recent major glaciations is preserved. During both the Bull Lake (∼160–130 ka) and Pinedale (∼30–12 ka) glaciations, lobes from the Yellowstone ice cap extended south through Jackson Hole, and valley glaciers from the Tetons formed tributaries to this larger ice stream. Ice extents were greatest during the Bull Lake glaciation, when the Jackson Hole ice lobe extended the entire length of the range. During the Pinedale glaciation the ice lobe reached the northern shores of present-day Jenny Lake, and valley glaciers further south spilled out into Jackson Hole, leaving extensive moraines [Pierce and Good, 1992; Love et al., 2003].
 Different climate modeling studies suggest alternative controls on the regional LGM ELA. Hostetler and Clark  combined a general circulation model (GCM) with a regional circulation model (RCM) and suggested that cold, dry katabatic anticyclonic easterlies from the Laurentide ice sheet dominated the LGM climate of the Tetons and surrounding areas, leading to cold summer temperatures, limited melting, and the depressed regional LGM ELA. In contrast, the GCM of Meyer et al.  indicates that moist Pacific westerly winds dominated, with anticyclonic easterlies having only a slight retarding influence. This interpretation is supported by thickness variations and grain size distributions in SRP loess [Bettis et al., 2003], and the spatial correspondence between LGM ELAs and modern precipitation patterns in the ranges surrounding the eastern SRP, where higher LGM ELAs correspond with presently arid regions [Meyer et al., 2004; Foster et al., 2008]. Prevailing modern and LGM wind systems appear, therefore, to be comparable, suggesting that precipitation patterns are likely to have been similar. The reinforcement of moist westerlies and increased precipitation following the recession of the Laurentide ice sheet may explain why extensive mountain glaciers were maintained following the LGM [Thackray et al., 2004; Licciardi and Pierce, 2008].
 The lithology of the Teton Range consists of a Precambrian crystalline basement of granites, pegmatites and gneisses, overlain unconformably by Paleozoic-Mesozoic sedimentary units and Neogene tuffs [Love et al., 1992]. Differential Laramide uplift and exhumation resulted in arching and southward tilt (2°–3°) of the unconformity during the late Cretaceous Period [Roberts and Burbank, 1993]. Late Cenozoic Basin and Range extension and normal faulting tilted and uplifted the Precambrian-Cambrian unconformity to its current westward dipping orientation and, in concert with erosive processes, is responsible for the spectacular modern topography [Roberts and Burbank, 1993; Byrd et al., 1994]. As a result, gently west dipping layered sediments cap the northern and southern ends and western flank of the Tetons, while competent crystalline granites and gneisses outcrop in the center of the range and comprise the major peaks of the shear eastern range front [Love et al., 1992].
 Both the total offset and timing of initiation of the Teton Fault are disputed. Some workers report a Miocene (5–13 Ma) initiation, with 6–9 km of offset [e.g., Smith et al., 1993]. Others suggest total displacement to be as little as 2.5–3.5 km [Byrd et al., 1994]. Gravity models, the emplacement and ∼10° westward tilting of the ∼2 Ma Huckleberry Ridge Tuff, and the absence of basement-sourced Precambrian clasts in Jackson Hole sediments all suggest that only minor displacement accumulated on the Teton Fault prior to ∼5 Ma and that the majority of offset has accrued since ∼2 Ma, contemporaneous with the arrival of the Yellowstone hot spot [Pierce and Morgan, 1992; Byrd et al., 1994; Machette et al., 2001; Love et al., 2003, p.85]. Assuming that approximately half of the displacement on the Teton Fault is taken up by footwall uplift [Byrd et al., 1994], consistent with observations from neighboring major range-bounding normal faults [e.g., Stein et al., 1988; Anders et al., 1993], the maximum long-term rock uplift rate is ∼0.6–0.9 mm yr−1 at the eastern range front, given 2.5–3.5 km of displacement since 2 Ma [Byrd et al., 1994]. Late Quaternary slip rate variations are relatively well defined along strike, with direct measurements of offset marker horizons yielding maximum slip rates of ∼1–2 mm yr−1 (Figure 2) [Machette et al., 2001; O'Connell et al., 2003]. Independent estimates of uplift from submerged Jackson Lake shorelines [Pierce and Good, 1992], tilting of the Huckleberry Ridge Tuff [e.g., Byrd et al., 1994], the close correspondence between slip rates and range crest topography, and expected normal fault displacement patterns [e.g., Cowie and Roberts, 2001] all suggest that these slip rates (Figure 2a) provide a good indication of the distribution of long-term fault displacement. Anomalously high postglacial slip rates (∼2.5–3 mm yr−1) [O'Connell et al., 2003] have been attributed to crustal unloading and rebound following the recession of the Yellowstone ice cap and Teton Range valley glaciers [Hampel et al., 2007].
3.1. LGM Glacier Reconstructions
 To understand the former distribution of glaciers, LGM ice extents are reconstructed following the methods of Brocklehurst et al.  and Foster et al. . In each valley, LGM trimlines are mapped using a combination of topographic maps, field observations, and aerial photography draped onto DEMs and viewed in virtual globes (NASA Worldwind™ and GoogleEarth™). The trimline is found using four main topographic features: (1) at the transition from smooth, glacially polished bedrock below the trimline to rougher, aerially exposed bedrock above; (2) at the break in slope that separates the lower valley wall, oversteepened by glacial erosion, and the shallower upper valley wall, dominated by periglacial processes, landslides, and debris flows; (3) at the top of talus cones, which, based on observations in the Teton Range and other formerly glaciated ranges, we infer to mark the break in slope separating subaerially exposed hillslopes from the oversteepened, glacially eroded valley [e.g., Gillespie, 1982; Brocklehurst et al., 2008; Foster et al., 2008]; and (4) where present, at the outermost ridges of lateral moraines in the ablation zone. For simplicity, where valley glaciers formed tributaries to the Jackson Hole ice lobe, we limit our LGM trimline reconstructions to the portion of the glacier confined within the range.
