Sediment deposition patterns in a tropical floodplain, Tana River, Kenya

Catena 143 (2016) 57–69

Contents lists available at ScienceDirect

Catena journal homepage: www.elsevier.com/locate/catena

Sediment deposition patterns in a tropical floodplain, Tana River, Kenya Fred Ochieng Omengo a,b,⁎, Tine Alleman a, Naomi Geeraert a, Steven Bouillon a, Gerard Govers a a b

Katholieke Universiteit Leuven, Department of Earth & Environmental Sciences, Celestijnenlaan 200E, 3001 Leuven, Belgium Kenya Wildlife Service, P.O. Box 40241-00100, Nairobi, Kenya

a r t i c l e

i n f o

Article history: Received 24 July 2015 Received in revised form 8 March 2016 Accepted 21 March 2016 Available online xxxx Keywords: Tropical river floodplain Sediment budgets Lead-210 and Ceasium-137

a b s t r a c t Floodplains exert strong controls on downstream sediment transport and are as such important in material budgets of river systems. To understand sediment budgets in the Tana River, we investigated sediment storage along a 380 km floodplain reach (Garissa-Garsen) in the lower Tana River (Kenya), using a combination of approaches: (i) measurements of sediment deposition after an important flood event and, (ii) quantification of sediment storage using fallout radionuclide activities (137Cs and 210Pbex). Event-based sediment deposition ranged between 215 mm vertical accretion, corresponding to an average of 0.58 ± 0.42 g cm-2 (dry weight). Average annual sediment storage based on fallout radionuclide activities were at a 50 yr mean of 1.15-1.21 g cm-2 yr-1 and a 100 yr mean of 1.01 g cm-2 yr-1 (using 137Cs and 210Pbex, respectively). Sediment deposition rates were mainly dependent on distance relative to the main river, flood height and microtopography. The deposited sediments originated from various sources including deeply mobilized, radionuclide-poor sediments. Shallow overbank deposits (b 2 m depth), dominated by silty-clay sediment fractions were observed in most depositional areas. Floodplain sediment stocks are controlled by annual overbank flooding, with a 73 yr flooding frequency of 1.05 flood yr-1. On average ca. 1-2 Mt (15-30%) of the river sediment load is deposited in the Garissa-Garsen floodplain reach. However, the overall channel dynamics including channel sediment storage and reworking are important and may have a very high impact on short-term sediment storage and release. © 2016 Elsevier B.V. All rights reserved.

1. Introduction River systems receive substantial quantities of sediment from their catchments, but rather than merely transporting the sediment load towards the coastal zone, they continuously deposit and rework sediments as they flow downstream (Day et al., 2008; Hoffman and Gabet, 2007). Both man-made structures such as reservoirs (Syvitski et al., 2005) and natural systems such as lakes and floodplains (Aalto et al., 2003; Noe and Hupp, 2009) are known to be key sites that exert strong controls on material storage and downstream mobilization. The exchange of sediment between the floodplain and the river channel is an essential component of the river system and even in relatively small catchments (b 1000 km²), sediments may reside for N 1000 yrs in floodplain environments before they are finally discharged at the river mouth (Hoffmann, 2015; Notebaert et al., 2009). While longitudinal studies of river sediment loads and biogeochemistry can provide important insights into the processing of sediment, carbon and nutrients during their residence in the river channel, a comprehensive understanding of the functioning of the system can only be obtained if the lateral

⁎ Corresponding author at: Katholieke Universiteit Leuven, Department of Earth & Environmental Sciences, Celestijnenlaan 200E, 3001 Leuven, Belgium. E-mail addresses: [email protected] (F.O. Omengo), [email protected] (T. Alleman), [email protected] (N. Geeraert), [email protected] (S. Bouillon), [email protected] (G. Govers).

http://dx.doi.org/10.1016/j.catena.2016.03.024 0341-8162/© 2016 Elsevier B.V. All rights reserved.

exchange between the river channel and its environment is accounted for (Day et al., 2008). Despite the fact that floodplain deposition is now widely recognised, data that would allow to us to quantify the relative importance are still scarce (e.g. Noe and Hupp, 2009; Swanson et al., 2008). This is especially true for non-temperate environments and this lack of understanding fundamentally limits our ability to predict the response of rivers to natural or man-made disturbances. In addition, current sediment yield models do not account for the buffering of sediments in alluvial or colluvial environments but relate sediment export to variables describing the catchment’s state such as topography and land use on the one hand and climatic drivers on the other hand (e.g. Syvitski et al., 2005). For many catchments, such a direct relationship does not exist as sediments may be buffered for centuries, if not millennia, and this buffering should be accounted for when considering land-ocean material transport. Understanding sediment erosion and transport within a catchment requires proper estimation of sources, sinks and outputs. For small catchments, quantification of these variables can be relatively easy to address leading to fairly accurate budgets (Hoffmann, 2015; Notebaert et al., 2009; Walling et al., 2002). However, erosion, transport and storage of materials in larger catchments is more difficult to quantify. In such catchments, fallout radionuclides offer a complimentary approach that allows estimation of sediments fluxes and budgets from various sources within the catchment (Wallbrink et al., 1998; Walling et al., 2002). Fallout radionuclides such as 137Cs and 210Pb are strongly sorbed

58

F.O. Omengo et al. / Catena 143 (2016) 57–69

by soil particles upon reaching the earth surface, their subsequent postdepositional redistribution is useful in tracing sediment mobilization. Assessment of their redistribution and storage within the catchment has been used to shed light on rates and patterns of sediment mobilization (Zapata, 2002). Complementary information may also be obtained from direct surveying of sediment deposits after a flood event (Gomez et al., 1995; Walling et al., 1997). The major objective of our study was to quantify the amount of sediment deposited in the Tana floodplain in order to understand the impact of floodplain deposition on the river’s sediment dynamics. We used a combination of these approaches to quantify the amount of sediment deposited and stored in a 380 km reach of the lower Tana River floodplains. We chose the lower Tana River for our study based on the fact that, the lower Tana River meanders for several 100s of kilometres, from the town of Mbalambala down to the town of Kipini, through extensive, semi-arid floodplains with no permanent tributaries, making it an interesting case to study river-floodplain interaction. Second, a number of recent studies in the Tana River have focused on the sources, transport, and processing of sediment and associated carbon in the stretch between the towns of Garissa and Garsen (Bouillon et al., 2009; Kitheka et al., 2005; Tamooh et al., 2013, 2014), providing a wealth of background data for this study. These earlier studies have demonstrated a substantial reduction in annual sediment and organic carbon fluxes between Garissa and Garsen, suggesting important floodplain storage (Tamooh et al., 2014). We used fallout radionuclide inventories (137Cs and 210Pbex) on a series of sediment cores collected throughout the lower Tana floodplains to obtain long-term sediment storage rates. Secondly, we quantified sediment delivery into the floodplain by measuring the thickness of recent sediments deposits along transects perpendicular to the river course at various sections along the river. Based on calculated radionuclide inventories, we made a first-order estimate of the overall sediment budget in the lower Tana between Garissa and Garsen (hereafter referred to as the GSAGSN reach) over the past decades and for the last major flood events, by combining this information with Landsat imagery and hydrological data. 2. Materials and Methods 2.1. Study area The Tana is the largest river in Kenya, and flows ca.1,000 km from Kenya’s central highlands of Mt. Kenya and Aberdare mountains (5800 m a.s.l) traversing the landscape to drain into the Indian Ocean at Kipini (Fig. 1). The catchment covers ca. 100,000 km² and can be divided into the headwaters (the mountainous streams from the Aberdare ranges, Mt. Kenya and Nyambene hills in the central parts of Kenya) and the lower Tana consisting of the section downstream of Kora where the river flows for ca. 700 km through semi-arid plains. The lower Tana has intermittent tributaries covering large catchments but only flowing in short pulses during the wet season (Maingi and Marsh, 2002). Three main seasonal rivers (Laga Tula, Laga Galole and Laga Tiva, Fig. 1), flow into the Tana River below the town of Garissa. The upper Tana catchment has five hydro-electric dams with a combined surface area of 150 km2, and which are known to have influenced the downstream hydrology (Maingi and Marsh, 2002; Saenyi and Chemelil, 2003). The water in the dams has a residence time that varies between 3 months and 2 yrs, (Pacini et al., 1999). Recently, the dams have been shown to affect the seasonality of the discharge (higher average dry season discharge and lower average wet season discharge) but they do not appear to have a major impact on net sediment fluxes in the lower Tana due to the importance of autogenic sediment reworking (Geeraert et al., 2015). 2.1.1. Climate The average annual precipitation varies from 2200 mm for the upper Tana catchment to 370 mm for the lower Tana downstream of Kora, with a gradual increase from 350 mm yr-1 at Garissa to about 470 mm yr-1 at Hola, and over 1000 mm yr-1 at areas downstream of Garsen

