TIME-SERIES VARIABILITY OF SOLUTE TRANSPORT AND PROCESSES

IN IN SEMI-ARID ENDORHEIC BASINS: THE OKAVANGO

DELTA

by

Kopo V. Oromeng

A thesis submitted to the Faculty of the University of Delaware in partial fulfillment of the requirements for the degree of Master of Science in

Summer 2020

© 2020 Oromeng All Rights Reserved

TIME-SERIES VARIABILITY OF SOLUTE TRANSPORT AND PROCESSES

IN RIVERS IN SEMI-ARID ENDORHEIC BASINS: THE OKAVANGO

DELTA

by

Kopo V. Oromeng

Approved: ______Eliot A. Atekwana, Ph.D. Professor in charge of thesis on behalf of the Advisory Committee

Approved: ______John A. Madsen, Ph.D. Interim Chair of the Department of Earth Sciences

Approved: ______Estella A. Atekwana, Ph.D. Dean of the College of Earth, Ocean and Environment

Approved: ______Douglas J. Doren, Ph.D. Interim Vice Provost for Graduate and Professional Education and Dean of the Graduate College ACKNOWLEDGMENTS

I thank my advisor, Dr. Eliot A. Atekwana (Dr. E) for his support and mentorship throughout my graduate career. Thank you for always reminding me to make science fun, your constant push for excellence and your passion for this discipline. You have helped me gain a deep appreciation for Earth Sciences and for the Okavango Delta

(Delta). I also thank the members of my committee, Dr. Holly Michael and Dr. Shreeram Inamdar for their guidance and valuable input. This study would not be possible without the support of Dr. Loago Molwalefhe,

Dr. E and student participants in the 2010-2012 National Science Foundation International Research Experience for Students (IRES) program who were instrumental in the experimental design, trip logistics and assisted in the deployment of data loggers across the Delta. Special thanks to the Botswana Department of Water and Sanitation and the proprietor of Crocodile Camp in Maun who gave us permission to install data loggers at their facilities along the in Mohembo and Maun, respectively. I give thanks to Goabaone Ramatlapeng and the members of Dr. Michael’s

Research group for their support and advice on my thesis research. And finally, I thank my mama and papa, Tshegofatso and Kenalesego Oromeng for instilling me with curiosity, grit, and integrity. To my brothers, Atang and Phenyo, thank you for your friendship, humor, and love.

iii TABLE OF CONTENTS

LIST OF TABLES ...... vi LIST OF FIGURES ...... vii ABSTRACT ...... ix

Chapter

1 INTRODUCTION ...... 1

1.1 Hypothesis,objectives and scope of study ...... 3

2 STUDY AREA AND BACKGROUND ...... 5

2.1 Geology and ...... 5 2.2 ...... 7 2.3 Hydrology ...... 7

3 METHODS ...... 9

3.1 Data Collection ...... 9

3.1.1 Water level, water temperature, specific conductance, air temperature and air pressure ...... 9 3.1.2 River and rainfall ...... 10

3.2 Data Analyses ...... 10

3.2.1 Total dissolved solids (TDS) estimates ...... 10 3.2.2 Discharge estimates ...... 11 3.2.3 Concentration-discharge (C-Q) analysis ...... 11 3.2.4 Solute load and solute flux estimates ...... 12

4 RESULTS ...... 14

4.1 Summary statistics ...... 14 4.2 Solute behavior at the inlet of the Delta ...... 14

4.2.1 TDS concentrations ...... 14 4.2.2 Air and water temperature ...... 16 4.2.3 Discharge and monthly rainfall ...... 16 4.2.4 Concentration-discharge relationships ...... 19

4.3 Solute behaviour at the outlet of the Delta- Maun ...... 20

iv 4.3.1 TDS concentrations ...... 20 4.3.2 Air and water temperature ...... 21 4.3.3 Discharge and monthly rainfall ...... 21 4.3.4 Concentration-discharge relationships ...... 22

5 DISCUSSION...... 23

5.1 Processes controlling variations in river solute concentrations ...... 23

5.1.1 Evapotranspiration (ET) as a driver for solute concentration variations ...... 23 5.1.2 Water column processes as drivers for solute concentration variations ...... 25 5.1.3 Hydrology as a driver for river solute concentration variations .. 26

