Earth and Planetary Science Letters 277 (2009) 213–222

Contents lists available at ScienceDirect

Earth and Planetary Science Letters

j ourna l homepage: www.e lsev ie r.com/ locate /eps l

Direct evidence of the feedback between climate and weathering

Sigurdur R. Gislason a,⁎, Eric H. Oelkers b, Eydis S. Eiriksdottir a, Marin I. Kardjilov c, Gudrun Gisladottir c,Bergur Sigfusson a,1, Arni Snorrason d, Sverrir Elefsen d,2, Jorunn Hardardottir d,Peter Torssander e, Niels Oskarsson a

a Institute of Earth Sciences, University of Iceland, Sturlugata 7, 101 Reykjavík, Icelandb Géochimie et Biogéochimie Experimentale — LMTG/Université Paul Sabatier/CNRS, UMR 5563, 14 rue Edouard Belin, 31400 Toulouse, Francec Department of Geography and Tourism, University of Iceland, Sturlugata 7, IS 101, Reykjavik, Icelandd National Energy Authority, Grensásvegi 9, 108 Reykjavík, Icelande Department of Geology and Geochemistry, Stockholm University, SE-10691 Stockholm, Sweden

⁎ Corresponding author. Tel.: +354 525 4497; fax: +35E-mail address: sigrg@raunvis.hi.is (S.R. Gislason).

1 Present address: Reykjavík Energy, Baejarhálsi 1, 1102 Present address: Mannvit, Grensásvegi 1, 108 Reykja

0012-821X/$ – see front matter © 2008 Elsevier B.V. Aldoi:10.1016/j.epsl.2008.10.018

a b s t r a c t

a r t i c l e i n f o

Article history:

Long-term climate moderat Received 29 April 2008Received in revised form 7 October 2008Accepted 18 October 2008Available online 5 December 2008

Editor: M.L. Delaney

Keywords:climate moderationchemical weatheringbasaltCO

2flux

global carbon cycle

ion is commonly attributed to chemical weathering; the higher the temperatureand precipitation the faster the weathering rate. Weathering releases divalent cations to the ocean viariverine transport where they promote the drawdown of CO2 from the atmosphere by the precipitation andsubsequent burial of carbonate minerals. To test this widely-held hypothesis, we performed a field studydetermining the weathering rates of 8 nearly pristine north-eastern Iceland river catchments with varyingglacial cover over 44 years. The mean annual temperature and annual precipitation of these catchmentsvaried by 3.2 to 4.5 °C and 80 to 530%, respectively during the study period. Statistically significant linearpositive correlations were found between mean annual temperature and chemical weathering in all 8catchments and between mean annual temperature and both mechanical weathering and runoff in 7 of the 8catchments. For each degree of temperature increase, the runoff, mechanical weathering flux, and chemicalweathering fluxes in these catchments are found to increase from 6 to 16%, 8 to 30%, and 4 to 14%respectively, depending on the catchment. In contrast, annual precipitation is less related to the measuredfluxes; statistically significant correlations between annual precipitation and runoff, mechanical weathering,and chemical weathering were found for 3 of the least glaciated catchments. Mechanical and chemicalweathering increased with time in all catchments over the 44 year period. These correlations werestatistically significant for only 2 of the 8 catchments due to scatter in corresponding annual runoff andaverage annual temperature versus time plots. Taken together, these results 1) demonstrate a significantfeedback between climate and Earth surface weathering, and 2) suggest that weathering rates are currentlyincreasing with time due to global warming.

© 2008 Elsevier B.V. All rights reserved.

1. Introduction

Atmospheric circulation models indicate that global climate isintimately linked to the atmospheric CO2 content; increased CO2 leadsto warmer temperatures, changing precipitation patterns, and anoverall increase in runoff (Labat et al., 2004; Gedney et al., 2006; Alleyet al., 2007). Increased CO2 content in the atmosphere, however, hasbeen limited over geologic time by weathering of silicate rocks(Walker et al., 1981; Berner et al., 1983; Berner and Kothavala, 2001;Wallmann, 2001; Berner, 2004). This process stems from the weath-ering of Calcium–Magnesium silicates, and over the last 100 millionyears mainly Ca silicates. For example, the weathering of Ca-rich

4 525 4499.

Reykjavík, Iceland.vík, Iceland.

l rights reserved.

plagioclase, a common silicate mineral, leads to carbonate mineralprecipitation according to the following reactions:

CaAl2Si2O8 þ 2CO2 þ 3H2OY ð1ÞPlagioclase

Al2Si2O5ðOHÞ4 þ Ca2þ þ 2HCO3− ð2Þ

Kaolinite

Ca2þ þ 2HCO3− → CaCO3 þ CO2 þ H2O ð3Þ

Calcite

These reactions proceed via the chemical and mechanical weath-ering of silicates on land coupled to the riverine transport ofsuspended matter and aqueous Ca2+ and HCO3

− to the oceans wherethey react to form calcite and aragonite (Aller, 1998; Gislason and

214 S.R. Gislason et al. / Earth and Planetary Science Letters 277 (2009) 213–222

Oelkers, 2003; Stefánsdóttir and Gislason, 2005; Gislason et al., 2006;Pogge von Strandmann et al., 2008). The feedback between climateand weathering has been attributed to a number of factors includingthe effect of temperature on 1) silicate dissolution rates, 2) runoff, 3)mechanical weathering, and 4) glacial melting (e.g. Peters, 1984;Meybeck, 1986; Sverdrup, 1990; Bluth and Kump, 1994; Meybeck,1994; Gibbs and Kump 1994; White and Blum, 1995; Gislason et al.,1996; Gaillardet et al., 1999a; Stefánsson and Gislason 2001; Tranter etal., 2002; Dessert et al., 2003; Millot et al., 2003; West et al., 2005;Anderson, 2007; Navarra-Sitchler and Brantley, 2007).

This study was initiated to provide direct evidence of the effectof climate change on chemical and mechanical weathering rates.This evidence was obtained by focusing on individual catchmentsrather than comparing the behaviour of different catchments atvarious temperatures and precipitation/runoffs. In this way wewere able to measure directly changes in weathering rates due to

Fig.1.Mapshowing the locationof thecatchments investigated in this study, thesamplingstationvariation in annual precipitation, ΔP from 1961–2005. The studied rivers fromwest to east are4) Jökulsá í Fljótsdal atHóll, 5) Fellsá, 6) Grímsá, 7) Lagarfljót and8) Fjardará. Theword Jökulsá in Iinblue letters. TheFellsá,Grímsá, andFjardaráare direct runoff rivers, the Jökulsá áDal, Jökulsá í Fmixture of spring, glacier and direct runoff waters. Water flow of the Lagarfljót and Grímsá rive

climate change in each catchment, limiting potential ambiguitiesassociated with relief, rock type, vegetation, and glacier cover, etc.,which could mask the effects of temperature and precipitation/runoff on weathering (Edmond et al., 1995).

