Sediment flux in proglacial fluvial settings is primarily controlled by discharge, which usually varies predictably over a glacial–interglacial cycle. However, glaciers can flow against the topographic gradient to cross drainage divides, reshaping fluvial drainage networks and dramatically altering discharge. In turn, these variations in discharge will be recorded by proglacial stratigraphy. Glacial-drainage capture often occurs in alpine environments where ice caps straddle range divides, and more subtly where shallow drainage divides cross valley floors. We investigate discharge variations resulting from glacial-drainage capture over the past 40 k.y. for the adjacent Ashburton, Rangitata, and Rakaia basins in the Southern Alps, New Zealand. Although glacial-drainage capture has previously been inferred in the range, our numerical glacier model provides the first quantitative demonstration that this process drives larger variations in discharge for a longer duration than those that occur due to climate change alone. During the Last Glacial Maximum, the effective drainage area of the Ashburton catchment increased to 160% of the interglacial value with drainage capture, driving an increase in discharge exceeding that resulting from glacier recession. Glacial-drainage capture is distinct from traditional (base level–driven) drainage capture and is often unrecognized in proglacial deposits, complicating interpretation of the sedimentary record of climate change.


Over glacial–interglacial timescales, rates of climate-driven erosion and sediment transport scale with drainage area (e.g., Brocklehurst and Whipple, 2007), leading to the expectation that drainage capture will affect sediment transport and be recorded by proglacial stratigraphy. However, understanding is limited of the timing of sediment transport during the glacial–interglacial cycle (e.g., Dühnforth et al., 2008; Shulmeister et al., 2010). At thousand-year timescales, greater than that taken to reach a hypothetical steady state (Cuffey and Paterson, 2010), glacial-drainage capture can change the effective drainage area of adjacent catchments, resulting in variations in discharge. At hundred-year timescales, when glaciers are at a transient state of adjusting to a change in climate, water is released by or stored within the ice mass during recession and advance (Dühnforth et al., 2008).

We investigate discharge variations resulting from variations in ice extent—by both glacial-drainage capture and advance or recession—over the past 40 k.y. for three adjacent catchments in the Southern Alps, New Zealand. The sedimentary record in New Zealand is an important archive of Southern Hemisphere climate change (Newnham et al., 1999), as the Southern Alps underwent multiple, extensive glaciations during the late Quaternary (e.g., Suggate, 1990). Gravel-bed braided rivers constructed a series of prograding, coalesced alluvial fans in three catchments—the Rakaia, Ashburton, and Rangitata—which are exposed at the coastal cliff of the Canterbury Plains (Fig. 1). Luminescence dating indicates that the majority of the visible stratigraphy was deposited during the Last Glacial Maximum (LGM, 24–18 ka) and a previous glacial maximum at 37–31 ka (Rowan et al., 2012) (Fig. 1D). Are the Canterbury sediments primarily a record of ice advance and recession within the current catchments, or has glacial-drainage capture had a substantial impact on sediment flux and deposition?


The modern Ashburton catchment has a significantly smaller drainage area (1239 km2) than both the Rangitata (1549 km2) and Rakaia (2372 km2) (Fig. 1A). However, the Ashburton proglacial fan appears to represent a greater amount of vertical deposition than the Rangitata (∼18 m cliff thickness compared to ∼6 m) (Fig. 1D). Furthermore, the stratigraphy is subdivided into depositional storeys, of which there are fewer in the Rangitata section than those representing the Ashburton and Rakaia (Leckie, 2003; Rowan et al., 2012). The headwater geomorphology indicates that transfluent ice flowed from both the Rakaia and Rangitata into the Ashburton, dramatically increasing the effective drainage area of this catchment during glacials (Mabin, 1980; Evans, 2008; Shulmeister et al., 2010). There are two key regions of glacier transfluence (Figs. 1A and 1C). First, southward-dipping lateral moraines at Prospect Hill and on either side of the Cameron Valley confirm that glacial occupancy reversed the Lake Stream tributary to flow south into the Ashburton (Pugh, 2008). Second, at Lake Clearwater, where Potts Stream currently flows into the Rangitata, part of the Pleistocene Rangitata Glacier diverted into the Ashburton, reversed the flow of the lower Potts Stream Glacier, and constructed extensive moraines and outwash fans (Mabin, 1980; Evans, 2008). The Canterbury bedrock is regionally homogeneous graywacke (Leckie, 2003) removing a lithologic control on drainage patterns. Uplift-driven drainage reversal seems highly unlikely. Specifically, reversal of Lake Stream by postglacial fluvial incision would have required ∼200 m of incision within ∼20 k.y., a mean incision rate an order of magnitude higher than the local rock uplift rate (Herman et al., 2009).