5.2 Solute transport regimes ...... 28

5.2.1 Solute transport regime of the upper Okavango River Basin ...... 28 5.2.2 Solute transport regime of the Okavango Delta ...... 31 5.2.3 Solute transport regime of humid vs. arid rivers ...... 32

5.3 Solute load and solute flux into and out of the Delta ...... 33

5.3.1 Temporal variations in the instantaneous solute load ...... 33 5.3.2 Temporal solute flux into and out of the Delta ...... 36

6 CONCLUSION ...... 38

7 FUTURE WORK ...... 40

REFERENCES ...... 41

Appendix

A SUPPLEMENTAL MATERIAL ...... 47

A.1 Levelogger and barologger data from Maun and Mohembo ...... 47 A.2 Discharge-water level rating curves ...... 47

v LIST OF TABLES

Table 1: Summary statistics of total dissolved solids (TDS), air temperature, water temperature, instantaneous discharge, instantaneous solute load and annual solute flux at the Mohembo in the proximal portion of the Okavango Delta and at Maun in the distal portion of the Okavango Delta...... 15

Table A1. Data from the barologgers and leveloggers collected at the proximal Mohembo station and the distal Maun station. The data is available online at Google Drive-Thesis Appendix Data ...... 47

vi LIST OF FIGURES

Figure 1. The Okavango River Basin (ORB) with headwater Cubango River and Cuito River converging to form the Okavango River which flows into the Okavango Delta in northwestern Botswana. Red stars show the location of sampling stations at the inlet (Mohembo) and outlet (Maun) portions of the Okavango Delta. (Map is adapted from King et al., 2016)...... 6

Figure 2. Temporal variations in total dissolved solids (TDS), air temperature, water temperature, discharge for the Mohembo station (a, b, c, and d) and the distal Maun station (f, g, h and i) and plots of average monthly rainfall rates in Menongue(e) and Maun (j)...... 18

Figure 3. of the inlet and outlet stations segmented into receding and rising discharge regimes (Fig 3a and c). Subplots of log concentration and log discharge (log C- log Q) for Mohembo (Fig 3b) and Maun (Fig 3d), also showing the receding and rising discharge regimes between 2010 and 2012...... 20

Figure 4. Hydrograph for the inlet (a) and outlet stations (f) as well as sub-plots of log TDS concentration (C) on the y-axis in mg/L and log discharge(Q) on the x-axis in m3/s. Sub-plots are categorized as receding or rising discharge regimes and distinguished by different colours. The upper plots (a-e) show the inlet Mohembo station, and the lower plots (f-j) show data for the outlet Maun station...... 27

Figure 5. for the inlet and outlet stations color differentiated based on receding and rising discharge. Subplots show the ratios of the coefficient of concentration and discharge (CVC/CVQ) vs. the slope of the linear regression of log concentration (C) and log discharge (Q) for different discharge regimes for the Delta inlet station in Mohembo ( b, c, d, and e) and for the for the outlet Maun station (g, h, i, and j). Red diagonal lines in delineate plot regions where there is dilution (slope < 0) and enrichment (slope > 0). The vertical black line at CVC/CVQ = 1 represents the limit of chemostatic region...... 30

Figure 6. Solute load and boxplots showing instantaneous solute load variations during different phases of the hydrograph. The boxplots are color differentiated to distinguish the receding and rising limbs of the hydrographs for different years. Notched boxplots show mean, and outliers are shown as red stars lying beyond the lower and upper bounds of the plot...... 35

vii Figure B1. Log discharge-log water level rating curve for the proximal Mohembo station. A linear rating equation is shown with the R2 as a measure of fitness. The equation was consequently used to calculate instantaneous discharge for corresponding water level data from the logger...... 47

Figure B2. Log discharge-log water level rating curve for the distal Maun station. A linear rating equation is shown with the R2 as a measure of fitness. The equation was consequently used to calculate instantaneous discharge for corresponding water level data from the logger...... 48

