Letter | Published:

Chemical weathering in active mountain belts controlled by stochastic bedrock landsliding

Nature Geoscience volume 9, pages 4245 (2016) | Download Citation

Abstract

A link between chemical weathering and physical erosion exists at the catchment scale over a wide range of erosion rates1,2. However, in mountain environments, where erosion rates are highest, weathering may be kinetically limited3,4,5 and therefore decoupled from erosion. In active mountain belts, erosion is driven by bedrock landsliding6 at rates that depend strongly on the occurrence of extreme rainfall or seismicity7. Although landslides affect only a small proportion of the landscape, bedrock landsliding can promote the collection and slow percolation of surface runoff in highly fragmented rock debris and create favourable conditions for weathering. Here we show from analysis of surface water chemistry in the Southern Alps of New Zealand that weathering in bedrock landslides controls the variability in solute load of these mountain rivers. We find that systematic patterns in surface water chemistry are strongly associated with landslide occurrence at scales from a single hillslope to an entire mountain belt, and that landslides boost weathering rates and river solute loads over decades. We conclude that landslides couple erosion and weathering in fast-eroding uplands and, thus, mountain weathering is a stochastic process that is sensitive to climatic and tectonic controls on mass wasting processes.

Main

Exposure of fresh rock surfaces by erosion promotes efficient chemical weathering8; however, in active mountain belts with ample bedrock outcrop and mobile sediment, chemical weathering is strongly limited by the kinetics of mineral dissolution9 and fluid travel times10. These are not strongly coupled with erosion rates, and little spatial variability in the solute load of mountain rivers is therefore expected. Nevertheless, mountain weathering rates vary widely2 and can reach high values2. This could be because locally high denudation rates can engender extremely fast soil production11 and/or slow water circulation, with efficient solute release; however, on fast-eroding mountains, soils are mostly thin and discontinuous, and steep bedrock hillslopes promote efficient drainage. Instead, we propose that localized weathering is associated with deep-seated landsliding, to which these slopes are prone. Exposure of fresh rock surfaces by erosion promotes efficient chemical weathering8. Bedrock landslides generate fresh surfaces in erosional scars, and potentially more importantly by intense fragmentation of mobile rock mass12. Moreover, they introduce concavity into hillslopes, which acts to catch and funnel precipitation and runoff into debris with limited hydraulic conductivity13, thus allowing percolating water to react efficiently with the unweathered rockmass.

To determine the role of landslides in localizing and facilitating chemical weathering, we have measured surface water chemistry in the western Southern Alps (WSA) of New Zealand. There, high erosion rates (9 ± 4 mm yr−1) are driven by landsliding3 due to rapid rock uplift14,15 on the Alpine fault (Fig. 1), combined with orographically forced precipitation averaging 7 m yr−1 (ref. 16). The Southern Alps rise to 3,754 m, with mean catchment elevations of approximately 900–1,000 m. Dense vegetation covers slopes below about 1,250 m (ref. 17), but glaciers extend lower in three central catchments. The anthropogenic imprint on this landscape is slight.

Figure 1: Study area in the western Southern Alps, with catchment outlines and sampling locations for rivers, landslides and hot springs.
Figure 1

Landslides were sampled in detail at three sites: A, Haremare Creek; B, Gaunt Creek; C, Jackson Bay. Catchments where a lack of imagery prevented complete mapping of landslides are shown with a star.

The distribution of landslides in the WSA is well documented from remotely sensed images. Airphoto-based landslide inventories cover the period from 1935 to 2002 (refs 6,18) in approximately ten-year increments. We mapped subsequent landslides from Landsat 8 images taken in 2014. Together, these inventories contain about 4,500 landslides, ranging in area from approximately 102 m2 up to 1.02 × 106 m2, with a mean of 1.18 × 104 m2. These landslides affected 2.1% of the total mapping area, of about 2,500 km2, and between 0.96% and 13.1% of individual catchment areas. Many landslides collect surface drainage from a larger upslope area. From a 30 m resolution ASTER digital elevation map (DEM), we estimate that on average the upslope area is four to five times greater than the landslide area (Supplementary Fig. 3) and the effective area drained through mapped landslides is 1–35% of total catchment area for landslides since 1980, and 2–40% for landslides since 1935.