(Brown and Schneider, 1996; see Fig. 1). Temperatures are on average above 30°C in the lower Tana, with a mean annual potential evapotranspiration (PET) between 1500–1700 mm (Dagg et al., 1970). 2.1.2. Geology Geologically, the upper Tana catchment is underlain by a Precambrian basement complex, with volcanic formations mainly of tertiary age that originate from Mt. Kenya and the Aberdares. These volcanic rocks cover almost two thirds of the upper catchment. Downstream of Kora, the Tana is considered to be an alluvial river and is not constricted by outcrops of bedrock: therefore most of the meandering is determined by morphological processes (DHV, 1986). 2.1.3. Floodplains The minimal elevation drop downstream of Kora has resulted in the formation of an extensive floodplain wherein the river is freely meandering. Flooding is common and usually associated with increased rainfall in the upper catchment and release of water from the dams. In the floodplain, a characteristic riverine forest fringes the banks of Tana River. The forest consists of a mosaic of deciduous and evergreen trees rich in endemic species (Hughes, 1988; Medley, 1992). The riverine forest extends for over 400 km of the Tana River from Mbalambala to Kipini and is dependent on flooding and ground water recharge. Based on satellite imagery and Google Earth, the total recently active floodplain extent is ca. 998 km2. The lateral extent of the GSA-GSN floodplain is rather narrow and varies between 0.2 km and 5 km, as compared to the Delta (below Garsen) where the floodplain extends up to 20 km from the main river. On the other hand, within the same GSA-GSN floodplain section, the river consists of one channel with a width that is fairly constant averaging between 50 m to 100 m. There are only minor changes in the valley slope between Garissa-Garsen, which averages ca. 0.5 m km-1 (DHV, 1986). 2.1.4. Hydrology and flooding Two annual precipitation cycles occur in the basin resulting in a bimodal peak discharge and potential flooding frequency. The first of the two peak flows occurs between April and June (long wet season) and a shorter high flow period occurs during November/December (short wet season). At Garissa, daily discharge data based on gauge height and rating curves are available from 1941 to present. Flood flows generally range between 300-700 m3 s-1 while base flow is at a mean of 75 m3 s-1. Occasionally extreme flood events with discharges over 1000 m3 s-1 have been observed (WRMA, Water Resources Management Authority Kenya), (Fig. 2). Garsen discharge data are similarly based on gauge height and rating curves, it dates back to 1950s but with numerous gaps in the 1970s, and late 1990s to 2000s (supplementary information, Fig. S5). 2.2. Fallout radionuclides Fallout radionuclides such as 210Pb and 137Cs have been widely used to calculate short-term (decadal time scale) sediment deposition and accumulation rates in different depositional environments. 137Cs is an artificial or ‘man-made’ radionuclide; it has a half-life of 30.2 yrs and was generated as a product of thermonuclear weapons testing in the mid-1950s to the early 1970s. Other emissions are associated with nuclear accidents (e.g. Chernobyl in 1986). The use of 137Cs as a sedimentation technique is based on the principle that sediment containing the nuclide derived from the upstream catchment is deposited downstream at a measurable rate. However, while 137Cs provides a strong signal in sediment in the northern hemisphere, total fallout in the equatorial and southern hemisphere was only 25% that of the north and low activities are often observed. Nevertheless, various studies have also successfully used 137Cs to estimate sedimentation in fluvial and lacustrine environments in the southern hemisphere (Amos and Croke, 2009; Hughes et al., 2009; Humphries et al., 2010).


59

Fig. 1. Study area showing the location of Tana River, Kenya, and Garissa-Garsen section of Tana River floodplains (enclosed box). Three ephemeral streams (Lagas) join the main river within this reach, downstream Garsen, there is a shorter but more expansive lower section (Tana River Delta) before the river flows into the Indian Ocean at Kipini.

To derive sedimentation rates based on 137 Cs data, two approaches are traditionally used; (i) the first appearance (yr. 1953) method and (ii) the total inventory method. To use the first

Fig. 2. Historical discharge for the Tana River (period 1941-2013) based on daily stage height reading at Garissa. There is a semi-annual cycle between base flow and peak flow which potentially influences flooding and floodplain sediment deposition.

appearance technique, detailed measurements of 137 Cs activities down the profile are made and a particular age is then assigned to the first detectable levels of 137Cs within the profile corresponding to ca. 1953 for the Northern hemisphere and ca. 1955 for the southern hemisphere (Leslie and Hancock, 2008). Data need to be interpreted with care, as downward migration of 137Cs may occur due to bioturbation and tillage. To compliment the first appearance method, the 137Cs activity peak can also be used. The peak activity of 137 Cs is typically associated with the yr. 1963 in the northern hemisphere and 1965 in the southern hemisphere. The second approach is the 137Cs inventory method where the amount of deposition or erosion at a site is determined by comparing the total 137Cs inventory of a sediment core to that from an undisturbed “reference site” above the floodplain (Walling and He, 1997). This method assumes that the 137Cs inventory measured from a point in the floodplain is in excess of the local reference inventory above the floodplain. In both the first appearance and inventory approaches, it is assumed that any sediment deposited on the floodplain since fallout, carries a measurable quantity of 137Cs (Walling and He, 1997). Furthermore, the total inventory method requires the establishment of a reference inventory and will only register sedimentation if the deposited sediments do contain a constant and measurable amount of 137Cs. Knowledge of the 137Cs concentration in river sediments may therefore be helpful in interpreting the results.