Chemical andmechanical weatheringwas quantified bymeasuringthe flux of dissolved and suspended materials, respectively, to theocean in 8 rivers in north-eastern Iceland. The location of these riversand their catchments are shown in Fig. 1; further physical and climaticcharacteristics of these catchments are presented in Table 1. The 8studied rivers are situated on basaltic rocks. The ages of these basaltsgenerally increase from east to west, the youngest basalts are found inthe Jökulsá á Fjöllum river catchment, and are on average 0.3 millionyears of age. The oldest basalts are found in the Fjardará rivercatchment and are on average 11.2 million years of age (Moorbathet al., 1968; Johannesson and Saemundsson, 1998). Five of the eightriver catchments are partially glaciated. Two of the rivers, the

s, themeteorological stations and thevariation inmeanannual temperature,ΔTandpercent1) Jökulsá á Fjöllum at Grímsstadir, 2) Jökulsá á Dal at Brú, 3) Jökulsá á Dal at Hjardarhagi,celandicmeansa glacier-fed river. Thenames of glacier fed rivers andcatchments are shownljótsdal, andLagarfljót are amixture of glacieranddirect runoff, and the Jökulsá á Fjöllumisars is influenced by dams. The catchments of these 8 rivers cover a total of 11,350 km2.

Table 1The size of the river catchments, percent glacial cover, areaweighted age of the catchments rocks, runoff and flux range, and the annual mean temperature and precipitation range ofclosest weather station

RiverCatchmenta

Size Glaciercover

Rockageb

Study peroid Runoff rangec Ca2+ flux range HCO3−flux range SIM flux range Nearest

weatherstation

Mean annualtemp. range

Annual precip.d

range

(km2) (Myr) (yr) (m/yr) (mol/m2/yr) (mol/m2/yr) (kg/m2/yr) (°C) (mm/yr)

Grímsstadir 5179 28% 0.299 1971–2003 0.865 to 1.25 0.131 to 0.177 0.772 to 0.968 0.614 to 3.95 Grímsstadir −1.8 to 2.5 278 to 444Brú 2089 68% 1.32 1971–2004 1.14 to 2.47 0.090 to 0.154 0.365 to 0.602 0.957 to 5.01 Brú −1.3 to 3.21 334 to 820Hjardarhagi 3338 43% 1.71 1970–2004 0.945 to 1.76 0.094 to 0.142 0.382 to 0.554 0.595 to 3.70 Brú −1.3 to 3.21 334 to 820Hóll 560 27% 2.14 1963–2004 1.12 to 2.34 0.203 to 0.356 0.593 to 1.035 0.294 to 1.202 Hallormsstadur 1.7 to 5.67 304 to 1916Fellsá 124 0% 5.9 1977–2003 1.25 to 2.77 0.057 to 0.110 0.267 to 0.519 0.0112 to 0.0287 Egilsstadir 1.2 to 4.68 476 to 1059Lagarfljót 2777 6% 5.99 1975–2003 1.07 to 1.90 0.136 to 0.233 0.466 to 0.785 0.0239 to 0.0429 Egilsstadir 1.2 to 4.68 444 to 1059Grímsá 507 0% 6.45 1961–1997 0.910 to 2.29 0.110 to 0.236 0.350 to 0.761 0.0064 to 0.0219 Egilsstadir 1.2 to 4.5 225 to 1059Fjardará 56 0% 11.2 1962–2004 1.30 to 2.54 0.047 to 0.080 0.221 to 0.377 0.0094 to 0.0233 Seydisfjördur 2.1 to 5.3 1023 to 2495

a The 4 glacier covered catchments are denoted by their respective sampling stations, but the others by their rivers (Fig. 1).b Area weighted average rock age.c Runoff range during study period.d Maximum and minimum precipitation.

215S.R. Gislason et al. / Earth and Planetary Science Letters 277 (2009) 213–222

Lagarfljót and Grímsá, have been dammed. Critical to demonstratingthe effect of climate on weathering is performing studies over asufficiently long time period to obtain data exhibiting distinct meanannual temperature and precipitation changes. To obtain the robustevidence needed to demonstrate unambiguously the feedbackbetween climate and weathering, this study covers the years 1961–2004 corresponding to most of the recent rapid global warmingperiod (Alley et al., 2007).

The north-east Icelandic rivers were chosen for this study for anumber of reasons. First, these river catchments are sparselypopulated, minimizing the effect of human activity on weatheringrates. Second, the river catchments are comprised of basalts. Due totheir composition and rapid dissolution rates, the weatheringof continental basalts has a far faster weathering rate and CO2

consumption capacity than other major continental silicate rocks(Meybeck, 1986; Bluth and Kump, 1994; Dessert et al., 2003; Wolff-Boenisch et al., 2004, 2006). Third, Iceland is a volcanic islandrepresentative of the high-relief, volcanic and tectonically activeislands that contribute over 45% of suspended material to the oceans,and the suspended material from these islands is reactive in seawater(Milliman and Syvitski,1992; Gaillardet et al., 1999b; Stefánsdóttir andGislason, 2005; Gislason et al., 2006). Fourth, future warming andrainfall increase is expected to be greatest over land situated at highlatitudes (Alley et al., 2007). Fifth, the studied rivers consist of 5glacier-fed and 3 non-glacial rivers, allowing assessment of the role ofglacial melting on weathering rates. Finally, modelling studies havesuggested that the effect of climate on weathering would be greatestin volcanic rocks (Wallmann, 2001) making this among the best sitesto observe this effect.

2. Materials and methods

2.1. Measurement of discharge

River discharge was measured by the Hydrological Service of theNational Energy Authority at distinct monitoring stations. Thelocation of each station is shown in Fig. 1. Discharge at each stationwas calculated by using a rating curve describing the relationshipbetween water level and discharge. Water level is continuouslymeasured throughout the year and recorded at 60 minute intervals.Discharge measurements are performed several times each year ateach station to validate the rating curve which depends on thestability of the river bed cross section at the monitoring site. Duringthe winter, the river water level at the measuring site can beinfluenced by the presence of ice. In such cases, river discharge wasestimated taking account of the air temperature, precipitation, anddischarge of nearby ice-free rivers.