If landslides are important seats of weathering in the WSA, then this should give rise to marked variability of local weathering, and a link between the degree of weathering and landsliding at the catchment scale. To test this, we collected water from seepage at the base of landslide deposits, runoff from nearby small (<1 km2) catchments without signs of recent mass wasting, and all major streams draining the mountain belt to the west, near the mountain front (Fig. 1). Spot sampling is probably representative of the river chemistry, as total dissolved solids (TDS) values in individual WSA rivers vary only by 25% across a large range of discharge2. Lack of rain before and during sampling meant that landslide scarps were dry, but persistent seepage from landslide debris suggested a low hydraulic conductivity of these deposits. Seepage from two larger landslides of 1.67 × 104 m2 (Haremare Creek, Supplementary Fig. 1) and 3.18 × 105 m2 (Gaunt Creek) and a series of smaller slides of 1.5–2.5 × 103 m2 (Jackson Bay) was sampled in detail. The first two sites are set in quartz–feldspar–biotite schists, in the immediate hanging wall of the Alpine fault19, representative of the carbonate-poor metasedimentary rocks of the Southern Alps. The Jackson Bay landslides occurred in quartz-rich greywacke/argillite of the Australian Buller terrane20 west of the fault.

In our samples, TDS concentrations are closely tied to geomorphic characteristics. In landslide seepage, TDS values range from 2,630 to 9,840 μmol l−1, much higher than the 293–800 μmol l−1 measured in runoff from catchments without landslides (Fig. 2) and also systematically higher than in the local stream water collected downstream, by a factor of at least 1.4 (Haremare Creek) and up to 13 (Gaunt Creek). The difference in lithology between the WSA and Jackson Bay is expressed as variable Mg2+/K+ in the dissolved load (Supplementary Tables 1 and 3), but the elevated TDS in landslide seepage is systematic across all sites.

Figure 2: TDS values for major rivers (River), runoff from catchments unaffected by landsliding (Runoff), landslide springs (S) and their local stream water (L) for A, Haremare Creek; B, Gaunt Creek; C, Jackson Bay.
Figure 2

Note starred river samples were taken where landslides were not mapped owing to lack of imagery.

Measured TDS values of WSA rivers plot on a mixing line between dilute runoff from soil-covered slopes and concentrated landslide seepage (Supplementary Fig. 2), suggesting that other solute inputs may be limited. Groundwater with residence times of weeks to months can contribute significantly to the solute flux from mountain belts21,22, and landslide seepage is likely to be an important source. More than 80% of visited landslides had running seeps despite a relative drought before sampling. The solutes in this seepage come from the landslides themselves; elemental ratios indicate that they are not due to either hydrothermal input (Na+/Ca2+ ratios are approximately 0.02–0.1 for landslide seeps, and approximately 10 for hydrothermal springs, see Supplementary Tables 2 and 3 and Supplementary Fig. 2) or other deep groundwater; monitored groundwater wells within the area23,24 have Na+/Ca2+ ratios between 0.29 and 0.8. Moreover, runoff in catchments without landsliding—which should also incorporate deep groundwater—have low TDS values (approximately 600 μmol l−1). Thus, our results suggest that weathering in the rapidly eroding WSA is significantly localized in recent landslides, giving rise to strong variations in solute concentrations in (near-) surface water on length scales as small as 101m.

