Biogeosciences Distribution , origin and cycling of carbon in the Tana River ( Kenya ) : a dry season basin-scale survey from headwaters to the delta

The Tana River basin (TRB) is the largest in Kenya (∼120 000 km2). We conducted a survey during the dry season throughout the TRB, analyzing a broad suite of biogeochemical parameters. Biogeochemical signatures in headwater streams were highly variable. Along the middle and lower river course, total suspended matter (TSM) concentrations increased more than 30-fold despite the absence of tributary inputs, indicating important resuspension events of internally stored sediment. These resuspended sediment inputs were characterized by a lower and 14C-depleted OC content, suggesting selective degradation of more recent material during sediment retention. Masinga Dam (a large reservoir on the upper river) induced a strong nutrient retention (∼50% for inorganic N,∼72% for inorganic phosphate, and∼40% for dissolved silicate). Moreover, while DOC pools andδ13C signatures were similar above, in and below the reservoir, the POC pool in Masinga surface waters was dominated by 13C-depleted phytoplankton, which contributed to the riverine POC pool immediately below the dam, but rapidly disappeared further downstream, suggesting rapid remineralization of this labile C pool in the river system. Despite the generally high turbidity, the combination of relatively high oxygen saturation levels, low δ18O signatures of dissolved O2 (all <+24.2‰), and the relatively lowpCO2 Correspondence to: S. Bouillon (steven.bouillon@ees.kuleuven.be) values suggest that in-stream primary production was significant, even though pigment data suggest that phytoplankton makes only a minor contribution to the total POC pool in the Tana River.

as biogeochemical "hotspots" on the land-ocean interface (McClain et al., 2003). Due to the overwhelming evidence for a high degree of biogeochemical processing of organic matter in freshwater systems (e.g., Wollheim et al., 2006;Cole et al., 2007;Battin et al., 2008), the original view of rivers as mere inactive conduits for organic matter and nutrients has thus significantly evolved. There is evidence that millennia-old organic 10 matter from soils can undergo a surprisingly high degree of remineralization on a time scale of weeks after entering the aquatic system (Raymond and Bauer, 2001; see also Cole and Caraco, 2001), even though a smaller pool of more recent material may dominate overall remineralization (Mayorga et al., 2005;Holmes et al., 2008). Considering the various organic matter inputs in rivers, and large differences in the potential for car- 15 bon processing and exchange (e.g. from temperate to tropical environments, presence of flood plains, etc.), carbon cycling in rivers is in reality much more complex than the "pipeline vs. reactor" view (see Cole et al., 2007) on river carbon cycling suggests: various fractions of dissolved and particulate organic carbon (DOC and POC) are likely to have different reactivities and thus, may be modified or remineralized very differently 20 (Mayorga et al., 2005;del Giorgio and Pace, 2008). Key factors in determining the overall degree of organic matter processing in river networks are the presence and extent of hydrological retention and storage events or zones such as floodplains, deposition/resuspension of suspended sediment (Meybeck and Vörösmarty, 2005;Battin et al., 2008). Although the important role of freshwater systems in carbon cycling is of Mount Kenya, 5 rivers were sampled roughly along the road between Meru and Embu (Thingithu, Mara, Nithi, Rugiti, and Thuchi River). These were sampled at an altitude between 1350 and 1600 m, a zone which is largely farmland. Two perennial rivers (Mutundu River and Rojewero River) originating in the Nyambene Hills were sampled in Meru National Park (NP), close to their confluence with the Tana River at an 10 altitude of ∼600 m. Vegetation in Meru NP varies from open to closed-canopy savanna grasslands, with narrow bands of riverine forests along the perennial tributaries and along the Tana River. The main Tana River was sampled at 10 stations ranging in altitude from 1110 m to ∼10 m above sea level, encompassing an overall transect of >1000 km. The most upstream sampling stations along the main Tana River were 15 located ∼5 km upstream and ∼500 m downstream of Masinga Reservoir, with further sampling locations distributed along its further course until close to the river delta. The most downstream sampling station was located near Chalaluma on the Matomba branch of the Tana River, the Tana River braches out in this section but Matomba had the majority of discharge during the period of sampling (O. Hamerlynck and S. Duvail, 20 personal communication, 2008). Below the Tana River sampling point on the border between Meru and Kora NP (i.e. over a length of ∼650 km), no tributaries delivered water to the main river during the period sampled. 4 surface water samples were taken along a transect on the Masinga Reservoir (1050 m altitude) from near the dam outflow to approximately the middle of the reser- 25 voir. Masinga is the largest of the reservoirs on Tana River (∼120 km 2 ) and has an estimated mean water residence time of ∼3 months, suggesting a potentially large impact on the sediment, carbon and nutrient transport further downstream. The sediment influx has been found to be much larger than anticipated and a significant part of the original storage capacity has been lost, with an estimated annual sediment deposition rate of ∼10-15 10 6 m yr −1 (Dunne and Ongweny, 1976;Walling, 1984;Mutua et al., 2005). According to Schneider (2000), the sediment trapping efficiency of the dam ranges between 75 and 98%, with most of the deposition occurring along the thalweg and little on the reservoir terraces (Saenyi, 2003). The suspended sediment load of 5 the Tana River has considerably decreased since the dam construction (Kitheka et al., 2005), although there is evidence that this decrease in sediment transport already set in earlier (Dunne, 1977). Moreover, the flow regulation has led to major changes to the downstream river ecology, including a reduction in flood events, a reduction in the river meandering rate, and a reduction in riverine forests (see Maingi and Marsh, 2002).