viii ABSTRACT

This study investigated hourly time-series of total dissolved solids (TDS), discharge, air temperature and water temperature in the Okavango at the inlet and outlet of the Okavango Delta (Delta) in semi-arid Botswana. We hypothesized that solute delivery from watersheds to rivers in arid environments is controlled by the temporal activation of surface and subsurface flow pathways by rain and flooding, and that hydrological perturbations are more important than evapotranspiration in solute cycling. The objectives were to (1) document the temporal variations of solutes (represented by TDS) in the Okavango River and determine the processes that control the variations in solute concentrations and (2) quantify the solute load into and out of the Delta. Solute concentrations in the Okavango River varied from sub-weekly to monthly time frames, with a semi-annual increase controlled by annual pulse flooding and local rains. Variable interaction of and rainwater with solute stores in the , islands and isolated saline wetland pools controlled temporal solute loading in the Okavango River. The concentration-discharge relationships show that solute delivery to the river varies across the Delta and is a function of the spatial variability of solute stores, solute availability in the stores and the heterogeneous activation of the hydrologic flow pathways connecting the river to the solute stores. Of the 266,160 Mg/y of dissolved solids delivered to the Delta, 46,980 Mg/y (18%) is removed via the main Maun-Boro river and outlet, and 220,000 Mg/y is retained. Annually, 87% of the dissolved solutes removal through the Maun occurred during the pulse, 4% during the local rainy season and 9% at other times. Our findings indicate that hydrologic perturbations and temporal river connectivity to watershed solute stores modulate solute transport in rivers in arid environments.

ix Chapter 1

INTRODUCTION

Knowledge of solute balance in river basins is essential for determining fluxes and investigating how watershed hydrochemistry responds to environmental stressors such as global warming and land use changes. Natural and anthropogenic solute sources at and near-surface are delivered to rivers by different hydrologic flow pathways (e.g., Godsey et al., 2009; Thompson et al. 2011; Rose et al. 2018). Although solute delivery to rivers from watersheds is chiefly governed by hydrology, the solute types, availability and concentrations are modulated by weathering and catchment specific characteristics such as topography and land use (Basu et al., 2010). Therefore, the spatiotemporal variability of solute transfer from watersheds to rivers depends on the (1) types and locations of solute reservoirs in the watershed and (2) mobilization and hydrologic transport of solutes from the different stores to the river (Runkel, 1998;

Destouni, 2010). Hydrologic flow pathways for solute transfer to rivers in humid watersheds are the same as those in arid watersheds. However, in rivers in humid watersheds, (groundwater) contribution significantly buffers river chemistry and modulates solute behavior especially during normal and low flows (e.g., Godsey et al., 2009; Rose et al., 2018). In contrast, rivers in arid watersheds are characterized by severed groundwater connections, ephemeral flows and downstream decreases in discharge due to transmission losses and evapotranspiration (Tooth, 2000; Nanson et

1 al., 2002; Costa et al., 2012; Parsons et al., 1999). The downriver decrease in discharge affects hydrologic connectivity, changes river flow and solute transport dynamics.

Higher variability in the timing, duration and magnitude of discharge in rivers in arid watersheds complicate solute transport and cycling studies in these watersheds.

Research of solutes in rivers in arid watersheds is challenged by low frequency data that are inadequate for characterizing hydrological changes and investigating solute transport processes; compounding challenges include limited long-term datasets on flow and solute concentrations. Solute concentrations and discharge data collected at sufficiently high sampling frequencies enables better assessments of the solute transport processes occurring at sub-daily to seasonal time scales (e.g., Aubert et al., 2013;

Dawson et al., 2008).

The lack of suitable models and metrics developed for rivers in arid watersheds further challenges studies investigating spatiotemporal solute behavior. For example, it is uncertain how models of solute behavior in rivers in humid watersheds determined from concentration-discharge (C-Q) metrics (e.g., Evans and Davies, 1998; Godsey et al., 2009; Lloyd et al., 2016; Moatar et al., 2017; Rose et al., 2018; Zhi, 2019) apply to rivers in arid watersheds. Concentration-discharge behavior in arid watersheds is complicated by the fundamentally different climate, weathering and hydrological conditions. As a result, the C-Q behavior of rivers in arid watersheds is expected to differ from those of rivers in humid watersheds remarkably and need to be investigated.