60


210 Pb on the other hand, is a natural geogenic radioisotope from the U decay series with a half-life of 22.3 years. It results from the decay of 226Ra which is found in most soils, rocks and sand that produce shortlived gaseous 222Rn (half-life 3.8 days) as its daughter. The majority of the 222Rn decays to 210Pb within the soil, producing “supported” 210Pb, which is essentially in equilibrium with the parent 226Ra. However, some of the 222Rn diffuses upwards into the atmosphere, where it rapidly decays to 210Pb. This 210Pb is deposited as fallout and since it is not in equilibrium with the parent 226Ra, it is commonly termed unsupported or excess 210Pb (210Pbex), to distinguish it from the supported 210 Pb (Gilmore, 2008). As opposed to 137Cs fallout, the fallout of 210Pb occurs permanently through time due to its natural origin. However, some studies have identified seasonal and longer-term variations in 210 Pb concentrations in the air and in atmospheric fallout i.e. rain, snow and dry deposition (Guelle and Balkanski, 1998; Preiss et al., 1996). When using 210Pbex measurements to document soil deposition, both total 210Pb and 226Ra have to be determined, in order to calculate 210 Pbex. The activity of 210Pbex in sediment samples is determined by the difference between total 210Pb and 226Ra (Gilmore, 2008). Alluvial sediments may not always contain radionuclides, this is hardly surprising, as rivers rework sediments over timescales that are far longer than the timescales over which these radionuclides have been deposited (e.g. Hoffmann, 2015; Notebaert et al., 2009). It is therefore expected that, some of the deposits will contain varying proportions of ‘old’ deposits reworked from depths where no measurable quantities of radionuclides were present. The resulting variations in radionuclide content do not reflect variations in sedimentation rates but in sediment provenance (Aalto and Nittrouer, 2012). This also means that care has to be taken when estimating sedimentation rates using conventional methods based on total radionuclide inventory or peak fallout. 238

2.3. Event based sediment deposition Direct post-event surveys of flood deposits provide a complementary approach that has been used to quantify deposition rates for sites where sampling or monitoring devices have not been installed prior to the depositional event (Gomez et al., 1995). The approach can be applied whenever the freshly deposited sediment can be clearly differentiated from the “older” underlying sediment. (Gomez et al., 1995; Walling et al., 1997). 2.4. Sampling and analytical techniques 2.4.1. Field work Fieldwork was conducted in October-November 2012 (1st campaign) and May-July 2013 (2nd campaign). During the first campaign, no substantial flooding occurred and sampling was limited to collection of sediment cores. Prior to the second campaign, extensive flooding

occurred lasting at least one month in most areas of the study reach. Immediately after the recession of the floods, post-flooding assessments were undertaken to measure sediment deposition at 76 locations within GSA-GSN floodplain reach (Fig. 1 & Fig. 3). The procedure involved running transects, the location of which was mainly dictated by accessibility and were oriented at an angle perpendicular to the main river. Each transect ran from the edge of the flooded area towards the river and sediment samples were collected along them at intervals of ca. 100 m. The sampling involved scraping of freshly deposited materials using a small spade and measuring their thickness (Fig. 4) with a small ruler, as well as the measurement of the associated flood height based on existing vegetation as flood height markers. The elevation and the distance of these points relative to the main river was recorded using a Garmin e-Trex 30 GPS which has an accuracy of ca. ± 3 m in standalone mode, or by use of a car odometer, with an accuracy of ca. ±10 m. Short distances were measured by physically walking through transects with a measuring tape. Upon return to the laboratory, the exact positioning of all locations were cross checked using Google Earth. The fresh sediment samples were air-dried and stored in zipped plastic bags for later analyses in the lab. During both campaigns, 11 sediment cores were collected at various locations within the floodplain and analysed for fallout radionuclide activities (Table 1). Coring sites were chosen so that the diversity of geomorphological features (depositional environments) within the floodplain was covered (see Fig. 3). Coring was performed with an Eijkelkamp hand sediment core sampler (Model 04.16, Ø 40 * l, 200 mm) tubing extendable to 2.2 m depth. Sediments were later pushed from the tubes with an Eijkelkamp hand pusher, sliced at 1 cm intervals, air dried and stored in zipped airtight polyethylene bags. During coring, some compaction inevitably occurred and this varied between sites and with depth. Samples were taken in 20 cm increments and the length of each core was accurately measured after each tube before the core was sliced. Samples were sliced in 1 cm intervals in the field and the depth covered by each slice was then corrected by linear extension of the length obtained to the actual depth of the core. By doing so we implicitly assumed that compaction was homogeneous within a 20 cm coring increment. Given the fact that the sediments were relatively homogeneous in terms of grain size no major errors were generated by this assumption. For each coring site, an extra core was dug using a soil auger and samples for bulk density taken at ca. 10 cm depth intervals using a Kopecky ring and soil auger corer. To check for local fallout, two cores for reference samples were taken close to Hola and Garsen at sites that were located well above the floodplain. 2.4.2. Laboratory analysis Upon return to the laboratory, bulk density samples were weighed to establish the wet weight and then oven-dried at 105°C for 24 hrs and re-weighed to establish the dry weight. The remaining samples

Fig. 3. Typical cross-section of Tana River floodplain and an illustration of the environments where cores were taken. An attempt was made to cover the diverse geomorphological features (Oxbow Lakes, Back-swamps, Open areas, Riparian forested areas) within the floodplain. For the post-event surveys, transects were run from the floodplain margin to as far as possible towards the main river, covering different depositional areas.


61

1996) where the average sedimentation rate over last 100 yrs is given by: R ¼ λPb

Ainv Ainv;At Cr

ð1Þ

Where:

Fig. 4. Part of the fresh sediment from post event surveys after the 2013 flooding. It was easy to differentiate and accurately measure the thickness of fresh sediment deposits above the old sediment layer.

were oven dried at 105°C for 24 hrs, and ground with a pestle and mortar. Particle size distribution was determined using a laser diffraction particle size analyser LS13320 (range: 0.02–2000 μm). For the analysis of fallout radionuclides 137Cs and 210Pbex, a wellmixed subsample of the core slices was put in pre-weighed 1 ml glass vial sealed with airtight plastic stoppers. The sealed vials were stored for at least 2 weeks to prevent 222Rn gas from escaping and to allow the unsupported 210Pb activities to equilibrate with 222Rn. Measurements of 137Cs and 210Pb were made by gamma ray spectrometry using a high purity Ge gamma well detector (model; Canberra GCW7021), with a well diameter of 18 mm and well depth of 35 mm). Counting times were typically more than 160,000 seconds and counting uncertainties less than 5%. Total 210Pb was measured by its emission at 46.5 keV and supported 210Pb by the weighted average decays of 226Ra daughters at 295, 351, and 609 keV. 210Pbex was calculated as the difference between the measured total 210Pb at 46.5 keV and the supported 210Pb activity. 137Cs was determined using its gamma emission at 661 keV. For both nuclides, errors were determined from counting statistics using a two standard deviation counting error (95% confidence interval) converted to activities. Bi-weekly background counts were taken and subtracted from sample counts to give net counts and actual activities. Further, for all the 137Cs results measured, activities of less than 0.5 mBq g-1 were considered to be insignificant and assumed to be zero as the uncertainty on these values exceeded the measurement uncertainty, which was primarily due to variations in background noise. 2.4.3. Calculation of sedimentation rates To quantify sediment deposition over time using 210Pbex, we applied the floodplain inventory model proposed by (He and Walling,

