Exploring the hydrology and biogeochemistry of the dam-impacted Kafue River and Kafue Flats (Zambia)

Physics and Chemistry of the Earth 36 (2011) 775–788

Contents lists available at SciVerse ScienceDirect

Physics and Chemistry of the Earth journal homepage: www.elsevier.com/locate/pce

Exploring the hydrology and biogeochemistry of the dam-impacted Kafue River and Kafue Flats (Zambia) J. Wamulume a,b,⇑, J. Landert b, R. Zurbrügg b,c, I. Nyambe a, B. Wehrli b,c, D.B. Senn b,c a

Integrated Water Resources Management Center, School of Mines, University of Zambia, P.O. Box 32379, 10101 Lusaka, Zambia Eawag, Surface Waters, Seestrasse 79, 6047 Kastanienbaum, Switzerland c Institute of Biogeochemistry and Pollution Dynamics, ETH Zurich, Universitätstrasse 16, 8092 Zurich, Switzerland b

a r t i c l e

i n f o

Article history: Available online 6 August 2011 Keywords: Carbon export Floodplain hydrology Nutrients River Water balance Tropical wetland

a b s t r a c t Wetland processes are strongly influenced by hydrologic factors such as precipitation, surface runoff, and flooding dynamics. Anthropogenic disturbances to flooding regimes can thus substantially alter wetland habitat and biogeochemistry. The Kafue Flats, a large floodplain (6500 km2) along the Kafue River in South-Central Zambia, is a wetland impacted by upstream and downstream hydropower dams. The main purpose of this study was to develop a water budget for the Kafue Flats under current conditions, quantify nutrient and organic carbon concentrations in the river, and use the combined information to estimate biogeochemical budgets. A water balance was developed for the Kafue Flats at a subcatchment scale for the years 2002–2009 using daily hydrological data. In addition, bi-monthly flow and chemical measurements were performed over 1 year (May 2008–May 2009) at multiple stations. Evapotranspiration was an important process in the Flats, accounting for up to 49% of total hydrologic outputs in some subcatchments. Direct precipitation contributes substantial to water inputs to the flats: runoff from the upstream catchment accounted for 45% of water inputs to the Kafue Flats, while the remaining 55% came from direct precipitation to the Kafue Flats from its subcatchment. Estimates from the wet season suggest that 75% of the water flowing in the river’s main channel as it exits the Flats spent some time within the highly productive floodplain. This exchange between the floodplain and the river appeared to play an important role in nutrient and carbon export to the river’s main channel and out of the wetland. The floodplain was a net source of phosphate (220 t/year), total nitrogen (1300 t N/year, of which 90% was organic nitrogen) and total organic carbon (50,000 t C/year) to downstream systems. Thus, when considering dam impacts and altered flooding dynamics in this system, potential changes to carbon and nutrient cycling also need to be taken in to consideration, which may have implications for nutrient availability within the Kafue Flats and nutrient export to downstream systems. Ó 2011 Elsevier Ltd. All rights reserved.

1. Introduction Hydrology is an important driver of biogeochemical processes in floodplain ecosystems (Junk et al., 1989; Tockner et al., 2000). Important floodplain processes such as particle deposition (OldeVenterink et al., 2006), organic matter mobilization, and nutrient turnover (Baldwin and Mitchell, 2000) are governed by the exchange between river and floodplain. In tropical areas with distinct rainy and dry seasons, floodplains are often seasonal wetlands with high productivity (Neue et al., 1997) and high biodiversity. While the effect of hydrological exchange between river and floodplain on biogeochemistry has received substantial attention in temperate systems (e.g., Tockner et al., 1999; Wiegner and Seitzinger, 2004; Hunsinger et al., 2010), studies in tropical systems ⇑ Corresponding author at: Integrated Water Resources Management Center, School of Mines, University of Zambia, P.O. Box 32379, 10101 Lusaka, Zambia. E-mail address: [email protected] (J. Wamulume). 1474-7065/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.pce.2011.07.049

are sparse (e.g., Bouillon et al., 2007; Nwankwor and Anyaogu, 2000). Studies in the Okavango delta have shown that organic matter mobilization and transport is a direct effect of hydrological river–floodplain exchange (Mladenov et al., 2005, 2007). We explored interactions between the river and the floodplain and their influence on organic carbon (OC) and nutrient exports in the Kafue River and Kafue Flats system in southern Zambia (Fig. 1). The Kafue River Basin (KRB) is the most economically active basin in Zambia and a major tributary of the Zambezi River (Fig. 1). Anthropogenic water uses in the basin are dominated by hydropower, mining, and irrigation. These activities, in particular mining and hydropower, have led to degraded water quality along some stretches as well as altered hydrology in the KRB (Kambole, 2002). Two large dams, Itezhi Tezhi (ITT) and Kafue Gorge (KG), regulate the flow through the lower KRB. The Kafue Flats, a large (6500 km2) and high-value ecosystem (Ramsar site), lies between these two dams. One of the major impacts of the Itezhi Tezhi and Kafue Gorge dams has been a change in the extent and duration

776

J. Wamulume et al. / Physics and Chemistry of the Earth 36 (2011) 775–788

Fig. 1. The Zambezi River Basin in Southern Africa and the location of the Kafue River Basin and the Kafue Flats.

of the flooding (Mumba and Thompson, 2005), and studies suggest that altered hydrology due to dam operation have substantially influenced this systems’ ecology (Dudley and Scully, 1980; Obrdlik et al., 1989). While the hydrology of the Kafue River Basin and Kafue Flats has been studied extensively, there is limited information on river–floodplain interactions and how this influences the biogeochemistry of Kafue Flats and the Kafue River (Salter, 1985). The main goals of this study were to (i) characterize the spatial and temporal variability in the relative importance of different water inputs (riverine input, direct precipitation, lateral inputs) and river–floodplain exchange in the Kafue Flats and (ii) characterize the temporal and spatial variability of N, C, and P concentration, speciation and loads over an annual cycle, in the Kafue River as it flows through the Kafue Flats. To explore these goals a network of river stations was established and bi-monthly field trips over the course of 1 year were conducted to measure discharge and collect water samples for chemical characterization. Using these measurements, combined with daily hydrological data collected by local agencies, we developed water balance, nutrient (N, P) and carbon loading estimates at subcatchments scale along the Kafue River. 2. Methods 2.1. Study area: the Kafue River and Kafue Flats floodplain system The Kafue River Basin (KRB) is one of the major sub-basins in the Zambezi Basin in Southern Africa (Fig. 1). The Kafue River is 1500 km in length and has a catchment area of 154,000 km2 (Obrdlik et al., 1989; Mumba and Thompson, 2005). Its drainage basin can be subdivided into three sub-basins: the upper half of the catchment area in the wetter northern part of Zambia, north

of the ITT dam; the central basin that includes the Kafue Flats and extends from ITT to the KG dam and the lower Kafue, downstream of KG dam, where the river drops steeply from the plateau and joins the Zambezi River (Pinay, 1988; Scott Wilson, 2003; Euro Consult Mott MacDonald, 2008). The KRB has a tropical climate with two distinct seasons, a wet season between November and March and a dry season between April and October. The daily mean temperature varies from 13 °C to 20 °C in July and from 21 °C to 30 °C in November (Yachiko, 1995). The average annual rainfall over the Kafue flats floodplain is 850 mm and is concentrated in the wet season (Yachiko, 1995). The study area for this project extends from Hook Bridge (K0, Fig. 2), upstream of ITT dam, up to the confluence with the Zambezi River (K7, Fig. 2). From ITT dam the river meanders for approximately 400 km through the Kafue Flats, over a west–east distance of 200 km. The river’s gradient through the Flats is extremely shallow, 0.04 m per km in the western half and 0.01 m per km in the eastern half, resulting in a total elevation drop of 10 m. (Ellenbroek, 1987; Pinay, 1988; Scott Wilson, 2003). The Kafue Flats wetland covers between 4400 and 6500 km2 (Dudley, 1979; Obrdlik et al., 1989; Scott Wilson, 2003; Mumba and Thompson, 2005; Ramsar, 2006), and the flooded area varies seasonally and interannually. There are seasonally flooded natural meadows, creeks, braided channels, back swamp levees, lagoons, and large flats (Welcomme, 1975; Ellenbroek, 1987; Ramsar, 1999). Most of the floodplain is covered with grass, and there are woodlands on higher grounds (Ellenbroek, 1987; Ramsar, 1999). The Kafue Flats floodplain extends 40–56 km at its widest point (Scott Wilson, 2003; Mumba and Thompson, 2005). Since the construction of the Itezhi Tezhi and Kafue Gorge dams in the 1970s the extent and duration of the flooding in the Kafue Flats has been altered (Mumba and Thompson, 2005).

