Interpretation and application of carbon isotope ratios in freshwater diatom silica

ABSTRACT Carbon incorporated into diatom frustule walls is protected from degradation enabling analysis for carbon isotope composition (δ13Cdiatom). This presents potential for tracing carbon cycles via a single photosynthetic host with well‐constrained ecophysiology. Improved understanding of environmental processes controlling carbon delivery and assimilation is essential to interpret changes in freshwater δ13Cdiatom. Here relationships between water chemistry and δ13Cdiatom from contemporary regional data sets are investigated. Modern diatom and water samples were collected from river catchments within England and lake sediments from across Europe. The data suggest dissolved, biogenically produced carbon supplied proportionately to catchment productivity was critical in the rivers and soft water lakes. However, dissolved carbon from calcareous geology overwhelmed the carbon signature in hard water catchments. Both results demonstrate carbon source characteristics were the most important control on δ13Cdiatom, with a greater impact than productivity. Application of these principles was made to a sediment record from Lake Tanganyika. δ13Cdiatom co‐varied with δ13Cbulk through the last glacial and Holocene. This suggests carbon supply was again dominant and exceeded authigenic demand. This first systematic evaluation of contemporary δ13Cdiatom controls demonstrates that diatoms have the potential to supply a record of carbon cycling through lake catchments from sediment records over millennial timescales.