• R(g cm-2 yr-1)- is the sedimentation rate per year • λPb is the decay constant for 210Pb ( 0.04492 yr-1) • Ainv (mBq cm-2) is total 210Pbex inventory for the point measured in the floodplain • Ainv,At (mBq cm-2) is the local fallout inventory from the reference site • Cr (mBq cm-2) is the concentration of catchment derived 210Pbex • The concentration of catchment derived sediment Cr was determined directly from average value of fresh sediment deposits collected during the 2013 flood event (5.7±0.9 mBq cm-3, n=10). • We used reference sites for comparison with our floodplain and specifically for calculation of sedimentation rates with 210Pbex. It is worth mentioning that we do not have a proper history of the reference sites but at the time of sampling the sites were un-disturbed and we assumed (being a remote area) they have been relatively intact over time. The 137Cs activities in the depth profiles were used to calculate deposition rates using the aerial loading method proposed by (Walling et al., 1999), where the average sedimentation rate since 1955 is given by Rðt Þ ¼

Mp ðt Þ V Tp

ð2Þ

Where: • R(t) (g cm-2 yr-1) is the average sedimentation rate since 1965/1955 • Mp(t) (g cm-2) is the cumulative mass to the lowest depth where 137Cs is detected. • Tp yr is the time elapsed between 1955 and the time of sample collection (t) • V the amount of soil in the mixing layer (g cm-2), i.e. the layer over which 137Cs is mixed due to tillage and/or bioturbation. We assumed a depth of 0.15 m for the mixing layer at all sites. Given that we observed erratic patterns in the radionuclide depth profiles, the conventional principles that govern Eqs. (1) and (2) above may be bound to uncertainties. We therefore in addition, used an alternative approach to derive sedimentation rates by plotting 137 Cs cumulative radionuclide inventories as a function of depth down from the surface. This allows for a much clearer identification of the point below which little or no 137Cs occurred for most of our profiles. We assumed that this point in the profile corresponded to the lower limit of the topsoil mixing zone at the moment when 137Cs started

Table 1 Summary of the cored sites. Coring was limited to a maximum depth of upto continuous sand layer (old River bed) and as shown below this depth was very shallow in most cores. Site/Core

Site Name

°S

°E

Local Use

Distance from river (m)

Max depth (cm)

Description

Site 1 Site 2 Site 3 Site 4 Site 5 Site 6 Site 7 Site 8 Site 9 Site 10 Site 11

Wenje Lake Gadia Kajitupe Kiluluni III Left Bank Makere IV Maroni II Kiozani II Mnazini II Kone II Mnazini I Maroni III

1.784 1.819 1.839 1.835 1.850 1.827 1.435 2.036 1.523 1.979 1.827

40.108 40.108 40.109 40.119 40.125 40.108 40.031 40.131 40.046 40.111 40.108

Partial Agriculture Forest Partial Agriculture Forest Forest Partial Agriculture Agriculture Forest Partial Agriculture Agriculture Forest

305 438 608 400 241 203 320 103 331 2660 202

147 100 140 220 180 220 182 221 102 121 180

sand reached sand reached sand reached sand not reached water table reached sand reached water table reached sand reached sand reached rocky sand reached

62


(1955). Again, we assumed the depth of the mixing zone to be 0.15 m. It is possible that the effective mixing depth may have been somewhat lower on sites which were permanently under forest since 1955 but such sites are difficult to identify, given that local farmers indicate that they often shift their fields and also cultivate below trees. The cumulative inventory methodology can also lead to errors, especially if the first sediments deposited at a location after 1955 did not contain any radionuclides, this will lead to an underestimation of sedimentation rates. Variations in mixing depth may lead to both under and overestimations. However, the methodology does allow to reliably identify the point above which radionuclides are systematically occurring within the profile more robustly than the identification of a single peak. Also, this approach is not sensitive to variations in total inventories which may result from variations in the properties of the sediments that are deposited or from variations in flooding frequency. The calculations described above allowed us to estimate net deposition, i.e. the difference between gross deposition and erosion. However, the topography in the Tana floodplain is near horizontal. Furthermore, during flood events, water flows into the floodplain and remains stagnant for a period of several days to several weeks. Under such conditions, it can reasonably be assumed that erosion by resuspension is negligible so that gross and net deposition rates are, for all practical purposes equal. 2.4.4. Image Analysis Representative Landsat images taken between 1990 and 2013 (Table 2) (LANDSAT-4&5/TM and LANDSAT-7/ETM+, NASA Earth Resources Observation and Science-EROS centre, http://eros.usgs.gov/) were used to quantify the inundation extent during different flooding events. The maximum floodplain extent was established by digitizing a clear image from 21st May 2012 (ETM+L1T) on which the contrast between the inundated floodplain and the surrounding drylands was clearly visible. Images were processed using ENVI 5.0 (Exelis Visual Information Solutions, 2012), corrected for sensor defects and sensor differences by converting top atmosphere reflectance using published postlaunch gains and offsets (Chander et al., 2009). Dark object subtraction was then performed using the band minimum from each image (Song et al., 2001). Following this correction, a supervised classification was done using pre-defined flooded area pixels as training areas. A summary of the satellite images that were used to calculate flooded areas and flooding frequencies for the GSA-GSN reach is given in Table 2. 3. Results 3.1. Single event floodplain sedimentation Deposition rates in the floodplain after the 2013 flooding event ranged between 2 and 15 mm corresponding to mass sedimentation rates of 0.27-2.04 g cm-2 (n = 91). There was a strong inverse

correlation between the overall sedimentation rates and the distance of the sites from the main river (Fig. 5). In addition, the flood height had a strong influence on amounts of deposited sediment (vertical accretion) at any point. The 137Cs activity in the fresh deposits were in most cases below detectable limits (n=9). Flood height was not related to the distance from the river, suggesting that local topographic variations exerted a major control on flood heights. 3.2. 3.2 Radionuclide inventories and sediment storage in the floodplain Sediment deposition rates calculated for the various coring locations and a representative sample of radionuclide and texture profiles are shown in Table 3 and Figs. 6-8. The floodplain sediments from all cores show a dominance of fine grain sizes with N 80% silt-clay in the upper 100 cm depth. At the majority of the sites, the deeper layers are similarly characterized primarily by fines (N 70% silt-clay) down to the depth where sandy river bed/bank deposits are reached. The thickness of overbank deposits was less than 2 m at most sites and we reached the old river bed (visible as a sudden transition to a pure sandy layer for the rest of the depth) at an average depth of 165±46 cm (Table 1). Despite the fine-grained nature of the floodplain deposits, total 137Cs inventories were relatively low and ranged between 3.6-21.8 mBq cm-2. The majority of the samples had activities below 3 mBq g-1 and in most profiles, several 137Cs peaks were observed. At some locations, a single peak of 137Cs could be observed well below the level where 137Cs was consistently present (e.g. site 2 & 3, Figs. 6 & 7). Conversely layers poor in radionuclides sometimes occurred above horizons containing significant amounts of 137Cs and 210Pbex (e.g., site 2, Fig. 6). We calculated an average sedimentation rate for the 11 cores over the past 50 yrs giving mean annual sedimentation rates of 1.21±0.35 g cm-2 yr-1 using the aerial loading method and 1.15±0.46 g cm-2 yr-1 using the change in cumulative inventory method. The resultant mean sedimentation rates showed a good correlation, except for site 10 (r²=0.64, n=11, pb 0.01). Total 210Pbex inventories were 46.9 mBq cm-2 and 64.2 mBq cm-2, respectively, for reference sites 1 and 2 with a peak 210Pbex concentrations occurring at a depth of 12 cm and 8 cm for reference site 1 and 2 respectively, (Fig. 9). At the depositional sites, 210Pbex activities showed an exponential decay in 5 of the 11 cores and a more erratic but still downcore reduction in the other profiles (e.g. site 1 & 2, Fig. 6). Overall, the activities varied between 20 and 80 mBq g-1, with a clear subsurface peak in most profiles. The total inventory for 210Pbex for the depositional sites was between 131 - 434 mBq cm-2. These values are much higher than those measured at the two reference sites, implying that they are a result of net deposition (He and Walling, 1996). The calculated average sedimentation rates for the 11 cores over the last 100yrs using the 210Pbex inventory approach was 1.01±0.50 g cm-2 yr-1. The total inventory for 210Pbex and estimated average sedimentation rates over the past 50 years and 100 years for 137Cs and 210Pbex respectively compare quite well with