3.2. Topographic Analyses
 We use MATLAB and ArcGIS to investigate a ∼10 m resolution USGS National Elevation Data set digital elevation model (DEM) of the Tetons, clipped to the range front so that only topography from the uplifted footwall of the Teton Fault is considered. We investigate range-scale topographic trends by plotting swath elevation profiles parallel and perpendicular to strike (Figures 2 and 3). The spectrum of physiographies characterizing east and west draining basins is explored by investigating a number of facets of catchment geomorphology. Watersheds of individual drainage basins are delineated using an upstream flow-routing method in ArcGIS, and trunk stream longitudinal profiles are found for all catchments with a drainage area >1 km2 by tracking the elevation of the channel thalweg upstream from the outlet [after Brocklehurst and Whipple, 2002; Wobus et al., 2006].
 We investigate the relief of individual basins to assess how spatial distributions of “missing” rock mass [e.g., Small and Anderson, 1998; Montgomery and Greenberg, 2000] relate to patterns of rock uplift and glaciation, following the methods of Brocklehurst and Whipple . First, an interpolated cubic spline surface is fitted over basin-bounding ridgelines and any intrabasin highs. The DEM is then subtracted from this surface, returning the subridgeline relief, an estimate of the volume within the catchment below the ridgelines. The subridgeline relief is then divided by the catchment area to yield the mean subridgeline geophysical relief, a length measure of the mean difference in elevation between the present topography and the interpolated ridgeline-enveloping surface. In the absence of datable geomorphic markers, the geophysical relief does not represent an explicit measure of erosion, as the initial surface topography, the volume of material removed from above summits and ridgelines, and the time scale over which valleys are incised, are unknown. We instead interpret the geophysical relief to represent a measure of relief production by fluvial, glacial and hillslope processes comparable between different catchments.
 The extent to which the underlying bedrock influences landscape form is addressed by exploring two different metrics of topography: hillslope gradients and local relief. A slope map is created from the base DEM and clipped around LGM glacier outlines to focus analyses to subaerially exposed hillslopes, and to mitigate any bias from the tendency for crystalline basement rocks to form subglacially eroded valley floors, and overlying sedimentary units to cap summits, ridges and plateaux. The slope data are then clipped to the outlines of four geologic units, based on the geological map of Love et al. . The geologic units we investigate are (1) Mesozoic and Cenozoic sedimentary cover; (2) layered and foliated gneiss units, including the Webb Canyon Gneiss; (3) a homogenous augen gneiss unit; and (4) the intrusive Mount Owen Quartz Monzonite that forms the bulk of Grand Teton. Slope histograms are plotted by finding the number of cells that fall within 2° slope bins. Lithological controls on hillslope relief are investigated by clipping the base DEM to the outline of each geologic unit and using a circular moving window to find the elevation range of all cells within a 100 m radius. Only cells completely surrounded to 100 m by the same rock type are considered in order to negate any edge effects. Relief data are then placed into 20 m bins to create histograms of hillslope relief.
3.3. Flexural Isostasy
 First-order estimates of the extent to which isostasy contributes to raising summit elevations in the Teton Range are obtained by modeling the flexural response to erosion and deposition [after Anderson, 1994; Montgomery and Greenberg, 2000]. An upper limit on differential erosion in the Teton Range is estimated by stitching together subridgeline relief grids of individual drainage basins. The subridgeline relief is an indicator of maximum differential erosion between peaks, ridges, and valley floors, therefore the flexural model provides a maximum estimate of the “extra” elevation of ridges and summits that is a flexural isostatic consequence of valley incision. Although we focus our interpretations of the model on isostatic uplift of the Teton Range, we also incorporate estimates of sediment loading in the adjacent Teton Basin and Jackson Hole to avoid potential edge effects that may come from their omission. Estimates of sediment thickness in Jackson Hole and Teton Basin are taken from the regional gravity survey of Mankinen et al. , who estimated the depth to the pre-Cenozoic basement using a gravity inversion method [after Jachens and Moring, 1990]. A mass redistribution grid is created by combining these two data sets (Figure 4), to which we apply a relatively simple three-dimensional model of flexure, assuming a thin elastic plate with a constant thickness, Te. The total deflection at a given point is calculated by finding the flexural response to individual point loads, q, applied at all other cells in the workspace. Point loads are negative in the Teton Range, where erosion removes rock mass, and positive in basins, where sediment is deposited:
where ρsed, ρc, zsed, and zrelief are the density of deposited sediments and eroded crustal rocks, the depth of sediment and the subridgeline relief, respectively; dx and dy are the horizontal grid spacing (1 km); and g the acceleration due to gravity. The vertical deflection, w, in response to the load q applied at a distance, r, away from the cell is found by
where ρm is the density of the ductile mantle, Kei is the Kelvin function, a damped sinusoid function that controls the wavelength of the flexural response, and α the flexural parameter. The total deflection at each cell is thus found by applying equation (2) for each of the other cells in the model space and summing the results. We treat the Teton Fault as a discontinuity in the thin elastic plate that mechanically separates Jackson Hole from the Teton Range and Teton Basin, therefore we do not consider loads applied to cells west of the fault in flexural calculations for cells east of the fault, and vice versa. The flexural parameter, α, is found by
where D is the flexural rigidity, which controls the distance away from the point load that an appreciable deflection will be experienced. The flexural rigidity is greatly controlled by Te, such that
where E is Young's modulus (typically 10–100 GPa) and v is Poisson's ratio (typically 0.25).