Introduction
Stable isotope analyses of the siliceous cell walls (frustules) of diatoms provide insights into a broad range of environmental processes tracked from the perspective of a single, ecologically well-constrained organism. To date, most diatom-based stable isotope studies have focused on the stable oxygen and silicon isotope composition (d 18 O diatom and d 30 Si diatom ) of diatoms from lacustrine and marine sediments (Leng and Barker, 2006;Swann and Leng, 2009;Leng and Henderson, 2013). Changes in d 18 O diatom are used as a proxy record of water source and hydrological balance in palaeolimnology (Barker et al., 2001;Rioual et al., 2001;Shemesh et al., 2001) and global ice volume, temperature and local effects in palaeoceanography Hodell et al., 2001). d 30 Si diatom in freshwater is used to understand changes in climate, weathering and soil processes through the balance of silicon supply and demand (De La Rocha et al., 2000;Ding et al., 2004;Street-Perrott et al., 2008). Within marine environments, utilization of dissolved silica can be reconstructed through the ratio of silicic acid uptake by diatoms to initial dissolved concentrations (De La Rocha et al., 1998;Varela et al., 2004;Cardinal et al., 2005).
Diatom frustules are also a host for carbon isotopes measured on organic molecules occluded within diatom frustule walls (d 13 C diatom ). This occluded organic matter comprises proteins and long-chain polyamines (Kr€ oger and Poulsen, 2008) and represents a source of carbon potentially protected from degradation over geological timescales (Singer and Shemesh, 1995;Crosta and Shemesh, 2002). During cell formation diatoms source this carbon via photosynthetic uptake from the surrounding water body. The fraction available for photosynthesis is dissolved inorganic carbon (DIC), which diatoms take up preferentially as dissolved CO 2 or as bicarbonate under conditions of high carbon demand (Giordano et al., 2005). d 13 C diatom provides a record of changes in this carbon pool, overcoming issues of sample heterogeneity and potential diagenesis associated with investigations of stable carbon isotopes of bulk organic material (d 13 C bulk ) from sediments.
Use of d 13 C diatom as a palaeoenvironmental proxy is already well established within palaeoceanography where d 13 C diatom is usually interpreted as a record of pelagic primary productivity as discrimination against 13 C by diatoms is reduced during periods of high carbon demand (Singer and Shemesh, 1995;Crosta and Shemesh, 2002;Schneider-Mor et al., 2005). However, care is required in interpretation of d 13 C diatom records, as further biological variables with potential to impact carbon fractionation and isotope composition are yet to be fully constrained (Jacot Des Combes et al., 2008). These variables include diatom species assemblage, carbon availability and carbon source.
Two key factors that determine the degree of fractionation during photosynthetic carbon uptake are the balance between internal and external CO 2 concentrations and discrimination by the enzyme RuBisCO (Jacot Des Combes et al., 2008). The impact of these factors is compounded by species-specific 'vital effects', including cell growth rate, geometry and growth environment. For example, high cell growth rates reduce the internal to external CO 2 ratio (Hill et al., 2008). Discrimination by RuBisCO to 13 C is theoretically proportional to this ratio (Korb et al., 1996;Hill et al., 2008), and fractionation of carbon will subsequently be less in faster growing cells. Conversely, where cell geometry maximizes the surface area to volume ratio, CO 2 is more efficiently absorbed leading to a relative increase in carbon fractionation (Popp et al., 1998). Planktonic species have also been associated with lower d 13 C values compared with benthic varieties, attributed to the more turbulent growth environment of the former, which reduces the impact of boundary layer thickness on carbon uptake (France, 1995;France and Cattaneo, 1998;Wang et al., 2013).
Confinement of carbon isotopic analysis to the initial protein matrix established during cell formation represents a carbon source less likely to be affected by such speciesspecific effects. In fact, tests of the impact of different diatom species composition on d 13 C diatom from freshwater Lake Challa, Mount Kilimanjaro, were within analytical error (Hurrell et al., 2011). To reduce the risk of any vital or species-specific effects, it is recommended within the more established field of palaeoceanography to sieve to the <20-mm fraction as it is here that most diatom material is found and assemblages tend to be dominated by fewer species .
Availability of DIC is also a key determinant of carbon fractionation, as slow diffusion of dissolved CO 2 through water risks transport limitation. To prevent this, carbon concentrating mechanisms (CCMs), which manifest as active uptake of dissolved CO 2 and/or bicarbonate, are believed to take place in almost all diatoms (Giordano et al., 2005). Utilization of bicarbonate by diatoms can result in a further increase in d 13 C of the photosynthate by approximately 9‰ (Finlay, 2004). Whilst interpreting d 13 C diatom records it is therefore important to consider whether an increase in value has been enhanced by carbon transport limitation.
Of particular relevance to interpretation of freshwater d 13 C diatom is the carbon isotope composition of DIC sources (d 13 C DIC ). A far greater variety of carbon sources is found within terrestrial environments compared with the open ocean. The associated d 13 C DIC is in turn diverse, ranging from 0 to þ1‰ for carbonate bedrock and from À26 to À20‰ for soil carbon in C 3 landscapes, for example (Clark and Fritz, 1997). The expression of these different origins is then modified by mixing and fractionation during carbon phase transformation and species changes in transfer from the catchment to the water body, or because of internal aquatic processing (Finlay, 2003).
As a possible consequence of greater complexity introduced by more diverse carbon sourcing, there have been far fewer studies of d 13 C diatom in palaeolimnology compared with palaeoceanography. Interpretations reached have also been inferential rather than reconstructions grounded in, and constrained by, modern environmental data (Hern andez et al., 2011(Hern andez et al., , 2013Barker et al., 2013). For example, in Lago Chungar a, Chile, diatoms deposited over the late glacial to early Holocene period have high d 13 C diatom values (À27.5 to À22.6‰) during arid stages compared with those of wetter, humid periods (À30.3 to À25.4‰) when greater input of 13 C depleted dissolved biogenic carbon to the lake from the catchment was likely (Hern andez et al., 2011(Hern andez et al., , 2013. Similarly, a sediment core from Lake Challa, Mount Kilimanjaro, displayed positive correlation between d 13 C diatom and d 13 C bulk during dry intervals as high diatom productivity depleted the lake DIC pool. This correlation largely broke down during wetter periods and it was hypothesized that increased catchment carbon loading satisfied demand from primary productivity . These lake sediment records demonstrate an application for d 13 C diatom in tracing catchment carbon cycling, and the importance of testing the down-core changes against contemporary environmental relationships. Here we (i) explore relationships between water chemistry and d 13 C diatom from contemporary data sets to more precisely determine the environmental controls of d 13 C diatom , and (2) apply the understanding gained to assess the utility of d 13 C diatom in unravelling temporal carbon dynamics. Firstly, diatom epilithon (diatoms extracted from submerged biofilm comprising other algae, bacteria, fungi and the products they secrete) from river reaches in north-west England were sampled to provide an understanding of detailed carbon dynamics at the catchment scale. Rivers were investigated as they represent active hydrological pathways, connecting a lake with its catchment. Secondly, d 13 C diatom from sediment samples of lakes situated across central, north-west and northern Europe were analysed. These sediments were collected as part of a broader sampling campaign designed to study the relationship between methane concentrations, d 13 C of DIC and methane, and the carbon isotopic composition of aquatic invertebrate fossils and other sediment components (e.g. Schilder et al., 2015;St€ otter et al., unpublished data). The lakes are geographically dispersed, spanning different climate zones ranging from temperate to boreal and incorporating diverse catchment geology with examples of calcareous and non-calcareous lithology (Rinta et al., 2015). Thirdly, we applied the method to core material from Lake Tanganyika, East African Rift Valley. This lake has a wellestablished palaeoenvironmental history combined with deposition of sediments in anoxic conditions allowing for ideal organic carbon preservation. Methodological refinements were made to reduce and assess any impact of species and vital effects on the resulting d 13 C diatom values.