Table 2 Historical flood events based on satellite imagery. The images are biased to include only cloud free images which in some cases did not exactly coincide with the actual flooding event. None the less, based on these images a relatively low percentage of the total floodplain has been flooded in the last 30 yrs Sediment flux estimates are based on actual measurements made after the 2013 flood event. 15 day flow (m3s-1)

Images

Flooded Area

Date

Type

Mean

Max

Total flow (*106 m3)

(km2)

% of floodplain

Sediment flux (Mt)

4/10/1985 1/23/1986 4/29/1986 5/18/1990 12/10/2001 4/6/2002 12/16/2003 12/2/2004 1/9/2007 12/8/2009 4/7/2010 8/17/2013

L5 TM L5 TM L5 TM L4 TM L7 ETM L7 ETM L7 ETM L7 ETM L7 ETM L7 ETM L7 ETM L7 ETM

187.4 81.3 191.3 450.5 129.6 329.8 337.0 508.1 661.3 154.1 148.7 920.0

642.3 96.4 345.7 547.2 229.8 456.6 441.4 628.7 1386.1 227.5 252.2 1332.7

242.9 105.4 247.9 583.9 168.0 427.4 436.8 658.5 857.0 199.7 192.8 1192.3

16.9 17.4 23.0 94.3 17.4 86.7 70.0 67.0 209.0 8.9 16.3 356.0

1.7 1.7 2.3 9.4 1.7 8.7 7.0 6.7 20.9 0.9 1.6 35.7

0.08 ±0.06 0.08 ±0.06 0.11 ±0.08 0.46 ±0.33 0.08 ±0.06 0.42 ±0.30 0.34 ±0.25 0.32 ±0.23 1.01 ±0.73 0.04 ±0.03 0.08 ±0.06 1.72 ±1.25


63

Fig. 5. Overbank sediment deposition rates determined from post event survey after the 2013 flood event. Despite the high variability in the sediment deposition rates within the floodplain, there is a general trend indicating decreasing sediment deposition rates with increasing distance from the main River channel.

inventory and sedimentation rates reported in other studies in equatorial and Southern hemisphere systems, e.g. inventories of between 11-70 mBq cm-2 for 137Cs and 433-616 mBq cm-2 for 210Pbex with sedimentation rates of 2-5 mm yr-1 in the Mkuze River in South Africa and Kaleya River in Zambia (Collins et al., 2001; Humphries et al., 2010). 3.3. Floodplain extent and flooding frequencies To obtain a first estimate of the inundated areas at different water levels, we examined the relationship between discharge (Q) in Garissa and the total flooded area as derived from satellite imagery (Fig. 10). An excellent positive correlation was found between the cumulative Q over the 15-days preceding the imagery date (Q15) and the total flooded

area (Table 2 and Fig. 11a). Classification of images from the recent extensive flooding of 2013 indicated that the flooded area reached 40% (i.e. 356 km²) of the active floodplain area (Table 2, Fig. 10b). We used this relationship to estimate (i) flooding frequency since 1941 and (ii) the area flooded in each flood since 1941. Based upon the relationship, we only accounted for cumulative discharge likely to flood more than 5% of the total floodplain area, corresponding to a Q15 of 0.4 km3. In total we identified 77 significant floods between 1941 and 2013, corresponding to a flooding frequency of 1.05 floods yr-1. The estimated flooded area varies widely and is at a mean of 168±144 km² with a maximum of 852 km² (Fig. 11b). The above data were used to make a first estimate of total sediment deposition within the Tana floodplain by multiplying the area flooded

Table 3 A summary of sedimentation rates calculated from 210Pbex, 137Cs activities of Tana River sediments and overbank fresh sediment deposits after the 2013 flood. We used the aerial loading (Walling et al., 1999), for 137Cs with first appearance, 1955 (137Cs-A) and change in cumulative inventory (137Cs-B) . 210Pbex was calculated using inventory method (He and Walling, 1996). Fresh sediment accretion rates are from post event surveys after the 2013 flood and only points that coincided with the coring locations are shown here. Site

Site 1 Site 2 Site 3 Site 4 Site 5 Site 6 Site 7 Site 8 Site 9 Site 10 Site 11

Wenje Lake Gadia Kajitupe Kiluluni III LB Makere IV Maroni II Kiozani II Mnazini II Kone II Mnazini I Maroni III Mean±SD

Inventories (mBq cm-2)

Sedimentation rates (g cm-2 yr-1)

137

137

Cs

13.3 10.1 13.6 10.5 21.8 10.9 14.1 11.6 3.6 5.3 9.8 11±5

210

Pbex

232.2 175.4 259.2 434.9 423.0 235.1 298.5 226.9 130.9 169.8 213.3 255±97

Cs-A

0.83 0.88 1.42 1.35 1.51 0.84 1.35 1.77 0.65 1.28 1.40 1.2±0.4

137

Cs-B

0.77 0.91 1.60 1.68 1.49 0.89 1.26 1.74 0.50 0.49 1.34 1.15±0.5

210

Pbex

Fresh sediment

0.89 0.60 1.03 1.94 1.88 0.91 1.24 0.86 0.37 0.57 0.79 1.01±0.5

1.50 0.82 1.02 1.50 1.16 1.02 0.82 0.41 0.54 0.54 0.34 0.88±0.4

64


Fig. 6. 137Cs, 210Pbex depth profiles and particle size distribution for site 1 & 2, Gadia & Wenje L. Error bars represent uncertainties at 95% confidence interval.

with an average sedimentation rate, and based on field observations of sediment deposition during the 2013 flood. We first assumed an average deposition rate of 0.58 g cm-2 (4.3 mm) during each flood (Fig. 5). This is evidently a simplification as it may be expected that, during larger floods, average deposition rates will be higher (due to greater average flooding heights and possibly higher sediment concentrations) but data to account for this variation are lacking. We obtained a total floodplain deposition of 75 Mt for the period 1941-2013 or an average sediment accumulation rate of 0.97 Mt yr-1. However, if the average sedimentation calculated from radionuclide inventories is used, total floodplain

sedimentation between 1941 and 2013 is estimated at 156 Mt with an average annual sedimentation of 2 Mt yr-1. 4. Discussion 4.1. 4.1 Event-based sediment delivery and floodplain sediment storage Fresh sediment deposits from the post-event survey indicate higher sediment accumulation rates at points closer to the main river (b 500 m) with sedimentation rates decreasing as one moves

Fig. 7. Site 2, 137Cs, 210Pbex depth profiles and particle size distribution for site 3 & 4, Kiluluni III & Kajitupe. Error bars represent uncertainties at 95% confidence interval.