We have used the ensemble volume of mapped landslides to assess their importance in setting whole-catchment solute outputs. The volume of landslides per catchment, calculated with a regional area–volume relationship25, was normalized by the catchment area, to yield a catchment landslide erosion rate, LSnorm, for a given interval. This has also allowed us to assess how landslide age can affect the river solute load. For 19 sampled rivers, the measured TDSmodern correlates with the logarithm of LSnorm in the catchment. The period 1980–2014 correlated best, with R (ref. 2) of 0.88 and Kendall’s-Tau (from here on K-Tau) of 0.74 (Fig. 3). For this interval, TDS values systematically range from 500 to 1,800 μmol l−1 for catchments with LSnorm from approximately 103 m3 km−2 up to 105 m3 km−2 (Fig. 3). Spot samples taken from some of these rivers in 2000 (ref. 26) show similar correlation with LSnorm for 1973–2002 (also three decades before sampling; Fig. 3), providing further support for our result.

Figure 3: Total dissolved solids concentrations measured in rivers in February 2014 plotted against the normalized landslide volumes in their catchment for the landslide interval 1980–2014.
Figure 3

Also shown is the river chemistry data from ref. 26, plotted against normalized landslide volumes for 1973–2000. The log-linear best fit (dashed line) and accompanying 1 s.d. range (dotted lines) are calculated from our data only. Horizontal error bars are 95% confidence intervals for volumes; vertical error bars represents the total uncertainty on measurement of the spot samples. Circular symbols are rivers without significant glaciers; squares include glaciers.

The TDSmodern–Lsnorm relation suggests that spatial variability in river solute flux is determined by landslide weathering, which overprints a background set by baseflow and surface runoff. A two-endmember mixing model suggests that landslide seepage on average supplies 42% of solutes (Supplementary Table 6 and Supplementary Fig. 4) from about 10% of the water flux in sampled rivers. This is in agreement with the calculated fraction of catchment area drained through landslides. At low-flow conditions such as those sampled we anticipate that slower seepage will form a greater part of the solute budget. For catchments with the greatest landslide volume, associated seepage contributes as much as 80% of the river dissolved load from 35% of the water flux, whereas in catchments with least landsliding, its contribution to the weathering flux can be negligible.

Data scatter may be caused by variable river dilution due to occasional precipitation during sampling, the range of landslide ages across the region, as discussed below, and the different relative distributions of landslide sizes, where some catchments have a few large slides and others have a similar ensemble landslide volume from many smaller slides. The log-linear relationship between catchment landslide volume and river solute load may result from a combination of effects; larger landslides have deeper scars and thicker deposits. Their volume does not scale linearly with the area of the scar25 or the drained area (see Methods) and, therefore, with the amount of precipitation collected. Percolation of water through a slide may also decrease with depth of deposit. Finally, deeper, fresher material, mined by larger landslides, is less likely to be fractured by near-surface effects27 and so the internal surface area may also not scale linearly with landslide volume.

The weathering boost caused by exposure of fresh mineral surfaces should decay as landslide deposits age. This is supported by a comparison of the chemistry of seepage from landslides with known age against the average composition of surface runoff from catchments without landslides. The high Ca2+/(Na++ Si + Ca2+) signature specific to landslides progressively decays from the youngest sampled slides, <5 yr, to values indistinguishable from surface runoff in a 60-year-old landslide (Supplementary Fig. 5). Also, the strength of fit between LSnorm and TDSmodern generally decreases for older landslide inventories (Supplementary Table 7), despite important temporal variations in landslide rates (Supplementary Fig. 6). The strongest correlation with river TDSmodern is obtained when all landslides since 1980 are combined. Thus, the very high landslide rates during 1980–1985 still disproportionately affect the river chemistry, and the effect of the degradation of these older slides on weathering is less than that of the order-of-magnitude drop in landslide rate since then. Nevertheless, the weaker correlation between river TDSmodern and landslide rates before 1980 indicates that the landslide weathering boost dissipates on decadal timescales. Both lines of evidence suggest that, in the WSA, the timescale over which landslides affect the weathering budget is about 30–60 yr.

High measured Ca2+/(Na++ Si + Ca2+) implies a high proportion of carbonate weathering in the landslides, but silicon concentrations approximately double over the observed range of catchment landslide volumes. This implies that, in steep, fast-eroding uplands where landsliding is dominant, erosion and silicate weathering-driven CO2 drawdown are coupled, albeit with limited efficiency. The impact of landsliding on silicate weathering fluxes is further constrained by the small fraction of landscape area impacted by the process at a given time.