J. Wamulume et al. / Physics and Chemistry of the Earth 36 (2011) 775–788

777

Fig. 2. Delineation of the Kafue Flats into three subcatchments CV1, CV2 and CV3, along with locations of stations for discharge measurement and water sampling (K0-K7).

2.2. Water balance For the purposes of the water balance, nutrient and carbon loading calculations, the area of the Kafue Flats and its lateral tributaries were divided into three subcatchments or control volumes (CV1, CV2, and CV3; Fig. 2). The subcatchments were delineated to coincide with existing flow gauging stations in the study area for which long term flow data or water elevation measurements were available. The stations were located at Itezhi Tezhi (K1), Namwala (K2), Nyimba (K3), Kasaka (K5) and Kafue Gorge dam (K6). Raster data for the study area was obtained from US Geological Survey Earth Resources Observation and Science (EROS). Stream network data was obtained from HYDRO 1 k and the USGS 3000 digital elevation model (DEM) of the world (GTOPO30) as well as the 300 SRTM DEM. The digital elevation model is corrected for flow and has a resolution of 1 km. The data layers were then imported into ArcGis 9.2 where delineation of the subcatchments was conducted. The created subcatchments were found to compare well with shape files obtained from GReSP (Ground Water Resources for Southern Province Project under Department of Water, Zambia). Each subcatchment was further divided into floodplain areas and nonfloodplain areas (Lehner and Döll, 2004) using the spatial analysis tool in ArcGis 9.2. A water balance was developed for each subcatchment for the period October 2002–May 2009 using daily precipitation, evapotranspiration and flow data. For each subcatchment the following water balance equation was used (Savenije and De Laat, 2002; Mitsch and Gosselink, 2007):

P  Aflooded þ Q gwi þ Q in þ Q lat  Et  Aflooded  Q out  Q gwo ¼ DS=Dt

ð1Þ

where DS/Dt is the change of storage per time (m3/d), P is the floodplain precipitation (m/d), Et is floodplain evapotranspiration (m/d), Qin and Qout (m3/d) are the flow into and out of the subcatchment respectively, along the Kafue River main channel; Aflooded (m2) is the maximum flooded area of the Kafue Flats within each subcatchment, and Qlat (m3/d) is the lateral flows from the non-flooded area of the control volume. Qgwi and Qgwo (m3/d) represent infiltration and exfiltration, respectively, and were assumed to be minor components and were not quantified. Daily measured estimates are available of P, Et, Qin and Qout.

We initially attempted to quantify Et over each subcatchment (flooded plus typically dry areas). However, large negative cumulative changes in storage during these initial modeling efforts indicated that evapotranspiration was being overestimated when the Et rate (even after applying land-cover correction factors) was applied over the entire subcatchment. The areas that experiences substantial flooding in the central KRB represent only 10–15% of each subcatchment. To address this problem, we divided each subcatchment into the maximum flooded area and the ‘‘dry’’ area. P and Et rates recorded at the Itezhi Tezhi (ITT) dam were applied directly to this maximum flooded area. Runoff (R) contributions draining from the dry areas to flooded areas were incorporated into the water balance as lateral flows, Qlat, calculated as

Q lat ¼ P  ðAsubcatch  Aflooded Þ  R

ð2Þ

where P is the precipitation rate measured at ITT dam, Asubcatch is the area of the entire subcatchment, and R is a runoff coefficient. R was assigned a value of 0.1, based on other studies in the area (Yachiko, 1995). The water balance was explored on monthly and annual time steps. Particular attention was directed toward evaluating the relative importance of the different input and output terms within each subcatchment because of the relevance of this information for understanding river–floodplain exchange and its effects on river chemistry. Water flowing along the main river channel is considered to be the only hydrologic pathway connecting the control volumes (CVs), e.g., water flowing from CV2 to CV3 through the floodplain is not explicitly considered. If the digital elevation model’s (DEM) resolution and accuracy were sufficiently accurate and the subcatchments were correctly drawn, this exchange should be zero. However, the available DEM had relatively coarse resolution and low accuracy considering the slight gradients in the Kafue Flats. Thus we cannot rule out some water exchange between the subcatchments within the floodplain. 2.3. Input data for water balance Data inputs for the water balance came from a variety of sources. Daily precipitation, pan evaporation, water elevation at multiple stations (K2, K3, K4), and flow rates at the dams, were obtained from the Zambia Electricity Supply Company (ZESCO) and Department of Water Affairs in Zambia (DWA). These agencies also

778

J. Wamulume et al. / Physics and Chemistry of the Earth 36 (2011) 775–788

had stage–discharge curves at several sites, which are discussed further below. Physically recorded precipitation data from ITT dam was used as the precipitation data. We also obtained remote sensed precipitation data from two sources, the Famine Early Warning System (FEWS; http://earlywarning.usgs.gov/adds/) and the Tropical Rainfall Measuring Mission (TRMM; http:// www.trmm.gsfc.nasa.gov/). FEWS and TRMM data were compared with physically recorded data at ITT dam to investigate spatial variations at the scale of the Kafue Flats (Wamulume, 2011). On a daily basis, FEWS data sampled near ITT (CV1) and ITT dam physically recorded data were not correlated (1-day FEWS) or weakly correlated (10-day and 30-day FEWS), and the slopes were substantially different from 1. TRMM was also poorly correlated with ITT. Interestingly, though, there was strong agreement between 10-day cumulative FEWS precipitation between all three control volumes, with slopes between 0.9 and 1.0; using 30-day data the slopes were all >0.95 and r2 > 0.9. Thus, the FEWS data could be interpreted as indicating low spatial variability in precipitation at time scales >10 days across the Kafue Flats. Given that there is limited ground-truthing of the FEWS data in this region, and the poor agreement between FEWS and ITT dam physically recorded data, we assert that the ITT dam data likely provides more accurate information. We therefore used physically recorded precipitation data from ITT dam for the water balance over the entire Kafue Flats. Since the FEWS data indicates that given the limited spatial variation indicated by the 10-day FEWS data at the scale of the entire Flats, we expect that applying the ITT dam precipitation measurements over the entire Kafue Flats is reasonably accurate and should introduce an acceptable level of uncertainty. Daily open surface evaporation (Eo) figures were obtained from physically measured records at ITT dam, performed by ZESCO. The Eo measurements were done using a class A evaporation pan. The assumption was made that the open surface evaporation measurements at ITT dam were the same as in the Kafue Flats, since there was no availability of daily surface evaporation in the study area. Using Penman relationship between Eo and potential evapotranspiration (Ep), Eo values were transformed to Ep.

