The stability of global soil carbon (C) represents a major uncertainty in forecasting future climate change. In the UK, substantial soil C losses have been reported, while at the same time dissolved organic carbon (DOC) concentrations in upland waters have increased, suggesting that soil C stocks may be destabilising in response to climate change. To investigate the link between soil carbon and DOC at a range of sites, soil organic matter, soilwater and streamwater DOC were analysed for radiocarbon (14C). DOC exported from C-rich landscapes appears younger than the soil C itself, much of it comprising C assimilated post-1950s. DOC from more intensively managed, C-poor soils is older, in some cases >100 years. Results appear consistent with soil C destabilisation in farmed landscapes, but not in peatlands. Reported C losses may to a significant extent be explained by mechanisms other than climate change, e.g. recovery from acidification in peatlands, and agricultural intensification in managed systems.
 Globally, soils contain more carbon (C) than either vegetation or the atmosphere, the majority held within organic-rich soils such as peats [Davidson and Janssens, 2006]. Accelerated decomposition of soil organic matter with rising temperatures could increase CO2 release to the atmosphere (a positive climate-change feedback) but the temperature-sensitivity of decomposition rates is debated [e.g., Fang et al., 2005; Davidson and Janssens, 2006; Knorr et al., 2005]. This debate is largely based on experimental and modelling studies; few long-term, large-scale measurements of soil C change are available, but recently reported large-scale soil C decreases in England and Wales [Bellamy et al., 2005] have been taken as evidence that climate change is causing soil C destabilisation [Davidson and Janssens, 2006; Roulet and Moore, 2006]. This repeat survey of 2179 sites, in 1978–1983 and 1995–2003, recorded a 0.6% per year average decrease in upper soil C content. With apparently greater C loss from organic-rich soils and no clear relationship between rate of C loss and land-use, a 0.5°C mean temperature increase over the period was invoked as a possible driver.
 Over a similar period, there have been widespread increases in DOC concentrations of UK upland surface waters [Freeman et al., 2001; Worrall et al., 2004], in some cases more than doubling between 1988 and 2003 [Evans et al., 2005]. Riverine DOC can represent a significant C loss pathway in organic-rich temperate and boreal ecosystems [Billett et al., 2004; Finlay et al., 2006], and Bellamy et al.  suggested that UK soil C losses and surface water DOC increases may be linked. There is now evidence that surface water DOC is increasing across much of Northern Europe and North America [e.g., Skjelkvåle et al., 2005; Stoddard et al., 2003], possibly representing first indications of soil C depletion across a much wider area of the Northern Hemisphere. This would have significant implications for global C budgets, particularly if climate-driven. Roulet and Moore  have highlighted the need for C isotope data in order to link observed DOC increases to possible sources and mechanisms. Here, we present radiocarbon (14C) isotopic data for soil organic matter, soil solution DOC, and stream DOC, for a representative UK upland area, and assess whether these data are consistent with a widespread, climate-induced destabilisation of soil C stores.
 Samples were collected from the 256 km2 Conwy catchment, North Wales, which drains a typical UK upland mixture of semi-natural and managed land on organic and mineral soils. The five major land classes [Evans et al., 2006a] are: (1) peat moorland (blanket bog, minimal grazing); (2) peaty gley moorland (acid grassland on poorly drained soils, low-intensity grazing); (3) montane grassland (acid grassland on thin, well-drained mineral or organo-mineral soils, moderate grazing); (4) improved grassland (managed permanent grassland on mineral soils, intensive sheep and cattle grazing); and (5) conifer forest (plantation forest, mainly on organo-mineral soils). For each land class we sampled (1) soil at 1–3 depths; (2) soil solution at the same locations and depths (using zero tension lysimeters in organic horizons, Prenart suction lysimeters in mineral horizons); and (3) 3–5 streams draining predominantly that land class, of which a subset were analysed for DO14C. Six larger rivers, draining a mixture of land classes, were sampled for both DOC and 14C analysis. Samples were collected during summer baseflow, and during an autumn high flow.