 A variety of methods have been applied to estimate Te (and D) in the western United States, yielding a range of values (Table 1). In general, Te is lower in regions undergoing crustal extension and experiencing high heat flow and recent thermal events [e.g., Bechtel et al., 1990; Lowry and Smith, 1995]. The position of the Teton Range at the junction of several physiographic and tectonic provinces (the Yellowstone volcanic field; the eastern SRP; the northern Basin and Range; the middle Rocky Mountains) complicates local estimates of Te. Values for Te of ∼5 km below Teton Basin and ∼40 km beneath Jackson Hole were presented by Lowry and Smith , based on coherence analyses of topography and Bouger gravity anomalies. Stein et al.  compared displacement profiles from recent seismic events to models of coseismic and interseismic deformation across three active normal faults, including the nearby Lost River Fault, and found that modeled displacement patterns that assumed Te ∼ 2–4 km best fit observed displacements. We address the uncertainty relating to the values of Te and E in the Tetons by modeling flexure for a range of plate thicknesses (Te = 4–16 km) and Young's moduli (E = 25–100 GPa). Here, we present results from the end-member case of a relatively thin, flexible crust (Te = 4 km; E = 25 GPa), which yields a maximum estimate for the magnitude of the flexural response to erosion and deposition within the Teton Range.
Table 1. Selected Estimates of Te in the Locality of the Teton Range
Coherence studies of topography and Bouger gravity anomalies (2,3); comparisons with heat flow, geology, lithosphere age, seismic properties and fault rupture depths (4).
Middle Rocky Mountains
Coherence studies of topography and Bouger gravity anomalies (2,3); comparisons with heat flow, geology, lithosphere age, seismic properties and fault rupture depths (4).
Nearby Local Estimates
Lake Bonneville, UT
23 ± 2 km
Paleolake shorelines track crustal rebound (5).
Wind River Range, WY
Flexural and gravity modeling (6).
Lost River Range, ID
Comparison of modeled and measured fault displacement profiles (7,8).
 An additional uncertainty arises from (1) the time scale over which relief is generated in the Teton Range, where the majority of uplift has occurred since ∼2 Ma [Byrd et al., 1994], and (2) the time over which sediment, which comprises Mankinen et al.'s  estimates of the depth to the pre-Cenozoic basement, has accumulated. We address this by calculating flexure for a range of different estimates of sediment thickness (10, 30, 50, 100% of the depth to the pre-Cenozoic basement) that may have accrued contemporaneously with Teton Range uplift. However, we do not consider this uncertainty to adversely affect our interpretations of the flexural model because (1) the Teton Fault acts as a discontinuity in the thin elastic plate, therefore loading of Jackson Hole does not affect isostatic uplift in the Teton Range in our model; (2) where Te is small, sediment loading of Teton Basin does not greatly alter maximum flexural uplift in the Teton Range; and (3) we are primarily interested in the role isostasy plays in contributing to summit elevations in the Teton Range, rather than subsidence of the surrounding basins.
3.4. Climate Data
 Patterns of modern precipitation are obtained from an 800 m resolution mean annual precipitation map, generated by the PRISM (Parameter elevation Regressions on Independent Slopes Model) climate model (http://www.ocs.orst.edu/prism), based on a 30 year (1970–2000) average precipitation data set. PRISM takes point climate measures from a network of weather stations, and, in conjunction with a landscape DEM, creates continuous grids of precipitation and temperature using a linear climate elevation regression function that reflects the strong tie between topography and climate. The PRISM methodology, algorithms, assumptions, and weaknesses are presented in detail by Daly et al. [2002, 2008]. In brief, the model uses a moving window algorithm to create a unique linear regression function for each cell within the grid, with precipitation values linearly tied to the grid cell elevation. The regression function examines measured precipitation data from the nearest 40 weather stations, and assigns a weighting to the reading from each station based on the distance between, and physiographic similarity of, the weather station and grid cell in question. Sixty-nine weather stations are situated within a 1° radius of the summit of Grand Teton (Figure 1b) [Daly et al., 2008]. Reviews of PRISM output maps by state and regional hydrologists and climatologists, cross-validation routines whereby predicted values of precipitation are compared to measured values, and a series of statistical uncertainty tests have been undertaken to assess the validity of model outputs [Daly et al., 2002, 2008]. PRISM does not explicitly consider factors such as wind direction or airflow dynamics. However, because PRISM uses a linear regression function, precipitation values can be extrapolated to cells that lie above the highest weather station considered in the regression equation. This is undesirable, though necessary in mountainous regions where weather stations are situated only in relatively accessible locations.
 Neither a high-resolution validation of PRISM outputs within mountainous topography, nor the development of a more detailed model of orogenic-orographic precipitation fall within the remit of this study. We simply use precipitation values generated by PRISM to illustrate how the topographic asymmetry in the Tetons can result in eastern basins receiving appreciably greater amounts precipitation, and to inform the conceptual model of range evolution. We compare precipitation values between individual drainage basins by resampling the PRISM 30 year average precipitation grid to a smaller 50 m cell size, clipping to the outlines of individual drainage basins, and averaging the value of all cells within each catchment. Two assumptions are implicit in our treatment of the modeled modern precipitation, which we suggest also to be broadly indicative of patterns of glacier accumulation at the LGM: (1) that it is reasonable for rates of precipitation to increase with elevation and (2) that because Pacific westerly winds dominated both at the LGM and the present-day, modern patterns of precipitation are broadly comparable to those at the LGM [Love et al., 2003, p. 95; Meyer et al., 2004]. The total catchment precipitation, rather than just that which falls within former glacier trimlines, is investigated as we consider the majority of snow that fell on supraglacial hillslopes to still contribute to glacier accumulation following transport downslope by avalanching.