Our data show that landslides—with associated expansive mineral surface area in debris and extended, slow hydrologic pathways—provide an optimal weathering environment, the volume of which is a first-order control on the dissolved load of rivers draining the WSA. Where bedrock-involved landsliding dominates, it provides an effective link between physical erosion and chemical weathering. The temporal and spatial stochasticity of landslide-driven erosion will be reflected in the weathering budget, and can explain the major part of spatial and temporal variability of solute transport in mountain rivers. Distributed weathering at sites not recently affected by landsliding seems to provide a steady input of solutes to mountain rivers, to which landslides add concentrated seepage for a period of decades and in proportion to their cumulative volume in a catchment. We anticipate that measurements of the weathering budget of the Southern Alps, and similar settings such as Taiwan and the Himalayas, will be susceptible to important influences from the stochastic drivers of mass wasting, such as intense rain28 or shallow earthquakes29, as the increased solute input from landslides after such events is likely to decay well within their return time.

Methods

Sampling and analytical methodology.

Samples were collected after three relatively dry weeks with approximately half the monthly average precipitation measured at a number of stations in the WSA30. Under these conditions, dilution of surface waters due to direct runoff of rainfall is limited. Despite the lack of significant recent rainfall, springs were found at the base of about 80% of visited landslides, suggesting that many landslides in the WSA are unlikely to ever be dry. Landsliding is prevalent across the mountain belt, but several of the sampled sites coincide with faults where springs could have a hydrothermal source. To evaluate this, hot spring water was sampled at three locations within the WSA (Fig. 1).

All water samples were collected using an HDPE syringe, and filtered on site using single-use 0.2 μm PES filters into several HDPE bottles thoroughly rinsed with filtered sample water for different analyses. Samples for cation analysis were acidified using ultrapure HNO3. pH values were measured in the field at the time of sample collection. Analysis of cations was carried out using a Varian 720 ICP-OES, using SLRS-5 as an external standard, and GFZ-RW1 as an internal standard and quality control (QC). QC samples were included for every ten samples to account for drift; no systematic drift was found, with random scatter less than 5%. Sample uncertainties were determined from calibration uncertainties, and were always lower than 10% (Ca2+: 4%; K+: 4%; Mg2+: 4%; Na+: 10%; Si: 8%, Sr2+: 2%). Anion analysis was performed using a Dionex ICS-1100 Ion Chromatograph, using USGS standard M206 (Spring 13) as external standard and QC. Uncertainties were always less than 10% for each of the major anions (Cl and SO42−). Bicarbonate (HCO3) was calculated by charge balance. Nitrate (NO3) was negligible in all the measured samples.

Correction for atmospheric input.

Cyclic salt dissolved in rain can impact the total dissolved load measured in river waters, and must be removed to observe true effects from weathering. This can be done using spot samples or seawater ratios as an approximate estimate26,31, but we preferred using volume-weighted average rainfall chemistry from the MaiMai catchment (also on the western side of the Southern Alps)32 owing to the similar setting and long-term average—single samples might not accurately represent the reach over which incoming storms had travelled (and therefore the level of concentration versus seawater).

Using the ratios of Cl to major cations (Ca2+, K+, Mg2+ and Na+) and assuming all Cl results from cyclic contribution, we removed cyclic cation contributions using the following standard equation: (Supplementary Equation 1), where [X] is any of the measured concentrations of cation. We appreciate that the assumption of all chloride resulting from seawater may be an overestimation, and we did not correct the hot spring samples as chloride is at very high concentrations. The previous study of the MaiMai catchment did not measure SO42− concentrations, so we followed an established protocol2 in using the seawater ratio divided by two.

The correction using the volume-weighted average for the MaiMai catchment yielded results close to those obtained by other authors working in the same area26,31 for the larger rivers within the mountain belt, but the samples collected in Jackson Bay were taken so close to the ocean (tens of metres) that this was not appropriate; the measured chloride was so high (between 211 and 454 μmol l−1) that in some cases negative results were obtained for concentrations of other solutes when rainwater ratios were used. For these samples we used seawater ratios for the major ions to correct for cyclic input.