Sampling and analytical techniques
Surface water for field measurements of dissolved O 2 , pH, temperature, and salinity were taken with a 1.7 L Niskin bottle from ∼0.5 m below the surface. On the main Tana River, these were taken from bridges in the middle of the river wherever possible. Oxygen saturation level (%O 2 ) was measured immediately after collection with a po- 15 larographic electrode (WTW Oxi-340) calibrated on saturated air, with an accuracy of ±1%. pH was measured using a combined electrode (Metrohm) calibrated on the NBS (US National Bureau of Standards) scale, as described by Frankignoulle and Borges (2001), with a reproducibility of ±0.005 pH units. Samples for determination of total alkalinity (TA) were obtained by pre-filtering 100 mL of water through precombusted 20 Whatman GF/F filters, followed by filtration through 0.2 µm Acrodisc syringe filters, and were stored in HDPE bottles until analysis by automated electro-titration on 50 mL samples with 0.1 mol L −1 HCl as titrant (reproducibility estimated at ±2 µmol kg −1 ). The partial pressure of CO 2 (pCO 2 ) and total dissolved inorganic carbon (DIC) concentrations were computed from pH and TA measurements with the thermodynamic constants de- 25 scribed in Frankignoulle and Borges (2001 Niskin bottle by gently overfilling 12 mL glass headspace vials, and poisoned with 20 µL of a saturated HgCl 2 solution. For the analysis of δ 13 C DIC a He headspace was created, and ∼300 µL of H 3 PO 4 was added to convert all inorganic carbon species to CO 2 . After overnight equilibration, part of the headspace was injected into the He stream of an elemental analyser -isotope ratio mass spectrometer (EA-IRMS, Ther-5 moFinnigan Flash1112 and ThermoFinnigan Delta+XL) for δ 13 C measurements. The obtained δ 13 C data were corrected for the isotopic equilibration between gaseous and dissolved CO 2 as described in Gillikin and Bouillon (2007 Gillikin and Bouillon (2007). Samples for CH 4 were collected directly from the Niskin bottle in 40 mL headspace 15 vials, poisoned with HgCl 2 , and closed with a rubber septum and aluminum cap. CH 4 concentrations were determined by gas chromatography, after creating a headspace with N 2 , as described in Abril and Iversen (2002). Dissolved CH 4 concentrations were calculated using the solubility coefficient of Yamamoto et al. (1976). Samples for ammonium, nitrate, phosphate, and silicate were similarly obtained, preserved with HgCl 2

20
(1 µL mL −1 sample), and analysed with automated colorimetric techniques. A subset of samples for SiO 2 were prepared by filtration on 0.45 µm membrane filters to check for possible artifacts caused by the pre-filtration procedure with glassfibre filters, but no deviations in concentrations were found. Samples for TSM were taken with sampling bottles at ∼0.5 m below the water sur-Introduction These filters were later decarbonated with HCl fumes under partial vacuum for 4 h, re-dried and packed in Ag cups. POC, PN, and δ 13 C POC were determined on the EA-IRMS using the TCD signal of the EA to quantify POC and PN, and by monitoring m/z 44, 45, and 46 on the IRMS. Acetanilide was used as a standard for POC and PN, while sucrose (IAEA-C6) was used to calibrate the δ 13 C POC data. Reproducibility of 5 δ 13 C POC measurements was better than 0.2‰. Samples for DOC and δ 13 C DOC were filtered as described above for nutrients, 40 mL of filtrate was preserved in glass vials with teflon-coated screw caps, by addition of 50-100 µL of H 3 PO 4 . DOC and δ 13 C DOC were measured with an OI-1010 TOC analyser coupled to a Thermo DeltaPlus IRMS (see St-Jean, 2003;Osburn and St-Jean, 2007). Typical reproducibility was in the order 10 of <5% for DOC, and ±0.2‰ for δ 13 C DOC . To obtain sufficient amounts of suspended material for surface area measurements and for δ 13 C analysis of phospholipid fatty acids (PLFA), 2-5 L of surface water was taken and pressure-filtered on 140 mm membrane filters (0.45 µm) on some of the sampling sites on the main Tana River. These filters were wrapped up in cryotubes and immediately stored on liquid N 2 until further 15 processing in the laboratory. Samples of surface soils (top 0-5 cm layer) were taken at all samples sites (except for Masinga Dam), slightly upstream from the water sampling site and ∼10 m from the river bank. Where possible, river sediments were sampled with a Van Veen grab (or directly in sample tubes for the shallow highland rivers). At some of the Tana River 20 sampling stations, however, no sediment material could be obtained due to the presence of stones or rocks deposited below the bridges from which we sampled. Soil and sediment samples were transferred to cryotubes and stored in liquid N 2 . All soil and sediment samples were freeze-dried and homogenized upon return in the laboratory, and analysed for bulk organic carbon content, particulate nitrogen, and δ 13 C of or-25 ganic matter after acidification in silver cups. Soils and sediments samples were also analyzed for δ 13 C of PLFA as described below for suspended matter samples. Surface area measurements (see below) for soils and selected sediment samples were performed on <63 µm fractions. samples were acidified on the vacuum line using H 3 PO 4 and the resulting CO 2 was similarly purified and processed. All results have been corrected for isotopic fractionation according to the conventions of Stuiver and Polach (1977). Surface area (SA) measurements of soils, sediments and suspended matter were made on 200-600 mg freeze-dried and homogenized samples, using multi-point BET 10 (Braun-Emmet-Teller) adsorption isotherms. Measurements were made using a Quantachrome NOVA 3000 surface area analyser, and verified with BCR-173 (Institute for Reference Materials and Measurements). While organic matter is frequently removed prior to SA measurements, the data in Mayer (1994) indicate that this does not systematically affect SA data -hence, no further sample pretreatment was performed. 15 Extraction and derivatization of PLFA was performed using a modified Bligh and Dyer extraction, silica gel partitioning, and mild alkaline transmethylation as described in Bouillon and Boschker (2006). δ 13 C values of the resulting FAMEs (fatty acid methyl esters) were determined on a HP6890 GC coupled to a ThermoFinnigan delta+XL via a GC/C III interface. All samples were run in splitless mode, using a HP-5 col-20 umn (30 m, 0.32 mm ID) and a He flow rate of 2 mL/min. The δ 13 C values of PLFAs are corrected for the addition of the methyl group by simple mass balance. Reproducibility of the PLFA δ 13 C analyses is estimated to be ±0.6‰ for PLFA from natural samples. For consistency with other datasets on bacterial PLFA δ 13 C (e.g., Bouillon and Boschker, 2006), we report here the δ 13 C values for the bacterial markers i15:0 25 and a15:0, which were obtained by manually integrating these 2 adjacent peaks on the IRMS chromatograms. Samples for pigment analysis by HPLC were obtained by filtering a known volume of surface water on glass fibre filters (0.7 µm, Whatman GF/F), immediately rolled 5970 Introduction

Conclusions
References Tables  Figures   Back  Close Full Screen / Esc

Printer-friendly Version
Interactive Discussion up in cryotubes and stored in liquid N 2 . Upon return to the home laboratory, these were stored at −80 • C until analysis. Pigments were extracted in 10 mL acetone:water (90:10), and a subsample separated by HPLC on a C18 reverse phase column. Calibration was performed with working standards prepared from commercially available pure compounds.