 We investigate how topographic shading may influence the distribution of glaciers by creating a map of summer insolation from the modern DEM using ArcGIS 9.2. Insolation values are obtained by calculating the total solar radiation received at each point in the landscape between 1 June and 1 October, dates which relate to the summer melt season. The insolation algorithm considers the latitude, elevation, slope and aspect for each cell in the DEM, along with the viewshed with respect to the surrounding topography, and tracks the position of the sun through the sky at half-hourly intervals throughout a daily cycle, repeated once a fortnight for the duration of the summer [Fu and Rich, 1999]. A constant transmittivity (fraction of radiation that passes through the atmosphere) of 0.5 and uniform diffusivity (fraction of the global normal radiation flux that is diffuse) of 0.3 are assumed, both equivalent to a generally clear sky [Fu and Rich, 1999]. Variations in solar radiation in response to equinoctial precession follow ∼20 ky cycles, and, all else being equal, modern values should approximate those at the LGM [Kutzbach and Guetter, 1986]. Reflected radiation is not incorporated into the insolation algorithm.
 Topographic shading mitigates melt in glacier ablation zones, therefore thicker ice buildup can be accommodated where high ridgelines bound glaciers. Steady state LGM ice extents should build up to broadly similar levels below bounding ridgelines and experience similar levels of ice surface insolation, irrespective of the total valley relief. Because it is shading of the unfilled valley floor that would have promoted ice buildup during glacial advances, we consider insolation values from valley floors, rather than from a reconstruction of the ice surface. We clip the insolation grid to the outline of reconstructed glaciers to find valley floor insolation within the LGM ice extents. Insolation values are considered only for topographically shaded portions of the landscape, and values from below piedmont portions of glaciers that spilled onto Jackson Hole are ignored. To compare insolation values along and across strike, mean insolation values for the landscape within individual LGM glacier outlines are calculated.
4.1. Morphometric Analyses
 Thirty-two east draining and eleven west draining basins are delineated (Table 2). Figure 1a shows drainage basin outlines and the trace of trunk streams for east draining basins. Figure 5a shows trunk stream longitudinal profiles, and Figure 5b shows drainage areas and geophysical relief for eastern basins plotted by the position of the catchment outlet along strike. Based on values for drainage area and geophysical relief, the steepness of trunk stream longitudinal profiles, and the geometric position of the basin in relation to the drainage divide and range crest, we separate eastern catchments into three morphologic groups.
Table 2. Basin Characteristics for Eastern and Western Catchments in Order of the Outlet Distance From the Southern End of the Rangea
Group 1 in bold, group 2 in plain, group 3 in italics.
Maximum distance separating the ridgeline-enveloping surface from the DEM.
Measured minus predicted relief values (from equation (6)). Positive values have more relief than predicted, negative values less.
If more than one LGM glacier occupied a single basin, the mean insolation value of all ice-occupied cells within the basin is taken. Where LGM glaciers breached present-day drainage divides and occupied more than one basin, the mean insolation value from beneath the whole glacier is the same for each basin. The LGM glacier outline was not reconstructed for North Bitch Creek due to poor preservation of the LGM trimline.
N. of Jensen Canyon
Rock Spring Canyon
N. of Rock Spring Canyon
Teton Ski Area
S. of Beaver Creek
N. of Glacier Gulch
N. of Leigh Canyon
S. of Skillet Glacier
S. of Moran Canyon
E. of Ranger Peak
South Leigh Creek
North Leigh Creek
South Badger Creek
South Bitch Creek
North Bitch Creek
 Group 1 catchments have drainage areas greater than ∼20 km2, shallow-gradient longitudinal profiles, and, with the exception of Phillips Canyon at the far southern range tip, geophysical relief in excess of 250 m. These large basins all extend beyond the range crest to the drainage divide, and form dramatic, deeply incised canyons (Figure 2b). Group 2 basins have drainage areas between 10 and 20 km2, with geophysical relief typically in excess of 150 m. Trunk stream longitudinal profiles are steeper than for group 1 basins (Figure 5a). Group 2 basins do not generally extend back to the main drainage divide, although Avalanche and Garnet Canyons, which drain the eastern flank of Grand Teton and occupy the one part of the range where the drainage divide and range crest coincide, are exceptions (Figure 1a). Group 3 catchments are the smallest basins; they directly drain the high, steep slopes of the eastern range front. Group 3 basins have drainage areas <10 km2, the steepest trunk stream long profiles, and are the least deeply incised, with geophysical relief typically less than 150 m (Figure 5; Table 2). Group 3 basins form relatively steep chutes with bare valley floors generally free of large boulder deposits (Figure 6), suggesting that debris flows, rockfalls and avalanches are important processes transporting material down the valley and out to the range front. The less steep valley floors associated with group 1 and group 2 basins mean that these colluvial processes deliver rock clasts (and snow) only as far as the valley floor, where they would be stored until transported out of the range by glacial or fluvial processes. Extensive mantling of lower valley floors by scree and talus deposits, and the prevalence of debris flow scars in the hillsides, attest to the importance of colluvial processes in denuding hillslopes in the larger drainage basins (Figure 6).