2.2. Determination of dissolved concentrations and fluxes

Each studied river was sampled regularly from the locations shownin Fig.1 eight to ten times per year from 1998 to 2003. These riverwatersamples were filtered on site immediately after sampling through0.2 µm Millipore cellulose acetate filters. Alkalinity was determined byGran titration (c.f. Stumm and Morgan, 1996) and Cl by ion chromato-graphy. Samples to be analysed by inductively coupled plasma atomicemission spectrometry (ICP-AES) were acidified by adding 1 ml ofconcentrated suprapure HNO3 to a 90 ml water sample. Total dissolvedCa and Si were determined by ICP-AES. The relative error of all thesemeasurements is less than 5%. The Ca concentration was corrected fordissolved Ca originating from precipitation by assuming all dissolved Clpresent inwater samples originated from rainwater and using the Ca/Clratio of Icelandic precipitation (Gislason et al., 1996). Fluxes werecalculated by multiplying corrected Ca concentrations by the measureddischarge at the time of sampling. Resulting fluxes are reported as Ca2+

fluxes sinceCa2+ dominateddissolvedCa species in all thewater samplesaccording to speciation calculations using PHREEQC (Parkhurst andAppelo, 1999). The total dissolved inorganic carbon (DIC) flux wascalculated from the measured alkalinity, pH, temperature, totaldissolved silicon and the discharge at the time of sampling (Gislasonet al., 2004). The DIC concentrations were corrected for the DICoriginating from precipitation by assuming that precipitation was inequilibriumwith air at 0 °C (Gislason et al.,1996). This flux is reported asHCO3

−flux in the figures and tables as it dominates the DIC flux and is

consistent with reaction (2).

2.3. Determination of SIM concentration and fluxes

Total suspended inorganic particulate matter (SIM) and particulateorganic matter (POM) were sampled from the main channel of therivers, at the sampling stations shown in Fig. 1 using suspendedmatter samplers (Guy and Norman, 1970). Two different samplerswere used. All samples from the Fellsá and Fjardará rivers were takenwith a DH48 hand sampler (Guy and Norman, 1970). This sampler wasfastened to a rod that was lowered by hand into the river. Samplesfrom the other rivers, with few exceptions were taken with a S49sampler (Guy and Norman, 1970). This sampler was attached to awinch which lowered and lifted the sampler to and from the riverbottom at a constant rate (Hardardottir et al., 2003).

The concentration of the suspended inorganic particulate matter(SIM) was measured at the Hydrological Service of the NationalEnergy Authority. The organic fraction of the suspended sediment wasremoved prior to analysis by boiling the sample in hydrogen peroxide(H2O2, Pálsson and Vigfússon, 1996). The instantaneous SIM flux wascalculated by multiplying measured SIM concentrations by the

3 These average annual data are available directly from the IMO website: http://andvari.vedur.is/vedurfar/yfirlit/medaltalstoflur/Arsgildi.html.

Fig. 2.Measured alkalinity versus discharge (Q) for all rivers of the present study. Note the different scales of the plots. Data from rivers draining the youngest catchment rocks are onthe top and those from rivers draining the oldest rocks are shown near the bottom. All the datawere fit with the power law (Eq. (4)). The equation and coefficient of determination (r2)of the resulting fit are provided in each plot.

216 S.R. Gislason et al. / Earth and Planetary Science Letters 277 (2009) 213–222

discharge at the time of sampling. The chemical composition of theSIM samples collected from 1998 to 2003 was measured as describedby Eiriksdóttir et al. (2008).

2.4. Mean annual temperature and annual precipitation measurements

There are 6 meteorological stations located within the studiedcatchment areas run by the Icelandic Meteorology Office (IMO). Thelocation of these stations is shown in Fig. 1. All stations recordedprecipitation and 5 have recorded temperature since the early 1960s(The Icelandic Meteorological Office, 2007). The IMO reports annualmean temperature in °C, and integrated annual precipitation is

reported in mm.3 The precipitation is measured by Hellaman-meterswith a wind shield (The Icelandic Meteorological Office, 2007). Theprecipitation data, however, is much less reliable than the tempera-ture data. This is due to 1) the high wind speed in Iceland, 2) the largefraction of the precipitation is in the form of snow, 3) local topographyand manmade structures, and 4) the planting of trees in vicinity of thestations (The Icelandic Meteorological Office, 2007).

Fig. 3. Fluxes measured at the Hóll station on the River Jökulsá í Fljótsdal: a) suspendedinorganic material (SIM) flux versus discharge (bottom) and runoff (top) at the time ofsampling for the period 1963 to 2004; b) and c) dissolved Ca2+ and HCO3

−fluxes as a

function of discharge (bottom) and runoff (top) for the period 1998 to 2003. Ca2+

dominated dissolved Ca species in all the waters. The symbols correspond toinstantaneous fluxes and the lines result from a linear regression of the data. Equationsof each regression are provided on the figure where Q represents discharge in units ofm3/s and r2 the coefficient of determination.

217S.R. Gislason et al. / Earth and Planetary Science Letters 277 (2009) 213–222

2.5. Statistical analysis

The statistical significance of linear relationships (p) betweenmean annual temperature and annual precipitation versus annualrunoff, mechanical weathering fluxes, and chemical weathering fluxeswere determined by calculating the statistical significance of thesecorrelations using STATISTICA 7.0 (STATISTICA, 2004). A correlation isassumed to be significant if the calculated statistical significance, p, isless than 0.05. Linear least squares regressions of these relations,correlation coefficients (r), coefficient of determinations (r2) andstandard errors were also determined using STASTICIA 7.0 (STATIS-TICA, 2004).

Table 2Fit equations and coefficients of determination describing the logarithm of river fluxes as a

Catchment a SIM r 2 log Ca2+

(kg/s) (mol/s)

Grímsstadir log SIM=3.11 log Q−4.93 0.742 log Ca2+=0.858Brú log SIM=2.21 log Q−2.82 0.939 log Ca2+=0.675Hjardarhagi log SIM=2.68 log Q−4.15 0.887 log Ca2+=0.659Hóll log SIM=1.99 log Q−2.24 0.849 log Ca2+=0.698Fellsá log SIM=1.18 log Q−2.25 0.849 log Ca2+=0.781Lagarfljót log SIM=1.00 log Q−1.65 0.695 log Ca2+=0.937Grímsá log SIM=1.32 log Q−2.60 0.757 log Ca2+=0.821Fjardará log SIM=1.45 log Q−2.49 0.827 log Ca2+=0.789

a The first four glacier covered catchments are denoted by their respective sampling statioNote that the linear correlations listed in this table are all strong correlations with pb0.01.

3. Results and discussion

3.1. Climate and rivers of north-eastern Iceland

The range in the mean annual temperature and annual precipita-tion of each weather station present in the study area is provided inTable 1. The mean annual temperature and annual precipitation at themeteorological stations varied by 3.2 to 4.5 °C and 80 to 530% over thestudy period. There is an overall temperature increase at allmeteorological stations during the past 40 years but precipitationincreased or decreased depending on the meteorological station (TheIcelandic Meteorological Office, 2007).