Contemporary UK river sites
Sampling of the north-west England river catchments for epilithon and spot water chemistry samples took place in May 2012 over consecutive days to minimize hydroclimate variability. Collection from riffles in 20 river reaches captured the late spring diatom bloom. Six major river catchments were targeted: the Wyre, Ribble, Lune, Derwent, Leven and Eden, situated in North Lancashire and Cumbria (Fig. 1). Land use was largely agricultural with land proximal to the river and stream sites dominated by rough and improved pasture for grazing. The geology underlying the sampling sites consisted of combinations of sedimentary rocks (sandstones, siltstones, mudstones and gritstones) with volcanic rocks at the most easterly sites.

Contemporary European lake sites
The European lake study consisted of surface sediments analogous to core tops, and spot water chemistry samples from 30 relatively small (0.3-303 ha) lakes situated in five countries; Switzerland, the Netherlands, Germany, Sweden and Finland (Fig. 2). Water samples were collected 0.5 m below the surface in late summer before the breakdown of water stratification in the autumn (Rinta et al., 2015). The Swiss, Dutch and German lakes are located in the temperate zone, whereas the Finnish and Swedish lakes are in the hemiboreal to boreal zones. The underlying lithology of the Swiss, Dutch and German lakes is dominated by Quaternary sediments and older limestones. In contrast, most of the Finnish and Swedish lakes are underlain by non-calcareous Precambrian bedrock covered by Quaternary deposits.

Lake Tanganyika sediments
The Lake Tanganyika down-core study comprised 14 sediment samples taken from Kullenberg piston core NP04-KH04-4A-1K collected in 2004 as part of the Nyanza Project (Felton et al., 2007). The pelagic zone of meromictic Lake Tanganyika is highly sensitive to catchment changes that alter the carbon and nutrient concentrations. The core was taken from the Kalya Horst, which is a structural high within the southern basin of Lake Tanganyika (Fig. 3). The coring location was situated below the oxycline, the anoxic state providing ideal conditions for organic carbon preservation. Sediments were dated to the last ca. 34 000 years by correlation to a second, directly radiocarbon-dated core (NP04-KH04-3A-1K) (Tierney et al., 2008) using 20 age/depth control points.