General physicochemical characteristics and nutrient concentrations
• C, with a clear altitudinal gradient (Table 1) concentrations showed a wide range in the different tributaries (1.9-64.9 µM) but were fairly stable along the main Tana River (17.4±2.9 µM), the only marked pattern being a reduction of >50% between pre-and post-Masinga Dam (Table 1). Surface waters of Masinga Reservoir were extremely depleted in both NO − 3 (0.59±0.14 µM) and NH + 4 (from 0.13 µM at the most upstream location to undetectable 5 levels near the reservoir outlet). As observed for nitrate, phosphate variations were highly variable in headwater streams (0.13-4.60 µM) and highly depleted in surface waters of Masinga Reservoir (0.13±0.1 µM). Along the main Tana River, however, phosphate concentrations showed a very distinct (more than 20-fold) and consistent downstream increase, from 0.14 µM below Masinga to >3 µM in the downstream part 10 of the river ( Table 1). As for NO − 3 and phosphate, silicate concentrations varied widely in headwater streams (164-959 µM, with highest concentrations in streams draining the Nyambene Hills). Along the main Tana River, a major reduction was observed between pre-and post-Masinga Reservoir (from 364 to 215 µM, i.e. a decrease of 41%), and further downstream, silicate concentrations increased gradually but were stable at 15 310±4 µM in the lower part of the river (below 150 m). In surface waters of Masinga Reservoir, silicate concentrations (208±1 µM) were markedly lower than in the inflowing water (364 µM). Mt. Kenya streams, but were markedly lower in the (high-DIC) streams draining the Nyambene hills (−11.5 to −11.2‰, Fig. 3). δ 13 C DIC along the main Tana River ranged between −7.8‰ above Masinga Reservoir to −10.2‰. δ 13 C DIC was significantly higher in the surface waters of Masinga Reservoir, ranging between −4.7 and −4.4‰ (Fig. 3). ∆ 14 C DIC were markedly more 14 C-depleted compared to ∆ 14 C POC : one sample from 5 Thuchi River (Mt. Kenya) had a ∆ 14 C DIC signature of −658‰ (i.e. ∼8600 yr), along the main Tana River the 5 available ∆ 14 C DIC data range from +178‰ (i.e., modern) upstream of Masinga Reservoir to −552‰ (i.e. ∼6400 yr) at the most downstream sampling site, with an overall decrease in 14 C ( Table 2). pCO 2 in headwater streams ranged between 110 and 1017 ppm and showed no 10 correlation with any of the other measured parameters (Fig. 4a). For the rivers on Mt. Kenya, there was a clear decreasing pattern in pCO 2 along the E-W gradient. pCO 2 in surface waters of Masinga Reservoir were slightly under-or oversaturated (313-443 ppm) and decreased towards the downstream part of the reservoir. Along the Tana River, pCO 2 was highest in the upper reaches, i.e. above (1085) and below Masinga

Total
Reservoir (3570 and 2170 ppm). Further downstream, pCO 2 was relatively low and ranged typically between 500 and 660 ppm, increasing again in the delta (1543 ppm). The latter would be consistent with earlier dry season data from the freshwater tidal estuary (4390±660 ppm, see Bouillon et al., 2007).

Dissolved oxygen and δ 18 O-O 2 20
Oxygen saturation levels (%O 2 ) were generally close to or above saturation levels in most of the headwater streams (99-118%), above saturation in surface waters of Masinga Reservoir (108-114%), and highly variable in the Tana River (41-102%, Fig. 5). In the latter, lowest %O 2 was observed just below Masinga Reservoir, which likely reflects the outflow of anoxic bottom waters from the reservoir (the outlet of the Introduction Tana River, oxygen levels were slightly undersaturated, but decreased markedly in the lowest sampling site in the delta (74%), consistent with the low %O 2 values recorded earlier in the freshwater end-member of the Tana estuary (64±5%, Bouillon et al., 2007). δ 18 O-O 2 values were consistently lower than the value expected for equilibrium with atmospheric O 2 (i.e., +24.2‰), and ranged between +19.0 and +23.0‰ (Fig. 5 (Fig. 5).

Methane
CH 4 concentrations ( Fig. 4b) in headwater streams ranged typically between 25 and 92 nM, except for Muringato River (314 nM) and Mutundu River (410 nM). In the main 15 Tana River, CH 4 concentrations were relatively high upstream of Masinga Reservoir (505 nM) but markedly lower below the reservoir (372 nM). In the surface waters of the Masinga Reservoir itself, CH 4 concentrations were the lowest observed in this study (51±7 nM). Along the middle and lower course, CH 4 concentrations ranged between 54 and 387 nM, with the highest concentration observed in the Tana Delta (Fig. 4b). 20 The observed CH 4 concentrations consistently represent high levels of oversaturation, ranging from 850 to 21,700%.

Carbon pools in surface soils and river sediments
The soil δ 13 C data encompass the full range of δ 13 C signatures expected for either C3or C4-dominated soils (−28.5 to −13.2‰, see Table 3). The most negative δ 13 C was 25 observed in the riverine forests of the Tana Table 3). Riverine sediment %OC decreased from values between 3 and 5% at >2000 m to very low values in the lower Tana River (generally <0.2%), with the exception of sediments in the Tana delta (Table 4). Elemental ratios (TOC/TN, atom) ranged between 6.3 and 13.4 for soils, and between 8.0 and 16.8 for riverine sediments (Tables 3, 4). TOC/TN ratios showed a decreasing pattern with altitude in suspended matter and riverine sediments, but this was less pronounced 15 for soils (Fig. 6b).

Suspended matter and aquatic organic carbon pools
TSM concentrations were generally low in the perennial headwater streams ( Interactive Discussion high %POC/TSM (32.6 ±3.6%) diverge from this pattern ( Fig. 8) because particulate matter there was predominantly composed of phytoplankton (see below). The contribution of POC to suspended sediment pool (%POC/TSM) was consistently higher than in corresponding bottom sediments, and higher than in soils in all but three stations in the lower river (Tables 1, 2 The specific surface areas (SA) measured on 5 of the Tana River suspended matter samples from the lower river are very high (63.6-82.2 m 2 g −1 ) compared to values 15 reported for coastal sediments, estuaries and rivers (typically <50 m 2 g −1 ), and OC:SA ratios found here (0.16-0.22 mg C m −2 ) are consequently in the lower range of those reported in the literature (e.g., Keil et al., 1997;Aufdenkampe et al., 2007). The SA values in suspended matter are also consistently higher than those observed in surface soils which ranged between 11.8 and 67.7 m 2 g −1 (Table 3).  Masinga Reservoir, with an average δ 13 C of −23.9±0.3‰ (Fig. 10b). Overall, δ 13 C DOC and δ 13 C POC were clearly uncoupled, showing a relatively weak relationship (Fig. 10c). ∆ 14 C data are available for a selected number of POC samples from the main Tana River and ranged between −116 and −35‰, corresponding to an estimated age of ∼935 and 230 years. ∆ 14 C POC showed a general decrease along the lower Tana River, 5 which coincides with the overall decrease in %POC/TSM in suspended matter (Fig. 11).