The discharge and concentration of the suspended inorganicmaterial (SIM) of all 8 studied rivers have been monitored for 30–44 years depending on the river. The chemical composition of thesuspendedmaterial and the concentration of dissolved elements weremeasured eight to ten times per year from 1998 to 2003. Thesuspended inorganic material from the rivers is basaltic and thechemical composition of this suspended material in each catchmentwas close to constant (Gislason et al., 2006; Eiriksdottir, et al., 2008).The contribution of carbonate dissolution to the dissolved load ofthese rivers is negligible as demonstrated by the strong correlationbetween alkalinity and silica concentrations shown in Fig. 1 of thesupplementary material. Alkalinity is relatively high in the riversdraining young rocks. The relationship between Si and alkalinity inJökulsá á Fjöllum at Grímsstadir is less clear than other young glacialfed rivers because a considerable fraction of its discharge comes fromgroundwaters and there is some geothermal component in the riverwaters (Eiriksdottir et al., 2008). The range in alkalinity and dissolvedsilicon concentration is diminished by the presence of dams andreservoirs in the Lagarfljót and Grímsá rivers. Note that the riversJökulsá í Fljótsdal, Fellsá and Grímsá are all tributaries to Lagarfljót.Particulate organic carbon concentration was low in all rivers, neverexceeding 3% of the total suspended matter.

3.2. River fluxes

Many authors (e.g. Walling and Webb, 1986; Gislason et al., 2006)have shown that river water concentrations change with waterdischarge according to a power law

C = aQb ð4Þ

where C represents concentration, a refers to a constant, Q denotesriver discharge, and b stands for a constant describing the powerdependence of river concentrationwith discharge. If b is equal to zero,the concentration is independent of discharge. If b is equal to −1, theriver water concentrations may be controlled by its dilution by purewater. Previous studies (e. g. Walling and Webb, 1986; Gislason et al.,2006) have shown that b is greater than zero for river suspendedinorganic matter (SIM) and b is between 0 and −1 for dissolved

function of the logarithm of discharge (Q)

r 2 DIC r 2

(mol/s)

log Q−0.510 0.788 Log DIC=0.675 log Q+0.659 0.786log Q−0.425 0.983 Log DIC=0.626 log Q+0.282 0.986log Q−0.259 0.961 Log DIC=0.589 log Q+0.496 0.937log Q−0.296 0.974 Log DIC=0.689 log Q+0.184 0.977log Q−1.11 0.942 Log DIC=0.792 log Q−0.446 0.963log Q−0.760 0.990 Log DIC=0.911 log Q−0.171 0.987log Q −0.686 0.970 Log DIC=0.838 log Q−0.206 0.978log Q−1.31 0.910 log DIC=0.789 log Q−0.642 0.954

ns, but the others by their rivers (Fig. 1). The fit equations are the logarithmic of Eq. (5).

218 S.R. Gislason et al. / Earth and Planetary Science Letters 277 (2009) 213–222

constituents, with majority of rivers falling in the range 0 to −0.4 withan overall mean of −0.17 (Walling and Webb, 1986). Such behavioursuggests that increased river discharge increases SIM but dissolvedconcentrations decrease, but less than if they were diluted with purewater. Measured alkalinity for all the rivers of the present study isplotted as a function of discharge at the time of sampling in Fig. 2. A fitof these data with the power law (Eq. (4)) is superimposed on eachplot together with the corresponding square of the coefficient ofdetermination (r2). The power dependence of these regression fits (b)is greater than −1, suggesting an increased dissolution of minerals andglasses with increasing discharge. Note that the high alkalinity–lowdischarge samples in Fig. 2. are from the winter whereas the lowalkalinity concentration–high discharge samples are from the sum-mer, underscoring the effect of climate on mineral dissolution andriverine transport of dissolved material (c.f. Eiriksdottir and Gislason,submitted for publication). Note the power dependence (b) is lowestfor the Lagarfljót and Grímsá rivers which are dammed.

The instantaneous flux of dissolved and suspended constituentscan be calculated by multiplying the river water concentrations bydischarge at the time of sampling (e.g. Gislason et al., 2006), increasingthe power in Eq. (4) by 1;

F = C Q = aQb + 1 ð5Þ

where F designates the river flux. The logarithm of Eq. (5) yields asimple linear equation with intercept equal to log a and slope of b+1.

The logarithm of the variation of the instantaneous SIM flux versusthe logarithm of discharge and runoff (Eq. (6)) at the time of samplingfor the Jökulsá í Fljótsdal River at the Hóll station from 1963 to 2003 isshown in Fig. 3a. Runoff was calculated according to the followingequation;

runoff =Q 3:15 × 107=A ð6Þ

The units for runoff are m/yr, discharge Q is in m3/s and A is thecatchment area in m2 (see Table 1). The corresponding measureddissolved Ca2+ and HCO3

−fluxes as a function of discharge and runoff

for the period 1998 to 2003 are shown in Fig. 3b and c. The resultsshown in Fig. 3 are representative of the studied rivers. Linearregressions of these data are provided in this figure. Correspondingregressions of the fluxes of all rivers investigated in this study areprovided in Table 2. All fluxes increase with runoff but SIM flux

Table 3Correlation coefficients (r) and statistical significance (p) of linear correlations among annutemperature and rainfall

Shaded correlations are not statistically significant.

increase more rapidly than dissolved fluxes with increasing runoff, asdemonstrated by the slope of the fit equations provided in Table 2.These fit equations were used together with daily dischargemeasurements to calculate the daily SIM and dissolved Ca2+ andHCO3

−fluxes; these daily fluxes were then integrated to generate

corresponding annual fluxes. Note that the annual dissolved fluxes,estimated for a ∼44 year period, are based on Ca2+ and HCO3

−

concentrations measured during 1998 to 2003. An inherent assump-tion of this estimation is therefore that the dissolved fluxes versusdischarge curves shown in Fig. 3b and c varied insignificantly duringthe 44 year study period. The high coefficients of correlation of thedissolved fluxes versus discharge fits shown in Table 2 supportstrongly this assumption. In contrast, the annual SIM fluxes weregenerated from SIM concentrations measured over the whole∼40 year study period.