We conducted this investigation in the Okavango River as it flows into and out of the Okavango Delta (Delta) located in semi-arid lower endorheic Okavango River Basin

2 (ORB) of northwestern Botswana. The Delta is the largest freshwater wetland in southern Africa (McCarthy and Ellery, 1998; Mendelsohn and el Obeid, 2004). The local riparian subsistence economies of Angola, Botswana and Namibia depend on the

Delta for fishing and water. Understanding its internal solute cycling mechanisms can inform conservation and management strategies. Many studies have investigated the processes that control the chemistry of the Okavango River in the Delta at different points of the Delta and at lower temporal frequencies (e.g., Akoko et al., 2013; Sawula and Martins, 1991; Zimmermann et al., 2006). However, no studies have examined the detailed river chemistry before the river enters or exits the Delta at multiple levels of temporal resolution (sub-daily and sub-seasonal aggregates). There is a lack of insights on how the upper watershed affects solute influx into the Delta, how solutes are modified during transit and before exiting the Delta, despite needing this understanding to inform effective conservation and water management practices in the ORB

(OKACOM, 2020).

1.1 Hypothesis,objectives and scope of study

This study investigated spatiotemporal variations of solute transport in a river in an arid climate. High-frequency data loggers were deployed to record hourly time-series data in the Okavango River as it flows in and out of the Okavango Delta (Fig. 1). We hypothesized that solute delivery from watersheds to rivers in arid environments is controlled by the temporal activation of surface and subsurface flow pathways by rain and flooding, and that hydrological perturbations are more important than evapotranspiration in solute cycling.

3 The objective of this study was to identify processes controlling the spatiotemporal variations of solutes at the inlet and the outlet of the Delta. This study addresses three research goals:

(1) To document the temporal variations of solutes in the Okavango River at the inlet and outlet of the Delta, (2) To determine the processes that control solute concentration variations in the Okavango River, and

(3) To quantify solute load into and out of the Okavango Delta.

High frequency investigation of the solute behavior of the Okavango River upon entry and exit of the Delta enables observation of microscale to macroscale chemohydrologic changes across the Delta. The placement of stations at the proximal and distal portions of the Delta facilitate the capture of upper catchment-scale solute transport processes before the river entered the Delta, allowing us to assess the local effects of the Delta watershed on river solutes.

4 Chapter 2

STUDY AREA AND BACKGROUND

The Okavango Delta is located between latitudes 19° - 20° S and longitudes 22° -

24° W in northwestern Botswana (Fig. 1). The upper catchment of the Okavango River

Basin (ORB) in the Bié Plateau is drained by the Cuito River and Cubango River which merge to form the Okavango River. The Okavango River enters the Delta at Mohembo and exits the Delta at Maun, before flowing towards the terminus in the Makgadikgadi

Pans Complexes.

2.1 Geology and geomorphology

In the upper ORB, the Cubango sub-catchment (103,800 km2), which is almost twice the size of the Cuito sub-catchment (57,300 km2) is underlain by a crystalline Precambrian bedrock comprising of granite, dolerite, and characterized by and (Jones, 2010). Thick unconsolidated Kalahari sands superimposed by regions of Karoo sands, silcretes and calcretes underlie the smaller Cuito sub-catchment, characterized by wide valleys, and extensive peatlands (Sawula and Martins, 1991: Jones, 2010).

In the lower ORB, the Okavango River funnels through a 95 km ‘Panhandle’ before branching into different channels that form a delta (McCarthy et al., 1993). The delta lies within the Okavango Rift graben (Reeves, 1972; Hutchins et al.., 1976; Modisi et al., 2000; Kinabo et al., 2007, 2008; Bufford et al., 2012; Leseane et al., 2015). The local watershed consists of ancient internal rivers that drained the surrounding and were partly responsible for the of the thick Kalahari sands (Thomas et al., 1991; Haddon and McCarthy, 2005). Hundreds of

5 thousands of tree islands, isolated wetlands, peaty floodplains and vegetated swamps are spread across the Delta (McCarthy et al., 1998; Gumbricht and McCarthy, 2003). The tree islands are active zones of chemical sedimentation. crust of silica and calcite precipitate have been found on island centers and fringes (McCarthy et al., 1991).

Figure 1. The Okavango River Basin (ORB) with headwater Cubango River and Cuito River converging to form the Okavango River which flows into the Okavango Delta in northwestern Botswana. Red stars show the location of sampling stations at the inlet (Mohembo) and outlet (Maun) portions of the Okavango Delta. (Map is adapted from King et al., 2016).

6 2.2 Climate

The ORB climate is seasonal and primarily driven by the annual oscillation of the Central Africa Inter-Tropical Convergence Zone (ITCZ) across Angola (Arbuszewski et al., 2013; Suzuki, 2011). The ITCZ moves from the Equator and southwards over Angola during the summer, initiating heavy rains from October to May.