Extent of landslide mapping.

Although we took spot samples of the rivers all along the mountain front, from Hokitika to Jackson Bay, not all of those catchments had landslides mapped owing to incomplete coverage of either aerial photographs or cloud-free satellite images; even since the advent of regular Landsat photography of the requisite detail level to map landslides (Landsat 5 and ASTER onwards) there are no cloud-free images of the southernmost catchments (Haast, Okuru, Turnbull, Waiatoto and Arawhata rivers) until very recently (Landsat 8 onwards—2013–2014), which precludes obtaining the local landslide rates over the period of interest.

Estimation of total landslide volume.

We have used published area–volume relationships25 to estimate the volume of landslides from their mapped areas. It was assumed that landslides with area >105 m2 involved bedrock, and that smaller landslides were mixed bedrock and soil failures. Our landslide maps do not distinguish between scar and deposit, lumping the two in one area measure. According to Larsen et al.25, scars and deposits have area–volume relations with the same power-law exponent, implying constant size ratios between scar and deposit areas of 1.1 and 1.9 for mixed and bedrock landslides, respectively. Hence, we have estimated the scar area by dividing the mapped landslide area by 2.1 and 2.9 for mixed and bedrock landslides, respectively, assuming that runout was equal to the length of the affected area. This may lead to an overestimation of landslide scar volume where runout was much longer, mostly for small slides, which do not contribute significantly to the total eroded mass. Conversely, some large landslides on gentle slopes have overlapping scar and deposit areas, meaning that our correction causes a significant underestimation of the scar size and thus the landslide volume. As a systematic way to constrain runout variations is not available, we have applied a blanket correction for every slide, thus obtaining a conservative total volume. We have calculated the volume of every individual landslide, and summed to obtain a total volume of landslides for a catchment and/or mapping interval. Uncertainties in this approach include the coefficient and exponent of the landslide area–volume relations, V = αAγ, for which standard deviations have been reported as σγ = σα = 0.005 for mixed bedrock–soil landslides and σγ = 0.02 and σα = 0.03 for bedrock landslide scars25, and mapping errors for which we have assumed a standard deviation of 20% of the mapped area. These uncertainties were propagated into our volume estimates using a Gaussian distribution. The standard deviation on the total landslide volume for a catchment or interval was calculated assuming that the volume of each individual landslide was unrelated to that of any other, thus, ignoring possible co-variance. Hence, the total volume uncertainty depends heavily on the size distribution of landslides. When the total landslide volume is dominated by the many medium-sized landslides in a population, then the uncertainty on the total volume estimate is small, because it is unlikely that all important individual landslides are biased in the same way. However, when the total volume is dominated by a few very large landslides, then the uncertainties on their volumes are less likely to cancel out, which leads to a large uncertainty on the total volume estimate.

Estimation of area draining into landslides.

We calculated the upslope areas draining into each landslide using the 30 m ASTER DEM of the WSA. The area calculated is the number of DEM cells from where the downstream flow path intersects a mapped landslide, plus the area of the landslide itself, calculated using the FLOWobj d8 routing algorithm available in the Topotoolbox-2.0.1 release for MATLAB. These areas are summed for each catchment and then compared with the total area of the catchment. To avoid double-counting overlapping upslope areas for landslide catalogues of different ages, each DEM pixel is counted only once for the purposes of total upslope area. The upslope area draining into landslide deposits constitutes 8.7% of the WSA landscape, ranging from 1 to 34% among the sampled catchments for landslides that occurred after 1980. This increases to 2.8–40% for the entire 70-year landslide catalogue. Supplementary Fig. 3 shows an example of the processed DEM. We acknowledge that any estimates of fluxes based on drained area are an upper limit on the water flowing through these landslide springs.

Estimation of endmembers and fluxes.