Pigment concentrations
Chl-a concentrations averaged 0.91±0.21 µg L −1 in the streams draining Mt. Kenya, were about twice as high (1.75±0.08) in the streams draining the Nyambene Hills, and increased from ∼0.80 µg L −1 in the upper Tana River to values between 5.23 and 10 6.96 µg L −1 in the lower Tana River (Fig. 12a, note that no pigment data are available for the streams in the Aberdare range). In Masinga Reservoir, chl-a concentrations were high (3.23-5.08 µg L −1 ) compared to the Tana River before and after the reservoir. The resulting POC/chl-a ratios ( Fig. 12b) were relatively high in the Mt. Kenya and Nyambene Hill streams (1156±129 and 580±31, respectively), and low in the 15 surface waters of Masinga Reservoir (206±35). Along the Tana River, POC/chl-a ratios showed no clear gradient but were relatively high overall (570-2081, average of 934±472). chl-b, indicative of green algae, was detected in most sampling sites, in particular in the lower Tana River (>500 µg L −1 , but <200 µg L −1 in other sites). The cyanobacterial pigment echinenone was only detected in Masinga Reservoir (0.15-20 0.29 µg L −1 ), and was absent in all other sites. Zeaxanthin (another cyanobacterial marker, but which can also be present in diatoms) was more widespread, but highest concentrations in Masinga Reservoir (0.29-0.39µg L −1 , compared to 0-0.25 µg L −1 elsewhere). Alloxanthin, which is very specific to cryptophytes, was present in surface waters of Masinga (0.14±0.06 µg L −1 ) and in the Tana River sites just upstream and Introduction

Stable isotope signatures in bacterial PLFA
δ 13 C data on bacterial PLFA (i+a15:0) from soils generally followed the gradient in δ 13 C signatures of bulk organic carbon (R 2 =0.75), and data from river sediments and suspended matter fell within this pattern, although their range of δ 13 C was much smaller (Fig. 13). On average, δ 13 C i+a15:0 was depleted by 4.4±2.4‰ relative to bulk 5 organic carbon, which is close to the isotope fractionation expected based on a large dataset of bacterial PLFA data from coastal sediments (Bouillon and Boschker, 2006). Although there is scatter around the overall regression, with a range of ∆δ values between −9.3 and +0.3‰, this suggests that in general, bacteria appear to use C3 and C4-derived carbon in the same proportions as they occur in the bulk organic matter 10 pool.

General physicochemical characteristics and nutrient concentrations
Given the large altitudinal gradient covered in this survey, water temperatures ranged widely from as low as 11.7 • C in some upper headwater streams to ∼30 • C in the lower 15 Tana River, and high temperatures in surface waters of Masinga Reservoir (Table 1). Although no metabolic rate measurements were made during this initial survey, we can expect that such large temperature gradients will influence aquatic respiration and photosynthesis rates.  Bouillon et al., 2007). The elevated water residence time in Masinga Reservoir and associated biological processes were responsible for a major reduction in DIN (Tables 1 and 5). Surface waters of Masinga Reservoir were extremely depleted in both NO − 3 (0.59±0.14 µM) and NH + 4 (from 0.13 µM at the most upstream location to undetectable levels near the 10 reservoir outlet), and the overall relative decrease in concentrations between the main inlet and the outlet of the reservoir was 53 and 55% for NO − 3 and NH + 4 , respectively ( Table 5). The recent review by Harrison et al. (2009) confirms the importance of reservoirs on N retention, and estimates that these systems are responsible to up to 33% of the global N removal from lentic ecosystems (i.e. lakes and reservoirs), despite 15 their relatively small areal extent. As observed for DIN, phosphate variations were highly variable in headwater streams (0.13-4.60 µM) and depleted in surface waters of Masinga Reservoir (0.13±0.1 µM), with an overall reduction in phosphate between the main inflow and outflow of the reservoir of 72% (Table 5). Along the main Tana River, however, phosphate concentrations showed a very distinct and consistent downstream 20 increase, from 0.14 µM below Masinga to >3 µM in the lower part of the river, i.e. an increase of more than 20-fold (Table 1). Phosphorus dynamics in aquatic systems are often to a large extent determined by interactions with suspended particles. Bound P associated with sediment particles (generally through adsorption on Fe or Al oxides) reversibly exchanges with the dissolved phase (Fox, 1989;Fox et al., 1986). Major pro-Introduction  (Evans et al., 2004, and references therein). The large and steady increase in phosphate observed here along the middle and lower Tana River matches very well with the concurrent increase in TSM concentrations (Fig. 7), which strongly suggests that phosphate concentrations are indeed mainly governed by dynamic exchange equilibria between dis-5 solved P and particle-associated P.
As for DIN and phosphate, dissolved silicate concentrations varied widely in headwater streams (164-959 µM), with highest concentrations in streams draining the Nyambene Hills. The latter region is lithologically distinct from the other headwater regions in being partly dominated by Quaternary volcanic rocks rather than Precambian or Ter-10 tiary volcanic rocks (King and Chapman, 1972). This difference in parent material and the likely lower soil thickness covering it can be expected to result in higher weathering rates of Si and other elements (Heimsath et al., 1997). In line with this, the 2 streams sampled in this area also have markedly higher concentrations of K, Ca, Mg, Na (data not shown) and total alkalinity (Fig. 2) than the headwater streams draining 15 the Aberdares and Mt. Kenya.
A major reduction between pre-and post-Masinga Reservoir was also observed for silicate (from 364 to 215 µM, i.e. a decrease of 41%, Table 5). Consistent with this large reduction in dissolved Si, the phytoplankton composition in Masinga Reservoir has been found to consist predominantly of the diatom genus Nitzschia (Uku and Mavuti, Introduction