 In general, the largest basins have the greatest geophysical relief. Functional relationships between drainage area and geophysical relief for western and group 1 and group 2 eastern basins are found from the reduced major axis analysis [Mark and Church, 1977] of logged area and relief data (Figure 7a), returning the following power law relationships:
where Re and Rw are the geophysical relief and Ae and Aw the drainage area for larger eastern (groups 1 and 2) and western basins, respectively. Data from Cascade, Leigh and Moran Canyons, which bound the high peaks of Grand Teton and Mount Moran, are not included in the functional relationship calculation, as the high elevations of these peaks suggest different histories for these basins than for valleys elsewhere. Group 3 eastern basins are discounted from the functional relation due to the very high scatter and low correlation between drainage area and relief. Group 1 and group 2 eastern basins have greater geophysical relief than western basins of the equivalent area. From equations (5) and (6), the expected geophysical relief of group 1 and 2 eastern basins is approximately 2.0, 1.85, and 1.65 times greater than for western basins at drainage areas of 10, 20, and 50 km2, respectively. Western basins and small group 3 eastern basins show apparently random scatter with no discernable along-strike trend in geophysical relief residuals, calculated as the observed minus the expected relief. However, group 1 and 2 east draining basins generally have greater relief than is predicted by equation (5) in the center of the range and less relief than predicted toward the north and south (Figure 7b).
 Two possible causes for the greater geophysical relief for eastern basins in the central part of the range are the greater rate of rock uplift and the greater competency of the basement bedrock. Slope angles and the local hillslope relief of different lithologic units are compared to investigate potential variability in the competency of different bedrock units. Figure 8 shows a bedrock geology map of the central part of the Teton Range, delineating those portions of the range formed in intrusive granites, augen gneiss, layered foliated gneisses, and Paleozoic-Mesozoic sediments [after Love et al., 1992]. Grand Teton is mostly composed of intrusive granitic rocks, while a band of augen gneiss forms much of Mount Moran. Mean slope and relief values and histograms reveal little difference between the three crystalline basement units, suggesting that threshold slopes and rock mass strength are essentially the same throughout the Precambrian basement (Figure 9 and Table 3). Shallower slope angles and lower relief characterize overlying sedimentary rocks, suggesting that the rock mass strength of the overlying sedimentary cover is significantly less.
Table 3. Mean Slope Angles and Hillslope Relief for Different Lithological Unitsa
Figure 10 shows vertical deflections in response to erosional unloading and sediment loading for a thin, flexible crust (Te = 4 km, E = 25 GPa), which we consider to be a maximum estimate of isostatic summit uplift in the Teton Range. Results are presented for end-member scenarios for the proportion of total sediment thickness deposited contemporaneous with Teton Range uplift; results with 10 and 100% of the depth to pre-Cenozoic basement [Mankinen et al., 2004] are shown. Flexural data are missing where the mass redistribution grid did not record either erosion or deposition (i.e., from portions of the landscape neither in the Tetons nor part of a depositional basin). Figure 11 presents a 5 km wide swath profile B–B′, showing both the topography and basin depths and the flexural response to sediment loading and erosional unloading.
 Maximum isostatic uplift of ∼150 m is focused at the central eastern part of the Teton Range, unsurprising given that this region includes the highest topography and greatest relief. Irrespective of the volume of basin sediments considered in the flexural calculation, the magnitude of maximum uplift does not vary by more than ∼10 m, as the thin, flexible crust does not readily transmit far-field signals of loading to the eastern part of the range. However, where the total depth of Teton basin sediments is considered, the westernmost Tetons subside while the eastern range front continues to uplift (Figure 10b). Maximum isostatic uplift decreases for larger values of Te and E (a thicker, more rigid crust). Because we consider only the subridgeline relief in our flexural calculations, rather than a total volume of eroded material (including that from above the peaks and ridges), Figures 10 and 11 do not represent an estimate of the total isostatic deflection of the Tetons since uplift and erosion initiated. Instead, the patterns of flexure presented here provide a maximum estimate of the extra elevation that flexural uplift contributes to summits as a result of differential erosion and relief production.
4.3. Precipitation and Insolation
Figure 12 shows modern precipitation and summer insolation maps for the Teton Range and Jackson Hole. Comparisons of mean basin-wide precipitation and mean insolation for western and eastern (groups 1 and 2) drainage systems are shown in Figure 13. Group 3 catchments show a high degree of scatter with no discernable along-strike trend, and are thus omitted.
 The strong tie between elevation and precipitation in the PRISM algorithm results in the greatest modeled precipitation values being tied to the highest topography, in particular, a bull's-eye surrounding the summit of Grand Teton and an elongate zone tied to the drainage divide (Figure 12a). The topographic asymmetry that characterizes the Tetons is thus mirrored by an asymmetry in the modeled distribution of precipitation, with, in particular, basins draining the flanks of Grand Teton receiving ∼50% more precipitation per unit area than adjacent western basins. At the northern and southern ends of the range, where ridge crest topography is more subdued, relatively little difference exists between mean basin-wide precipitation values for east and west draining valleys (Figure 13a).
 Areas that receive the lowest values of summer insolation tend to be steep slopes with northerly aspects; conversely, south facing slopes receive the greatest insolation (Figure 12b). Valleys surrounding the high, steep peaks of the eastern range flank are characterized by generally lower insolation, notably on the northern flanks of Grand Teton and Mount Moran. Mean summer insolation is consistently 235 ± 5 W m−2 for portions of the valley floor that were covered by large LGM glaciers in west draining basins; slightly higher values occur where small, thin glaciers occupied valleys. In group 1 and 2 east draining basins, valley floors beneath LGM glaciers show greater variation. At the southern end of the range insolation is similar to that found below adjacent west draining glaciers; however, where glaciers drained the higher peaks to the north, mean insolation values are consistently lower (220 ± 5 W m−2) (Figure 13b).