To test the hypothesis that glacial-drainage capture resulted in discharge variations of sufficient magnitude to be reflected in the stratigraphic record, we used a glacier model (Plummer and Phillips, 2003) to simulate ice volumes since 40 ka. Advance and recession of mountain glaciers are controlled primarily by changes in temperature and mesoscale meteorology (Rother and Shulmeister, 2006; Anderson and Mackintosh, 2006). High precipitation levels in the Southern Alps increase glacier sensitivity to temperature (Golledge et al., 2012), which is the primary control on glacier mass balance (Anderson and Mackintosh, 2006). Previous glacier modeling (Golledge et al., 2012), climate modeling (Drost et al., 2007), and sea-surface temperature (SST) proxy records (Barrows et al., 2007) demonstrate that LGM temperatures in eastern South Island were 6–8 °C cooler than present with little change in precipitation amount. We consider the uncertainty associated with LGM precipitation variability as within the overall uncertainty assigned to the relationship between simulated glacier extent and the primary driving force for glacial advance and recession (i.e., temperature change).

The glacier model was implemented in a series of experiments that varied mean annual temperature from −8 to +2 °C relative to modern conditions in 0.5 °C increments (hereafter referred to as temperature change, ΔT) to simulate steady-state ice volume and flow. We tested all available rainfall distributions for the Southern Alps and found the New Zealand National Institute of Water and Atmospheric Research Ltd. (NIWA) data (Tait et al., 2006) to be most appropriate (see the GSA Data Repository1). The ice volumes calculated for each ΔT were given a chronology using the timing of changes in temperature relative to modern conditions indicated by SST data (Barrows et al., 2007). Simulations to describe steady state at different ΔT are achieved via transient ice-flow calculations that constrain the time-dependent transfer of mass through each glacier, the end point of which is when the system is balanced. Modeled glaciers for different ΔT were compared to mapped terminal and lateral moraines (Mabin, 1980; Evans, 2008; Pugh, 2008; Shulmeister et al., 2010), and equilibrium-line altitude (ELA) depressions calculated via the accumulation-area ratio method (Porter, 1975).

The glacier model is two-dimensional in the quasi-horizontal plane and uses an energy balance calculation to drive ice flow. Temperature was calculated using a lapse rate of −6 °C km−1. The ice-flow model provides the primary control on discharge variations associated with glaciation. At steady state, discharge is equal to precipitation and scales with effective drainage area, modulated by the nonuniform precipitation distribution. For glacial-drainage capture to affect proglacial discharge, either subglacial meltwater flow follows the direction of ice flow across catchment boundaries or glacial melt occurs downstream of transfluence. Effective drainage areas were defined from subglacial meltwater flow directions for a given ΔT. Above the ELA, minimal melting is assumed. Below the ELA, the direction of subglacial meltwater flow (the hydraulic potential) is controlled by the surface slope of the ice rather than the topographic surface at the base of the glacier. The magnitude of the reverse bed slope must be greater than the magnitude of the glacier surface slope by a factor of 11 for meltwater to not follow the glacier flow direction (Shreve, 1972) (a description of this model is provided in the Data Repository).