3.3. Correlations between river fluxes and climate

The effect of runoff on mechanical weathering rates dependsstrongly on the identity of the catchment despite the fact that the 8catchments have the same lithology and are adjacent to each other.This effect is exemplified by the slope of the fit equations presented inTable 2. The effect of runoff on mechanical weathering rates isinfluenced by glacial cover, catchment rock age, and the presence ofdams (e.g. Tómasson, 1990; Hardardottir and Snorrason, 2003;Gislason et al., 2006). The mechanical weathering rate dependenceon runoff increases due to the presence of glaciers because of grinding,decreases with catchment rock age because older rocks contain ahigher proportion of finer grained particles, and decreases by thepresence of dams because of coarse suspended material trapping. As aconsequence, the mechanical weathering rate dependence on runoffis greatest for pristine glacier catchments containing young rocks asevidenced by the fit equations in Table 2. An order of magnitudeincrease in runoff in the Jökulsá á Fjöllum river at Grímsstadir, whichdrains a pristine, glaciated catchment containing young rocks, leads toa 3.11 order of magnitude increase in SIM flux. In contrast, an order ofmagnitude increase in runoff in Fjardará River, which drains a non-glaciated catchment containing the oldest rocks, leads to only a 1.45order of magnitude increase in SIM flux.

The effect of runoff on chemical weathering rates is less than thatof mechanical weathering rates. In contrast to the SIM flux fitequations, the slopes for chemical weathering flux versus discharge

al runoff, mechanical weathering fluxes, and chemical weathering fluxes versus average

Fig. 4. Annual river fluxes from the Jökulsá í Fljótsdal at Hóll versus the annual mean airtemperature at the Hallormsstadur station: a) annual runoff in m/yr, b) annualsuspended inorganic material (SIM) flux in kg/m2/yr, c) annual dissolved Ca2+ flux inmol/m2/yr, d) annual dissolved HCO3

−flux in mol/m2/yr. The symbols correspond to

measured fluxes but the line represents a least squares fit of the data. The equation andcoefficient of determination (r2) of the fit are provided in each plot.

Table 4Fit equations and coefficients of determinations of annual river fluxes as a function ofannual mean air temperature (°C) at the respective meteorological stations, percentchange in the respective fluxes per one degree C change in temperature and theapparent activation energy

aStandard errors are provided in parentheses.Shaded results are not statistically significant.

219S.R. Gislason et al. / Earth and Planetary Science Letters 277 (2009) 213–222

fit functions are less than one as shown in Table 2. As a consequence,changes in chemical weathering fluxes are smaller than thecorresponding change in runoff. In addition the combined effects ofglacial cover, rock age, and the presence of dams has much smallereffect on the chemical weathering flux dependence on runoff than theSIM flux dependence on runoff. Note, however, that in contrast toother rivers, where increases in chemical weathering fluxes couldpotentially be attributable to land use changes (e.g. Raymond andCole, 2003), the rivers of north-eastern Iceland are sparsely populatedand land use has changed negligibly over the study period.

A number of studies have proposed a direct link betweentemperature and chemical weathering (e.g. White and Blum, 1995;Millot et al., 2003; West et al., 2005). This link stems, at least in part,from the variation of mineral and glass dissolution rates withtemperature (c.f. Oelkers and Schott, 1995; 2001; Oelkers 2001). Theactivation energy for basaltic glass dissolution in the laboratory has

been shown to range from 25–35 kJ/mol (Gislason and Eugster, 1987;Gislason and Oelkers, 2003). Models of climate–chemical weatheringfeedback, taking account of such experimental data, suggest a 2 to 10%increase in chemical weathering rates for each degree of temperatureincrease (Berner and Kothavala, 2001; Wallmann, 2001). These modelresults are tested in the present study by calculating the correlationmatrix between annual runoff, mechanical weathering fluxes, andchemical weathering fluxes and the mean annual temperature of theclosest weather station. Runoff was used in this correlation matrixrather than precipitation because the measurements are moreaccurate and because of the large contribution of glacial melt waterin these systems. The result of this statistical analysis is presented inTable 3. Significant positive linear correlations (pb0.05) are foundbetween mean annual temperature and annual chemical weatheringfluxes in all 8 catchments. Significant positive linear correlations arefound between mean annual temperature and runoff and betweenmechanical weathering (SIM flux) and temperature for all but thenon-glacial Fjardará River. The correlations between the annual meanair temperature at the Hallormsstadur meteorological station andannual runoff, mechanical weathering fluxes, and chemical weath-ering fluxes of the Hóll catchment is shown in Fig. 4. The slope of theregression lines in Fig. 4 yields a 16% increase in runoff, a 21% increasein SIM flux, and a 14% increase in Ca2+ and HCO3

− chemical weatheringfluxes per degree temperature increase. This chemical weathering fluxchange corresponds to apparent activation energy of 86 kJ/mol. Thecorresponding flux versus temperature fits for all studied rivers is

Fig. 5. The temporal evolution of annual suspended and dissolved fluxes of the Jökulsá íFljótsdal at the Hóll station: a) the suspended inorganic matter (SIM) flux, b) dissolved Ca2+

flux, and c) dissolved HCO3−flux. All fluxes were generated by combining the regression

equations provided inTable 2with the continuous discharge records to estimate the total SIMand chemical weathering fluxes for each day during the study period. The daily fluxes werethen numerically integrated to generate annual fluxes. The symbols correspond to annualfluxes and the lines result froma linear regressionof thedata. Equations of each regressionareprovided on the figure where t represents time in calendar years and r2 corresponds to thecoefficient of determination.

Fig. 6. Annual percent change in a) runoff, b) suspended inorganic matter (SIM) flux, c)dissolved Ca2+ flux, and d) dissolved HCO3

−flux in all of the studied rivers from 1964 to

2004. The glacier fed rivers are denoted by their respective sampling stations. Error barsrepresent 95% confidence levels. Rivers are presented in order from west to east,corresponding to an increase in average rock age from 0.3 to 11.2 million years. Glaciersare present in the four westernmost rivers and the Lagarfljót River. Results weregenerated from the linear regression of annual fluxes versus time as shown in Fig. 5.

220 S.R. Gislason et al. / Earth and Planetary Science Letters 277 (2009) 213–222

presented in Table 4. These fit equations indicate a 6 to 16% increase inrunoff, a 5 to 29% increase in SIM flux, and a 4 to 14% increase in Ca2+

and HCO3− chemical weathering fluxes for each degree temperature

increase. These chemical weathering flux increases correspond toapparent activation energy of 24–86 kJ/mol. The apparent activationenergy is largest in catchments where runoff increases fastest withtemperature.