 The 14C content of organic matter reflects the isotopic composition of atmospheric CO2 at the time it was trapped by photosynthesis. C fixed prior to 1950 can be ‘radiocarbon dated’, based on the rate of radioactive decay of cosmogenic 14C. Since the 1950s, atmospheric testing of nuclear devices has provided a ‘tracer’ pulse of enhanced atmospheric 14CO2, peaking in the 1960s [Levin and Kromer, 2004] (Figure 1). Since soil and dissolved organic matter generally represent a mix of compounds of varying ages, no single age can be ascribed to any one sample, but the following general statements can be made: (1) samples with 14C < 100% ‘modern’ (i.e. atmospheric 14CO2 before 1955) must contain predominantly pre-bomb C, and an average age can be assigned; (2) samples in the range 100 to 106% modern (just below atmospheric 14CO2 levels when samples were collected) must contain a substantial fraction of C fixed before 1957 (the last time atmospheric 14CO2 was this low); and (3) samples with 14C > 106% modern must contain a substantial fraction of C fixed since 1957. Due to the nature of historic 14C variations, precise proportions of old, bomb-peak and post-bomb C in samples with 14C greater than current atmospheric 14CO2 cannot be determined. However, the most probable explanation for a DO14C value above 106% (the higher the value, the greater the probability) is that most of this DOC is derived from plant material formed since 1957.
 Samples were transferred directly to the NERC Radiocarbon Laboratory, Scotland, for 14C analysis. Soils were homogenised, oven-dried (60°C), combusted to CO2 in a high pressure bomb in the presence of oxygen, and converted to benzene. 14C content was determined using a Quantulus 1220 liquid scintillation counter. Water samples were filtered (0.7 μm glass fibre), acidified to pH 4 with 2M HCl and purged with helium, then neutralised to pH 7 with 1M KOH, rotary evaporated, frozen and freeze-dried. Weighed aliquots were combusted to CO2 at 900°C in vacuum sealed silica quartz tubes containing copper oxide and silver foil. The gas was converted to graphite by Fe/Zn reduction [Slota et al., 1987]. 14C content was determined by Accelerator Mass Spectrometry at the Scottish Universities Environmental Research Centre [Xu et al., 2004]. 14C results were normalised to a δ13C of −25‰ (CO2 sub-samples having been measured for 13C) and expressed as %modern [Stuiver and Polach, 1977].
 Results for the unforested land classes (Table 1, Figure 2) show clear differentiation between peaty organic soils and more mineral soils. In peats, soil solution DOC is generally younger than the peat itself, containing post-1950s C in near-surface horizons. If no pre-bomb C were present in these samples, observed 14C values would suggest an average DOC age of 10–15 years. DOC at greater depths contains older C, and the input of this older material is evident at baseflow, when peat stream DO14C is below current atmospheric 14CO2. During autumn high flow, however, stream DO14C is close to that of shallow soil waters, again suggestive of a relatively recent origin. Moderate-to-high flow measurements collected from peat streams at other times were almost identical (110.9–113.7 %modern), suggesting that a ‘modern’ DOC signal is characteristic of non-baseflow samples throughout the year (M. Billett, unpublished data, 2006). Similarly ‘modern’ DO14C dates have been observed in other studies of northern peatland rivers [Schiff et al., 1997; Palmer et al., 2001; Benner et al., 2004] and Amazonian rivers [Mayorga et al., 2005]. Exceptions occur due selective in-stream degradation [Raymond and Bauer, 2001] and water table drawdown [Schiff et al., 1997]. Lower baseflow DO14C values in the current study could be explained by either mechanism, and do confirm the presence of a proportion of old, soil-derived C in peat DOC export which (given high DOC concentrations) could represent a substantial loss pathway for soil C. However, since DO14C is well above current atmospheric 14CO2 during higher flows (conditions which account for most of the total DOC export) our results suggest that recent plant material, rather than old soil organic matter, provides the dominant source of peatland DOC loss. Furthermore, with average DOC concentrations having almost doubled since the 1980s, it is impossible to explain more than a small part of this increase through accelerated loss of old soil C; if this were the dominant source of the additional DOC, present-day DO14C levels would have to be much lower than those observed.
Table 1. 14C Isotopic Composition of Soil and Dissolved Organic Carbon, and Stream DOCa
Stream DOC, mg/l
For 14C values, bold font indicates samples containing mainly post-1950s carbon. Underlined italic font indicates samples containing mainly pre-1950s carbon. Normal font indicates samples containing a proportion of pre-1950s carbon. Stream data are presented as mean ± s.d. (no. of samples). Soil and soil-water samples were collected at one location per landscape class.