Estimation of the proportional landslide weathering input to river solute load requires definition of endmembers. As stated in the main text, we use landslide springs and runoff, both of which have a wide range of measured values, the choice of which will influence the relative fractions of the sources in the final result. We therefore calculate mixing proportions for each river using the mean of our measured samples. The range is one standard deviation. Landslide springs: 5,820 ± 2,293 μmol l−1. Runoff: 576 ± 150 μmol l−1.

Note that we did not use samples collected in Jackson Bay for these endmember estimates, as although they exhibit the same systematic increase in TDS in the landslide springs, the different lithology of the Australian plate has a strong control on the actual values of TDS; therefore actual values of TDS are not comparable. However the samples collected from the rest of the WSA all drain the same lithology and, as such, are comparable.

We can compare the estimates of solute flux from these endmembers with the estimated water flux through landslides from the calculated upstream collection areas. The proportion of runoff that passes through the landslides tends to be higher when the endmembers and measured river concentrations are used to calculate proportions than when using upstream area of landslides to estimate the flux; however, the same pattern emerges. The period of sampling was during a drier-than-average period in the WSA, and it is not surprising that the landslide component should be over-represented in the rivers during the sampling period, as this is a slower path for solutes, as evidenced by the continued flux from these sources despite the lack of rainfall. Supplementary Table 6 contains all calculated data for solute flux, water flux, landslide areas and volumes for all catchments in question.

Although the estimates of water fluxes based on endmember contributions and those based on calculations of upstream areas broadly agree, there are important differences; in particular, the upstream-area calculations for several of the catchments where landslide volumes are low exceeds the required flux based on endmember calculations, whereas for some of the rivers with higher TDS values, the flux based on upstream-area underestimates the endmember requirement. We probably overestimate the water entering the smaller landslides, which proportionally make up a greater part of the volumes in the catchments with lower TDS, whereas we probably underestimate the flux from large landslides where more water may be stored, and at low-flow sampling conditions would give rise to higher concentrations. Very large landslides dominate the volumetric contribution from landslides in the Cook, Poerua and Waitaha rivers, and thus upstream area to volume ratios are smaller. We also note that the McDonald Creek has a higher flux from landslides than expected from concentration, but this river drains mostly glacial till; the landslides are systematically smaller within the catchment, but generate a disproportionally large upstream area. This leads to an overestimate of flux. Despite these caveats, the estimates of fluxes still broadly agree. This is demonstrated in Supplementary Fig. 4, where upstream area is calculated for each catchment and compared with the TDS. The relationship is similar to the volume/TDS relationship, with the notable exception of the catchments mentioned above. No error bars are plotted for the upstream-area calculation as we have not quantified the error on these measurements.

Landslide ages and chemical decay time.

At 13 sampling points split between eight locations (Supplementary Table 5), landslide seepage has been sampled where the age of the landslide is well constrained from satellite images or aerial photographs. With these samples, we are able to show how the increased solute output from landslides decays with time. We have compared the chemistry of the landslide seepage with a mountain-belt-wide average of surface runoff composition—an average of water from small streams draining sub-catchments without recent landslide activity. This is the same surface runoff endmember signature we defined in the mixing calculations above. Simple TDS values might not account for local differences between landslides, for example, how much rainfall might have fallen on each one recently. Therefore, we used specific chemical ratios to isolate the chemical effect of landslide ageing. Because trace carbonate in landslides is weathered out first, resulting in an excess in Ca2+ versus other weathering products, it is informative to look at the ratio of calcium to other major weathering products (as calcium is also the major cation in the solute budget, this represents the majority of the excess weathering from the landslides)—that is, Ca2+/(Ca2++ Na++ Si)—in the landslide seepage as compared to the ambient surface runoff.

Landslide size is not considered in this preliminary analysis of chemical landslide ageing, but the trend in Supplementary Fig. 5 independently supports our observations on the time-dependent strength of the relationship between TDS and LSnorm, from which we infer that landslides boost weathering for 30–60 yr after failure. The strength of this boost decreases with time, meaning that the impact of older landslides on river chemistry is likely to be swamped by that of younger landslides. Full data for these landslides is found in Supplementary Table 5.