The apparent activation energy generated from for this field studystems from both 1) the direct effect of temperature on dissolution ratesof glasses and minerals and 2) the influence of temperature on a)increasing runoff, b) enhancing primary production by vegetation and c)decomposition of organic matter in soil that in turn all could affect thedissolution rate and saturation state of minerals and glasses in soil andriver waters. Experimentally measured, solution composition indepen-dent, activation energy for the dissolution rate of basaltic glass at highundersaturation is 25.5 kJ/mol (Gislason and Oelkers, 2003) and forother natural glasses range from 27 to 57 kJ/mol (Wolff-Boenisch et al.,2004). These activation energies are on average lower than that deducedfrom this field study, suggesting that other factors influenced bytemperature tend to enhance weathering rates. One of these factors isincreased runoff. Higher temperatures lead to increased runoff (c.f.Fig. 4)which leads to increased dissolution rates for a number of reasonsincluding1) the loweringof pHwhere the concentration of organic acidsis low, 2) a decrease in the concentration of total dissolved Al, andMg

(c.f. Oelkers and Schott, 2001; Oelkers and Gislason, 2001), 3) increasingundersaturation of primary minerals, 4) decreasing degree of super-saturation of secondary minerals, and 5) increasing the reactive surfacearea between rock and water.

An additional factor that could be contributing to the observationthat measured field weathering rates have higher apparent activationenergy than laboratory studies is increasing atmospheric CO2

concentrations. Over the past 4 decades, atmospheric pCO2 hasrisen, presumably in close tandem with the temperature increasesobserved for the field site in Iceland. Higher atmospheric CO2 content

221S.R. Gislason et al. / Earth and Planetary Science Letters 277 (2009) 213–222

may accelerate chemical weathering rates, as weathering is partlydriven by carbonic acid. In addition, increased atmospheric CO2

content could increase primary production and thus biologicalproduction of organic acids. Such acids have been shown to increasesignificantly mineral and glass dissolution rates in the laboratory andin the field (Oelkers and Schott, 1998; Moulton et al., 2000; Oelkersand Gislason 2001; Sigfusson et al., 2008).

In contrast to the significant correlations found betweenweatheringand annual mean temperature calculations summarized in Table 3,calculations indicate significant positive correlations between theannual precipitation at the nearest weather station and annual runoff,between the annual precipitation and mechanical weathering fluxes,and between the annual precipitation and chemical weathering fluxesonly in the Jökulsá í Fljótsdal at Hóll, Grímsá River, and Fjardará River(Table 3). The poor connection between precipitation andweathering inthe studied catchments result likely from1) the large influence of glacialmelting on runoff in the studied catchments, 2) the effect of localtopography on precipitation and 3) uncertainties associated with theprecipitation measurements as discussed in Section 2.4.

3.4. Weathering rates and climate change from 1964 to 2004

This study has focussed on the variation of weathering rates in 8Icelandic river catchments over the past 40 years. This 40 year periodalso corresponds to the most rapid recent global warming period(Alley et al., 2007). As is the casewithmuch of the Earth, temperatureshave been increasing in north-eastern Iceland. Although there isimportant scatter, the average annual temperature has increased at allmeteorological stations shown in Fig. 1 (The Icelandic MeteorologicalOffice http://andvari.vedur.is/vedurfar/yfirlit/medaltalstoflur/Arsgildi.html). A relevant question is, therefore, can an increase in weatheringrates with time be observed in the 8 studied Icelandic catchments? Toassess this possibility, the total annual mechanical and chemicalweathering fluxes of the studied rivers have been plotted as a functionof time. The clearest connection between these fluxes and time wasobserved in the River Jökulsá í Fljótsdal at the Hóll monitoring station.These results are shown in Fig. 5. A summary of the changes in annualsuspended and dissolved fluxes in all 8 studied rivers over this timeperiod is shown in Fig. 6. The annual mechanical and chemicalweathering fluxes of all eight catchments have increased since 1964,but the 95% confidence limits of 6 of the 8 studied catchmentsencompass zero change. In contrast, the mechanical and chemicalweathering fluxes of the Rivers Jökulsá á Fjöllum at Grímsstadir andJökulsá í Fljótsdal at Hóll show a statistically significant positivecorrelation with time during the 40 year study period.

Some additional insight into the effect of climate onweathering canbe seen in Fig. 6. The Jökulsá í Fljótsdal at Hóll and Fellsá catchmentsexperienced the greatest increase in runoff, more than double theincrease experienced by any other studied river. As a result these riversexperienced the greatest increase in chemical weathering rates.Mechanical weathering rates, however, increased fastest in the Jökulsáá Fjöllum catchment, the catchmentwhich contains the youngest rocks.This observation is consistent with previous conclusions that themechanical weathering rate dependence on runoff is greatest forcatchments containing young rocks (e.g. Gislason et al., 2006).

3.5. The effect of glaciers on weathering rates

This study provides insight into the role of glacial melting on thefeedback between climate change and weathering. Four of the studiedcatchments (Jökulsá á Fjöllum at Grímsstadir, Jökulsá á Dal at Brú,Jökulsá á Dal at Hjardarhagi, and Jökulsá í Fljótsdal at Hóll) havesignificant glacial cover, and four are essentially glacier-free (Fellsá,Grímsá, Lagarfljót and Fjardará). Results inTable 4 show that the effect oftemperature onmechanicalweathering is greater in glacial catchments;SIM flux increases by 17 to 30% for each degree C in the glaciated

catchments, but by only 5 to 16% in the non-glaciated catchments. Thisobservationmay stem from the input to the rivers of newly exposed finesediments as the glaciers melt. In contrast, the presence of glaciersapparentlyhas a smallereffectonchemicalweathering. ThedissolvedCaflux increases by 5.6 to 14.0%/°C in the glaciated catchments and 5.7 to13.6%/°C in the non-glaciated catchments.

The present study focused on the variation of chemical andphysical weathering rates over a ∼40 year time span. The degree towhich the climate/weathering flux relations obtained from this studyare invariant and could thus be extrapolated over the longer term isuntested. Such relations could be influenced by a large number offactors including changes in 1) catchment hydrology, 2) anthropo-genic inputs, 3) glacial cover, and 4) quantities of sediments present,each of which could vary over the long term in the study area. It isanticipated that future studiesmay illuminate in detail the role of eachof these factors on climate/weathering flux relations.

4. Conclusions

This study shows, by direct measurement, that chemical andmechanical weathering fluxes depend on climate via changingtemperature and runoff. The measured feedback betweenweathering,temperature, and runoff is consistent with both the results of previouspostulated models and laboratory measured dissolution rates. Thiscoherence of evidence adds considerable confidence to our ability topredict long-term climate changes stemming from atmospheric CO2

variations across geological timescales. Further analysis shows thatboth mechanical and chemical weathering fluxes have increased withtime over the past 40 years, though due to scatter in annual runoffversus time these correlations are only statistically significant for 2 ofthe 8 studied catchments.