Indicators of aquatic metabolism
In headwater streams, pCO 2 ranged between 110 and 1017 ppm (Fig. 4a) and showed no correlation with any of the other measured parameters. Surface waters of Masinga Reservoir were slightly under-or oversaturated in CO 2 (pCO 2 ranging from 313-443 ppm) and decreased towards the outflow of the reservoir. Even along the main 5 Tana River, pCO 2 was relatively low, ranging typically between 500 and 660 ppm along the lower course of the river, with the exception of the outflow of Masinga Reservoir (<3500 ppm) and towards the tidal floodplains in the delta (1543 ppm (Fig. 13), suggesting that other processes besides pelagic photosynthesis and pelagic respiration are significant in influencing O 2 dynamics (e.g. benthic O 2 con-20 sumption which is accompanied with little fractionation (Brandes and Devol, 1997)). Data on %O 2 and δ 18 O-O 2 data can be used to estimate the ratio of primary production to respiration (P:R) in freshwater systems (Quay et al., 1995)  Interactive Discussion mates range from 1.1 to 1.3 for Masinga Reservoir and between 0.76 and 1.08 for the main Tana River. Although the assumption of steady-state conditions can not be verified in the absence of data on diurnal variations, the clear negative correlation between P:R estimates and pCO 2 data (R 2 =0.89 for an inverse first order fit, data not shown) suggests that our approach provides a relative indication of P:R in this system. In conclusion, although direct metabolic process rates are currently unavailable for any of the sampling sites, the combination of pigment concentrations and indirect indicators of aquatic metabolism (pCO 2 , %O 2 , δ 18 O-O 2 ) all suggest that in situ primary production cannot be ignored as a potential contributor to carbon sources in the system -even in the highly turbid lower Tana River. 10 Considering all sampling sites, δ 13 C DIC was negatively correlated with DIC concentrations, i.e. highest DIC concentrations were associated with lowest δ 13 C DIC (Fig. 3b).
This pattern is similar to that observed in the Rhône basin (France) by Aucour et al. (1999). Furthermore, for the headwater streams, we found δ 13 C DIC to be negatively correlated with silicate concentrations, and TA was correlated with silcate concentra-15 tions. Silicate weathering results in a DIC pool with a δ 13 C signature expected to range between −18 and −5‰ (for C3-and C4-dominated sites, respectively), i.e., similar to that of the organic matter remineralized within the soil but taking into account additional fractionation during diffusion of CO 2 into the soil water phase (see Brunet et al., 2005). Streamwater DIC from atmospheric inputs or carbonate mineral disso-20 lution would result in more enriched δ 13 C signatures (>−5‰), since DIC produced by carbonate dissolution is composed of both soil CO 2 (δ 13 C depending on C3/C4 abundance, ∆ 14 C variable) and carbonates (δ 13 C∼0‰ and 14 C-free assuming ancient marine carbonates). Thus, a further contrast between silicate weathering and carbonate weathering can be expected in the resulting ∆ 14 C−DIC signatures, which are much 25 more 14 C-depleted in the case of carbonate weathering (Fig. 14). Radiocarbon data on DIC from our survey (  Mayorga et al., 2005), but similarly low ∆ 14 C DIC data have been reported from the Strickland River (Papua New Guinea, see Alin et al., 2007) and are 5 more typical of higer-altitude streams (see references in Alin et al., 2007) due to e.g. contributions by carbonate weathering or groundwater contributions. In particular, the downstream decrease observed here along the Tana River is suggestive of increased contributions of groundwater containing 14 C-depleted inorganic carbon. In-stream production of DIC along this stretch of the river does not appear to be significant in the 10 overall DIC transported downstream: the DIC inputs from respiration are expected to similar to those of POC (δ 13 C∼−22‰, ∆ 14 C between −40 and −120‰) which would result in an opposite downstream trend as that observed (Table 2). Our combined δ 13 C and ∆ 14 C data thus suggest that alkalinity in the Tana headwater streams results from a variable combination of both silicate and carbonate weathering inputs, and stress the 15 role of these processes in setting baseline δ 13 C DIC across a range of settings. Finally, an important implication for these low ∆ 14 C DIC values for future radiocarbon studies in this system is that in-situ primary production in the lower Tana River will also be characterized by low ∆ 14 C POC values, i.e. that recent phytoplankton production is likely to have an "old" 14 C age.

Methane
Methane concentrations were highly variable, ranging between 25 and 505 nM, and always highly oversaturated (850-21,700%), in accordance with data from other temperate and tropical streams and rivers (e.g., Jones and Mulholland, 1998;Middelburg et al., 2002;Abril et al., 2005Abril et al., , 2007Guérin et al., 2006 and references therein). CH 4 concentrations in rivers result from the balance between inputs from groundwaters followed by transport with water masses, production in river sediments, and losses due to oxidation in waters and surface sediments and to the outgassing to the atmosphere. The relatively low (below 100 nM) CH 4 concentrations in the low-order tributaries of the Tana River, at high and intermediate elevations, suggest that soil CH 4 inputs were moderate in this area at that season. In some US temperate rivers, Jones and Mulholland (1998) have found a CH 4 maximum in small headwater streams at highest elevation, 5 but also at highest organic soil content, they attributed to large CH 4 groundwater inputs from soils. This trend does not appear in the tributaries of the Tana River. Along the main course of the Tana River, CH 4 concentrations were also relatively low (around or below 100 nM), with the exception of the vicinity of the Masinga Reservoir and in the delta (Fig. 4b). Such generally low CH 4 concentrations suggest moderate production in Hydroelectric reservoirs have been shown to be substantial sources of CH 4 emission to the atmosphere (St. Louis et al., 2000). Our data from Masinga Reservoir 15 show relatively high CH 4 concentrations in the inflowing Tana River (505 nM) and in the outflow of the reservoir (372 nM), but much lower concentrations in the upper water column (51±7 nM). High concentration at the entrance of the reservoir is probably due to intense local sedimentation in this area where water current decreases. High sedimentation promotes methane production and diffusion in the water column, which is 20 likely still well mixed in this area. High methane production probably occurs also in the deeper reservoir, but as the water column appears stratified, it affects only the deeper, anoxic waters. The lower concentrations in surface waters are most likely the result of high oxidation rates of methane diffusing from anoxic bottom waters into the oxic upper water layer, as reported in other tropical reservoirs (e.g., Guérin and Abril, 2007). The 25 relatively high CH 4 concentrations at the outflow of the reservoir is due to the deep location of the outlet where CH 4 -rich anoxic water are passing. Part of the CH 4 escapes to the atmosphere immediately downstream of the dam due to pressure change and high turbulence, while further degassing may occur downstream in the river (Guérin  , 2006). CH 4 concentrations further downstream are much lower (Fig. 4b), consistent with important further CH 4 degassing and/or oxidation within the river. Towards the Tana Delta, however, an increase in CH 4 concentrations is observed with highest concentrations in the most downstream site (387 nM), which suggests increased lateral or sediment CH 4 inputs in the delta wetlands.