Ep ¼ fp  Eo

ð3Þ

where fp is a reduction factor which varies according to location and season. We used fp = 0.8 (Ettrick, 1990) assuming high rates of evapotranspiration in the study. Additional correction factors Kp of 0.70 and 1.0 were adopted from Ellenbroek (1987) for non-floodplain and floodplain areas respectively. The correction factors were derived from Ellenbroek’s work on evapotranspiration in the Kafue Flats. Thus potential evapotranspiration was calculated using these correction factors shown below:

Ep ¼ fp  K p  Eo

2.4. Water quality measurements, nutrient and carbon loading A sampling network was designed to cover the central and lower KRB. The following criteria were used for selecting the station: location and co-occurrence with an existing discharge/stage measuring station; well mixed river; and accessibility. Seven sites were chosen (Fig. 2): Kafue River at Hook Bridge (K0), Itezhi Tezhi (K1), Namwala Pontoon (K2), Nyimba (K3), upstream Sable Farm (K4), Kasaka (K5) and Chiawa Pontoon (K7). Surface water samples were collected approximately bi-monthly over a period of 1 year (May 2008–June 2009) into sample-rinsed polyethylene bottles; acidified to pH 2 with 2 M HCl and filtered for the dissolved fraction (0.45 lm). The samples were placed on ice for 8–24 h and then frozen when brought to the laboratory. Nitrite, ammonium and ortho-phosphate were measured spectrophotometrically (DEW, 2002). Nitrate was measured using an Antek 745 Vanadium reduction unit, coupled to an Antek 9000 chemoluminescence detector (Braman and Hendrix, 1989). Total nitrogen (TN) and total phosphorus (TP) were determined by peroxidisulphate digestion and subsequent spectrophotometric nitrate/phosphate detection. A Shimadzu 5000 TOC analyzer was used to determine total organic carbon (TOC). Loads (i.e., mass per time) of nutrients and other compounds are a function of both concentration and river discharge, with stream discharge often being the dominant factor (Li et al., 2003). For this reason, a mass balance approach was used to estimate chemical loads to and from the Kafue River subcatchments. The same subcatchments or control volumes were used as for the water balance, control volumes (CV1, 2 and 3; Fig. 2). The mean of bi-monthly loads of nutrients and other compounds entering and leaving the control volumes via the river were calculated using the following general expression:

Q in  C in  Q out  C out ¼

X

lateral inputs X X  lateral outputs  sinks X þ sources

ð5Þ

where Qout and Qin are downstream and upstream flows, Cout and Cin are downstream and upstream concentrations in the river. Field data or calculated data (e.g., using a stage–discharge relationship) were available for the terms on the left side of the equation. On most dates, ADCP discharge measurements were available and used for Qin and Qout; when ADCP measurements were not available (K1, K2 and K3: May and June 2008; K3: February 2009; and K5: August and October 2008), a stage–discharge relationship was used to estimate the flow. No measured data are available to distinguish between the terms on the right side of Eq. (5). Thus, when there is an imbalance between river inputs and outputs (i.e., the left side of the equation) possible explanations based on various sources, sinks or lateral inputs/outputs will be discussed qualitatively.

ð4Þ

Between May 2008 and May 2009, discharge measurement were performed using an Acoustic Doppler Current profiler (ADCP, River Surveyor, SonTek v4.60) at water sampling stations and at stations with existing rating curves (Fig. 2). Flow measurements were conducted approximately bi-monthly at each of the flow monitoring stations in order to check the accuracy of the existing rating curves. For every station 3–6 single transects were measured with a precision of 2–10%. ADCP data were processed using River Surveyor v4.60. All transects were corrected for a vertical bank slope. Daily stage data was obtained from DWA and ZESCO for all the sampling points. Rating curves that existed on the stations had been developed either by DWA and ZESCO. Newly measured flows (using ADCP) were compared against calculated discharge (rating curves) and where necessary new rating curves were calculated.

3. Results 3.1. Evaluation of stage–discharge relationships Measured discharge and gauge plate readings for measurements conducted in 2008 and 2009 are presented in Table 1. Comparison of measured discharge with the discharge calculated using the rating curves (where curves were available) shows that the rating curves are still accurate at some stations such as K0, K1, and K2 (Figs. 2 and 3). However, conditions at K3 and K5 have changed considerably (Fig. 3). These changes are most likely due to backwater effects from the installation of Kafue Gorge dam, which other studies have suggested can be observed as far upstream as K3 (Mumba and Thompson, 2005).

779

J. Wamulume et al. / Physics and Chemistry of the Earth 36 (2011) 775–788 Table 1 Discharge measurements conducted during the study.

1

Station

Date

Gauge plate reading (m)

Measured m3/s (ADCP)

Kafue River at Hook Bridge (K0)

18.08.2008 19.10.2008 10.12.2008 23.02.2009

1.8 1.5 1.9 3.2

77 30 96 650

Kafue River at ITT ZESCO (K1) (Datum1 980 m.a.s.l.)

18.08.2008 20.10.2008 10.12.2008 24.02.2009 06.05.2010

3.4 4.8 5.2 7.7 8.0

78 207 238 489 531

Kafue River at Namwala Pontoon (K2) (Datum 980 m.a.s.l.)

19.08.2008 20.10.2008 11.12.2008 24.02.2009 07.05.2010

3.3 n/a 4.9 8.1 8.1

58 217 251 752 728

Kafue River at Nyimba (K3) (Datum 975 m.a.s.l.)

21.08.2008 28.10.2008 19.12.2008 11.05.2010

4.3 4.9 5.6 7.0

54 166 165 217

Kafue River upstream Sable Sugar (K4) Pumping point (Datum 971 m.a.s.l.)

25.06.2008 20.08.2008 18.12.2008

no GP no GP no GP

328 153 169

Between Sable and Mazabuka

25.06.2008

no GP

306

Kafue River at Kasaka (K5) (Datum 969 m.a.s.l.)

27.06.2008 28.05.2008 20.12.2008 26.02.2009 30.05.2009 29.05.2010

8.4 8.0 7.4 7.4 8.1 8.6

440 720 257 260 505 849

Kafue River at Chiawa Pontoon (K7)

30.05.2008 27.06.2008 20.12.2008

6.0 5.7 5.4

574 444 267

The Datum’s are based on FAO/UNDP (1968), Kafue Basin.

The poor agreement between measured flows and existing stage–discharge curves at K3 and K5 indicate that new stage–discharge relationships needed to be determined before existing daily water elevation data could be used to calculate discharge at these stations. At Nyimba a reasonably good linear fit was obtained for measured discharge vs. elevation (Fig. 3, r2 = 0.91). For these flow measurements, water elevations ranged from 4.73 m to 6.72 m. We compared this range with water elevations over the modeling period (2002–2008) and found that over 85% of the water elevations during this time period fall within this range. While under ideal circumstances additional flow measurements would be used to develop a more accurate stage–discharge relationship, given the altered flow conditions, given the reasonably good fit for the new stage–discharge relationship and the representativeness of the water elevation range, the new stage discharge relationship was used for quantifying flows at Nyimba for the model period. A new stage–discharge relationship was also calculated for K5 (Fig. 3). Approximately 20% of the water levels observed from 2002 to 2009 were less than the minimum value for which we have new discharge measurements; only 3% exceeded the maximum level we observed. While a simple linear regression and best-fit power equations yielded reasonable fits to the data (r2 = 0.74 for both), both have the potential to substantially overestimate flows when stage is less than 7 m. Therefore, another curve that provided a sharper decrease in Q below water levels of 7 m was also fit to the data and provided a reasonably good fit (r2 = 0.64). Measured discharge was quite sensitive to small differences in water elevation and given the modest r2 of the new stage–discharge relationship, estimated flows likely have a higher degree of uncertainty than at other stations. Kasaka is relatively close to the outlet of the Kafue Gorge dam, where flow is known with a higher degree of

certainty. In Fact, water balance estimates suggest that the calculated discharge at K5 substantially overestimated the amount of water leaving the system, and thus flows at K6 were used for subsequent calculations. 3.2. Precipitation and discharge time series in the middle Kafue River (2002–2009) Precipitation and discharge time series for several stations along the middle and lower Kafue River are presented in Fig. 4. The discharges at K1 and K6 were based on recorded daily flows by ZESCO, the dam operators. At the other river stations (K2, K3, K5), discharges were calculated from daily water level data using previously established stage discharge curves that were verified with ADCP measurements (K0, K2), or using newly established stage– discharge relationships (Section 3.1; Fig. 3). Continuous water level data sets were available for all stations from 2002 to 2009, except for 6 months of missing data at K3 (28.06.2006 to 01.12.2006). Comparing K0 and K1, it can be seen that in general between December and May the dam was filling up, and from June to December the ITT dam was releasing excess flows downstream (Fig. 4B). The peak flows at K0 and K1 were also well synchronized. The graph shows, as expected, that ITT dam has greatly increased the low flows downstream of the dam (Mumba and Thompson, 2005). The minimum flows at K0 were as low as 20 m3/s while the flows at the dam outlets were always above 110 m3/s over the same period (2002–2009).The very similar flows at K1 and K2 indicate that most of the water at K2 originates from the dam, except for modest differences (max  100 m3/s) in the wet season that must arise from lateral inflows (Fig. 4C). This also shows that channel flow is the dominant flow type in this stretch