Acknowledgements

We thank S. Callahan, O. Pokrovsky, J. Schott, and K. Burton forinsightful discussion and encouragement throughout this study.Halldór Björnsson and Trausti Jónsson at The Icelandic MeteorologicalOffice are thanked for information on the Meteorological data. We aregrateful for constructive reviews by J. West, P.M. Delaney and ananonymous reviewer. This work was supported by the Landsvirkjun,Icelandic Ministry for the Environment, National Energy Authority,Icelandic Science Foundation RANNÍS, the Centre Nationale de laRecherche Scientifique, the European Commission Marie Curie MIN-GRO research and training network (Contract MRTN-2005-031464),and Institute of Earth Science, University of Iceland.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.epsl.2008.10.018.

References

Aller, R.C., 1998. Mobile deltaic and continental shelf muds as suboxic, fluidized bedreactors. Mar. Chem. 61, 143–155.

Alley, R., et al., 2007. Climate change 2007: the physical science basis, sum-mary for policymakers. Contribution of Working Group I to the FourthAssessment Report of the Intergovernmental Panel on Climate Change.Intergovernmental Panel on Climate Change, Geneva. http://ipcc-wg1.ucar.edu/wg1/docs/WG1AR4_SPM_PlenaryApproved.pdf.

Anderson, S.P., 2007. Biogeochemistry of glacial landscape systems. Annu. Rev. EarthPlanet. Sci. 35, 375–399.

Berner, R.A., 2004. The Phanerozoic Carbon Cycle. Oxford University Press, Oxford.Berner, R.A., Kothavala, Z., 2001. GEOCARB III. A revised model of atmospheric CO2 over

Phanerozoic time. Am. J. Sci. 301, 182–204.Berner, R.A., Lasaga, A.C., Garrels, R.M., 1983. The carbonate–silicate geochemical cycle

and its effect on atmospheric carbon dioxide over the past 100 million years. Am.J. Sci. 283, 641–683.

Bluth, G.J.S., Kump, L.R., 1994. Lithologic and climatologic controls of river chemistry.Geochim. Cosmochim. Acta 58, 2341–2359.

222 S.R. Gislason et al. / Earth and Planetary Science Letters 277 (2009) 213–222

Dessert, C., Dupré, B., Gaillardet, J., Francois, L.M., Allégre, C.J., 2003. Basalt weatheringlaws and the impact of basalt weathering on the global carbon cycle. Chem. Geol.202, 257–273.

Edmond, J.M., Palmer, M.R., Measures, C.I., Grant, B., Stallard, R.F., 1995. The fluvialgeochemistry and denudation rates of the Guyana shield in Venezuela, Colombiaand Brazil. Geochim. Cosmochim. Acta 59, 3301–3325.

Eiriksdottir, E.S., Gislason, S.R., submitted for publication. Climatic effects on chemicalweathering of basalts in NE-Iceland: saturation state and dissolution rates.Geochim. Cosmochim. Acta.

Eiriksdottir, E.S., Louvat, P., Gislason, S.R., Óskarsson, N., Hardardóttir, J., 2008. Temporalvariation of chemical and mechanical weathering in NE Iceland: evaluation ofa steady state model of erosion. Earth Planet. Sci. Lett. 272, 78–88.

Gaillardet, J., Dupré, B., Louvat, P., Allègre, C.J., 1999a. Global silicate weathering and CO2

consumption rates deduced from the chemistry of large rivers. Chem. Geol. 159,3–30.

Gaillardet, J., Dupré, B., Allègre, C.J., 1999b. Geochemistry of large river suspendedsediments: silicate weathering or recycling tracers? Geochim. Cosmochim. Acta 63,4037–4051.

Gedney, N., Cox, P.M., Betts, R.A., Boucher, O., Huntingford, C., Stott, P.A., 2006. Detectionof a direct carbon dioxide effect in continental river runoff records. Nature 439,835–838.

Gibbs, M.T., Kump, L.R., 1994. Global chemical erosion during the last glacial maximumand the present: sensitivity to changes in lithology and hydrology. Paleoceanography9, 529–543.

Gislason, S.R., Eugster, H.P., 1987. Meteoric water–basalt interactions: I. A laboratorystudy. Geochim. Cosmochim. Acta 51, 2827–2840.

Gislason, S.R., Oelkers, E.H., 2003. The mechanism, rates and consequences of basalticglass dissolution: II. An experimental study of the dissolution rates of basaltic glassas a function of pH and temperature. Geochim. Cosmochim. Acta 67, 3817–3832.

Gislason, S.R., Arnórsson, S., Ármannsson, H., 1996. Chemical weathering of basalt inSouthwest Iceland: effects of runoff, age of rocks and vegetative/glacial cover. Am.J. Sci. 296, 837–907.

Gislason, S.R., Snorrason, Á., Eiriksdottir, E.S., Sigfússon, B., Elefsen, S.Ó., Harðardóttir, J.,Gunnarsson, Á., Hreinsson, E.Ö., Torssander, P., Óskarsson, N.Ö., Oelkers, E.H., 2004.Chemical composition of dissolved and suspended riverine constituents inNortheastern Iceland, V. The database of the Science Institute, University of Icelandand the National Energy Authority, Iceland. Science Institute, RH-05-2004, Reykjavík.

Gislason, S.R., Oelkers, E.H., Snorrason, Á., 2006. The role of river suspended material inthe global carbon cycle. Geology 34, 49–52.

Guy, H.P., Norman, V.W., 1970. Techniques of Water-Resources Investigations of theUnited States Geological Survey. Field Methods for Measurement of FluvialSediment 3 (C2). United States Government Printing Office, Washington, DC.

Hardardottir, J., Snorrason, Á., 2003. Sediment monitoring of glacial rivers in Iceland:new data on bed load transport. Erosion and Sediment Transport Measurement inRivers: Technological and Methodological Advances (Proceedings of the OsloWorkshop, June 2002), vol. 283. IAHS Publ., pp. 154–163.

Hardardottir, J., Gunnarsson, Á., Thorlaksdottir, S.B. 2003. Mælingar á rennsli, svifaur ogskridaur í Jökulsá á Dal árid 2002. OS-2003/001 38 pp (in Icelandic).

Johannesson, H., Saemundsson, K., 1998. Geological Map of Iceland — Bedrock Geology.The Icelandic Institute of Natural History, Reykjavik.

Labat, D., Goddéris, Y., Probst, J.L., Guyot, J.L., 2004. Evidence for global runoff increaserelated to climate warming. Adv. Water Resour. 27, 631–642.

Meybeck, M., 1986. Composition chimique des ruisseaux non pollués de France.Scientifique Géologie Bulletine 39, 3–77.

Meybeck, M., 1994. Origin and Variable Composition of Present Day RiverborneMaterial. Studies in Geophysics, Material Fluxes on the Surface of the Earth.National Academy Press, Washington, D.C., pp. 61–73.