Suspended sediment in headwater streams and along the Tana River
The most striking pattern in the suspended matter data is the gradual but large increase in TSM observed along the middle and lower course of the main Tana River, from low values of 15.2 and 12.2 mg L −1 before and after Masinga Reservoir, respectively, to 483 mg L −1 in the most downstream sampling station (Fig. 3). As sampling was conducted during the dry season, no tributary inputs reach the Tana River below the Nyambene Hills tributaries (Fig. 1), suggesting that this increase in TSM is caused by resuspension of internally stored sediment. In various other tropical river systems, sediment delivery is observed to be highly episodic or concentrated in shortterm periods with highest precipitation (e.g., Townsend-Small et al., 2007;Hilton et al., 15 2008). According to Dunne (1979), 35-75% of sediment delivery to Kenyan river systems is thought to occur during periods when water flow is in the highest 1% range of flow rates, depending on the dominant land use in the catchments (e.g., with a much higher fraction of sediment delivery during such events for grazing lands compared to forest-dominated catchments). With this in mind, substantial within-channel sediment 20 storage appears highly likely. Kitheka et al. (2005) provide some data on the seasonality of discharge and TSM concentrations at Garsen (corresponding to our sampling station 24) and found that peak TSM concentrations typically preceded discharge peaks -suggesting the release of relatively mobile sediment during the first stages of river discharge increases. The latter is consistent with observations in other large 25 river systems such as the Amazon, Mississippi and Orinoco where a seasonal cycle of sediment storage and resuspension has been observed (Meade et al., 1985;Meade, 1988;Mossa, 1996 Mossa (1996) observed both downstream increases and decreases of TSM levels depending on river discharge stage. The deposition/resuspension cycles in large river networks appear to be governed mainly by seasonal changes in river discharge and river slope (Meade, 1988), but the presence of floodplains allowing temporal deposition and retention also play a role. To our knowledge, the ∼30-fold downstream increase in 5 TSM concentrations observed along the Tana River during our sampling period is much more pronounced than that observed in any other river system so far. Seasonal altitudinal profiles of TSM concentrations during different stages of river discharge will be required to further document deposition and resuspension cycles in this river system, and its consequences for C and nutrient cycling. 10 An important implication of such pronounced resuspension events is that much of the suspended matter observed in the lower Tana River has likely experienced a much longer residence time within the river system compared to the (smaller) pool of TSM in the higher reaches of the Tana River, i.e. the gradient in TSM concentrations is expected to coincide with a gradient of lower to higher residence time within the system. 15 The longer residence time creates opportunities for biogeochemical processing of the associated organic matter (Battin et al., 2008) which become evident when comparing the observed biogeochemical characteristics of particulate OC between the upper and lower Tana River (see below). 20 Soil %OC showed a marked altitudinal gradient, with high values (6.2 to 17.2 %, Table 1) in the >3000 m sites in the Aberdare range, and decreasing to values averaging 2.2±0.9% on the slopes of Mt. Kenya (1350-1590 m altitude), and typically low values in sites along the lower Tana River, but higher values in the riverine forest of the TRPR (1.8%) and in the grassland soils of the Tana delta (2.9%, Table 1). This altitudinal 25 gradient in soil organic carbon is in accordance with previous and more large-scale datasets on soil organic carbon stocks in Kenya (Batjes, 1996(Batjes, , 2004, and is also observed in other large-scale studies such as in the Amazon basin (e.g., Townsend-Small 5986 Introduction  , 2007). A similar gradient was observed in riverine sediment %OC (Fig. 6a). Such gradients are typically explained by the associated temperature gradient, which leads to less efficient soil organic matter degradation in colder, high-altitude regions. In contrast with expectations and similar datasets from other regions (e.g., Townsend-Small et al., 2005), soil TOC/TN ratios did not show 5 a pronounced altitudinal gradient, although such a pattern was discernable in riverine particles and sediments (Fig. 6b, R 2 =0.48 and 0.54, respectively).

Carbon pools in surface soils and river sediments
The soil δ 13 C data encompass the full range of δ 13 C signatures expected for either C3-or C4-dominated soils (−28.5 to −13.2‰, see Table 1). Areas with extensive savanna grasslands and where we thus expected high C4 contributions include the 10 sites in Meru NP (sites 12, 13) and those in the Tana River delta (site 25), and these soils indeed showed relatively high δ 13 C signatures (−14.3 to −17.4‰). In contrast to our expectations, however, most of the higher altitude sites in the Aberdare range also showed relatively strong C4 contributions (δ 13 C between −20.9 and −19.5‰) and one of the sites was even dominated by C4 inputs (δ 13 C of −13.2‰). The distribution of C3 15 and C4 grasses in Kenya has been documented by Tieszen et al. (1979), they found a clear altitudinal shift of a complete C4-dominance below 2000 m to a dominance of C3 grass species above 3000 m. A few notable exceptions, however, occur in this pattern, and particularly relevant here is the occurrence of the tussock-forming C4 grass Andropogon amethystinus in the Aberdare range and on Mt. Kenya, even above 20 3000 m altitude. Our own δ 13 C measurements of A. amethystinus from the Aberdares (−12.1‰) are within the range observed for other C4 grasses in other eastern and southern African ecosystems (e.g., Tieszen et al., 1979;Codron et al., 2005). Most negative δ 13 C signatures were observed in the in the lower floodplains of the Tana River, where riverine forests (i.e. C3-dominated) still persist. It should be stressed 25 that estimates of C3 and C4 contributions using soil δ 13 C data refer to carbon inputs from these 2 vegetation types, and do not necessarily reflect their relative standing biomass, due to potential differences in their relative productivity and degradability of litter. Gillson et al. (2004), for example, demonstrated that soil δ 13 C data in Kenyan

Carbon pools and sources in dissolved and particulate organic carbon: relationships with soil carbon pools
Riverine carbon is either produced within or delivered to the river. The contribution of phytoplankton production to stock of POC can be estimated from C/chl-a ratios. The 5 observed ratios in the river Tana and its tributaries are much higher than those typically found in phytoplankton (>500; Fig. 12b, typical ratios for phytoplankton ∼50), implying that more than 90% of the particulate organic carbon is detrital. Soil organic carbon content appeared to be an important factor influencing riverine %POC/TSM, this relationship was particularly pronounced when only considering data from the tributaries 10 (p<0.001, Fig. 15a). %POC/TSM data, however, were consistently higher than soil %OC determined in the top 5 cm of soil profiles, indicating that suspended particulate organic carbon was also derived from direct litter inputs and/or from topsoil layers, where %OC is likely higher than in the top 5 cm profile which was sampled here. In particular, the very high %POC/TSM values in some of the highland rivers (15.8 to 49.8%) 15 indicate that the TSM was in some cases predominantly organic rather than mineral in nature. Streamwater DOC/POC ratios were also found to be well correlated to soil %OC, with higher DOC/POC ratios in tributaries with higher soil OC contents (Fig. 15b). DOC concentrations in our tributaries were not correlated with soil TOC/TN ratios, in contrast to 20 what would be expected based on strong correlations observed between soil TOC/TN and riverine DOC fluxes in large-scale datasets from diverse river systems (e.g., Aitkenhead et al., 1999;Aitkenhead and McDowell, 2000). There is no consensus, however, on the mechanisms responsible for the latter relationship, and without seasonal data or actual DOC flux estimates, we cannot speculate on whether the data for the Tana River 25 tributaries are an exception to such broader-scale patterns, or whether the limited time frame of sampling renders our DOC concentration data not representative for the DOC fluxes from these tributaries. For the tributaries draining the three perennial catchments (n=11), δ 13 C signatures in suspended matter were significantly correlated to those in soil organic matter (p=0.02, Fig. 15c), but the δ 13 C range was much narrower in suspended matter (−26.5 to −22.7‰) than in soils (−26.5 to −13.2‰). Based on measured end-member values for C4 and C3 vegetation in the Aberdares (−12.1‰ and −28.1±0.8‰, respectively), 5 the contribution of C4-derived materials in POC was estimated to range between 10 and 35%, while corresponding soils show a range between 10 and 98%. The lower variation in stream/river POC is not unexpected given that soil δ 13 C values in these subcatchments (which all appear to have mixed C3 and C4 vegetation) are heterogeneous and vary spatially depending on the dominant vegetation, resulting in an "aver-10 aged" mixed δ 13 C signature in riverine suspended matter. In contrast, δ 13 C DOC and δ 13 C of riverine sediment organic carbon were not significantly correlated to soil δ 13 C signatures (p=0.27 and p=0.12, respectively) for the tributary data. Longitudinal mixing of river sediments will inevitably lead to a decreased variation in δ 13 C values of river sediments compared to soils: river sediment δ 13 C signatures should therefore be 15 considered as a weighted average of δ 13 C values of the different sources contributiong the the sediment load, i.e. from local soils as well as from upstream sediment sources.