780

J. Wamulume et al. / Physics and Chemistry of the Earth 36 (2011) 775–788

Rating curve Itezhi Tezhi

4

8

9

7

8

3 3

Q = 127.96*(h-1.036) 1.988

2 2 1

6 5 4

2 1

0

0 400

600

800

Q = 45.458*(h-1.777) 1.30

3

1 200

Q = 73.631*(h-2.379) 1.30

7 6 5 4 3 2 1 0

0

1000

gauge height (m)

10

0

150

Discharge (m3/s)

300

450

600

Discharge (m3/s)

Rating curve Nyimba

0

100 200 300 400 500 600 700 800

Discharge (m3/s)

Rating curve Kasaka

8

10 y = 0.0154x + 3.2996 R 2 = 0.9185

9

gauge height (m)

7

gauge height (m)

Rating curve Namwala Pontoon

9

gauge height (m)

gauge height (m)

Rating curve Hook Bridge 4

6 5 4 Q = 3.798*(h-0.440) 2.531

3 2 1

Q=0.8261*(h-3.226)4.057

8 7 6 5

Q = 50.704*(h-3.226)2.015

4 3

0

2 0

100

200

300

Discharge (m3/s)

400

0

200

400

600

800

1000

Discharge (m3/s)

Fig. 3. Discharge measurements (red dots) performed at Hook Bridge (K0), Itezhi Tezhi (K1), Namwala (K2), Nyimba (K3) and Kasaka (K5) during the study and the original rating curves that were obtained from DWA and ZESCO. At Kasaka, alternative rating curves are shown that were fitted to the discharge data (orange and green line). The rating curve shown in green (Q = 0.8261  (h  3.226)4.057) was applied to the Kasaka stage data. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

of the river. The flood plain area in this section is also relatively small (544 km2) compared to the total subcatchment area of 11,000 km2. The relative importance of lateral inflows to the overall flow depends on operating decisions at Itezhi Tezhi dam; for example, if releases from Itezhi Tezhi dam are small during a particularly rainy period, the lateral inflows between K1 and K2 would be relatively more important. While flows at K2 ranged from 110 to 880 m3/s (Fig. 4D), flows downstream Nyimba (K3) were substantially lower throughout the year and limited to a range of 100–300 m3/s. There is a large subcatchment (13,400 km2) between K2 and K3 that should contribute additional water; thus the dramatically lower flow rates at Nyimba were unexpected. After exploring this issue further (Zurbrügg et al., in preparation), we suspected that the reduction in the flow rate in the main channel can be attributed to the channel’s morphology in this section of river. Flows at K6 were relatively constant from 2002 to 2005, exceeding those at K3 throughout most of the year (Fig. 4E). During the 2002–2009 period, years that had higher precipitation, flows at K6 substantially exceeded those at K3 during the wet season. 3.3. Water balance As a starting point for exploring the water balance of the individual subcatchments, long term (2002–2009) mean annual water balances were calculated for the control volumes depicted in Fig. 2. Because of the relatively high uncertainty associated with some of the variables (Et, Qlat, DS/Dt), the balances were approached in three ways. First, best estimates were used for all input and output terms and the water balance equation (Eq. (4)) was solved for the

change in storage within the control volume, DS/Dt (Fig. 5A). For the second and third approaches (Fig. 5B and C), we assumed that the change in storage within each control volume over an annual cycle was small relative to the major input and output terms and used the water balance equation to solve either for Et or Qlat. While the assumption of zero annual change in storage is not entirely accurate (e.g., change in storage could differ substantially from zero due to relatively dry or wet years), it is a reasonable first approximation because it may have less uncertainty associated with it than the uncertainty associated with some of the water inflow or outflow parameters. The water balance parameters that are known with the highest degree of confidence are Qin, Qout, and P. However, there is no measured data available on lateral inputs from tributaries, Qlat, to the flooded areas of the Kafue Flats. Nonetheless, Qlat can be reasonably well-constrained, especially over an annual time-scale, using a realistic runoff coefficient (e.g., 10–20%). The amount of water lost from the CV due to Et is also subject to substantial uncertainty due to temporal variations in flooded area and spatial variations in landcover. The mean annual magnitudes (2002–2009) of the principal inflows and outflows in the three subcatchments of the Kafue Flats are presented in Fig. 5. In most cases, evapotranspiration exceeds direct precipitation over the floodplain on an annual time step. Depending on the approach used to close the water balance, evapotranspiration from the floodplain accounts for 4–8%, 42–56% and 19–42% of total outflows in CV1, CV2 and CV3 respectively. The different approaches to closing the water balance identified some physical limitations that can better constrain the Et and Qlat. In CV1, the results do not change substantially across the three approaches, because Et and Qlat are of only minor impor-

781

J. Wamulume et al. / Physics and Chemistry of the Earth 36 (2011) 775–788

(A) 400

P (mm)

300

200

100

Apr-09

Jan-09

Jul-08

Oct-08

Jan-08

Apr-08

Jul-07

Oct-07

Apr-07

Jan-07

Jul-06

Oct-06

Apr-06

Jan-06

Jul-05

Oct-05

Jan-05

Apr-05

Jul-04

Oct-04

Apr-04

Jan-04

Jul-03

(B) 2000

Oct-03

Apr-03

Jan-03

Oct-02

0

HB K0 ITT K1

Q (m3/s)

1500

1000

500

Apr-09

Jan-09

Jul-08

Oct-08

Jan-08

Apr-08

Jul-07

Oct-07

Apr-07

Jan-07

Oct-06

Jul-06

Jul-05

Oct-05

Jul-05

Oct-05

Jul-06

Apr-05

Jan-06

Jan-05 Jan-05

Apr-05

Apr-06

Oct-04 Oct-04

Jan-06

Jul-04 Jul-04

Apr-06

Apr-04

Jan-04

Apr-04

Oct-03

Jul-03

Apr-03

Jan-03

Oct-02

0

(C) 2000 ITT K1 Nam K2

Q (m3/s)

1500

1000

500

Apr-09

Jan-09

Jul-08

Oct-08

Apr-08

Jan-08

Oct-07

Jul-07

Jan-07

Apr-07

Oct-06

Jan-04

Oct-03

Jul-03

Apr-03

Jan-03

Oct-02

0

Fig. 4. (A) Cumulative monthly precipitation at Itezhi Tezhi dam site, and (B–E) daily discharge at Hook Bridge (K0), Itezhi Tezhi (K1), Namwala (K2), Nyimba (K3) and Kafue Gorge (K6). Note the different range of the discharge axes: (B and C) 0–2000 m3/s, (D and E) 0–1200 m3/s.

tance to the budget. However, when solving for DS/Dt, CV2 and CV3 have relatively large annual changes in storage that unrealistically imply that CV2 and CV3 experience net retention and net loss of water, respectively, each year. This imbalance suggests that Et or Qlat were being inaccurately quantified. The value for negative Qlat in Fig. 5C for CV2 suggests Et was being underestimated and may be 50% higher than suggested by the original calculated value (i.e., Et in Fig. 5A). On the other hand, in CV3 if only Qlat is allowed to vary(Fig. 5C) to compensate for the negative change in storage found in Fig. 5A, the calculated Qlat requires a runoff coefficient >30% to achieve. Thus, for CV3, a more realistic balance may represent higher Qlat (8 Mm3/year) and lower Et (9 Mm3/year).