104.4 ± 0.5 (2)
112.3 ± 1.3 (6)
8.4 ± 4.0 (5)
15.3 ± 3.7 (9)
Peaty gley moorland
107.3 ± 1.0 (2)
109.7 ± 0.8 (2)
9.1 ± 3.7 (3)
25.1± 3.0 (3)
94.3 ± 12.4 (2)
105.4 ± 2.7 (2)
2.0 ± 1.0 (5)
2.7 ± 1.2 (5)
95.3 ± 3.0 (2)
100.5 ± 1.2 (3)
3.4 ± 1.7 (5)
14.5 ± 1.8 (5)
104.5 ± 3.1 (4)
111.1 ± 1.7 (3)
3.8 ± 2.6 (5)
9.5 ± 6.9 (5)
102.9 ± 2.0 (5)
109.0 ± 1.7 (6)
2.9 ± 1.5 (6)
10.3 ± 6.1 (6)
 Results for peaty gleys were similar to those for peats; deep soil solution DOC appears almost as old (∼1000 years) as the soil C at the same depth, but due to the low hydraulic conductivity of this clay-rich subsoil, the contribution of this aged DOC to runoff appears negligible. Instead, stream DO14C is almost identical to organic horizon DO14C, varies little with flow, and contains post-1950s material. Were all the observed DOC of post-bomb origin, an average age of <10 years would be obtained.
 For improved and montane grasslands, there is clear and contrasting evidence of an older DOC component. The only samples with a predominantly modern 14C signature are from surface soils, and from shallow soil solution collected under improved grassland during high-flow conditions. All other soil, soil water and stream DOC samples contain a definite proportion of pre-1955 C. At low flows, the apparent average age of DOC draining these land types is ∼350–700 years. Even at high flows, there is a large pre-1955 component to DOC export; this is clearest in streams draining improved grasslands, which are also characterised by high DOC concentrations at high flow. The total loss of older soil C as DOC from these systems is thus likely to be considerable.
 DOC concentrations in streams draining forested catchments were variable, particularly at high flows (Figure 2). This land class, as defined, encompasses a wide range of soils from brown podzols to shallow peats; results suggest that soil type, rather than vegetation, may provide the main spatial control on DOC leaching. However, the 14C signature of DOC in forest streams is reasonably consistent, suggesting a mixture of old and recent C at low flow (DO14C < current atmospheric 14CO2), and predominantly post-1950s C at high flow (DO14C > current atmospheric 14CO2). Unfortunately, DOC samples from the relatively dry forest soils were insufficient for 14C analysis, but the organic horizon itself was highly 14C-enriched, at 121.7%modern, implying a large input of litter derived from forest biomass accumulated over the course of the atmospheric 14CO2 peak. The data, derived from mature forest stands, are consistent with soil solution DO14C measurements from a Swedish spruce forest chronosequence [Karltun et al., 2005], which showed evidence of old soil C mobilisation in young forest stands, but increasing 14C with stand age due to an increasing new litter contribution. Fröberg et al.  also recorded fairly high (approximately 115–118% modern, samples collected 2000–2001) 14C levels for DOC in soil solution draining the organic horizon of a mature Swedish forest stand.
 Finally, concentrations and 14C values for samples from larger (mixed land class) rivers were all within the range of values observed for single land class tributaries, exhibiting the same pattern of increasing DOC, and transition from mostly pre-1955 to mostly post-1955 DO14C at high flows (Table 1). It seems reasonable to infer that DOM in the lower river system represents an approximately conservative mix of DOM from contributing sources. In contrast to the results of Raymond and Bauer , there is no evidence of selective in-river degradation of younger DOM resulting in an older 14C age, although it should be noted that water residence times in the Conwy are much lower than in the large American river systems they studied.