Glacier advance reduces discharge as water is taken into storage as ice, whereas recession produces brief discharge peaks as extra meltwater is released. The change in discharge volume decays exponentially over the response time (Cuffey and Paterson, 2010). Transient ice-flow calculations show that steady state is reached within 400 yr of a specified perturbation, consistent with the response time estimated from analytical solutions (Jóhannesson et al., 1989) and that measured for the modern Tasman Glacier (20–200 yr) (Herman et al., 2011). The transient data provide a means of estimating the change in discharge that would occur over short (hundred-year) timescales. The discharge resulting from the transient glacier response to climate change was quantified from the change in balance within the effective drainage area for a given ΔT, and combined with steady-state discharge for the duration of the glacier response to climate change. Sediment flux was estimated using the power-law relationship derived for Ivory Glacier in the central Southern Alps (Hicks et al., 1990).


Simulated glaciers provided a good estimate of mapped ice extents (Fig. 2), although we note a 0.25 °C offset of the ΔT required for the Rakaia to reach LGM limits (at ΔT = −6.5 °C) relative to the Ashburton and Rangitata Glaciers (at ΔT = −6.25 °C) (quantification of the model-data fit is given in the Data Repository). Based on modeled ice extents, ΔT of at least −5 °C (Fig. 2B) was required for glacial-drainage capture at Lake Clearwater, where part of the Pleistocene Rangitata Glacier diverted into the Ashburton (Fig. 3). With ΔT = −5.5 °C (Fig. 2C), a lobe of the Rakaia Glacier diverted to the south at Prospect Hill through Lake Stream, reversing the flow of the lower Cameron Glacier (Fig. 1A). Under LGM conditions (ΔT > −6.0 °C), ice occupying Lake Stream blocked further ice flow southward and the Rakaia Glacier reverted to flow exclusively along the Rakaia Valley, effectively “switching off” drainage capture (Fig. 2D). Simultaneously, a much larger volume of ice diverted into the Ashburton at Lake Clearwater (Fig. 3), dramatically increasing effective drainage area at the expense of the Rangitata. Animations of model output illustrating glacial-drainage capture are presented in the Data Repository. LGM drainage capture increased the effective drainage area of the Ashburton relative to the modern drainage area from 1239 km2 to 1981 km2 (to 160% of the modern drainage area), at the expense of the Rangitata, which decreased from 1549 km2 to 975 km2 (to 63%), making the Ashburton the larger basin (Fig. 4B). The Rakaia catchment had a slight decrease in effective drainage area from 2372 km2 to 2204 km2 (to 93%), but was the largest basin through the LGM and postglacial.

Recession from the LGM was rapid: the Rakaia terminus receded from the range front to Prospect Hill with warming (from ΔT = −6.5 °C to −3 °C) (Fig. 2), indicated by SST to have occurred within four thousand years of the LGM peak (Barrows et al., 2007) and in agreement with results from terrestrial cosmogenic nuclide dating of moraines in the Rakaia Valley (Shulmeister et al., 2010) when calibrated using the local production rate (Putnam et al., 2010). Moreover, relatively minor climate variations drove dramatic drainage capture during the glacial: oscillations in temperature of 0.5 °C occurred over ∼500 yr (Barrows et al., 2007) (Fig. 4A), driving variations in effective drainage area of ∼20% between the Ashburton and Rangitata. Temperature change from −4.5 °C to −6.5 °C drove transfluent ice flow at Lake Clearwater (Fig. 3). The domains captured by the Ashburton include areas close to the main drainage divide where precipitation levels are highest, so the relationship between drainage area and discharge is not linear. Therefore, discharge from the Ashburton increased more rapidly than effective drainage area during the LGM to 218% of modern discharge, whereas discharge from the Rangitata and Rakaia decreased to 53% and 95% (Fig. 4C).