Local precipitation, in the form of Qlat and P, plays an important role in the hydrology of the Kafue Flats. On average Qlat and P accounts for 15, 29 and 46% of the inflows in to CV1, CV2 and CV3 respectively (Fig. 5). At the scale of the entire Kafue Flats, the inflow from ITT dam (21 Mm3/year) accounts for 45% of the total inflow to the Flats with the remaining 55% coming from direct precipitation to the floodplain and lateral inflows (Fig. 5). Estimates of river inputs, direct precipitation, and lateral inflows to each of the subcatchments are presented on a monthly basis from 2002 to 2009 in Fig. 6 to illustrate how the relative importance of these inputs vary seasonally, interannually and by subcatchment. Monthly data was used instead of daily data because the water tra-

782

J. Wamulume et al. / Physics and Chemistry of the Earth 36 (2011) 775–788

(D)

1000

K2 K3

Q (m3/s)

800

600

400

200

0

(E) 1200 K3 K6

1000

Q (m3/s)

800

600

400

200

Oct-08

Feb-09

Jun-08

Feb-08

Oct-07

Jun-07

Oct-06

Feb-07

Jun-06

Oct-05

Feb-06

Jun-05

Oct-04

Feb-05

Jun-04

Oct-03

Feb-04

Jun-03

Feb-03

Oct-02

0

Fig. 4 (continued)

vel times along the main channel (3–5 days for each subcatchment; vavg  0.5 m/s). River inputs at K1 (ITT dam releases) are the dominant source of water to CV1 (Fig. 6A), with only limited contributions from direct-P and lateral inflows. CV2 (Fig. 6B) has slightly larger contributions from direct precipitation and lateral inflows to the floodplain. CV3 (Fig. 6C) appears quite different from CV2 and CV1, with direct precipitation and lateral inflows to the floodplain taking on relatively larger importance. 3.4. Water quality measurement Discharge and various water quality measurements as a function of distance on the seven sampling dates are presented in Fig. 7, with distance = 0 km corresponding to the Itezhi Tezhi dam (K1), and 120 km corresponding to Kafue Hook Bridge (K0). The station at  475 km is Chiawa Pontoon (K7), which is downstream of Kafue Gorge dam (K6) and just upstream of the Kafue and Zambezi confluence. Data from K7 was not used for subsequent calculations but are presented here nonetheless for completeness. Site K6 was not accessible for water sample collection, but it is noted here because daily discharge data is available there from the Kafue Gorge dam operators. On most sampling dates there was a trend of increasing orthophosphate ðPO3 4 -PÞ as the Kafue River flowed downstream from Itezhi Tezhi dam (K1) through the Kafue Flats (Fig. 7B). The downstream increase in orthophosphate concentrations could be partly related to its release during organic matter mineralization in the floodplain, which variations in dissolved oxygen and pH suggested are important (Zurbrügg et al., in preparation). Although there was

substantial variation between months, it is difficult to discern a clear seasonal pattern of high or low concentrations for orthophosphate. The highest variability was observed in the samples collected at K3, where concentrations varied by a factor 10 (Fig. 7B). Nitrate and ammonium were the dominant forms of dissolved inorganic nitrogen (DIN), and both species varied substantially in space and time (Fig. 7C and D). Overall, the levels of nitrate and ammonium were low and indicative of a fairly uncontaminated system. Nitrite was generally less than 2 lg NO 2 -N=L and always less than 10 lg NO -N=L, and thus was a minor portion of DIN. Rel2 atively elevated ammonium ðNHþ 4 Þ levels were measured in December 2008 and at several locations, with the suggestion of an ammonium peak around stations at K3 and K4. On other dates, ðNHþ 4 Þ concentrations (Fig. 7D) were 20 times lower than these peak values, and on some dates (August 2008, October 2008, May 2009) there was a slight trend toward increasing ðNHþ 4 Þ downstream of K2. Nitrate concentrations ranged from 10 to 110 lg NO 3 -N=L, with most values falling between 10 and 50 lg NO 3 -N=L. The highest values were recorded during peak flooding in May 2009. While clear patterns in nitrate levels are not evident on most dates, in May 2009 the recorded values decreased from a  maximum of 110 lg NO 3 -N=L at K1 to 25 lg NO3 -N=L at K4, and then increased further downstream at K5. These observations are supported by spatially intensive measurements in May 2009 (Zurbrügg et al., in preparation). In general, total nitrogen (TN) concentrations (200–600 lg N/L; Fig. 7E) exceeded DIN (nitrate + ammonium + nitrite) by more than a factor of 10. Spatially intensive measurements along the river in May 2009 found that 50% of TN were dissolved (passing a 0.7 lm glass fiber filter).

J. Wamulume et al. / Physics and Chemistry of the Earth 36 (2011) 775–788

783

Fig. 5. Long-term (2002–2009) annual average water balances performed using three different approaches, solving for: (A) DS/Dt, (B) DS/Dt = 0; and solving for Et; and (C) DS/Dt = 0, and solving for Qlat.

These observations indicate that most of TN in the river was dissolved organic nitrogen (DON) and particulate organic nitrogen (PON), as opposed to DIN. There was a tendency for TN concentration to increase over the stretch from K1 to K5. TOC exhibited clearer temporal and spatial patterns than the nitrogen species or phosphate (Fig. 7F). TOC concentrations were relatively high (12–23 mg C/L) in May 2008, June 2008 and May 2009, and increased between K1 and K5. In October 2008, December 2008 and February 2009, TOC concentrations were lower (3–8 mg C/L) and fairly constant over the 400 km of river downstream of ITT dam (K1). These values are similar for dissolved organic carbon (DOC) in the Okavango Delta on average 10.7 mg C/L (Mladenov et al., 2005). Mladenov et al. (2005) also noted that dissolved organic matter transport was controlled by slow movement of an annual flood pulse. This also seems to be the case for the Kafue Floodplain. 4. Discussion Under this section we discussion the water balance, river floodplain exchange and the implications of the water quality, nutrient and carbon loading measurements done under this study.

catchments. The relative importance of different water sources to the overall water budget exhibit considerable spatial, seasonal, and interannual variability. While river inputs are the dominant source of water year round to CV1, direct precipitation and lateral inputs are increasingly more important sources of water to CV2 and CV3. This is evident in the long-term annual water balances (Fig. 5), where direct precipitation and lateral inputs comprise nearly 50% of all inputs to CV3. Therefore, although the altered flow regime at ITT dam may change flooding patterns in the Kafue Flats, there remains a strong natural component to the system’s hydrological loading. The importance of direct precipitation and lateral inputs in CV2, and especially CV3, are also clear from monthly data over 2002– 2009 (Fig. 6), where they represent substantially more than 50% in certain months. The large relative importance of direct precipitation in CV3 is partially due to the low flows at K3 (low flows at Nyimba). This dominance of water originating in and spending a substantial amount of time in the floodplain, as opposed to traveling along the river’s main channel, appears to substantially influence the characteristics of the water leaving the system (Section 4.2). To further characterize river–floodplain exchange in each CV using water balance constraints, monthly flow rates were used to calculate a fractional exchange ratio, FE:

4.1. Assessing exchange between the river and the floodplain The Kafue River and Kafue Flats system has a complex hydrology, with strong influences from both the inflow of the upstream catchment and direct precipitation to the floodplain and its sub-

If Q out  Q in  0 FE ¼ ðQ out  Q in Þ=Q out

ð6aÞ

If Q out  Q in < 0 FE ¼ ðQ out  Q in Þ=Q in

ð6bÞ

784

J. Wamulume et al. / Physics and Chemistry of the Earth 36 (2011) 775–788

input volume (109 m3/month)

(A)

4

Q-lat

CV1

Direct-P Qriver-in

3

2

1

input volume (109 m3/month)

(B)

Jan-09

Apr-09

Jul-08

Oct-08

Jan-08

Apr-08

Jul-07

Oct-07

Apr-07

Jan-07

Jul-06

Oct-06

Apr-06

Jan-06

Jul-05

Oct-05

Apr-05

Jan-05

Jul-04

Oct-04

Apr-04

Jan-04

Jul-03

Oct-03

Apr-03

Jan-03

Oct-02

0

4

CV2

Q-lat Direct-P Qriver-in

3

2

1

Jan-09

CV3

Q-lat

input volume (109 m3/month)

Apr-09

Jul-08

(C) 4

Oct-08

Jan-08

Apr-08

Oct-07

Jul-07

Jan-07

Apr-07

Oct-06

Jul-06

Apr-06

Jan-06

Oct-05

Jul-05

Jan-05

Apr-05

Oct-04

Jul-04

Jan-04

Apr-04

Oct-03

Jul-03

Apr-03

Jan-03

Oct-02

0

Direct-P Qriver-in

3

2

1

Apr-09

Jan-09

Jul-08

Oct-08

Apr-08

Jan-08

Oct-07

Jul-07

Jan-07

Apr-07

Oct-06

Jul-06

Jan-06

Apr-06

Oct-05

Jul-05

Jan-05

Apr-05

Oct-04

Jul-04

Jan-04

Apr-04

Jul-03

Oct-03

Apr-03

Jan-03

Oct-02

0

Fig. 6. Contribution of different water inputs to (A) CV1, (B) CV2, and (C) CV3 using monthly average river discharge at the upstream station, direct precipitation to the flooded area, and lateral inputs. Lateral inputs were calculated using a 10% runoff coefficient of precipitation deposited in the ‘‘dry’’ areas of each subcatchment and assuming no time-lag.

The ratio FE varies between the ranges of 1 and + 1. When monthly outflow exceeds monthly inflow (FE > 0), the difference is divided by Qout and FE quantifies the fraction of Qout that must have entered the river from the floodplain along that stretch, either through disperse lateral inputs or tributaries draining the floodplain. When outflow is less than inflow (FE < 0), the difference is divided by Qin and FE quantifies the fraction of Qin that must have entered the floodplain along that stretch of river, as opposed to

flowing directly out along the river channel downstream. Because travel times along the main channel of the river that are in the order of several days, monthly flows were used for this calculation in order to account for short-term fluctuations in flows. As the Kafue River flows through CV1, it generally gained some water (average 15%, FE = +0.15; Fig. 8A). A substantial portion of this water likely comes from the Nanzhila River, a tributary that enters the Kafue River 30 km downstream from K1. The effect

J. Wamulume et al. / Physics and Chemistry of the Earth 36 (2011) 775–788

785

Fig. 7. (A) Discharge and (B–E) water quality parameters measured along the Kafue River between May 2008 and May 2009.

of peak discharges at K1 (Fig. 4) can be seen as local minima in FE (0.2 to 0.3; Fig. 8A) in September 2006, February 2006, February 2007 and March 2009. During this time period, 20–30% of the discharge from the dam entered the floodplain, likely through overbank flow. Some of this water would have undergone evapotranspiration in the floodplain, while the remainder would have flowed back into the Kafue River during flood recession, contributing to a positive FE during subsequent months. In contrast, CV2 had an average FE of approximately 0.3 (Fig. 8B), indicating that 30% of the water that entered at K2 left the river and entered the floodplain before the river reached K3. During peak flow periods at K3, (e.g., January–May), FE was typically between 0.4 and 0.6, indicating that 40–60% of the flow was forced out into the floodplain during this time. Positive peaks in FE from (June–October 2006, June–July 2007, and June–August 2008) suggest that some of the water that was forced into the floodplain during peak flows flowed back into the river channel during flood recession. Chimatiro (2004) observed similar trends in the Lower Shire Floodplain where water seems to be retained

in the floodplain. In CV3, FE oscillated between positive and negative from October 2002–September 2005 (Fig. 8C). The average over this time period was +0.13, with peaks of +0.6 during the high-flow season, indicating that on the order of 60% of the water leaving CV3 was draining from the floodplain. After 2005, FE remained positive throughout the entire year (except for a few modest excursions into negative values) with values as high as +0.68. While the average over the entire period was +0.22 it was much higher (+0.42) over the period May 2007–May 2009, indicating relatively wet years. While there was an initial assumption that the only route for water moving between CVs is via flow along the Kafue River main channel, this may not be entirely true considering the extremely flat subcatchments and uncertain boundaries (in particular during high water levels), and some water may cross between CVs through the floodplain. Thus, some of the positive FE for CV3 could have resulted from water forced into the floodplain within CV2, traveling through the floodplain to CV3, and entering the river downstream.

786

J. Wamulume et al. / Physics and Chemistry of the Earth 36 (2011) 775–788

(A) 1

CV1

0.8 0.6 0.4

FE

0.2 0 -0.2 -0.4 -0.6 -0.8

Jul-03

Oct-03

Jan-04

Apr-04

Jul-04

Oct-04

Jan-05

Apr-05

Jul-05

Oct-05

Jan-06

Apr-06

Jul-06

Oct-06

Jan-07

Apr-07

Jul-07

Oct-07

Jan-08

Apr-08

Jul-08

Oct-08

Jan-09

Apr-09

Oct-03

Jan-04

Apr-04

Jul-04

Oct-04

Jan-05

Apr-05

Jul-05

Oct-05

Jan-06

Apr-06

Jul-06

Oct-06

Jan-07

Apr-07

Jul-07

Oct-07

Jan-08

Apr-08

Jul-08

Oct-08

Jan-09

Apr-09

Oct-03

Jan-04

Apr-04

Jul-04

Oct-04

Jan-05

Apr-05

Jul-05

Oct-05

Jan-06

Apr-06

Jul-06

Oct-06

Jan-07

Apr-07

Jul-07

Oct-07

Jan-08

Apr-08

Jul-08

Oct-08

Jan-09

Apr-09

Apr-03

Jul-03

Oct-02

(B) 1

Jan-03

-1

CV2

0.8 0.6 0.4

FE

0.2 0 -0.2 -0.4 -0.6 -0.8 Apr-03

Oct-02

(C) 1

Jan-03

-1

CV3-KG

0.8 0.6 0.4

FE

0.2 0 -0.2 -0.4 -0.6 -0.8

Jul-03

Apr-03

Jan-03

Oct-02

-1

Fig. 8. Fractional exchange ratio (Eqs. (6a) and (6b)) on monthly time scales for A CV1, B CV2, and C CV3.