Pretreatment and measurement of d 13 C diatom
Successful determination of d 13 C diatom relies on the removal of all sources of both inorganic and organic carbon external to the frustule inclusions. To produce clean diatom material from the samples with variable organic carbon content we adapted the method described by Hurrell et al. (2011), which was based on that of Singer and Shemesh (1995). All samples were first passed through a 1-mm gauge sieve then heated to   70˚C in 10% HCl for 2 h to remove inorganic carbon. For the Lake Tanganyika material a sub-sample was removed at this stage for d 13 C bulk .
For the determination of d 13 C diatom , organic carbon (excluding the occluded material) was removed through oxidation by heating samples in 30% H 2 O 2 for 15 h at 70˚C and a further 2 h at 100˚C. Persistent organic carbon was eliminated through heating samples to 70˚C for 1 h in concentrated HNO 3 . Large mineral grains were separated by differential separation and discarded. Clay and silt particles with similar densities to diatoms were reduced by sieving to 20 mm. Samples with no more than 1% carbon content were considered free from contamination following Hurrell et al. (2011). Sieving to 20 mm was also completed to reduce potential influence of species effects attributed to cell size and geometry . Permanent slides were made of the processed samples and outline diatom counts were made based on 150 valves. For the contemporary samples, diatom genera were categorized by growth habitat as planktonic, benthic or colonial after Bellinger and Sigee (2010) as a further check for possible confounding vital effects.
13 C/ 12 C ratios for diatom material and bulk organic material from sediments were determined using an Elementar vario PYRO cube elemental analyser linked to an Iso-Prime100 isotope ratio mass spectrometer at Lancaster University for contemporary UK river samples and Lake Tanganyika sediments, and at Isoprime UK in Cheadle for contemporary European lake samples. Analysis was by combustion within tin capsules at 950˚C. 13 C/ 12 C ratios were corrected against VPDB using within-run analysis of standards IAEA-CH-6 (sucrose), Low Organic Content Soil Standard OAS and High Organic Content Soil Standard OAS [assuming d 13 C values of À10.45‰ (International Atomic Energy Agency, 2011), À27.46 and À26.27‰ (Elemental Microanalysis, 2011), respectively]. Data are reported in the usual delta notation; within-run replication of standard materials was <0.2‰ (1 SD, n ¼ 10). To ensure consistency between laboratories and conditions of analysis, external precision was monitored by use of a standard material analysed between all run sequences <0.2‰ (1 SD, n ¼ 164). Precision of sample analysis was <0.2‰ (1 SD, n ¼ 3). Where n ¼ 2, sample replicates did not vary by more than 0.5‰.

Water analysis
In-stream spot measurements of river site pH and electrical conductivity (EC) were taken using a WTW Multi 340i multiparameter water meter. Measurement accuracy was to 0.03 pH units and 1 mS cm À1 . Analysis of river water samples was completed at Lancaster University. Total phosphorus (TP) was measured following an acid-persulphate digest (O'Dell, 1993) using a Seal Analytical AQ2þ discrete colorimetric analyser (Seal Analytical, 2005). Total dissolved nitrogen (TDN) was measured using an Analytical Sciences Thermalox analyser (BS EN, 2003). Detection limits (standard deviation of blanks multiplied by 3) for TP and TDN analysis were 0.005 and 0.13 mg L À1 , respectively.
For determination of d 13 C of the DIC pool (d 13 C DIC ) at the river sites, 10 mL of river water was injected into 12 mL preevacuated exetainers containing 150 mL of de-gassed, concentrated phosphoric acid after Waldron et al. (2007). d 13 C values of the product CO 2 were measured at the NERC Centre for Ecology and Hydrology, Lancaster, using a GV Instruments Tracegas Pre-concentrator coupled to an IsoPrime isotope ratio mass spectrometer. The isotope ratio of the resultant CO 2 was compared with pulses of known reference CO 2 and expressed relative to VPDB. Data are reported in the usual delta notation; within-run standard replication (1 SD) was better than or equal to AE0.15‰.
Lake water spot samples and measurements were taken in the deepest part of each lake basin using a 5-litre water sampler approximately 0.5 m below the surface as described in detail by Rinta et al. (2015). pH and EC were measured in the field using a pHScan 2 and WTW LF 330 device with TetraCon conductivity measuring cell, respectively. TP, total nitrogen (TN) and d 13 C DIC were determined using laboratory methods as described by Rinta et al. (2015).

Contemporary UK river sites
At least 90% of each assemblage consisted of the same nine benthic diatom genera. Achnanthidium was present in all assemblages, generally as A. minutissimum, and was typically dominant alongside Gomphonema, Cocconeis and Cymbella species. No systematic correlation was found between d 13 C diatom and species composition in these data.
River water pH values ranged from 6.1 to 8.5 and EC from 16 to 331 mS cm À1 (Table 1). The nutrient concentrations confirmed these streams to have low-to-moderate trophic status (EA, 1998), with several TP measurements below detection and maximum TP and TDN values of 0.052 and 1.10 mg L À1 , respectively. This is consistent with low-intensity farming practices that dominated the sampled area of northwest England.