This combines well with the comparison of changes in correlation between TDS and volumes; as discussed in the main text, the correlation peaks for the past 35 years of landsliding, primarily owing to the high rates between 1980 and 1985; the fall in correlation later on indicates that the older landslides are no longer having a strong effect. The values for R (ref. 2) and K-Tau are shown in Supplementary Table 7, and the changing rate of landsliding over the past 70 years is shown in Supplementary Fig. 6. Data collected by Jacobson et al. 26 in 2000 also show a peak in correlation 30 years before its sampling, although the peak correlation is when all of the landslides are considered as a whole. The landslides mapped between 1940 and 1948 show a similar spatial distribution to those between 1980 and 1985, so this increase in correlation for the full landslide catalogue is likely to reflect this rather than any long-lasting impact of the landslides between 1940 and 1948.

Data.

The data reported in this submission are fully available in the online supplementary material.

References

  1. 1.

    , & Erosional and climatic effects on long-term chemical weathering rates in granitic landscapes spanning diverse climate regimes. Earth Planet. Sci. Lett. 224, 547–562 (2004).

  2. 2.

    Chemical weathering in high-sediment-yielding watersheds. New Zeal. J. Geophys. Res. 110, F01008 (2005).

  3. 3.

    & Effects of physical erosion on chemical denudation rates: A numerical modeling study of soil-mantled hillslopes. Earth Planet Sci. Lett. 272, 591–599 (2008).

  4. 4.

    , , , & Competition between erosion and reaction kinetics in controlling silicate-weathering rates. Earth Planet. Sci. Lett. 293, 191–199 (2010).

  5. 5.

    , , , & Chemical weathering response to tectonic forcing: A soils perspective from the San Gabriel Mountains, California. Earth Planet. Sci. Lett. 323-324, 40–49 (2012).

  6. 6.

    , & Sediment flux from a mountain belt derived by landslide mapping. Geology 25, 231–234 (1997).

  7. 7.

    , , & Links between erosion, runoff variability and seismicity in the Taiwan orogen. Nature 426, 648–651 (2003).

  8. 8.

    & The fluvial geochemistry of the rivers of Eastern Siberia: III. Tributaries of the Lena and Anabar draining the basement terrain of the Siberian Craton and the Trans-Baikal Highlands. Geochim. Cosmochim. Acta 63, 967–987 (1999).

  9. 9.

    , & Tectonic and climatic controls on silicate weathering. Earth Planet. Sci. Lett. 235, 211–228 (2005).

  10. 10.

    & Hydrologic regulation of chemical weathering and the geologic carbon cycle. Science. 343, 1502–1504 (2014).

  11. 11.

    et al. Rapid soil production and weathering in the Southern Alps, New Zealand. Science 343, 637–640 (2014).

  12. 12.

    & The role of rock fragmentation in the motion of large landslides. Eng. Geol. 109, 67–79 (2009).

  13. 13.

    , , , & Using borehole prospecting technologies to determine the correlation between fracture properties and hydraulic conductivity : A case study in Taiwan. J. Environ. Eng. Geophys. 17, 27–37 (2012).

  14. 14.

    & Fission track analysis of the late Cenozoic vertical kinematics of continental Pacific crust, South Island, New Zealand. J. Geophys. Res. 98, 16119–16148 (1993).

  15. 15.

    & Uplifted marine terraces along the Alpine Fault, New Zealand. Science 234, 1225–1228 (1986).

  16. 16.

    & Extreme rainfalls in the Southern Alps of New Zealand. J. Hydrol. N.Z. 38, 309–330 (1999).

  17. 17.

    & Tree seedling growth and survival over 6 years across different microsites in a temperate rain forest. Can. J. For. Res. 918, 910–918 (2006).

  18. 18.

    , , , & Landslide impact on organic carbon cycling in a temperate montane forest. Earth Surf. Process. Landf. 36, 1670–1679 (2011).

  19. 19.