4.2. Quantifying inputs and exports of C, N, and P It is difficult to attribute many of the observed spatial and temporal variations of P, N, or C concentrations in the Kafue River to individual processes, especially when considering the spatial and temporal variations in discharge (and therefore loading, i.e., kg/ d). Nitrogen undergoes complex natural cycling in river–floodplain ecosystems, with numerous routes for transformations, loss and gain (Mitsch and Gosselink, 2007). Sinks for ammonium include

 plant or microbial uptake, oxidation of NHþ 4 to NO3 (nitrification), and immobilization by negatively charged soil particles. Nitrate losses can also occur by microbial or plant uptake, and by reduction of nitrate to N2 under low oxygen conditions (denitrification). Mineralization of organic matter is a potential source of DIN, as is nitrogen fixation. In addition, anthropogenic inputs, such as human wastewater, fertilizer and livestock excrement, can be substantial in certain systems; although the generally low DIN levels in the lower Kafue River suggest these human activities are not currently

787

J. Wamulume et al. / Physics and Chemistry of the Earth 36 (2011) 775–788

PO43- -P net export (kg/d)

CV1

500 0 M

J

J

A

S

O

N

D

J

F

M

A

M

5000 0 -5000

J

400000

10000

kg C/d

1000

200000 0

M

J

J

A

S

O

N

D

J

F

M

A

M

J

M

J

J

A

S

O

N

D

J

F

M

A

M

J

-200000

-500

-10000

-1000

-15000

2500

-400000

800000

25000

CV2

2000

CV1

600000

15000

kg N/d

CV2

20000

CV2

600000

15000

1000 500 0 M

J

J

A

S

O

N

D

J

F

M

A

M

5000 0 -5000

J

-500

-10000

-1000

-15000

2500

400000

10000

kg C/d

kg N/d

1500

J

J

A

S

O

N

D

J

F

M

A

M

J

M

J

J

A

S

O

N

D

J

F

M

A

M

J

-200000 -400000

25000

CV3

2000

200000 0

M

800000

CV3

20000

CV3

600000

15000

kg N/d

1500 1000 500 0 M

J

J

A

S

O

N

D

J

-500

F

M

A

M

J

kg C/d

kg P/d

CV1

20000

1500

kg P/d

800000

25000

2000

kg P/d

TOC net export (kg/d)

TN net export (kg/d)

2500

10000 5000 0 -5000

M

J

J

A

S

-10000

-1000

-15000

Fig. 9. Phosphate

ðPO3 4 -PÞ,

O

N

D

J

F

M

A

M

J

400000 200000 0 M

J

J

A

S

O

N

D

J

F

M

A

M

J

-200000 -400000

total nitrogen (TN) and the total organic carbon (TOC) net export from each subcatchment (May 2008–May 2009).

having a major impact. Mumba and Thompson (2005) noted that one of the major changes due to the construction of Kafue Gorge dam has been the permanent flooded areas (1100 km2) on the eastern part of the Kafue Flats due to the back water effects from KG. These conditions, along with additional flooding throughout the Kafue Flats during the wet season, would thus favor denitrification and tend to decrease nitrate concentrations (Mitsch and Gosselink, 2007). Organic matter mineralization in the floodplain would lead to gross production of partially-degraded DON and DOC, along þ with NO 3 , NH4 , or phosphate. The fraction of these compounds that were not recycled or otherwise lost in the floodplain would undergo seasonal flushing from the floodplain into the river channel, as suggested by the discharge measurements and dissolved oxygen (Fig. 7A and Zurbrügg et al., in preparation). To explore whether the Kafue Flats is a net source or sink of phosphate, TN and TOC to the river and to downstream systems, and how loadings vary in space and time, we used a basic box model approach that combined discharge data with chemical concentrations. The same control volumes were used for the chemical box models as used above for the water balance (CV1, CV2, CV3; Fig. 2). Input loads, output loads (kg/d), and net exports via the main river channel were calculated for phosphate, TN, and TOC. Given the relatively few chemical measurement dates (seven over the course of 1 year), these load calculations have considerable uncertainty associated with them. Nonetheless, it is possible to make some valuable inferences from the results. CV1 had small, positive net phosphate export rates in May– December 2008, and small negative net export rates between January and June 2009 (Fig. 9). Over the entire year, the cumulative net export (i.e., area under the curve) was 680 kg, but this was less than 1% of total phosphate entering CV1 from ITT dam, and thus inputs and outputs appeared to be approximately balanced during that year. In CV2, inputs and outputs were also fairly well-balanced from May to December 2008 (Fig. 9). From January to June 2009, however, there were more substantial negative net export rates of phosphate (200–800 kg P/d). Summed over the en-

tire year, there was a negative net export of 78 t of PO3 4 -P, which was approximately 30% of total phosphate inputs into CV2. This negative net export likely resulted in large part due to water leaving the river and entering the floodplain in CV2 (Fig. 8B). CV3 exhibited relatively large positive net export rates throughout most of the year that resulted in a cumulative net export of +360 t PO3 4 -P, with annual gross export exceeding input by a factor of 2. At the scale of the entire Kafue Flats (input at K1, export at K6), the floodplain appears to be a net source of phosphate, exporting 220 t/year. In CV1, TN and TOC net exports appear similar to that of phosphate, with fairly well-balanced inputs and outputs (Fig. 9). In January through May, the negative TN net exports are relatively larger than for phosphate or TOC. However, the cumulative net TN export from CV1 for the year was in the order of 500 t, which was only 10% of the total input at K1 (4000 t/year). The negative net exports of both TN and TOC were more pronounced from CV2. From May 2008 to May 2009, TN output from CV2 was 1100 t lower than TN input, which represented a decrease of approximately 30% of the input load. TOC output was also 30% lower than input, with a net export of 27,000 t. In CV3, the opposite was true for both TN and TOC. Outputs of TN and TOC were double inputs, with net exports of +2900 t and +72,000 t, respectively. At the scale of the entire Kafue Flats (input at K1, export at K6), TN and TOC export exceeded input by 30% and 50%, respectively, and net exports at the entire Kafue Flats scale were +1300 tN/year and +50,000 t C/year. The similar shape for phosphate, TN, and TOC net export curves for each of the control volumes, especially CV2 and CV3, suggests that the balances were largely dictated by the flow rate. Large exports of OC and nutrients from the Kafue Flats as a result of mobilization and transport from the river is consistent with previous understanding of river–floodplain interaction (Junk et al., 1989). Studies in other seasonally flooded wetlands like the Okavango delta (Mladenov et al., 2005, 2007) or the Amazon floodplains (Villar et al., 1998) report similar findings. However, the situation in the Kafue Flats is unique because of dam impacts on flooding