Fig. 8. Site 2, 137Cs, 210Pbex depth profiles and particle size distribution for site 5 & 6, LB Makere & Maroni II. Error bars represent uncertainties at 95% confidence interval.

Fig. 9. 210Pbex depth profile for the reference sites. Only particles b63μm were analysed. Error bars represent uncertainties at 95% confidence interval.

65

66


Fig. 10. A flooding extent map based on satellite imagery A- 01/09/2007 and B-08/17/2013, the total flooded area based on the images strongly correlated with 15-day total flow and mean discharge at Garissa (preceding the date of the image).

further from the river (Fig. 5). This points to a rapid sediment deposition during flood events and limited sediment conveyance to more distal portions of the floodplain. No consistent variation in deposition rates was observed along the downstream gradient in sampling locations. However, at each floodplain location, sedimentation rates were found to be spatially variable. The observed variability is due

to a number of factors. The microtopography in the floodplain is diverse and strong differences in the type and density of vegetation occur over short distances. The existence of old abandoned river beds which may become active during flood events can also be linked to transport of sediments over great distances across the floodplain as has been observed by other studies (Day et al., 2008).

Fig. 11. (a) A regression analysis between the flooded areas and the 15-day total flow at Garissa (trend line is a polynomial fit). (b) Historical flooding records (Garissa-Garsen) based on a polynomial regression of flooded areas from satellite imagery and stage height data as recorded at Garissa (1941-2013).


Relatively low 137Cs inventories were prevalent in most floodplain locations. This is expected given the location at the equator, a low fallout area (Foster et al., 2007; Owens and Walling, 1996). Another notable pattern is that despite the fact that grain size distributions were relatively homogenous down the profile at most depositional areas, we observed an erratic pattern and the existence of multiple peaks in some profiles and the absence of detectable radionuclide activity in some “fresh” deposits. This pattern is not limited to the Tana system, other studies also noticed that radionuclide deposits in alluvial settings can be irregular, even in the absence of significant erosion (Aalto and Nittrouer, 2012). This variability cannot be explained by vertical mobility of radionuclides within the profile but is related to temporal variations in the radionuclide content of the deposited sediments. The latter is due to the fact that the sources of Tana River sediment can be variable and while a significant part of the river sediment may, on average, be sourced from surface soils (and therefore contain detectable loadings of radionuclides), it has recently become clear that autogenic processes (such as bank erosion and river bed deepening) contribute to sediment mobilisation in the lower Tana River system (Geeraert et al., 2015). The sediment mobilised by these processes may not have a detectable radionuclide loading as they were isolated from atmospheric radionuclide deposition. Given that sediment sources may vary from one event to the next, the variations in the radionuclide loading of deposited sediment are to be expected. We were not able to assess this variability at the level of individual floods, as some vertical mixing following subsequent flood events had occurred in majority of the sites. Despite these issues, the average sedimentation rates calculated from the radionuclide profiles using three different methods are very similar (1.15, 1.21 and 1.01 g cm-2 yr-1, Table 3). Overall, the average event based sedimentation rates measured after the 2013 flood are lower than the radionuclide-derived sedimentation rates (0.58 g cm-2 vs. 1.01-1.21 g cm-2 yr-1 respectively). On the other hand, the radionuclide-derived sedimentation rates were very similar to the average sedimentation rates derived from fresh deposits at the coring locations (0.88 g cm-2 vs. 1.01-1.21 g cm-2 yr-1, see Table. 3). The fact that average deposit-derived sedimentation rate at the coring sites was significantly higher than the overall average deposit-derived sedimentation rate (0.88 vs 0.58 g cm-2) can be partially explained by the fact that the coring sites were, on average closer to the river than the postflood survey points (528 m vs 716 m respectively), suggesting that the coring locations may be predominantly located at locations where sedimentation rates may indeed be higher than the average floodplain sedimentation rate. When the weighted flood frequency based on 6 flood events between 1990 and 2013 (Table 2) is correlated with distance to the main river, a mean flooding distance of 1223 m was obtained. This his suggests that average sedimentation rates within the floodplain may indeed be lower than those observed at the coring sites, and the average sedimentation rates are likely to be closer to the estimate derived from the post-flood survey. If our historical sedimentation rates are compared with historical estimates of annual fluxes at Garissa, which are at 6.6 Mt yr-1 (1950-2015, Geeraert et al., 2015), it becomes clear that between 15% (0.97 Mt yr-1) and 30% (2.0 Mt yr-1) of the fluxes at Garissa are stored in the floodplain. These high levels of sediment deposition observed in the Tana River floodplain are not unique. High retention rates have been noted in other large river systems with extensive floodplains and deltas e.g. deposition rates of 6-9 mm yr-1 and up to 40% of the total load within a 1 km corridor for Strickland and Fly River floodplains (Aalto and Dietrich, 2005; Swanson et al., 2008). 4.2. Sediment stocks It is interesting to compare our retention estimates with those obtained in earlier studies. If we consider only the 2013 flood event, a

67

sediment flux of 4.61 Mt was recorded at Garissa over a 61-day flood pulse (April 1st to 31st May 2013), while during the same period, only 1.31 Mt of sediment was measured at Garsen, the other gauging station downstream (Geeraert et al., 2015). This net loss of 3.30 Mt of sediment would therefore correspond to an average sedimentation rate of ca. 0.9 g cm-2 which is higher but of the same order of magnitude with the average deposit-derived sedimentation rate we observed in the 2013 flood event, (0.58±0.42 g cm-2, 4.3±3.1 mm vertical accretion), translating to 2.08 ± 1.50 Mt of sediment, based upon a flooded area of 356 km2. Considering that on average there is one flood per year in the Tana, the recorded floodplain deposition appears to be lower than overall sediment retention as estimated by Tamooh et al. (2014) from a bi-weekly sampling (between 2.7 and 5.6 Mt yr-1of sediment) between Garissa and Tana River Primate Reserve, which is located ca. 80 km upstream of Garsen. It should be kept in mind that our measurements do not account for sedimentation in areas where water is permanently present such as oxbow lakes. Sedimentation rates in such environments are expected to be higher than the average floodplain sedimentation rates as these locations are subject to higher than average flood heights. Another important issue is that sediment storage is not limited to the floodplain. Geeraert et al. (2015) observed that during the recession phase of a flood, the river cross-section at Garissa became progressively smaller and was reduced by ca. 60 m² as it was adjusting to the lower discharges. If such a reduction was to occur over the entire channel length between Garissa and Garsen (ca. 380 km), large amounts of sediment could be stored within the channel in the post-flood phase. Channel dynamics are therefore important in Tana River and may have a very high impact on short-term sediment storage and release. 4.3. Sediment residence time Although our current data allow to constrain average floodplain sedimentation rates in the lower Tana to a range between ca. 0.5-1.21 g cm2 -1 y (4-8 mm vertical accretion). These estimates obviously carry some uncertainty, given that they are based on an average sedimentation rate for the entire floodplain, or at specific floodplain location and they assume that sedimentation is evenly distributed across the floodplain, which is clearly a simplification. The calculated budgets cannot be assumed to represent net accretion as no field measurements were made to quantify sediment reworking. As it is acknowledged that accretion in the floodplain cannot continue unabated (Dietrich et al., 1999; Lauer and Parker, 2008), various internal mechanisms exist to limit the vertical growth of floodplains. Ndlovu (2013) studied lateral river migration upstream of Garissa concluding that the Tana River was migrating at a long-term average rate of 2.3 m yr-1. Previous studies have similarly demonstrated that there is frequent and rapid meandering of Tana River (DHV, 1986). The river migrations and the autogenic processes play an important role in compensating for some of the lost sediment. At the average migration rate of 2.3 m y-1 for example, it would take the 380 km long Tana River ca. 1000 years to rework its entire floodplain (ca. 900 km²) between Garissa and Garsen over a depth of 2-3 m (the average bank height), thereby effectively limiting the time period over which sediments can accumulate. 5. Conclusions The interplay between active sediment deposition and alternating flooding and dry periods has shaped the geomorphology of the Tana River floodplain. The floodplain is highly dynamic with shallow depositional profiles but reflecting imprints of relatively old past and current erosion processes. Based on estimates of local sedimentation rates from radionuclide inventories, post-flood surveys, we estimate current floodplain sedimentation rates to be, on average, ca. 1-2 Mt yr-1, corresponding to local sedimentation rates of ca. 1-1.2 g cm yr-1 and resulting