Results for calculations of the transient glacier response to temperature change indicate short-lived peaks in discharge corresponding to the reduction in glacier volume during recession. During larger advances, discharge was effectively reduced to zero for short intervals (Fig. 4E). However, precipitation falling as rain would not contribute to glacier mass gain, and during advances seasonal discharge due to melting was still considerable, as indicated by outwash deposits (Shulmeister et al., 2010). While the magnitude of peaks produced by the post-LGM loss of ice mass was greater than discharge variations resulting from drainage capture (Fig. 4E), the duration of discharge variations due to the transient response did not exceed 400 yr, and the volume of water released during recession was less. When integrated over the LGM and the post-LGM discharge peak (24–15.5 ka), the Ashburton discharge is nearly doubled relative to the interglacial value due to drainage capture (from 1.3 × 109 to 2.2 × 109 m3 yr−1), whereas the increase in discharge due to loss of ice volume was 112% (to 1.4 × 109 m3 yr−1) (Fig. 4E). Hence, over longer periods than the response time, glacial-drainage capture exerts a greater control on the magnitude of discharge variations than recession.

Sediment production rates in the Southern Alps are among the highest on Earth (Hales and Roering, 2005). Therefore, if the proglacial rivers are transport-limited, one would expect glacial-drainage capture, rather than recession, to control sediment transport in the Ashburton and Rangitata. The existing geochronology indicates that deposition of the fluvial sediments at the coast occurred from 40 to 18 ka. Over this interval, the total sediment flux to the Canterbury Plains was 26% from the Ashburton and 20% from the Rangitata; the remainder was sourced from the Rakaia (54%) (Fig. 4D). However, if we integrate sediment volumes for the LGM (24–18 ka), the contribution from the Ashburton (29%) outweighs that from the Rangitata (18%), although an opposing trend for the penultimate glacial maximum (37–31 ka) gives a greater total sediment flux from the Rangitata (27%) than the Ashburton (17%). When LGM sediments were deposited at the modern coastline (Rowan et al., 2012), aggradation rates in the Ashburton were higher than during the previous glacial maximum due to drainage capture, as observed by the greater sedimentary thickness at the coast in this catchment relative to the Rangitata (Fig. 1D). The lesser sedimentary thickness in the Rangitata represents either deposition during only the penultimate glacial maximum when drainage capture was less extensive or low rates of sediment flux coincident with two phases of deposition in the Ashburton and Rakaia.


Glacial-drainage capture resulted in a sustained increase in discharge from the Ashburton during the LGM, which outweighed the short-lived increase in discharge due to glacier recession post-LGM, and maintained rates of sediment transport when glacier expansion reduced the transport capacity in adjacent basins. We conclude that effective drainage areas varied substantially during the glacial maxima and thus drove changes in discharge and sediment flux. While variations in the extent of glaciation alter the annual discharge within each catchment, the magnitude of the variations in discharge produced by glacial-drainage capture have a greater effect on long-term discharge and deposition in affected basins, and complicate the interpretation of proglacial fluvial stratigraphy as an archive of climate change.

This work was undertaken while Rowan was in receipt of Natural Environment Research Council studentship NE/F008295/1. Additional funding was gratefully received from the Dudley Stamp Memorial Fund, the Quaternary Research Association New Research Workers Grant, and the British Sedimentological Research Group Harwood Fund. The National Institute of Water and Atmospheric Research Ltd. (NIWA) provided the gridded rainfall data, and Land Information New Zealand provided the 50-m digital elevation model. Three anonymous reviewers provided constructive comments.

*Current address: Institute of Geography and Earth Sciences, Aberystwyth University, Aberystwyth SY23 3DB, UK.
1GSA Data Repository item 2013049, a description of the application of the glacier model, is available online at www.geosociety.org/pubs/ft2013.htm, or on request from editing@geosociety.org or Documents Secretary, GSA, P.O. Box 9140, Boulder, CO 80301, USA.