788

J. Wamulume et al. / Physics and Chemistry of the Earth 36 (2011) 775–788

and drainage of the floodplain. When considering dam impacts and altered flooding dynamics in the Kafue system, potential changes to organic carbon and nutrient cycling also need to be taken into consideration, and may have implications for nutrient availability within the Kafue Flats and nutrient export to downstream systems. 5. Conclusions The ecology and biogeochemistry of the Kafue Flats are inextricably linked to the system’s hydrology, and in particular with river–floodplain exchange. To assess the extent of river–floodplain exchange and its influence on nutrient and carbon budgets, we conducted a 1 year field investigation, measuring discharges and water quality at key locations, and combined this field data with existing flow and water elevation data to develop water and chemical budgets for subcatchments within the Kafue Flats. Water balance estimates (2002–2009) showed that during the high-flow periods (February–May) in the central region of the Kafue Flats (CV2) a substantial fraction (up to 60%) of the river flow was forced from the river into the floodplain, apparently due to small channel capacity in this area. Conversely, in the eastern region of the Kafue Flats (CV3), water moved from the floodplain into the river. About 40–60% of the water that left the system at K5 or K6 (Fig. 2) during high flow periods originated from (or traveled through) the floodplain. This exchange had an important influence on nutrient and organic carbon loading, with the floodplain being a net source of phosphate (220 t/year), total nitrogen (1300 t N/year, of which 90% was organic nitrogen) and total organic carbon (50,000 t C/year) to downstream systems. Thus, since upstream and downstream dams have changed flooding patterns in the Kafue Flats, they may also have altered nutrient and carbon cycling in and net export from the system. Acknowledgements We gratefully acknowledge the assistance of the Zambian Department of Water Affairs (DWA), Zambia Electricity Supply Company (ZESCO) for sharing data and reports. From DWA special thanks go to Mr. A. Hussen, Mr. P. Chola, the late Mr. A. Mporokoso, Mr. K. Nyundu and Mr. C. Nthobolo. From ZESCO special thanks go to Mr. M. Mbuta, Mr. W. Sakala, and the late Mr. H. Sinyangwe for providing data. We thank the Zambia River Authority (ZRA) for lending us the ADCP and in particular Mr. E.M. Siamachoka, Mr. W.K. Sakala and Mr. S. Mwalefor assistance with ADCP measurements. Thank you to M. Kunz and R. Stierli for help with sample measurements. Funding this project came from the Competence Center for Environment and Sustainability (CCES) at ETH, and the Eawag Partnership Program (EPP). References Baldwin, D.S., Mitchell, A.M., 2000. The effects of drying and re-flooding on the sediment and soil nutrient dynamics of lowland river–floodplain systems: a synthesis. Regulated Rivers – Research & Management 16, 457–467. Bouillon, S., Dehairs, F., Schiettecatte, L.S., Borges, A.V., 2007. Biogeochemistry of the Tana estuary and delta (northern Kenya). Limnology and Oceanography 52, 46– 59. Braman, R.S., Hendrix, S.A., 1989. Nanogram nitrite and nitrate determination in environmental and biological materials by vanadium (III) reduction with chemiluminescence detection. Analytical Chemistry 61, 2715–2718. Chimatiro, S.K., 2004. The Biophysical Dynamics of the Lower Shire River Floodplain Fisheries in Malawi. PhD dissertation, Rhodes University, Grahamstown, SA. DEW, 2002. German standard methods for the examination of water, wastewater, and sludge. Wiley VCH, Weinheim. Dudley, R.G., 1979. Changes in growth and size distribution of SarotherodonMacrochir and arotherodon-Andersoni from the Kafue Floodplain, Zambia, since construction of the Kafue Gorge dam. Journal of Fish Biology 14, 205–223.

Dudley, R.G., Scully, R.J., 1980. Changes in experimental gillnet catches from the Kafue floodplain, Zambia, since construction of the Kafue Gorge dam. Journal of Fish Biology 16, 521–537. Ellenbroek, G.A., 1987. Ecology and Productivity of an African Wetland System, The Kafue Flats. Dr. W. Junk Publishers, Dordrecht, Netherlands. Ettrick, T.M., 1990. An Introduction to Hydrology, first ed. TU Delft. Euro Consult Mott MacDonald, 2008. Rapid Assessment – Final Report, Integrated Water Resources Management Strategy for the Zambezi River Basin. SADC-WD, Zambezi River Authority, SIDA/DANIDA, Norwegian Embassy, Lusaka. FAO/UNDP, 1968. Multipurpose Survey of the Kafue River Basin, Zambia. Final Report: Climatology and Hydrology, vol. III, FAO, Rome. Hunsinger, G.B., Mitra, S., Findlay, S.E.G., Fischer, D.T., 2010. Wetland-driven shifts in suspended particulate organic matter composition of the Hudson River estuary, New York. Limnology and Oceanography 55, 1653–1667. Junk, W.J., Bayley, P.B., Sparks, R.E., 1989. The Flood Pulse Concept in river– floodplain systems. Canadian Special Publication of Fisheries and Aquatic Sciences 106, 110–127. Kambole, M.S., 2002. Managing the water quality of the Kafue River. In: 3rd Waternet/Warfsa Symposium, and Dares Salaam. Lehner, B., Döll, P., 2004. Development and validation of a global database of lakes, reservoirs and wetlands. Journal of Hydrology 296, 1–22. Li, H., Lee, J.H.W., Cai, M., 2003. Nutrient load estimation methods. International Journal of Sediment Research 18, 346–351. Mitsch, W.J., Gosselink, J.G., 2007. Wetlands, fourth ed. John Wiley and Sons, Hoboken, New Jersey, USA. Mladenov, N., McKnight, D.M., Wolski, P., Ramberg, L., 2005. Effects of annual flooding on dissolved organic carbon dynamics within a pristine wetland, the Okavango Delta, Botswana. Wetlands 25, 622–638. Mladenov, N., McKnight, D.M., Macko, S.A., Norris, M., Cory, R.M., Ramberg, L., 2007. Chemical characterization of DOM in channels of a seasonal wetland. Aquatic Sciences 69, 456–471. Mumba, M., Thompson, L., 2005. Hydrological and ecological impacts of dams on the Kafue Flats floodplain system, southern Zambia. Physics and Chemistry of the Earth 30, 442–447. Neue, H.U., Gaunt, J.L., Wang, Z.P., Beckerheidmann, P., Quijano, C., 1997. Carbon in tropical wetlands. Geoderma 79, 163–185. Nwankwor, G.I., Anyaogu, C.N., 2000. Hydrologic characteristics of a small tropical riverine wetland at Ulakwo, Imo State, Nigeria. Hydrogeology Journal 8, 646– 653. Obrdlik, P., Mumeka, A., Kasonde, J.M., 1989. Rivers in Zambia – the case of the Kafue River. Regulated Rivers: Research and Management 3, 371–380. OldeVenterink, H., Vermaat, J.E., Pronk, M., Wiegman, F., van der Lee, G.E.M., van den Hoorn, M.W., Higler, L.W.G.B., Verhoeven, J.T.A., 2006. Importance of sediment deposition and denitrification for nutrient retention in floodplain wetlands. Applied Vegetation Science 9, 163–174. Pinay, G., 1988. Hydro biological assessment of the Zambezi system: a review. WP88-089. Ramsar, 1999. A Directory of Wetlands of International Importance. Wetlands International and Ramsar Convention Bureau. Ramsar, 2006. The Ramsar Convention on Wetlands. Information Sheet on Ramsar Wetlands (RIS): Zambezi Floodplains, 2006–2008 version. Compiled by the Zambian Wildlife Authority ZAWA, Chilanga, Zambia. . Salter, L.F., 1985. Preliminary-study of water-quality in the Kafue Flats, Zambia. South African Journal of Science 81, 529–531. Savenije, H.H.G., De Laat, P.J.M., 1992. Principle of Hydrology, Lecture Notes, Revised 2002. Delft. Scott Wilson, Piésold, 2003. Integrated Kafue River Basin Environmental Impact Assessment Study. State of Environment Report, Ministry of Energy and Water Development, OPPPI/ZESCO Limited. Tockner, K., Pennetzdorfer, D., Reiner, N., Schiemer, F., Ward, J.V., 1999. Hydrological connectivity, and the exchange of organic matter and nutrients in a dynamic river–floodplain system (Danube, Austria). Freshwater Biology 41, 521–535. Tockner, K., Malard, F., Ward, J.V., 2000. An extension of the flood pulse concept. Hydrological Processes 14, 2861–2883. Villar, C.A., de Cabo, L., Vaithiyanathan, P., Bonetto, C., 1998. River–floodplain interactions: nutrient concentrations in the Lower Parana River. ArchivFürHydrobiologie 142, 433–450. Wamulume, J., 2011. Water Balance, Nutrient and Carbon Loading of the Kafue Floodplain, a Wetland in South Central Zambia. Maters Thesis, University of Zambia, Lusaka. Welcomme, R.L., 1975. The Fisheries Ecology of African Flood Plains. CIFA Tech. Pap., [3] 51p. Wiegner, T.N., Seitzinger, S.P., 2004. Seasonal bioavailability of dissolved organic carbon and nitrogen from pristine and polluted freshwater wetlands. Limnology and Oceanography 49, 1703–1712. Yachiko Engineering Corporation, 1995. National Water Resources Master Plan, vol. 2 JICA, Lusaka. Zurbrügg, R., Wamulume, J., Wehrli, B., Senn, D.B., in preparation. River–floodplain exchange and its effects on river biogeochemistry in a large subtropical floodplain wetland (Kafue Flats, Zambia).