68


in deposition of upto 30% of sediments entering the river reach at Garissa. Comparison of floodplain deposition rates with differences between incoming and outgoing sediment fluxes suggests that important deposition also occurs in other environments within the river, including old river branches (oxbow lakes) as well as the current river channel (especially during flood recession). Despite the fact that floodplain deposits are texturally quite homogeneous, the source of the sediments that is deposited in the Tana floodplain is variable, which is illustrated by the erratic variations in their radionuclide content with depth. Important controls on local sedimentation rates are the distance to the river, the flood height and local microtopography. While the total amount of sediment deposited by the Tana is important, other floodplain processes such as lateral river migration and within-channel deposition and remobilisation are also relevant for the overall sediment budget of the lower Tana River. Acknowledgements This work was supported by the Research Foundation Flanders (FWOVlaanderen, project G024012N), ERC-StG 240002 (AFRIVAL, http://ees. kuleuven.be/project/afrival/) and the KULeuven Special Research Fund (BOF). We especially thank Lore Fondu and Zita Kelemen for the assistance with laboratory work, Kenya Wildlife Service (KWS) and Water Resources Management Authority (WRMA), Kenya for allowing access to different sites and availability of hydrological data, and the many field assistants who assisted in the coring and preparation of samples. Finally much appreciation to the journal editor and two anonymous reviewers whose suggestions helped to substantially improve the manuscript. Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.catena.2016.03.024. References Aalto, R., Nittrouer, C. a, 2012. 210Pb geochronology of flood events in large tropical river systems. Philos. Transact. A Math. Phys. Eng. Sci. 370, 2040–2074. http://dx.doi.org/ 10.1098/rsta.2011.0607. Aalto, R., Dietrich, W., 2005. Sediment accumulation determined with 210Pb geochronology for Strickland River flood plains. Papua New Guinea. Sediment budgets, pp. 1–7. Aalto, R., Maurice-bourgoin, L., Dunne, T., 2003. Episodic sediment accumulation on Amazonian flood plains influenced Southern Oscillation by El Nin. Adv. Catal. 425, 493–497. http://dx.doi.org/10.1038/nature01990.1. Amos, K., Croke, J., 2009. The application of caesium-137 measurements to investigate floodplain deposition in a large semi-arid catchment in Queensland, Australia: a low-fallout environment. Earth Surf. 529, 515–529. http://dx.doi.org/10.1002/esp. Bouillon, S., Abril, G., Borges, a.V., Dehairs, F., Govers, G., Hughes, H.J., Merckx, R., Meysman, F.J.R., Nyunja, J., Osburn, C., Middelburg, J.J., 2009. Distribution, origin and cycling of carbon in the Tana River (Kenya): a dry season basin-scale survey from headwaters to the delta. Biogeosciences 6, 2475–2493. http://dx.doi.org/10. 5194/bg-6-2475-2009. Brown, T., Schneider, H., 1996. Multi-scale estimates of erosion and sediment yields in the Upper Tana basin. IAHS Publications-Series of, Kenya, pp. 49–54. Chander, G., Markham, B.L., Helder, D.L., 2009. Summary of current radiometric calibration coefficients for Landsat MSS, TM, ETM+, and EO-1 ALI sensors. Remote Sens. Environ. 113, 893–903. http://dx.doi.org/10.1016/j.rse.2009.01.007. Dagg, M., Woodhead, T., Rijks, D.a., 1970. Evaporation in East Africa. International Association of Scientific Hydrology. Bulletin 15, 61–67. http://dx.doi.org/10.1080/ 02626667009493932. Day, G., Dietrich, W.E., Rowland, J.C., Marshall, A., 2008. The depositional web on the floodplain of the Fly River, Papua New Guinea. J. Geophys. Res. 113, F01S02. http:// dx.doi.org/10.1029/2006JF000622. DHV, 1986. Tana River Morphology Studies, Delft Hydraulics Laboratory. Delft, Netherlands. Dietrich, W.E., Day, G., Parker, G., 1999. The Fly River. Papua New Guinea: Inferences about river dynamics, floodplain sedimentation and fate of sediment. Varieties of Fluvial Form, pp. 345–376. Foster, I.D.L., Boardman, J., Keay-Bright, J., 2007. Sediment tracing and environmental history for two small catchments, Karoo Uplands, South Africa. Geomorphology 90, 126–143. http://dx.doi.org/10.1016/j.geomorph.2007.01.011. Geeraert, N., Omengo, F.O., Tamooh, F., Paron, P., Bouillon, S., Govers, G., 2015. Sediment yield of the lower Tana River, Kenya, is insensitive to dam construction: sediment mobilization processes in a semi-arid tropical river system. Earth Surf. Process. Landf. (n/a–n/a. doi:10.1002/esp.3763).