Milliman, J.D., Syvitski, J.P.M., 1992. Geomorphic/tectonic control of sediment dischargeto the ocean: the importance of small mountainous rivers. J. Geol. 100, 525–544.

Millot, R., Gaillardet, J., Dupré, B., Allègre, C.J., 2003. Northern latitude chemicalweathering rates: clues from the Mackenzie River Basin, Canada. Geochim.Cosmochim. Acta 67, 1305–1329.

Moorbath, S., Sigurdsson, H., Goodwin, R.,1968. K–Ar ages of the oldest exposed rocks inIceland. Earth and Planet. Sci. Lett. 4, 197–205.

Moulton, K.L., West, J., Berner, R.A., 2000. Solute flux and mineral mass balanceapproaches to the quantification of plant effects on silicate weathering. Am. J. Sci.300, 539–570.

Navarra-Sitchler, A., Brantley, S., 2007. Basalt weathering across scales. Earth Planet. Sci.Lett. 261, 321–334.

Oelkers, E.H., 2001. An experimental study of forsterite dissolution rates as a function oftemperature and aqueous Mg and Si concentration. Chem. Geol. 175, 485–494.

Oelkers, E.H., Schott, J., 1995. An experimental study if anorthite dissolution and therelative mechanism of feldspar hydrolysis. Geochim. Cosmochim. Acta 59,5039–5053.

Oelkers, E.H., Schott, J., 1998. Does organic acid adsorption effect alkali feldspardissolution rates? Chem. Geol. 151, 235–246.

Oelkers, E.H., Gislason, S.R., 2001. The mechanism, rates, and consequences of basalticglass dissolution: I. An experimental study of the dissolution rates of basaltic glassas a function of aqueous Al, Si, and oxalic acid concentration at 25 °C and pH=3 and11. Geochim. Cosmochim. Acta 65, 3671–3681.

Oelkers, E.H., Schott, J., 2001. An experimental study of enstatite dissolution rates as afunction of pH, temperature, and aqueous Mg and Si concentration and themechanism of pyroxene/pyroxenoid dissolution. Geochim. Cosmochim. Acta 65,1219–1231.

Pálsson, S., Vigfússon, G.H., 1996. Results of suspended load and discharge measure-ments 1963–1995. Reykjavik, National Energy Authority. OS-96032/VOD-05 B.

Parkhurst, D.L., Appelo, C.A.J., 1999. User's Guide to PHREEQC (Version2) — A ComputerProgram for Speciation, Batch-reaction, One Dimensional Transport, and InverseGeochemical Calculations, pp. 99–4259.

Peters, N.E., 1984. Evaluation of Environmental Factors Affecting Yields of MajorDissolved Ions of Streams in the United States. United States Geological SurveyWater Supply Paper, p. 2228.

Pogge von Strandmann, P.A.E., James, R.H., Van Calsteren, P., Gislason, S.R., Burton, K.W.,2008. Lithium, magnesium and uranium isotope behaviour in the estuarineenvironment of basaltic islands. Earth Planet. Sci. Lett. 274, 462–471.

Raymond, P.A., Cole, J.J., 2003. Increase in the export of alkalinity from North America'slargest river. Science, 301, 88–91.

Sigfusson, B., Gislason, S.R., Paton, G.I., 2008. Pedogenesis and weathering rates of aHistic Andosol in Iceland: field and experimental soil solution study. Geoderma144, 572–592.

STATISTICA 2004. Version 7. StatSoft, Inc. www.statsoft.com.Stefánsdóttir, M.B., Gislason, S.R., 2005. The source of suspended matter and suspended

matter/seawater interaction following the 1996 outburst flood from the VatnajökullGlacier, Iceland. Earth Planet. Sci. Lett. 237, 433–452.

Stefánsson, A., Gislason, S.R., 2001. Chemical weathering of basalts, SW Iceland: effect ofrock crystallinity and secondaryminerals on chemical fluxes to the ocean. Am. J. Sci.301, 513–556.

Stumm, W., Morgan, J.J., 1996. Aquatic Chemistry. Chemical Equilibria and Rates inNatural Waters, Third Edition. John Wiley & Sons, Inc., New York. 1022 p.

Sverdrup, H.U., 1990. The Kinetics of Base Cation Release Due to Chemical Weathering.Lund University Press, Lund, Sweden. 246 pp.

The Icelandic Meteorological Office, 2007. Annual average climate information 1961–2005. The Icelandic Meteorological Office, Reykjavík. http://andvari.vedur.is/vedurfar/yfirlit/medaltalstoflur/Arsgildi.html2007.

Tómasson, H., 1990. Suspended matter in Icelandic rivers. In: Sigbjarnarson, G. (Ed.),Vatnid og Landid. Orkustofnun, Reykjavík, pp. 169–174.

Tranter, M., Huybrechts, P., Munhoven, G., Sharp, M.J., Brown, G.H., Jones, I.W., Hodson,A.J., Hodgkins, R., Wadham, J.L., 2002. Direct effect of ice sheets on terrestrialbicarbonate, sulphate and base cation fluxes during the last glacial cycle: minimalimpact on atmospheric CO2 concentrations. Chem. Geol. 190, 33–44.

Walker, J.C.G., Hays, P.B., Kasting, J.F., 1981. A negative feedbackmechanism for the long-term stabilization of Earth's surface temperature. J. Geophys. Res. 86, 9776–9782.

Walling, D.E., Webb, B.W., 1986. Solutes in river systems. In: Trudgill, S.T. (Ed.), SoluteProcesses. Wiley, Chichester, pp. 251–327.

Wallmann, K., 2001. Controls on the Cretaceous and Cenozoic evolution of seawatercomposition, atmospheric CO2 and climate. Geochim. Cosmochim. Acta 65,3005–3025.

West, A.J., Galy, A., Bickle, M., 2005. Tectonic and climate control on silicate weathering.Earth Planet. Sci. Lett. 235, 211–228.

White, A.F., Blum, A.E., 1995. Effects of climate on chemical weathering in watersheds.Geochim. Cosmochim. Acta 59, 1729–1747.

Wolff-Boenisch, D., Gislason, S.R., Oelkers, E.H., Putnis, C.V., 2004. The dissolution ratesof natural glasses as a function of their composition at pH 4 and 10.6, andtemperatures from 25 to 74 °C. Geochim. Cosmochim. Acta 68, 4843–4858.

Wolff-Boenisch, D., Gislason, S.R., Oelkers, E.H., 2006. The effect of crystallinity ondissolution rates and CO2 consumption capacity of silicates. Geochim. Cosmochim.Acta 70, 858–870.