Organic matter changes along the river continuum
The strong downstream increase in TSM concentrations along the main Tana River (Fig. 7a) coincides with a strong increase in POC (Fig. 7b), but the latter is less pro-20 nounced, as a result, %POC/TSM decreased from 4.6% below Masinga Reservoir to ∼1.1% in the most downstream sampling stations (Fig. 8). Given that the increase in TSM concentrations can be linked to resuspension of internally stored sediment (see above), this significant decrease in %POC/TSM can be interpreted as an important loss of mineral-associated organic carbon during particle residence within the river system. Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version Interactive Discussion that ∼75% of particle-associated organic carbon was remineralized during its riverine transit. This estimate should be considered indicative only, since future seasonal data are needed to confirm our assumption that %POC/TSM of sediment inputs during high sediment influx events are similar to those observed upstream here during the dry season.

5
The specific surface areas measured on five of the Tana River TSM samples from the lower river are very high (63.6-82.2 m 2 g −1 ) compared to values reported for coastal sediments, estuaries and rivers (typically <50 m 2 g −1 ), and OC:SA ratios found here (0.16-0.22 mg C m −2 ) are consequently in the lower range of those reported in the literature (e.g., Keil et al., 1997;Aufdenkampe et al., 2007). The SA values in TSM 10 are also consistently higher than those observed in surface soils (<63 µm fraction) which ranged between 11.8 and 67.7 m 2 g −1 (Table 1). Soil OC:SA ratios (0.12 to 3.76 mg C m −2 , one exceptionally high value of 13.6 mg C m −2 ) were highly variable and generally higher those of riverine particles (0.220 to 0.143 mg C m −2 ), indicating that riverine particles have been subjected to extensive degradation losses. Moreover, 15 the downstream decrease in %POC/TSM observed in the main Tana River appears, based on the few SA measurements made on these TSM samples, to match also with a decreased surface loading of organic carbon on these particles (OC:SA from 0.220 to 0.143 mg C m −2 between station 20 and 24).
δ 13 C POC signatures below Masinga Reservoir are 13 C-depleted (−25.2‰) relative to 20 those observed upstream of the reservoir (−22.8‰, Fig. 10). In combination with the low δ 13 C signatures in Masinga Reservoir surface waters (−29.0±0.3‰), which appear to are dominated by phytoplankton biomass (based on the low POC:chl-a ratios, Fig. 12b), suggest that part of the POC export from Masinga Reservoir is derived from phytoplankton production in the reservoir. The subsequent downstream sampling site 25 on the Tana River, however, shows a distinctly higher δ 13 C POC (−21.5‰, i.e., a shift of 3.7‰) and a large concurrent decrease in %POC/TSM (from 4.6 to 2.0%). The combination of these data allow us to estimate the expected δ 13 C signature of the Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version Interactive Discussion depleted fraction (which likely consists of phytoplankton biomass from the reservoir) was rapidly and preferentially degraded, consistent with earlier observations in other tropical reservoirs (De Junet et al., 2009). Further downstream, δ 13 C POC shows relatively minor variations (−22.9 to −22.2‰) until the most downstream station in the Tana delta where δ 13 C POC increases (−21.2‰). Since marine inputs can be excluded 5 in the freswhater part of the delta, the latter trend could suggest an increased contribution of local C4 material (which would be consistent also with the slight increase in %POC/TSM). In contrast to the marked changes in the composition of POC before and after Masinga Reservoir, DOC concentrations and δ 13 C signatures are markedly uniform between the Tana River before and after the reservoir, and in surface waters 10 of Masinga Reservoir. None of the δ 13 C POC variations along the Tana River between Meru and Garsen ( Fig. 1) are correlated with changes in %POC/TSM; interpreted in the context of progressive degradation of POC, this suggest that little selectivity in remineralization of C3 and C4-derived carbon occurs, a conclusion consistent with our interpretation of δ 13 C data on bacterial PLFA (see below).
Radiocarbon data on POC (∆ 14 C POC ), available for limited number of sites along the Tana River, ranged between −116 and −35‰ (i.e., 14 C age of ∼935 to ∼230 yr). Raymond and Bauer (2001) were among the first to demonstrate that rivers may deliver relativey old, 14 C-depleted POC to the coastal zone, while DOC is often younger in age. Both experimental work (e.g., Raymond and Bauer, 2001) and field data (e.g., 20 McCallister Mayorga et al., 2005) have suggested that intensive bacterial remineralization can occur selectively, and hence substantially alter the age and composition of organic ultimately exported. In our dataset, there is a general decrease in ∆ 14 C POC along the lower Tana River, which coincides with the overall decrease in %POC/TSM (Fig. 11). Interpreting the gradual decrease in %POC/TSM in terms of 25 remineralization of POC during the residence period of particles within the river system, this decrease in 14 C age of associated POC indicates selective remineralization of a modern POC fraction (estimated ∆ 14 C of +56 to +140‰, depending on whether or not the data from station 11 are included in the regression used), imprinting the residual Interactive Discussion carbon delivered to the estuary and delta with an old, 14 C-depleted signature. Alternatively, within river primary production would add 14 C-old organic carbon because of the very low ∆ 14 C DIC numbers (Table 2), but this freshly produced material is likely readily remineralised and will thus not contribute strongly to POC pools. DOC concentrations ranged between 0.3 and 2.5 mg L −1 , with the majority of data 5 in a fairly narrow range between 0.6 and 1.2 mg L −1 (Fig. 9a). DOC was the largest organic carbon pool in high altitude tributaries (DOC/POC ratios of 1.9±0.2, Fig. 9b) while POC dominated in tributaries draining Mt. Kenya and Nyambene Hills (DOC/POC ratios of 0.8±0.3). Along the main Tana River, DOC concentrations varied within a very narrow range (0.98±0.14 mg L −1 , Fig. 9a), contrasting sharply with the steep downstream increase in POC concentrations (Fig. 7b). DOC concentrations in the lower Tana River are markedly lower than those observed in other large African rivers such as the Congo, Nile, Gambia and Niger (Martins and Probst, 1991), but consistent with average values for semi-arid climates presented in Spitzy and Leenheer (1991). As a result of the contrast in DOC and POC concentration profiles, DOC/POC ratios declined from ∼1.6 in the vicinity of Masinga Reservoir to 0.2 along the lower Tana River (Fig. 9b), and were inversely related to logarithm of TSM concentrations (Fig. 9c). An inverse relationship between log(TSM) and DOC/POC ratios has been observed in a wide range of river systems and estuaries (e.g., Ittekkot and Laane, 1991;Middelburg and Herman, 2008;Ralison et al., 2008). When comparing 20 our data from headwater streams, the main Tana River, and data from the freshwater part of the Tana estuary (Bouillon et al., 2007, Fig . 9c), however, different patterns in the DOC/POC vs. log(TSM) relationship can be discerned. While the slope of the relationships appear fairly similar, DOC/POC ratios are lower in the headwater streams compared to the main river for similar TSM values, and DOC/POC ratios from the 25 freshwater estuary are higher than those found here on the main Tana River (Fig. 9c).
The DOC/POC ratios in the most downstream stations are markedly lower than those observed in the freshwater and oligohaline part of the estuary (1.4±1.3, Bouillon et al. (2007)), suggesting either pronounced seasonal variations or substantial additional on surface water δ 13 C DOC . This suggests that either little excretion of DOC occurs during primary production in this system, or that labile algae-derived DOC is rapidly remineralized or photo-oxidized in the reservoirs surface waters, with little net effect on overall DOC concentrations and δ 13 C signatures. Overall, δ 13 C DOC and δ 13 C POC were clearly uncoupled and showed a relatively weak relationship (Fig. 10c), suggesting that 10 despite clear evidence for partitioning of OC between POC and DOC (i.e., the relationships between DOC/POC and TSM described above), a large fraction of the POC pool does not readily or dynamically exchange with dissolved C pools.