From June to September the ITCZ shifts northwards towards the equator resulting in cold dry winters (Huntley et al., 2018; Lars, 2004). Annual rainfall in the ORB is unevenly distributed. The upper Cuito-Cubango watersheds receive the most rainfall, while the lower ORB and Delta region receives less rainfall. The ORB’s upper catchment is subtropical, with a mean annual rainfall of 1100 mm (Pombo et al., 2015). Mean annual rainfall in the catchments of Cuito and Cubango are 876 mm and 983 mm, respectively (Cooke, 1980). Temperatures in the Angolan plateaus range from 13 - 17°C in June/July to 20 - 25°C in January (Baumberg et al., 2014). The rainy season occurs from October to May with higher temperatures and the dry season spans from June to September with lower temperatures (Preiswerk, 2013).

The ORB’s lower catchment is semi-arid, with a mean annual rainfall range of 455 - 480 mm over the Delta region (ODMP, 2008). The highest daily maximum temperatures of 34 - 35 °C are recorded in October and the lowest temperature of 25 °C is recorded in

July (Mendelsohn and el Obeid, 2004). Rain typically falls from October to May and the dry season is from June to September (McCarthy et al., 2000).

2.3 Hydrology

River flow in the ORB is generated mainly in the upper catchments of sub- tropical Angola where the Cuito River and Cubango River contribute 45% and 55% of the discharge, respectively to the Okavango River (Mendelsohn and el Obeid, 2004).

7 The Cubango River has rapid discharges with high peaks and low baseflow, while the

Cuito River shows significantly lower peaks and a generally higher baseflow (Baumberg et al., 2014). The Cuito River discharge peaks one month after the peak discharge of the Cubango River (McCarthy et al., 2000). The hydrology of the

Okavango River and Delta is driven by an annual flood pulse and local rains. The mean annual inflow into the Delta from the upper catchment is 9.2 x 109 m3/y as a flood pulse between February and April (Ramberg et al., 2006), and gradually spreads southeastward in a span of 4-6 months to reach the distal portion of the Delta in

July/September (McCarthy and Metcalfe, 1990). Rainfall in the Delta contributes an additional 6 x 109 m3/y (FAO-UN, 1968; Merron, 1991; McCarthy and Ellery, 1998).

8 Chapter 3

METHODS

3.1 Data Collection

3.1.1 Water level, water temperature, specific conductance, air temperature and air pressure The Solinst™ leveloggers (LTC) deployed at Mohembo (18° 16' 32.6388'' S, 21° 47' 14.3232'' E) in the proximal portion of the Delta and Maun (20° 0' 17.0604'' S,

23° 25' 35.0796'' E) in the distal portion of the Delta (Fig. 1) recorded river water level (m), water temperature (°C) and specific conductance (SC; µS/cm). The temperature sensor has an accuracy of +0.1 °C and a resolution of 0.1 °C. The water level sensor is equipped with an automatic temperature compensation for normalization, and records water level readings with an accuracy of +0.1% percentage of full scale and a temperature compensation range of 10 °C to 40 °C. The 4-electrode conductivity sensor has a resolution of 1 µS, a full range of 0 to 80,000 µS/cm and normalized readings to

25 °C. The conductivity sensor operates in temperatures ranging from -20 °C to 80 °C. The levelogger was set to an altitude of 1006 m above sea level (asl) for Mohembo station and 929 m asl for Maun station based on spot elevation.

The Solinst™ barologgers were also deployed at each station to record air temperature (°C) and barometric pressure (kPa) using air pressure-based algorithms to record accurate atmospheric pressure fluctuations. The barologger pressure sensor has an accuracy of + 0.05 kPa and was set to convert and record barometric pressure in the same units as water level (m) by using a conversion factor of 0.101972 m/kPa. The air temperature sensor has an accuracy of + 0.05 °C and a resolution of 0.003 °C. At each station, the levelogger was deployed in the water column in a 5.2-cm diameter

9 perforated PVC tube and the barologger was set in the same PVC tube above the water surface. Both the leveloggers and the barologgers were programmed to collect data in a linear sampling mode at hourly intervals from July 2010-March 2012.