Collins, a.L., Walling, D.E., Sichingabula, H.M., Leeks, G.J.L., 2001. Using 137Cs measurements to quantify soil erosion and redistribution rates for areas under different land use in the Upper Kaleya River basin, southern Zambia. Geoderma 104, 299–323. http://dx.doi.org/10.1016/S0016-7061(01)00087-8. Gilmore, G.R., 2008. Practical Gamma-Ray Spectrometry. John Wiley & Sons, Ltd, Chichester, UK. http://dx.doi.org/10.1002/9780470861981. Gomez, B., Mertes, L.A.K., Phillips, J.D., Magilligan, F.J., James, L.A., 1995. Sediment characteristics of an extreme flood: 1993 upper Mississippi River valley. Geology 23, 963–966. Guelle, W., Balkanski, Y.J., 1998. Wet deposition in a global size-dependent aerosol transport model 1. Comparison of a I year løpb simulation with ground measurements, p. 103. He, Q., Walling, D.E., 1996. Use of Fallout Pb-210 Measurements to Investigate LongerTerm Rates and Patterns of Overbank Sediment Deposition on the Floodplains of Lowland Rivers. Earth Surf. Process. Landf. 21, 141–154. Hoffman, D.F., Gabet, E.J., 2007. Effects of sediment pulses on channel morphology in a gravel-bed river. Geol. Soc. Am. Bull. 119, 116–125. http://dx.doi.org/10.1130/ B25982.1. Hoffmann, T., 2015. Sediment residence time and connectivity in non-equilibrium and transient geomorphic systems. Earth Sci. Rev. 150, 609–627. http://dx.doi.org/10. 1016/j.earscirev.2015.07.008. Hughes, F.M.R., 1988. ecology of African floodplain forests in semi-arid and arid zones: a review. J. Biogeogr. Hughes, A.O., Olley, J.M., Croke, J.C., Webster, I.T., 2009. Determining floodplain sedimentation rates using 137Cs in a low fallout environment dominated by channel- and cultivation-derived sediment inputs, central Queensland, Australia. J. Environ. Radioact. 100, 858–865. http://dx.doi.org/10.1016/j.jenvrad.2009.06.011. Humphries, M.S., Kindness, A., Ellery, W.N., Hughes, J.C., Benitez-Nelson, C.R., 2010. 137Cs and 210Pb derived sediment accumulation rates and their role in the long-term development of the Mkuze River floodplain, South Africa. Geomorphology 119, 88–96. http://dx.doi.org/10.1016/j.geomorph.2010.03.003. Kitheka, J.U., Obiero, M., Nthenge, P., 2005. River discharge, sediment transport and exchange in the Tana Estuary, Kenya. Estuar. Coast. Shelf Sci. 63, 455–468. http://dx. doi.org/10.1016/j.ecss.2004.11.011. Lauer, J.W., Parker, G., 2008. Net local removal of floodplain sediment by river meander migration. Geomorphology 96, 123–149. http://dx.doi.org/10.1016/j.geomorph. 2007.08.003. Leslie, C., Hancock, G.J., 2008. Estimating the date corresponding to the horizon of the first detection of 137Cs and 239+ 240Pu in sediment cores. J. Environ. Radioact. 99, 483–490. http://dx.doi.org/10.1016/j.jenvrad.2007.08.016. Maingi, J.K., Marsh, S.E., 2002. Quantifying hydrologic impacts following dam construction along the Tana River, Kenya. J. Arid Environ. 50, 53–79. http://dx.doi.org/10.1006/ jare.2000.0860. Medley, K.E., 1992. Patterns of forest diversity along the Tana River, Kenya. J. Trop. Ecol. 8, 353–371. Ndlovu, S.A., 2013. The Hydro-morphological Evolution of the Lower Tana River : 19752010. UNESCO-IHE, The Netherlands. Noe, G.B., Hupp, C.R., 2009. Retention of Riverine Sediment and Nutrient Loads by Coastal Plain Floodplains. Ecosystems 12, 728–746. http://dx.doi.org/10.1007/s10021-0099253-5. Notebaert, B., Verstraeten, G., Rommens, T., Vanmontfort, B., Govers, G., Poesen, J., 2009. Establishing a Holocene sediment budget for the river Dijle. Catena 77, 150–163. http://dx.doi.org/10.1016/j.catena.2008.02.001. Owens, P.N., Walling, D.E., 1996. Spatial variability of caesium-137 inventories at reference sites: An example from two contrasting sites in England and Zimbabwe. Appl. Radiat. Isot. 47, 699–707. http://dx.doi.org/10.1016/0969-8043(96)00015-2. Pacini, N., Harper, D., Mavuti, K., 1999. Hydrological and ecological considerations in the management of a catchment controlled by a reservoir cascade: the Tana River, Kenya. Preiss, N., Melieres, M.A., Pourchet, M., 1996. A compilation of data on lead 210 concentration in surface air and fluxes at the air-surface and water-sediment interfaces. J. Geophys. Res. 101, 28847–28862. Saenyi, W.W., Chemelil, M.C., 2003. Modelling of Suspended Sediment Discharge for Masinga Catchment Reservoir in Kenya. J. Civ. Eng. JKUAT 8, 89–98. http://dx.doi. org/10.4314/jce.v8i1.18997. Song, C., Woodcock, C., Seto, K.C., Lenney, M.P., Macomber, S. a, 2001. Classification and change detection using Landsat TM Data- When and how to correct atmospheric effects? Remote Sens. Environ. 75, 230–244. http://dx.doi.org/10.1016/S00344257(00)00169-3. Swanson, K.M., Watson, E., Aalto, R., Lauer, J.W., Bera, M.T., Marshall, A., Taylor, M.P., Apte, S.C., Dietrich, W.E., 2008. Sediment load and floodplain deposition rates: Comparison of the Fly and Strickland rivers, Papua New Guinea. J. Geophys. Res. 113, F01S03. http://dx.doi.org/10.1029/2006JF000623. Syvitski, J.P.M., Vörösmarty, C.J., Kettner, A.J., Green, P., 2005. Impact of humans on the flux of terrestrial sediment to the global coastal ocean. Science (New York, N.Y.) 308, 376–380. http://dx.doi.org/10.1126/science.1109454. Tamooh, F., Borges, a.V., Meysman, F.J.R., Van Den Meersche, K., Dehairs, F., Merckx, R., Bouillon, S., 2013. Dynamics of dissolved inorganic carbon and aquatic metabolism in the Tana River basin, Kenya. Biogeosciences 10, 6911–6928. http://dx.doi.org/10. 5194/bg-10-6911-2013. Tamooh, F., Meysman, F.J.R., Borges, A.V., Marwick, T.R., Van Den Meersche, K., Dehairs, F., Merckx, R., Bouillon, S., 2014. Sediment and carbon fluxes along a longitudinal gradient in the lower Tana River (Kenya). J. Geophys. Res. Biogeosci. (n/a–n/a. doi: 10.1002/2013JG002358). Wallbrink, P.J., Murray, A.S., Olley, J.M., 1998. Determining sources and transit times of suspended sediment in the Murrumbidgee River, New South Wales, Australia, using fallout 137Cs and 210Pb. Water Resour. Res. 34, 879–887. http://dx.doi.org/ 10.1029/97WR03471.

F.O. Omengo et al. / Catena 143 (2016) 57–69 Walling, D.E., He, Q., 1997. Use of fallout 137Cs in investigations of overbank sediment deposition on river floodplains. Catena 29, 263–282. http://dx.doi.org/10.1016/S03418162(96)00072-0. Walling, D.E., Owens, P.N., Leeks, G.J.L., 1997. The characteristics of overbank deposits associated with a major flood event in the catchment of the River Ouse, Yorkshire, UK. Catena 31, 53–75. http://dx.doi.org/10.1016/S0341-8162(97)00034-9. Walling, D.E., Owens, P.N., Leeks, G.J.L., 1999. Rates of contemporary overbank sedimentation and sediment storage on the Floodplains of the main channel systems of the Yorkshire Ouse and River Tweed , UK. p. 1009.

69

Walling, D., Russell, M., Hodgkinson, R., Zhang, Y., 2002. Establishing sediment budgets for two small lowland agricultural catchments in the UK. Catena 47, 323–353. http://dx. doi.org/10.1016/S0341-8162(01)00187-4. Zapata, F., 2002. Handbook for the assessment of soil erosion and sedimentation using environmental radionuclides. Springer.