 The consistent pattern of reported soil C losses across semi-natural uplands and agricultural lowlands in England and Wales has been considered evidence that such changes are caused by climate change [Bellamy et al., 2005]. Our data are difficult to reconcile with this ‘single driver’ hypothesis. 14C data suggest, counter-intuitively, that peaty soils, containing large stores of aged C, mainly export recently fixed C. Relatively C-poor mineral soils, under moderate-to-intensive grazing, export DOC derived from older soil C. If soil C is decreasing in both peatland and farmland soils, it may be that such changes represent a single response to two or more separate drivers. In peatlands, high DO14C levels and increasing DOC concentrations could be explained by two (non-exclusive) mechanisms. First, export of ‘new’ DOC may be increasing due to rising primary production, e.g. due to elevated plant growth in response to rising nitrogen deposition or atmospheric CO2 [Freeman et al., 2004]. This process alone would not, however, account for a concurrent decrease in soil C. ‘Priming’ of microbial activity by increased plant production of labile C could accelerate soil degradation [Kuzyakov, 2002], but modern DO14C dates do not support large-scale soil destabilisation via this mechanism. Simultaneously increasing ‘new’ DOC, and decreasing soil C could, however, be explained by declining soil DOC retention. Solubility of humic substances is strongly influenced by acidity, ionic strength and aluminium concentration [e.g., Tipping and Hurley, 1988; Mulder et al., 2001], and increased DOC mobility due to recovery from acidification may have contributed to DOC increases where acid loadings have decreased [Evans et al., 2006b]. Decomposition rates are also suppressed under high acidity [Mulder et al., 2001; Sanger et al., 1994]. It is therefore possible that, during the (1960–1980) peak of soil acidification, organic matter that would normally have been lost as DOC or CO2 could instead have accumulated in the soil. Remobilisation of this organic matter during recovery from acidification could have led to both an increase in DOC loss to streams, and a decrease in soil C storage. This mechanism is supported by data from a Czech forest [Oulehle et al., 2006], which attributed a 26% reduction in organic horizon mass since 1994 to accelerated microbial activity as soil pH rose. Substantial soil pH increases (0.2–0.4 pH units) have been recorded between 1978 and 1998/9 across all environmental zones in Britain [National Expert Group on Transboundary Air Pollution, 2001; National Soils Research Institute (NSRI), 2004]. These changes are concurrent with (and indeed, for the NSRI  study, based on the same soil samples as) the losses of soil C reported by Bellamy et al. .
 We suggest that the reported C losses from agricultural soils require a different explanation. The large pre-1955 component of DOC draining managed grasslands appears consistent with loss of old C from grassland soils, although the data do not exclude the possibility that new C inputs to the soil could equal or exceed outputs of older soil C as DOC. In either case, if there are doubts as to whether recorded C losses from semi-natural ecosystems (dominated by organic soils) are linked to climate change, the existence of such a link for more intensively managed systems (dominated by mineral soils) must also be questioned. For a 0.5°C temperature rise since 1978, such a high Q10 would be required to account for reported rates of C loss that a warming-related explanation seems unlikely. The C balance of agricultural grasslands is highly dependent on management practices such as fertilisation, ploughing, liming, grazing and grass cutting [Soussana et al., 2004; Franzluebbers et al., 2000], in particular the effect of fertilisation on the balance of production and decomposition [Soussana et al., 2004]. Since grassland (and other agricultural) management has progressively intensified in the UK since the 1970s, we consider this a likelier driver for soil C depletion in the lowlands, which might also explain the apparent loss of older C in DOC. 14C data for forests, from this and other studies, also point to land-use as an important influence on C loss.
 The lowlands and uplands of Britain are, like those of many other developed countries, far from being ‘natural’ landscapes; the lowlands are affected by high-intensity agriculture, the uplands by local management of varying intensity, including afforestation and sheep grazing, and long-range factors such as atmospheric pollution. Climatic changes do not act on these landscapes in isolation, but as part of a suite of anthropogenic influences. Taking into account the inherent limitations of 14C data, we believe that our results are inconsistent with climate change as the only driver of soil C losses. In farmlands, agricultural activity may be an important cause of soil C loss. Although this could exacerbate climate change by releasing CO2, it does not represent the positive climate-change feedback previously suggested. In upland systems, apparent soil C losses and export of relatively young DOC could be explained by remobilisation of a transient C store accumulated over the course of the acidification peak. If so, with acid deposition declining and DOC increasing across much of Europe and North America, it is possible that recent C flux studies in these regions might not be representative of systems in long-term balance with climatic conditions, but of systems undergoing transient change for other reasons. This should be considered in any assessment of the causes of long-term changes in terrestrial C fluxes.
 This study was supported by a NERC studentship (NER/S/A/2002/10912); NERC Radiocarbon Dating Allocations 1017.0303, 1123.0405, 1093.1004; the EU Framework Programme 6 Eurolimpacs project (GOCE-CT-2003-505540); and the Scottish Executive Environment and Rural Affairs Department/Welsh Assembly Government (FF/03/08). We are grateful to Ed Tipping, Don Monteith, Jo Clark, and two anonymous reviewers for helpful comments and suggestions.