5.1. Catchment Geomorphologies and Implications for Glacial Erosion
 With the notable exceptions of Grand Teton and Mount Moran, topography in the Teton Range appears tied to the ELA, suggesting that long-term glacial erosion has largely been capable of keeping pace with rock uplift [Foster et al., 2008]. However, the contrasting morphologies of east draining basins are consistent with observations that a threshold drainage area dictates glacial landscape response to rock uplift [Montgomery, 2002; Brocklehurst and Whipple, 2007]. Group 1 drainage basins (A > 20 km2) are deeply incised and have shallow valley floor longitudinal gradients, despite the fact that many drain the central part of the range where the hardest bedrock and most rapid rock uplift rates are found. That shallow downvalley gradients have been maintained suggests that group 1 basins hosted glaciers capable of long-term erosion rates that kept pace with rock uplift (∼0.5–1 mm yr−1) without significantly steepening the long profile. In comparison, glaciers and their contributing accumulation zones are smaller in group 2 and group 3 basins, which are also less deeply incised. The steepening of valley long profiles facilitates faster ice sliding and allows erosion by smaller glaciers to better offset the rock uplift. Thus, the long-profile form and extent of valley incision are an expression of the long-term balance between rock uplift, glacier mass balance and erosion.
 An analogous response to rapid uplift is well documented in fluvial landscapes, where stream profile gradients steepen, increasing stream power at a given runoff so that erosion and entrainment thresholds are more readily overcome, thus returning the system toward steady state [e.g., Snyder et al., 2003a, 2003b; Duvall et al., 2004]. Brocklehurst and Whipple  have previously shown that smaller glacial basins in the Southern Alps, New Zealand, and Nanga Parbat, Pakistan, show a similar response to rapid rock uplift, though less marked than in the fluvial system. Only basins that exceed 30 km2 and 100 km2 in area, respectively, were found to have held glaciers capable of eroding such that shallow downvalley gradients were maintained. Other studies have also noted an apparent link between glacier size and erosional efficacy. For example, Hallet et al.  found that sediment yields increased with areal ice coverage in glaciers in southern Alaska, while Montgomery  noted that in the Olympic Mountains, Washington, glaciated valleys >50 km2 have up to 500 m greater ridge crest–to–valley floor relief and 2 to 4 times the volume of rock removed as similar sized fluvial valleys, while little difference exists at drainage areas <10 km2.
 The question of why larger glaciers are apparently more capable than smaller glaciers of keeping pace with rock uplift can be addressed on two time scales. In the short term: (1) theoretical models suggest that both abrasion and quarrying rates are tied to the ice sliding velocity [e.g., Hallet, 1979, 1996], therefore larger glaciers will be more efficient at eroding due to the generally greater ice flux within the valley, as several tributaries converge to form the main trunk stream; and (2) ice thicknesses are generally greater in large, shallow longitudinal profile basins, as basal shear stresses (τb) beneath wet-based glaciers consistently approximate 0.1 MPa [Paterson, 1994]. Since τb = ρighS (where ρi is the ice density and g the acceleration due to gravity), the ice thickness, h, varies inversely with the ice surface slope, S, all other things remaining equal. This results in enhanced subglacial water pressure fluctuations and quarrying, as the recurrence interval between ice cavity growth and collapse cycles downstream of bedrock steps shortens at greater ice thicknesses [Hallet, 1996; Anderson et al., 2004]. In the longer term, large glaciers also tend to have greater longevities as they are generally fed by multiple cirques and tributary ice streams and are able to extend further down valley than smaller glaciers.
 Although the emerging paradigm suggests that erosion is generally more efficient beneath larger glaciers, it should be noted that local circumstances and multifaceted interactions between glaciers, climate and landscape can add complexity over long time scales and undermine general formulations of glacial landscape development. Berger and Spotila , for example, found no discernable correlation between glacier size and exhumation rates defined from (U-Th)/He apatite ages in the St. Elias Range, Alaska, despite variations in drainage area that exceed 2 orders of magnitude. Berger and Spotila  proposed a variety of mechanisms to explain this anomaly: that a coupling similar to that seen in the fluvial regime allows small tributary glaciers to rapidly erode valley floors in response to base level incision by trunk stream glaciers; that exhumation is focused at structural topographic fronts within the range, where more rapid advection of rock mass through the rock column may occur, rather than at the long-term ELA; that abundant precipitation could allow both small and larger glaciers to keep pace with rock uplift; and that process thresholds may limit maximum erosion rates beneath large ice streams and result in glaciers being graded, similar to rivers. For example, once overdeepenings have developed in bedrock long profiles, supercooling within the overdeepening may lead to subglacial streams being shut off, allowing till accumulation and the protection of bedrock from further erosion [e.g., Alley et al., 2003].
 In the Teton Range, geophysical relief increases with drainage area, consistent with observations from other western U.S. mountain ranges [Brocklehurst and Whipple, 2002; Brocklehurst et al., 2008]. In less tectonically active ranges, volumes of subridgeline relief are tied to basin drainage area and appear insensitive to either the degree of glacial modification or the position of the basin in regards to lithologic or tectonic structures within the range [Brocklehurst et al., 2008]. In contrast, it appears that rock uplift rates in the Teton Range influence relief, given that (1) eastern basins have one and one half to two times the relief of western basins at a given drainage area (Figure 7a); (2) relief residuals (the observed minus the expected relief, obtained from equation (5)) in group 1 and group 2 basins show the greatest positive values close to the center of the range (Figure 7b and Table 2); and (3) deeply incised large glaciers in the center of the range maintain shallow downvalley gradients and must have removed more material than glaciers at the range tips, where rock uplift is less. However, the high degree of scatter and lack of discernable pattern in the distribution of residuals for group 3 basins suggests that tectonic uplift plays only a minor role in influencing the amount of material removed from drainage basins smaller than ∼10 km2.
5.2. High Peaks
 Grand Teton and Mount Moran rise conspicuously several hundred meters above the surrounding topography and appear relatively immune to direct lowering by a glacial buzz saw. Grand Teton and Mount Moran are formed in crystalline basement lithologies in the central part of the eastern range front (Figure 2), the region of most rapid fault slip and greatest long-term rock uplift, and lie on major interfluves separating large, deeply incised group 1 glacial valleys. Cirques are relatively widely spaced on their flanks, so that undercutting of the high-level topography by lateral erosion does not seem to be a dominant process [e.g., Mitchell and Montgomery, 2006; Foster et al., 2008]. Figure 14 delineates portions of the landscape more susceptible to avalanching by highlighting gradients steeper than 30°. Although a variety of factors control when and where avalanching takes place, approximately two-thirds of slab avalanches occur at slope angles of between 30°–45°, while at steeper slopes sloughing (loose snow avalanching) mitigates thick snow accumulation [McClung and Schaerer, 2006]. The majority of the landscape encompassing Grand Teton and Mount Moran shows slope angles >30°, with relatively widely spaced cirques providing the only sites susceptible to snow and ice accumulation. Ice buildup and initial glacier formation are discouraged across much of the high peak flanks by the steep slope angles, while avalanching provides an efficient transport mechanism for delivering snow to neighboring valley floor glaciers, bolstering positive mass balance and the erosional efficacy.
 Grand Teton comprises granitic intrusive rocks, and Mount Moran is formed in a band of augen gneiss, both homogenous units [Behrendt et al., 1968; Reed and Zartman, 1973] that appear from our field observations to be less fractured and friable than the layered and foliated gneisses that comprise the rest of the basement. While it is perhaps tempting to attribute the higher elevations to a perceived greater competency of homogenous crystalline basement rock over the neighboring layered gneiss units, slope and hillslope relief histograms suggest that all basement units have essentially the same functional rock mass strength (Figure 9 and Table 3). In the absence of an efficient glacial buzz saw limit to elevations, the height of the teflon peaks is therefore likely controlled by a combination of the threshold hillslope gradient (∼40°) [e.g., Schmidt and Montgomery, 1995], and the spacing of large, group 1 glaciers that efficiently incised valley floors. Given threshold hillslopes of 40° and the spacing between adjacent valley floors, taken parallel to the strike of the range, the expected heights of Grand Teton (4197 m) and Mount Moran (3842 m) are ∼4390 and 4180 m, respectively, ∼140 and 340 m higher than the actual summits. These estimates suggest that the secondary influence of lateral erosion at relatively isolated cirques is more important in limiting the elevation of Mount Moran. This is likely due to the configuration and relatively denser spacing of cirques on the flanks of Mount Moran when compared to Grand Teton.
 In the case that hillslope wasting processes primarily govern the elevation of the high peaks, the preglacial drainage configuration places a first-order control on the site and elevation of any high peaks in the crystalline basement, as presumably only the largest fluvial valleys were precursors to group 1 catchments, whereas the smallest valleys developed into group 3 basins. We speculate that some form of dynamic topographic steady state may have developed in the Tetons, whereby, once threshold hillslopes dominated the high peaks, summit elevations have remained relatively constant. Any oversteepening of hillslopes beyond threshold slope angles during glacial periods will result in landsliding and colluvial failure following glacier retreat and ice buttress removal during interglacials [e.g., Whipple et al., 1999]. The extensive scree deposits and debris flow scars that currently mantle the tall hillslopes attest to the important role of colluvial processes in the modern landscape. Such a situation may be considered analogous to that reported by Burbank et al.  for the large massifs adjacent to the Indus River in the northwestern Himalayas, where landsliding is the dominant mechanism by which the landscape responds to rapid river incision.
5.3. Conceptual Model of Topographic-Climatic Feedbacks
 In the Tetons we suggest that three topographic characteristics have been most culpable in affecting local climate and influencing the distribution, size, and erosive capacity of glaciers. First, the cross-range tectonic gradient and offset between the drainage divide and range crest provide greater precipitation and shading to deeply incised east draining basins. Second, Grand Teton and Mount Moran penetrate the constraining envelope placed on maximum elevations by glacial erosion and control where this enhancement of precipitation and shading is most pronounced. Third, the preglacial drainage configuration (specifically the size and steepness of catchments) would have dictated the size of potential accumulation areas and the length, thickness and erosive capacity of valley glaciers, and thus the strength of coupling between topography and climate (Figure 15).
 Snow has a slower terminal falling velocity (∼1 m s−1) than rain (∼10 m s−1), and precipitation may be advected further into mountain ranges during glacial periods [Roe, 2005]. Anders et al.  used a coupled model of orographic precipitation and fluvial erosion in a hypothesized small developing mountain range to show that precipitation and erosion focus further into the range at slower falling velocities; however, they did not address how altering the precipitation phase would affect the relief structure of the range when glacial erosion is considered. Although links between orogenesis and orography are complex [e.g., Galewsky, 2009], we suggest that in the Teton Range a greater proportion of the precipitation would be transported over the drainage divide and into eastern basins during glacial, rather than interglacial, periods, as a result of prevailing westerly winds [e.g., Meyer et al., 2004] and the slower terminal falling velocity of snow. This would make the discrepancy in precipitation between eastern and western basins more pronounced, so that PRISM model outputs underestimate the across-range differences. Furthermore, because snow, once settled, is liable to remobilization by wind, snow that does fall onto the broad, open basins, and bedding-parallel plateaux that characterize the headwaters of many western basins would be liable to redistribution across the drainage divide [Anderton et al., 2004; Brocklehurst and MacGregor, 2005; Lehning et al., 2008].
 Abundant, closely spaced cirques border the drainage divide in eastern catchments and provide sites topographically suited to ice build up, given that they are sheltered settings where wind-blown snow can collect. Lateral erosion of the headwall, an important component in the evolution of cirques [e.g., Brook et al., 2006; Evans, 2006] may impose a “glacial buzz saw” limit on the elevation of the drainage divide and drive its westward migration [cf. Oskin and Burbank, 2005; Naylor and Gabet, 2007], accentuating the topographic asymmetry. We suggest that this represents an important feedback in the glacial-topographic system, increasing accumulation zones and the volume of snow influx to eastern glaciers.
 Valleys draining the eastern side of the range are proximal to the highest topography, and therefore receive the least solar radiation, allowing larger glaciers to build up by limiting melt (Figures 12b and 13b). Larger, more erosive glaciers would have incised deeper, lengthened hillslopes and further increased topographic valley floor shading. Longer hillslopes increase source areas for snow avalanching off the surrounding valley sides, and also provide an abundant source of rock debris that could further insulate valley floor glaciers from melt [Kayastha et al., 2000], as evidenced by the ubiquitous talus cones that presently mantle lower valley walls. The topography of west draining basins is more subdued, and as such provides less shade, avalanching snow or rock debris. Eastern basins are therefore topographically more suited to hosting larger, longer-lasting glaciers, which can continue to erode and maintain the tall, steep hillslopes amenable to promoting glacier health. The accumulation-enhancing and ablation-limiting mechanisms hypothesized will clearly be most significant in the largest (group 1) eastern basins adjacent to the highest topography, as these basins are incised most deeply into the range and extend west beyond the range crest to the drainage divide.
 Differential erosion between peaks and valley floors augments summit elevations by a maximum of ∼150 m as a result of flexural isostasy (Figures 10 and 11). Slow (∼0.01 mm yr−1) erosion of summit flats in neighboring Laramide Ranges has resulted in ∼20 m of summit lowering since ∼2 Ma [Small and Anderson, 1998]. If similar summit erosion rates apply to the Teton Range, this results in isostasy adding a maximum of ∼130 m to summit elevations. However, given the dominance of threshold hillslopes, and the sharpness of the high peaks, it seems more likely that summits are currently eroding at rates commensurate with valley floors and that any “extra” elevation that isostasy may contribute will be negated by summit lowering. Earlier in the development of the range, before the steady state case of threshold hillslopes reaching all the way to the tops of narrow ridges had developed, flexure could have augmented summit elevations by a small amount as relief was produced. The modest flexural response (Figure 11) can be attributed to the relatively small areal dimensions of the range, which limit the volume of rock mass that can be removed from below catchment ridgelines. Studies previously suggesting a significant isostatic response to relief generation [e.g., Small and Anderson, 1995; Montgomery and Greenberg, 2000; Champagnac et al., 2007; Medvedev et al., 2008] have focused on significantly more extensive ranges (the Sierra Nevada, California; the Olympic Mountains, Washington; the European Alps, East Greenland fjords), with range widths typically >100 km. Because deep incision characterizes basins in the Teton Range, we suggest that Molnar and England's  hypothesis that efficient valley erosion can increase summit elevations is only applicable to large orogens. Instead, our findings confirm Small and Anderson's  assertions that, when realistic flexural wavelengths are considered, isostasy does not greatly contribute to summit elevations in smaller ranges. The small size of the range can therefore be considered to be a limiting constraint on the strength of feedbacks between topography, climate and erosion by diminishing the extent of coupling between erosion and rock uplift.
 The Teton Range are a relatively small, structurally simple mountain range, yet multifaceted interactions between topography, climate and erosion complicate simple conceptual models of glacial erosion. While topography appears tied to LGM and mean Quaternary ELAs for much of the range, the presence of superelevated peaks that are apparently decoupled from the ELA precludes a literal interpretation of the “glacial buzz saw.” Rather, the elevations of the highest peaks are governed by a combination of the rock mass strength and the spacing of large, erosionally efficient glaciers. Back tilting of the normal fault block causes eastern basins to be characterized by higher topography and greater relief, a topographic asymmetry further accentuated by divide migration as a result of erosion at east draining cirque headwalls. Prevailing westerly winds and the topographic form of the landscape conspire to boost the health of valley glaciers in larger eastern basins. The strength of feedbacks between topography, climate and erosion in eastern basins is strongly modulated by the valley size. The largest valleys are the most deeply incised, the most shaded and receive the greatest amount of snow and rock debris avalanched from tall valley sides. The largest basins also extend to the drainage divide, making them more receptive to the extra input of snow blown from adjacent western basins, and better able to expand at the expense of western basins. In the eastern Teton Range, glaciers draining basins >20 km2 appear capable of maintaining shallow downvalley gradients in the face of relatively rapid rock uplift, whereas smaller glaciers have been steepened in long profile until glacial erosion was able to keep up with rock uplift.
 Insights from the Teton Range highlight the level of complexity inherent in the landscape evolution of tectonically active glaciated mountain ranges. Even greater difficulties likely exist where, for example, ranges encompass greater lithologic heterogeneity, increased climatic variability, or more convoluted structural histories. The example of the Teton Range highlights that a clear appreciation of how topography, climate, and erosion interact in glaciated mountain ranges is a vital precursor to gaining clearer insights and developing more realistic numerical models of glacial landscape evolution.
 This manuscript was greatly improved by the thoughtful, constructive reviews of Kelly MacGregor, Jim Spotila, Andrew Meigs, Emmanuel Gabet, and Associate Editor Alex Densmore. We would like to thank Wayne Gibson for providing us with PRISM weather station locations and Ivan Fabuel-Perez for his help with field work. Mark Schmitz, Karen Viskupic, and Ben Crosby provided valuable support and accommodation during field seasons. D. Foster was funded by NERC studentship NER/S/A/2004/12353A.