3.1.2 River discharge and rainfall

We used river discharge data collected at the inlet Mohembo station and outlet Maun station from the archives of the Okavango Research Institute (http://okavangodata.ub.bw/ori/monitoring/water/). The hydrometric data is collected by the Botswana Department of Water Affairs and Sanitation. Rainfall data for the Menongue and Maun stations were obtained from World Online (https://www.worldweatheronline.com/). In the upper ORB, we selected data from Menongue, Angola (14° 39' 30.6'' S, 17° 41' 27.564'' E) to represent rainfall in upper watershed, and selected data from the Maun station (19° 59' 43.08'' S, 23° 25' 5.16'' E) to represent rainfall in the Okavango Delta watershed.

3.2 Data Analyses

3.2.1 Total dissolved solids (TDS) estimates The total dissolved solids were estimated from the specific conductance measurements from the Leveloggers. The relationship between TDS and specific conductance was based on the assumption that TDS consist mainly of ionic elements that will conduct electricity (Lloyd and Heathcote, 1985). The TDS was computed by multiplying specific conductance with a lumped TDS/EC ratio multiplier(k):

푇퐷푆 = 푘 × 푆퐶 (1) The k used was 0.45 which was previously determined in the lower channels of the

Okavango River (Sawula and Martins,1991).

10 3.2.2 Discharge estimates

Water levels recorded by the leveloggers were corrected for barometric pressure differences by subtracting the time equivalent barometric pressure values recorded at each station. Barometric pressure compensation is necessary because storms and air pressure changes can introduce errors to water level recordings. After barometric pressure compensation, we constructed discharge-rating curves using the compensated water levels and time equivalent discharge obtained from Okavango Research Institute (ORI). We used linear equations from the site-specific regression of discharge vs. river levels (Appendix Fig. A1) to determine the hourly instantaneous discharge for each station

3.2.3 Concentration-discharge (C-Q) analysis

Concentration and discharge data were log-transformed before developing Concentration-Discharge (C-Q) relationships. Two C-Q matrices (1) the linear slope (훽) of the log C-log Q regression and (2) the ratio of the coefficients of variation of concentration and discharge (CVC/CVQ) were computed following the methods of Musolff et al. (2015), Godsey et al. (2009) and Thompson et al. (2011). The C-Q relationships are defined by a power law function (Eqn. 2):

퐶 = 푎푄훽 (2) C is the instantaneous solute concentration and Q is the concurrent instantaneous discharge. The intercept (a) and the slope (β) of the power-law are derived from linear regressions of the log C vs. log Q.

The CVC/CVQ was estimated as a ratio of the coefficients of variation (CV) of concentration and discharge. CV is the standard deviation (휎) of a variable normalized by its mean (휇), as used by Thompson et al. (2011). The CV are dimensionless values

11 that estimate the percentage variation in the mean of concentration and discharge. This metric compares the variability of concentration and discharge without assuming a relationship between the two.

퐶푉 휎 휇 퐶 = 퐶 푄 (3) 퐶푉푄 휇퐶휎푄 The slope (β) and the ratio of the coefficients of variation of concentration and discharge (CVC/CVQ) were estimated on a daily timescale (every 24 hours), assuming that it is a sufficient span to observe temporal changes in solute concentrations. C-Q data were processed in Microsoft Excel™ and MatLab™. To facilitate classification and discussions, chemostatic regime lies in the plot region where CVC/CVQ <1 with slopes near zero, and chemodynamic regime lies where CVC/CVQ >1.

3.2.4 Solute load and solute flux estimates

We employ an averaging method to estimate the instantaneous solute load (Li) which is the mass of dissolved solutes transported per unit time, in Mg for every ith hour measurement. Instantaneous solute load was estimated by multiplying TDS concentration Ci in mg/L with the time equivalent discharge Qi in l/s (Dolan et al., 1981; Ferguson, 1987).

퐿푖 = 퐶푖 × 푄푖 (4)

The annual solute fluxes (Lt) were estimated by summing the instantaneous solute load Li in (Mg) over the entire year for each location:

퐿푡 = ∑ 퐿푖 × ∆푡 (5)

Inorder to obtain solute load (Li), we assumed that Ci and Qi were constant between hourly measurements. Since solute concentration measurements were not

12 varied by depth, we assume a well-mixed water column where solute concentrations at the monitoring point are representative of the entire river segment.

13 Chapter 4

RESULTS

4.1 Summary statistics The temporal measurements of solute concentrations (TDS), air temperature, water temperature and discharge at the Mohembo station and the Maun station are presented in Appendix Table A1. Summary statistics of TDS, air temperature, water temperature, instantaneous discharge and the instantaneous and annual solute loads are presented in Table 1.

4.2 Solute behavior at the inlet of the Delta

4.2.1 TDS concentrations

The TDS at Mohembo ranged from 3.5 mg/L to 57.6 mg/L, with a mean concentration of 11.6 +7.9 mg/L (Table 1). The TDS concentrations displayed seasonal to sub-weekly anomalies (Fig. 2a). At the seasonal scale, there is a six months steady increase in TDS concentrations from August to January. A marked decrease in TDS follows the seasonal concentration increases, and a nearly six-month-long period of low solute concentrations (“TDS slump”) develops after that. During the “TDS slump,” solute concentration slowly decreased from late January and began to increase again in

August, marking the beginning of the next seasonal cycle. Despite interannual variability in the magnitude of seasonal perturbations, the timing and duration of seasonal anomalies remained similar from year to year (Fig. 2a).

14 At the sub-seasonal level, TDS concentrations were characterized by sharp upswings over 1-2 weeks from August to January. The most substantial weekly variation was in December, when solute concentrations increased by 2.4 times over 11 days. These sub-weekly perturbations occurred throughout the year but, became pronounced during the seasonal increase in TDS from August to January compared to the “TDS slump”.

Table 1: Summary statistics of total dissolved solids (TDS), air temperature, water temperature, instantaneous discharge, instantaneous solute load and annual solute flux at the Mohembo in the proximal portion of the Okavango Delta and at Maun in the distal portion of the Okavango Delta.

Mohembo Maun

Mean + SD Min Max Mean+ SD Min Max Total dissolved solids 11.6 + 7.9 3.5 57.6 48.9+6.4 27.4 67.0 (mg/L) Air temperature (°C) 22.7 + 6.9 0.5 45.6 22.9+7.1 2.0 42.6 Water temperature (°C) 23.5 + 4.3 13.8 30.5 23.9+4.2 13.0 31.1 Instantaneous discharge 381.0 + 217.4 77.4 943.7 19.8+16.7 1.0 60.1 (m³/s) Monthly Rainfall (mm) 70.8+72.2 0 214.9 34.9+46.1 0 150.2

Hourly solute load 15.4 +12.6 1.2 85.6 3.4+2.8 0.14 12.0 (Mg/h) Annual solute flux 266,160 46,980 (Mg/y)

Observations (n)=13,975 per station, SD = Standard Deviation, Min = Minimum, Max = Maximum

15 4.2.2 Air and water temperature

Air temperatures ranged from 0.5 - 45.6 °C with a mean of 22.7 +6.9 °C, while water temperatures ranged from 13.8 - 30.6 °C, with a mean of 23.5 +4.3 °C (Table 1). Similar seasonal to sub-weekly patterns were recorded for air (Fig. 2b) and water temperatures (Fig. 2c). The rainy summer season from September to April had high air and water temperatures, while low air and water temperatures were observed during the dry winter season between May and August. Air temperature peaked at 45.6 °C in October and was lowest at 0.5 °C in June. In contrast, water temperatures peaked at 30.6

°C in January and decreased to a low of 13.8 °C in July. At shorter intervals, daily, weekly and monthly air temperature variations were more pronounced during the rainy season from December through April. Water temperature variations were more pronounced during the rainy season between August and January, compared to the dry season between February and July.

4.2.3 Discharge and monthly rainfall

The instantaneous hourly river discharge ranged from 77.0 - 943.7 m³/s, with a mean of 381.0 +217.4 m³/s (Table 1). The annual hydrograph was asymmetric and multipeaked, with the rising limb beginning in November and peaking in May, followed by secondary peaks in February and April (Fig. 2d). Receding discharge lasted for six months from May to November, with the lowest flows in November. The rainy season in the upper watershed in Menongue was from October to May and rainfall peaked in January (Fig. 2e). The dry season lasted from June to September.

The timing of local rainfall and the differences in the arrival time of floodwaters from the Cuito and Cubango watersheds caused the differentiation in the magnitude and

16 timing of the three discharge peaks observed on the rising limbs of the hydrograph (Fig.