Contemporary European lake sites
The lake sediment diatom assemblages comprise planktonic life forms alongside benthic and colonial examples making them more diverse than the river assemblages. Dominant genera included planktonic Cyclotella, Aulacoseira and Stephanodiscus species. Despite this diversity, no systematic correlation was found between d 13 C diatom , the species data or the proportions of different life forms.
In comparison with the UK river study set, larger ranges in water chemistry values were measured in the lake waters. The pH values ranged between 5.4 and 8.9 and EC values from 24 to 462 mS cm À1 (Table 2). Trophic conditions varied from ultra-oligotrophic to hypertrophic (OECD, 1982), reflecting a wide range of nutrients with maximum TP and TN values of 0.12 and 2.30 mg L À1 , respectively. d 13 C DIC values varied from À23.6 to À2.7‰, and two distinct clusters were observed: the Swiss, Dutch and German lakes had d 13 C DIC > À10‰ (group 1), and the Swedish and Finnish lakes d 13 C DIC < À10‰ (group 2). These groupings corresponded to differences in catchment lithology, with lakes situated in hard water catchments containing calcareous bedrock (group 1) associated with d 13 C DIC > À10‰.
The range in d 13 C diatom values from À33.4 to À25.4‰ is lower than the range in d 13 C DIC . Statistical comparison using a Mann-Whitney test (IBM SPSS) found the d 13 C diatom values of each group of lakes differed significantly (U ¼ 42.5, z ¼ À2.85, P < 0.01, r ¼ À0.52). Generally, lakes with calcareous J. Quaternary Sci., Vol. 31(4) 300-309 (2016) catchments (group 1) had more positive d 13 C diatom (median: À26.7‰) compared with lakes in group 2 situated in noncalcareous catchments (median: À27.8‰). In addition a smaller range in values of 2.3‰ was present in group 1 compared with 7.3‰ in group 2. As seen within the UK river data, a significant positive relationship was present between European lake d 13 C DIC and d 13 C diatom (r s ¼ 0.59, P < 0.01) (Fig. 6). As was found for d 13 C DIC , two groupings of d 13 C diatom values emerged, with the Swiss, Dutch and German data (group 1) forming a cluster of higher isotope values, and the Swedish and Finnish data points (group 2) spread along a linear gradient of lower isotope values. There was a strong relationship between d 13 C DIC and d 13 C diatom in group 2 (r s ¼ 0.63, P < 0.01) but not between the equivalent values for lakes in group 1 (r s ¼ À0.25, P ¼ 0.41). No significant relationships were identified between d 13 C diatom and either TP (r s ¼ 0.13, P ¼ 0.49) or TN (r s ¼ 0.23, P ¼ 0.22) concentrations. Diatom communities reconstructed from the sediments were dominated by planktonic taxa including Cyclotella, Aulacoseira and Stephanodiscus species. d 13 C diatom varied from À30.0 to À22.4‰, equating to a range of 7.6‰ (Fig. 7a). In comparison, d 13 C bulk values were higher with a range of 7.3‰ from À28.2 to À20.9‰ (Fig. 7b). The offset between the two data sets varied from 0.4 to 4.2‰ with a median value of 2.7‰ (Fig. 7c). A significant positive relationship was present between d 13 C diatom and d 13 C bulk (r s ¼ 0.73, P < 0.01). At this coarse millennial scale, lowest values for both d 13 C diatom and d 13 C bulk occurred between 14.8 and 5.5 ka, the period broadly recognized as the African Humid Period (deMenocal et al., 2000) (light shading in Fig. 7). Conversely, both isotope proxies had their maximum values in sediments dating to the end of the last glacial period and again in the late Holocene. Here also the smallest offset between the two records was measured. The corresponding trends in d 13 C diatom and d 13 C bulk closely track that of higher plant leaf waxes (d 13 C wax ) (Fig. 7d), a terrestrial vegetation change proxy extracted from Lake Tanganyika sediments by Tierney et al. (2010).

Discussion
Previous studies of freshwater environments have identified (i) d 13 C of carbon sources, (ii) the relative contributions of these sources and (iii) 13 C enrichment by preferential 12 C uptake by lake primary producers (including diatoms) as key variables determining the d 13 C DIC available to diatoms for assimilation as d 13 C diatom . Investigation into environmental controls of contemporary d 13 C diatom over the different spatial and temporal scales reported here pinpoints the relative influence of carbon supply and demand factors, aiding the interpretation of palaeolimnological records as well as contemporary carbon cycling.

Translation of the carbon cycling history
Positive relationships identified between d 13 C DIC from waters and contemporary d 13 C diatom represent the primary control of catchment carbon source on the carbon isotopes in the diatom frustules. Principal carbon sources within a freshwater catchment have both geological and biotic origins and are associated with differing carbon isotopic ranges. Weathering of calcareous rock releases bicarbonate, which has a relatively high d 13 C value of 0 to þ1‰ (Clark and Fritz, 1997). In contrast, CO 2 released into soils from plant root respiration and vegetation decay has a lower d 13 C value, ranging from À26 to À20‰ in C 3 and À12 to À6‰ in C 4 landscapes, respectively (Mook et al., 1974;O'Leary, 1988). Oxidation of methane, associated with d 13 C values between -80‰ and -50‰, provides a carbon source that is even further 13 C depleted (Whiticar, 1999). These catchment carbon signatures are transported via infiltrating flows to freshwater bodies where they mix with dissolved carbon from autochthonous sources, including macrophytes with recorded d 13 C values of À50 to À11‰ (Keeley and Sandquist, 1992) and phytoplankton with bulk values ranging between À42 and À26‰ (Leng and Marshall, 2004). Oxidation of such materials during decomposition enables further release of 13 C depleted CO 2 , but if waters are stratified this process is slowed and the potential biotic carbon source is stored within anoxic sediments (Leng and Marshall, 2004).
The resulting d 13 C DIC signature of a water body is further impacted by atmospheric exchange leading to the preferential loss of 12 C. In addition, primary productivity (including diatoms) results in discriminatory uptake of 12 C. The pH of a water body is also significant as it determines the proportioning of different DIC species; each of which has a contrasting carbon isotope signature. At pH 8 the percentage of DIC present as dissolved CO 2 is close to 0, and the hydration and disassociation of dissolved CO 2 into bicarbonate causes an increase in d 13 C of approximately 9‰ (Clark and Fritz, 1997).
d 13 C DIC is therefore a record of catchment carbon cycling history reflecting sourcing and further changes to the isotope value related to fractionation during carbon phase and species changes. Final translation of the d 13 C DIC signature to d 13 C diatom is dependent on the photosynthetic pathway used, and any species-specific or vital effects that determine the degree of fractionation on uptake. The positive correlation identified between d 13 C DIC and d 13 C diatom in the UK rivers (Fig. 4) and European lakes (Fig. 6) suggests that even in sites of highly varying environmental characteristics, d 13 C DIC is a significant control in the determination of d 13 C diatom .

The role of catchment productivity
Within the contemporary UK rivers sampled, DIC probably had a biotic origin as no significant areas of calcareous geology were present in any of the catchments. The negative relationship identified between d 13 C DIC and TDN concentrations (Fig. 5) supports this conclusion and indicated that river DIC pool characteristics were probably controlled by catchment productivity at this scale. This relationship showed that Figure 7. d 13 C values from Lake Tanganyika sediments plotted against sample age. d 13 C diatom (a), d 13 C bulk values (b) and the offset between the two records (c) were determined for Lake Tanganyika sediment core NP04-KH04-4A-1K. d 13 C wax data are re-plotted from Tierney et al. (2010). Pale grey shading represents the African Humid Period as defined by deMenocal et al., (2000). Dark grey shading represents the likely peak in arid conditions experienced in East Africa during the last glacial (Barker and Gasse, 2003 (2016) dissolved biotic carbon supply was enhanced within more productive catchments and aquatic primary productivity did not lead to relative enrichment in 13 C. This agrees with the findings of Maberly et al. (2013) who found a greater availability of DIC associated with more productive catchments in the English Lake District. Increased loadings of dissolved biogenic carbon with a relatively low d 13 C signature were attributed to catchment land use, with greater availability of nutrients resulting in enhanced dissolved carbon release within the catchment. The absence of a correlation between d 13 C diatom and species composition suggests sieving to control for species-specific and vital effects was successful, or at least limited any significant impact on diatom isotopic value. In addition d 13 C diatom values were consistently lower than d 13 C DIC and fell within a small range of 3.6‰. This suggests possible uptake of bicarbonate via CCMs has not had a pronounced effect on the resulting diatom carbon values. These findings show that d 13 C diatom can be used to investigate catchment productivity, highlighting the close coupling between a water body and its catchment conceptualized as the balance between carbon supply and demand. For specific catchments the strength of this relationship is dependent on the multiple and interacting controls on d 13 C DIC in lake waters and its translation to the diatoms.

The role of catchment geology
In comparison with the UK river epilithon, the European lake sediments represented a potentially more integrated temporal record of carbon cycling, with seasonal variability masked by sediment accumulation during several annual cycles. In addition, the much broader geographical range was clearly manifest in the heterogeneity of water chemistry variables and diatom assemblage compared with that of the rivers. Most striking was the identification of two groupings in d 13 C diatom values, which coincided with both d 13 C DIC and major geological differences in catchment carbon source characteristics.
At the continental scale of analysis, no relationships were identified between d 13 C diatom and productivity indicators TP or TN, either in the complete data set or within sub-groups. It is likely that highly varied catchment carbon processing attributed to the diverse climate, land use and geology represented obscured any record of pelagic carbon demand differences between the lakes. As seen with the UK rivers, the impact of species effects on d 13 C diatom appears to have been limited successfully through sieving to <20 mm. Instead, DIC characteristics appear to be the principal environmental controls of d 13 C diatom at this scale. Higher d 13 C diatom signatures (À27.7 to À25.4‰) in lakes with calcareous catchments (group 1) compared with those without (À33.4 to À26.1‰) (group 2) reflected the contribution of carbonate geology (associated with d 13 C between 0 and þ1‰) to respective lake carbon pools. This demonstrated the significance of DIC sourcing to the production of an initial carbon signature, which is transferred to d 13 C diatom . Also influential was the relative availability of DIC for uptake by diatoms. The absence of a relationship between d 13 C DIC and d 13 C diatom in group 1 is probably a result of high background levels of geologically sourced dissolved carbon. In addition, the influence of geological sourcing on d 13 C diatom in group 1 may have been compounded by likely enhanced uptake of bicarbonate via CCMs due to the near 0% contribution of CO 2 to DIC at pH values over 8. It is only in the absence of significant geological carbon sources (group 2) where the transfer of a dissolved biogenic carbon signature reflecting catchment productivity can be determined in d 13 C diatom .

Interpretation of palaeoenvironmental records
Advancement of the findings by Maberly et al. (2013) concerning lake catchment productivity to include rivers has important implications for interpretation of palaeoenvironmental records. The close coupling between freshwater networks and catchment carbon cycling, and in particular the relative availability of dissolved biogenic carbon in response to land use, has been clearly demonstrated. Analysis of d 13 C diatom from highly varied lake sites demonstrated the difficulties associated with developing a universal model of catchment and water productivity relationships. Nevertheless, successful extraction of d 13 C diatom from lake sediments highlights the potential for obtaining palaeoenvironmental archives of changes in catchment carbon cycles from lake sediment cores. In particular, the fundamental principal of a carbon supply and demand balance can be applied to the interpretation of freshwater d 13 C diatom extracted from lakes situated in contrasting environmental and climatic settings.
The sediments from Lake Tanganyika provided an opportunity to test these conclusions on a lake with a wellestablished palaeoenvironmental history (e.g. Gasse et al., 1989;Scholz et al., 2003;Talbot et al., 2006) that would be expected to respond to changes in carbon cycling at the landscape scale. Within the Lake Tanganyika sediments there is a close coupling between d 13 C diatom and d 13 C bulk throughout the 34 000-year record with an offset no greater than 4.2‰ (Fig. 7). Diatom values were consistently lower than bulk carbon, suggesting diatoms were using the lighter isotope from dissolved carbon inputs. Using the conceptual relationships developed above, these isotope changes are thought to indicate that the lake carbon pool principally reflected changes in the quantity and nature of carbon supplied from the catchment with modifications by lake primary productivity as a secondary factor. This finding is attributed to the great size of Lake Tanganyika and its catchment where, particularly during wet periods, dissolved and particulate biogenic carbon produced in the catchment would have significantly contributed to the lake carbon pool.
Even during dry intervals of the last glacial period (Barker and Gasse, 2003) and the late Holocene (Haberyan and Hecky, 1987), signified by high d 13 C signatures in both d 13 C diatom (À24.7 to À22.4‰) and d 13 C bulk (À21.6 to À20.9‰), maintenance of a correlation and a constant offset indicates primary productivity did not significantly deplete the carbon pool. This is despite a probable decrease in carbon delivery from the catchment and potential enhancement of lake mixing processes leading to nutrient recycling (Scholz et al., 2003). In addition to variability in carbon loading, the coinciding measurement of high d 13 C signatures for diatoms and bulk sediments suggests a change in carbon source. Corresponding high d 13 C wax values (À29.0 to À26.4‰) at this time (Fig. 7d) indicate increased prevalence of C 4 -dominated savanna grassland within the Lake Tanganyika catchment (Tierney et al., 2010). Because of these ecosystem changes, greater contributions of dissolved carbon with higher d 13 C entered the lake during dry periods, and were thus translated into higher d 13 C diatom and d 13 C bulk values. The discovery of pervasive terrestrial supply domination over aquatic demand suggests large lakes are likely to have been substantial carbon sources to the atmosphere over centennial to millennial timescales. This contrasts with smaller lakes such as Lake Challa on the flank of Kilimanjaro where diatom and bulk carbon isotope records became periodically decoupled by enhanced in-lake productivity . Copyright  The Lake Tanganyika study demonstrates that comparison of d 13 C diatom with d 13 C bulk extracted from sediment cores enables catchment and lake carbon cycles to be disentangled, overcoming inherent ambiguities in the interpretation of bulk d 13 C. If geological carbon sources of lakes can be assumed to be constant and modifications to the soil carbon pool from vegetation changes can be understood, individual site histories can be reconstructed. Lake sediment d 13 C diatom therefore represents a largely under-exploited resource with the potential to provide highly insightful carbon cycling chronologies over millennial timescales.

Conclusions
Stable isotope analysis of diatom organic molecules, occluded in silica, constrains uncertainties associated with measurements of undifferentiated sedimentary carbon. The occluded organic matter also provides a carbon archive largely protected from degradation, oxidation and diagenesis. The application of the d 13 C diatom method in freshwaters requires adjustments to the standard inferences concerning pelagic productivity established by early marine studies. In freshwaters the controls of d 13 C diatom are more complex due to the high degree of connectivity between terrestrial vegetation, soils, bedrock and aquatic ecosystems. Within these environments carbon supply-side characteristics, including relative abundance of DIC from differing sources and associated d 13 C DIC , are important controls that change as a function of catchment characteristics at various spatial scales.
The concept of inorganic carbon supply and demand offers a useful framework through which to develop the interpretation of d 13 C diatom . Analysis of contemporary d 13 C diatom from UK river epilithon demonstrates the close linkages between carbon cycling in freshwater networks and their catchments. In support of findings by Maberly et al. (2013), more productive catchments are associated with greater availability of dissolved biogenic carbon. Investigation of contemporary European lake surface sediments confirms palaeolimnological inferences concerning catchment control of lake carbon supply made by Barker et al. (2013) and Hern andez et al. (2011Hern andez et al. ( , 2013. Consequently variations within a single site may be readily interpreted in terms of land use as the lithological template is held constant. It is presently not possible to produce a globally relevant quantitative relationship between d 13 C diatom and specific environmental variables. However, as demonstrated by the Lake Tanganyika study, great potential lies in the use of d 13 C diatom to inform interpretation of lake sediment records. Palaeoenvironmental interpretation could be further enhanced by modelling the transfer of carbon through specific catchments. Of particular significance is improved understanding of dissolved carbon cycling from diatom frustules, independent of particulate carbon compositional changes normally associated with lake sediment d 13 C bulk analysis, to evaluate changes in freshwater ecosystems and palaeoenvironments.