Carbon sources driving remineralization
δ 13 C data on bacterial PLFA (i+a15:0) from soils generally followed the trends in bulk 15 δ 13 C (Fig. 13, R 2 =0.75), and data from river sediments and suspended matter fell within this pattern (however, the range of δ 13 C was much smaller). On average, δ 13 C i+a15:0 was depleted by 4.4±2.4‰ relative to bulk particulate organic carbon, which is close to the isotope fractionation expected based on a large dataset of bacterial PLFA data from coastal sediments (Bouillon and Boschker, 2006). Although there is quite 20 some scatter around the overall regression, with a range of ∆δ values between −9.3 and +0.3‰, this suggests that in general, bacteria appear to use C3 and C4-derived carbon in the same proportions as they occur in the bulk organic matter pool. The pattern observed in our gradient of mixed C3-C4 soils, sediments and suspended matter samples matches very well with data from 2 intertidal mangrove systems (Fig. 13) where riverine C4 inputs were found to be significant, i.e. the Tana River delta (data from Bouillon and Boschker, 2006) and the Betsiboka estuary (Ralison et al., 2008), although the scatter in the latter two datasets appears to be smaller than in our current (larger-scale) data. The absence of clear evidence for selective remineralization of either C3 or C4-derived C is also consistent with the lack of changes in δ 13 C POC with decreasing %POC/TSM in our data on the lower Tana River suspended matter (see above).

5
Studies using microbial PLFA stable isotope signatures to differentiate use of C3 and C4-derived organic carbon in soils are very scarce, and rather than covering a natural gradient in soils with different relative inputs of C3 and C4-derived carbon, are derived from experimental plots where C3 vegetation was replaced by C4 vegetation (Burke et al., 2004;Gleixner, 2006, 2008). While these studies suggest that C4derived carbon is preferentially used by bacterial communities, the distinction between C3 to C4-organic matter coincides with a different age (in the order of several decades) of these carbon sources (former C3 vegetation, current C4 vegetation). This may obviously bias resulting conclusions regarding general patterns in C3/C4 degradability. A second caveat when attempting to find general patterns in C3-C4 decomposition 15 is that the distinction between C3 and C4 vegetation in many cases coincides with trees/shrubs (C3) vs. C4 grasses, and hence, that many of the observed differences may be related to vegetation types (grasses vs. dicotyls), rather than between C3 and C4 vegetation. Data from a gradient of grasslands (C4-dominated to C3-dominated) would therefore be an ideal supplement in this discussion. Detecting selective rem-20 ineralization using our approach, however, is not straightforward: even if selective C4 degradation occurs, the effect of selective degradation on the relationship between bulk and bacteria-specific PLFA δ 13 C signatures will not only depend on the relative contribution of C4 and C3-derived carbon present in the bulk C pool, but also on their relative age and input rates.  Kenyan Geogr., 2, 109-126, 1976. Dunne, T.: Sediment yield and land use in tropical catchments, J. Hydrol., 42, 281-300, 1979. Dunne, T.: Evaluation of Erosion Conditions and Trends, in: Guidelines for Watershed Management, FAO Conservation Guide n • 1, 1977 Evans, D. J., Johnes, P. J., and Lawrence, D. S.: Physico-chemical controls on phosphorus cycling in two lowland streams. Part 2 -the sediment phase, Sci. Total Environ., 329, 165-182, 2004. Fox, L. E.: A model for inorganic control of phosphate concentrations in river waters, Geochim.
Cosmochim. Ac., 53, 417-428, 1989.                   Profiles of (A) chl-a, and (B) POC/chl-a ratios in headwater streams, Masinga Reservoir, and along the main Tana River. Note that no data are available for streams draining the Aberdares, but the X-axis scale is similar to that of other profile plots for easier comparison.  . 13. Plot of δ 13 C signatures of bacterial PLFA (i+a15:0) vs. those of bulk organic matter fractions, including data from both soils, riverine sediments and suspended matter from this survey, and comparative data from intertidal sediments in the Tana River delta and the Betsiboka estuary (from Boschker, 2006 andRalison et al., 2008, respectively). Regression line and equation refers only to data collected in this survey.