    & Metamorphism and uplift of Alpine schist in the Franz Josef-Fox Glacier area of the Southern Alps, New Zealand. J. Metamorph. Geol. 10, 171–180 (1992).

  20. 20.

    Rb-Sr age and strontium isotope characteristics of the Greenland Group, Buller Terrane, New Zealand, and correlations at the East Gondwanaland margin. N.Z. J. Geol. Geophys. 47, 189–200 (2004).

  21. 21.

    et al. Contribution of deep groundwater to the weathering budget in a rapidly eroding mountain belt, Taiwan. Earth Planet. Sci. Lett. 303, 48–58 (2011).

  22. 22.

    et al. Impact of transient groundwater storage on the discharge of Himalayan rivers. Nature Geosci. 5, 127–132 (2012).

  23. 23.

    GNS Science Geothermal and Groundwater Database (GNS Science, accessed 3 August 2015);

  24. 24.

    et al. Changes in hot spring temperature and hydrogeology of the Alpine Fault hanging wall, New Zealand, induced by distal South Island earthquakes. Geofluids 15, 216–239 (2015).

  25. 25.

    , & Landslide erosion controlled by hillslope material. Nature Geosci. 3, 247–251 (2010).

  26. 26.

    , , , & Climatic and tectonic controls on chemical weathering in the New Zealand Southern Alps. Geochem. Cosmochim. Acta 67, 29–46 (2003).

  27. 27.

    , , & In situ stress control on microcrack generation and macroscopic extensional fracture in exhuming bedrock. J. Geophys. Res. 119, 594–615 (2014).

  28. 28.

    Establishing the frequency and magnitude of landslide-triggering rainstorm events in New Zealand. Environ. Geol. 35, 160–174 (1998).

  29. 29.

    et al. Prolonged seismically induced erosion and the mass balance of a large earthquake. Earth Planet. Sci. Lett. 304, 347–355 (2011).

  30. 30.

    New Zealand National Climate Database (NIWA, accessed 15 January 2015);

  31. 31.

    , , & Tracking the relationship between mountain uplift, silicate weathering, and long-term CO2 consumption with Ca isotopes: Southern Alps, New Zealand. Chem. Geol. 341, 110–127 (2013).

  32. 32.

    , , & A comparison of the chemical composition of fog and rainwater collected in the Fichtelgebirge, Federal Republic of Germany, and from the South Island of New Zealand. Theor. Appl. Climatol. 38, 210–221 (1987).

Download references

Acknowledgements

We thank the New Zealand Department of Conservation for permission to sample the field sites (Authority number 38154-RES), A. Golly for assistance in the field, and R. Hilton, A. J. West and C. France-Lanord for discussion. Sample analysis was carried out in the GFZ HELGES lab with assistance from J. Schuessler and C. Zorn. A. Heimsath provided advice which greatly improved the manuscript.

Author information

Affiliations

  1. GFZ Deutsches GeoForschungZentrum, 14473 Potsdam, Germany

    • Robert Emberson
    • , Niels Hovius
    •  & Odin Marc
  2. Institute of Earth and Environmental Science, University of Potsdam, 14476 Potsdam, Germany

    • Robert Emberson
    • , Niels Hovius
    •  & Odin Marc
  3. CRPG-CNRS-UL, 54500 Nancy, France

    • Albert Galy

Authors

  1. Search for Robert Emberson in:

  2. Search for Niels Hovius in:

  3. Search for Albert Galy in:

  4. Search for Odin Marc in:

Contributions

R.E., N.H. and A.G. conceived the study and collected the samples. R.E. carried out lab analysis and data processing of chemical samples. O.M. completed the landslide data and calculated volumes. R.E. wrote the paper with significant input from all other authors.

Competing interests

The authors declare no competing financial interests.

Corresponding author

Correspondence to Robert Emberson.

Supplementary information

PDF files

  1. 1.

    Supplementary Information

    Supplementary Information

About this article

Publication history

Received

Accepted

Published

DOI

https://doi.org/10.1038/ngeo2600

Further reading

Newsletter Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing