Thermochronometry Reveals Headward Propagation of Erosion in an Alpine Landscape

See allHide authors and affiliations

Science  01 Apr 2011:
Vol. 332, Issue 6025, pp. 84-88
DOI: 10.1126/science.1198401


Glacial erosion of mountain ranges produces spectacular alpine landscapes and, by linking climate with tectonics, influences a broad array of geophysical phenomena. Although the resultant landforms are easily identified, the timing and spatial pattern of topographic adjustment to Pleistocene glaciations remain poorly known. We investigated topographic evolution in the archetypal glacial landscape of Fiordland, New Zealand, using (U-Th)/He thermochronometry. We find that erosion during the past 2 million years removed the entire pre-Pleistocene landscape and fundamentally reshaped the topography. Erosion focused on steep valley segments and propagated from trunk valleys toward the heads of drainage basins, a behavior expected if subglacial erosion rate depends on ice sliding velocity. The Fiordland landscape illustrates complex effects of climate on Earth’s surface morphology.

The characteristic landforms and large relief of many alpine landscapes indicate that the effect of glacial erosion over the past ~2.5 million years (My) has been profound (1, 2). Understanding how this erosion progressed at the landscape scale over millions of years is essential for analyzing the connections between climate change, topography, and tectonic processes. Most quantitative studies of glacial landscape evolution rely on model simulations that calculate glacial erosion from poorly validated parameterizations [e.g., (3, 4)]. Alternatively, direct observational constraints could reveal the evolution of topography and so guide model development [e.g., (57)]. Such constraints are difficult to obtain, however, because erosion itself effaces evidence of past topography.

We used the isotopic legacy of evolving crustal temperature conditions to constrain the history of relief development in a mountain landscape. We collected low-temperature thermochronometric measurements of 33 bedrock samples from along valley axes and up valley walls in high-relief drainage networks near Milford Sound in Fiordland, New Zealand (8). In this setting, patterns of topographic evolution over the past ~2 My are clearly decipherable because of a fortuitous correspondence between the temperature sensitivity of He isotopic techniques and the overall magnitude of Pleistocene exhumation. All samples were taken from a ~21-km–by–38-km region (Fig. 1). The valleys exhibit classic glacial forms, including U-shaped cross sections, concave longitudinal profiles dominated by low slopes, and deeply incised valley-head cirques with exceptionally steep walls. The examined valleys are incised into strong plutonic rocks, primarily of the Arthur River and Darran complexes (9). Mean slopes of valley sides and cirque headwalls commonly exceed ~45° over horizontal and vertical scales of ~1.5 km. The dominant wavelength and relief of the regional surface topography are ~4 km and ~2 km, respectively (Fig. 1). This region currently receives abundant precipitation (>6 m/year) and lies along a tectonically active plate boundary. Oblique convergence of the Australian and Pacific plates, the source of mountain uplift, began about 6 My ago (10). Convergence has been accommodated primarily by deformation of weaker crust east of the Fiordland block.

Fig. 1

Modern Fiordland topography and sample locations, together with cooling constraints from apatite 4He/3He thermochronometry. (A) Topography and bedrock sample locations; red denotes cirque-floor samples. White rectangles outline model domains (see Figs. 3 and 4). (B to G) Cooling paths from 4He/3He thermochronometry of select samples from the North Branch Cleddeau [(B) to (D)] and the Neale Burn drainages [(E) to (G)]. The colored sets of randomly generated cooling paths predict each observed (U-Th)/He age to within analytical uncertainty (±1σ); the gray paths do not. Yellow and red paths are progressively inconsistent, respectively, with the 4He/3He data (figs. S1 and S2), whereas green paths are most consistent (12). Cooling paths shown as solid black curves were predicted from the topographic evolution models shown in Figs. 3 and 4; the dashed black curves are predictions for steady-state topography, from models calibrated to match lowest-elevation samples (Fig. 2).

Comparing measured apatite (U-Th)/He ages (8) to elevations for all sample sites reveals a broadly crescentic pattern (Fig. 2A); the youngest ages, ~1 My, generally occur at cirque floor elevations (~500 to 700 m), whereas older ages occur both at sea level (where ages approach ~2 My) and on summits (~2.5 My). This crescent-shaped age/elevation relationship appears in multiple valley systems, some of which drain NW from the main divide (to Milford Sound) and others to the SE (inland to Lake Te Anau). With one exception, all median (U-Th)/He ages are <2.5 My. Thus, these data primarily reflect processes active in the Pleistocene. The similarity between longitudinal age/elevation relationships of individual valleys (Fig. 2, B to D) and the regional age/elevation relationship (Fig. 2A) suggests both a consistent pattern of valley development across the region and an absence of localized tectonic influences on the age patterns (8).

Fig. 2

Observed and modeled apatite (U-Th)/He ages and elevation relationships. (A) Observations for all Fiordland sample sites (Fig. 1). Both observations and modeled values are shown within (B) the North Branch Cleddeau, (C) the North Clinton, and (D to F) the Neale Burn drainages. Small black points are individual crystal ages; triangles and horizontal lines are the median age and range, respectively, observed for each sample. Squares are model ages predicted for the transient topographies illustrated in Figs. 3 and 4, calculated following (13) and (14), using each sample’s crystal dimensions and U and Th concentrations. Open circles are model ages calculated instead with steady-state (modern) topography. By adjusting free parameters, steady-state models can match a subset of samples (the lowest-elevation ones in this case) but not all. Predicted ages from three alternative model classes are shown in (E) for samples along the axis of the Neale Burn drainage (NB15, NB16, NB18, NB21, NB04, and NB01), and (F) from its southern valley-wall transect (NB15, NB14, NB13, and NB12). To illustrate patterns, all models were calibrated to match the lowest-elevation sample; relative performance of the models is not sensitive to this choice (8). Open diamonds represent a proportional decrease in relief from 2.5 My ago to the present; fig. S3A shows the corresponding longitudinal profile evolution. The open downward triangles represent a continuous transition from a linear longitudinal profile (2.5 My ago) to the modern form; fig. S3B shows the profile evolution. The open upward triangles represent an extreme transition, between 1.75 and 0.75 My ago, from a 3-km plateau to the modern form; fig. S3C shows the profile evolution. MF is a misfit statistic calculated from the error-weighted differences between predicted and observed (U-Th)/He ages (8).

We also analyzed samples using 4He/3He thermochronometry (8). Unlike (U-Th)/He ages, which are calculated from the total abundance of radiogenic 4He relative to U and Th, the 4He/3He method constrains the spatial distribution of 4He within an apatite crystal via controlled stepwise degassing of samples containing synthetic proton-induced 3He (11). Such information delimits a sample’s continuous cooling history through the temperature range ~80° to ~20°C (12). The cooling history, in turn, reveals aspects of topographic development not captured by the He ages alone.

Thermal histories derived from 4He/3He thermochronometry differ greatly across the landscape, even over small distances (Fig. 1). At the head of the North Branch Cleddeau valley, for example, rocks at the floor of an 1100-m-deep cirque cooled continuously over the past ~1 My, from temperatures of ~75° to 110°C to the present surface temperature (Fig. 1D). In contrast, rocks currently at the ridge crest resided at temperatures <25°C throughout this time interval. Superficially, these results suggest that much of the cirque relief developed over the past ~1 My. In this landscape of closely spaced valleys and high relief, however, the geothermal gradient depends strongly on position in the landscape; different thermal histories are expected at different sites, even if topography and mean exhumation rate were steady. To disentangle these factors, and to account for the effects of both surface erosion and rock uplift, we interpreted our data in the context of a three-dimensional thermokinematic finite-element model for subsurface temperature evolution (13).

This model allowed us to prescribe both the surface topography and the rock uplift as arbitrary functions of time and space. The sum of surface lowering and rock uplift then determined exhumation, the primary control on thermal histories. Uplift was assumed to be spatially uniform and steady, although we explored the effect of time-varying uplift rate in some sensitivity tests (8).

Using a variety of topographic histories, designed to represent idealized styles of morphological change, we calculated evolving temperatures over the past 4.0 My for the three regions delimited by white boxes in Fig. 1. Each calculation began at a steady state. We recorded the time series of temperatures along the particle paths leading to our sample sites on the modern surface. From such cooling histories, we calculated model (U-Th)/He ages and model 4He/3He release spectra as a function of sample characteristics (14); these results were compared to observations (8). We sought models that could simultaneously reproduce the key features of the data: the minimum of ages at cirque-floor elevations and the very steep age-elevation gradients on the sides of trunk valleys and at some valley heads.

We attempted first to predict the observed (U-Th)/He ages by assuming steady topography and exhumation. These models predict a monotonic age/elevation relationship and can never reproduce the key features of the observations (Fig. 2). We then sought to fit the data with progressively more complex models (Fig. 2, E and F, and fig. S3). Rejected model classes include decreasing relief without changes of landform shape (i.e., exhumation proportional to present elevation); transition from an initially flat landscape to the present form (i.e., exhumation inversely related to present elevation); and deepening of valleys as their longitudinal profiles change from linear to concave.

The only class of models capable of matching the observations involves headward progression of erosion, so that deep exhumation of headwater regions (including cirques and drainage divides) occurred roughly 1 My after deep exhumation of the downstream segments of trunk valleys (Figs. 2 to 4). In the best-fitting models, prominent kilometer-scale topographic steps migrated up-valley (Figs. 3 and 4). Although idealized, this morphology resembles modern analogs in Norway and Antarctica, where steep valley ramps descend to level floors (8). Best-fitting models of this class imply basin-wide exhumation rates of 0.76 mm/year, partitioned between background exhumation and rock uplift of about 0.6 mm/year and maximum local bedrock lowering rates of ~4 mm/year (Fig. 3, D to F, and table S3). These models also succeed in predicting the local cooling paths constrained by 4He/3He data (Fig. 1, B to G).

Fig. 3

Best-fit transient topographic scenario of headward erosion propagation along the Neale Burn drainage (domain outlined in Fig. 1). (A to C) The input topography at three representative model times (1.75, 1.25 and 0.75 My before the present, respectively). White points indicate sample locations, and white curves locate the longitudinal profiles shown in (D to I). (D) to (F) The local lowering rate (erosion rate minus uplift rate) along the profile, calculated for the preceding 250,000 years of each model time frame. (G) to (I) Model longitudinal profile evolution from 2.5 My ago to the present in 250,000-year intervals; red profiles correspond to time frames shown in (A) to (C). The high-elevation surfaces may represent the level of valley bottoms beneath a nearly isothermal relief.

Fig. 4

Best-fit transient topographic scenarios of the North Branch Cleddeau and the North Clinton drainages. (A and B) The input topography shown at 1.00 My before the present for the model domains outlined in Fig. 1. As in Fig. 3, these models involve headward propagation of erosion along the longitudinal profile of each valley: the North Cleddeau (C and E) and the North Clinton (D and F). Panels and curve colors are as in Fig. 3.

In all cases, headward progression implies rapid erosion of steep valley segments as compared to adjacent flat segments downstream. Such behavior occurs in river systems, because boundary shear stress and the rate of energy expenditure increase with the slope (15). In Fiordland, however, the erosion pattern almost certainly arose from glacial action. Currently, glaciers survive on high slopes in this region of abundant snowfall. For most of the past 2 My, landscapes worldwide were considerably colder and icier than in the present interglacial period (16, 17), and New Zealand glaciations generally followed world trends (18). In Fiordland, moreover, elevations upstream of the erosion fronts would have been substantially higher than at present (≥1.5 km according to our estimates; Figs. 3 and 4), placing a large portion of the drainage basins above the snow line even during interglacial periods. Furthermore, fluvial modification of glacial forms during the Holocene has been minimal; features such as the abrupt step where the ~20-km2 Bowen River basin joins Milford Sound indicate that rivers, even with large discharges and steep slopes, lack the capacity to incise the hard crystalline bedrock at high rates.

Glacial erosion by abrasion and quarrying occurs only if the ice slides along its bed. Theoretical treatments (19, 20) indicate that erosion rate ė increases with sliding velocity ub; most simply, take e˙ubr, with r an unknown positive number. This suggests one origin for headward-propagating erosion in glacial valleys: Compared to more gently sloped sections upstream and downstream, a glacier flowing down a steep incline is thinner and faster, and hence more erosive. Simple calculations (8) imply that 10-fold contrasts in ė along valley profiles (Figs. 3 and 4) most likely correspond to a parameter r value in the range ~1 to ~3. Others have proposed that feedbacks between quarrying, water pressure variability, and glacier geometry should cause up-valley propagation of subglacial bedrock steps with dimensions on the order of 10 to 100 m (21, 22). The steep reaches associated with headward trending of erosion in Fiordland valleys are an order of magnitude larger. Valleys in our study region do display many steep bedrock facets with heights on the order of 100 m, but these features are too small to leave a record of propagation in the (U-Th)/He system.

Our thermochronometric data imply that large spatial variations of erosion rate determined the development of the Fiordland landscape. Our interpretation suggests that such variations arise from a glaciologically governed correlation between topography and sliding rate, together with a sliding-rate dependence of glacial erosion. A simple alternative approach to modeling glacial erosion—to use ice discharge as a proxy for erosion rate (23)—would not predict our observations because it precludes large contrasts of erosion rate over distances that are small compared to a glacier’s length (except at special places such as tributary junctions).

Our results also suggest a landscape-scale target for model simulations. In Fiordland, the transition to Pleistocene climate initiated a massive reconfiguration of mountain range topography. Trunk valleys eroded rapidly at the outset, but their downstream portions have changed little in the past 1.5 My. By the mid-Pleistocene, erosion focused largely on drainage divides and valley heads, with decreasing area of high-elevation catchments. By 0.5 My ago, fundamental change had ceased, and average exhumation rates were the lowest in the Pleistocene. In essence, after the onset of glaciation, the mountain range was denuded from its flanks into its core and now changes only gradually. Thus, glaciation not only sculpted distinctive landforms but also systematically reshaped broad features of the range. Concerning debates about the response of mountain relief to glacial conditions (24), our study suggests that range-scale relief decreased during the Pleistocene, while local peak–to–valley floor relief increased transiently as erosion shifted headward.

Supporting Online Material

Materials and Methods

Figs. S1 to S7

Tables S1 to S3


References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. We thank Milford Helicopters, B. Lum, and the Caltech Noble Gas Lab for field and laboratory support; J. Braun and F. Herman for modeling guidance; and three anonymous referees for constructive comments. This work was supported by NSF grants EAR-0642869 (to D.L.S.) and EAR-0642830 (to K.M.C.) and by the Ann and Gordon Getty Foundation.

Stay Connected to Science

Navigate This Article