2d).

17 2. Temporal variations in total dissolved solids (TDS), air temperature, water temperature, discharge for the Mohembo station (a, b, c, and d) and the distal Maun station (f, g, h and i) and bar plots of average monthly rainfall rates in Menongue(e) and Maun (j).

18 4.2.4 Concentration-discharge relationships

The hydrograph at the inlet Mohembo station is color differentiated based on receding (red and green) and rising discharge (blue and black) regimes (Fig. 3a). During the hydrograph recession, there is an initial general decrease in TDS concentrations and then a shift to a general increase of the TDS concentrations with decreasing discharge

(e.g., Fig. 3b). Remarkable departures from the overall rise in the TDS concentrations are identified as major positive and negative perturbations. For instance, the TDS concentrations increase markedly from Log TDS of 0.8 to 1.6 mg/L and then back to

~0.8 mg/L as discharge decreased from Log 2.4 to 2.3 m3/s. On the rising limb, the TDS concentrations generally increase with increasing discharge until near peak discharge, when the TDS concentrations drastically decrease and stay nearly constant (Fig. 3b).

There is major variability in the TDS concentrations during discharge increases. For example, there is a decrease in the TDS concentrations between a Log discharge of ~2.4 to 2.6 m3/s before the major decrease in the TDS concentrations (Fig. 3b). The general direction of progression for the C-Q plot for Mohembo is clockwise, with threshold responses in TDS concentrations at discharges above log 2.5 m3/s on the rising limb and discharges below log 2.5 m3/s on the receding limbs of the hydrograph. The C-Q plot is diamond shaped and closed with an open center (Fig 3b).

19

3. Hydrograph of the inlet and outlet stations segmented into receding and rising discharge regimes (Fig 3a and c). Subplots of log concentration and log discharge (log C- log Q) for Mohembo (Fig 3b) and Maun (Fig 3d), also showing the receding and rising discharge regimes between 2010 and 2012.

4.3 Solute behaviour at the outlet of the Delta- Maun

4.3.1 TDS concentrations

TDS concentrations at Maun station ranged from 27.4 - 67.0 mg/L, with a mean of 48.9 +6.4 mg/L (Table 1). Temporal TDS concentrations varied at seasonal and sub- weekly periods and displayed high frequency oscillations (Fig. 2f). The seasonal pattern was characterized by a steady and gradual four months increase in TDS concentrations from July to October, followed by decreases to February, an increase to March and a general decrease to June. Weekly and sub-weekly perturbations in TDS are superimposed on the seasonal TDS responses. The TDS concentrations exhibit higher

20 frequency and higher magnitude fluctuations from July to peak concentrations in

December in contrast to more subdued oscillations during the remaining time.

4.3.2 Air and water temperature

Air temperatures ranged from 2.0 - 42.6 °C, with a mean of 22.9 +7.1 °C, while water temperatures ranged from 13.0 - 31.1 °C, with a mean of 23.9 + 4.2 °C (Table 1). The air temperature (Fig. 2g) and water temperature (Fig. 2h) displayed a seasonal response, characterized by temperature increases from July to January followed by decreases from February until July. Superimposed on the seasonal response, are variations in maximum and minimum temperatures that occur over sub weekly, weekly and monthly periods.

4.3.3 Discharge and monthly rainfall River discharge at Maun station ranged from 1.0 m³/s to 60.1 m³/s, with a mean of 19.8 +16.7 m³/s (Table 1). The annual hydrograph is symmetric with a rising limb that lasts for 6 months, beginning in March and peaking in August (Fig. 2i). The recession period lasts for the remainder of the year. A minor peak in the hydrograph is observed between January and February caused by a minor discharge increase from local rainfall. Superimposed on the seasonal hydrograph are sub-weekly discharge variations of higher oscillation magnitude near the peak discharge between June and October. The rainy season lasted from October to May with peak rainfall in January, while the dry season lasted from May to September (Fig. 2j). Higher mean rainfall for 2011- 2012, compared to 2010-2011 indicate interannual variability, which in turn affects the magnitude of discharge peaks for that year (Fig. 2i). River discharge is out of phase with

21 the seasons, as the rainy season occurs during low discharge, while the dry season coincides with the arrival of flood pulse and thus corresponds with high discharge.

4.3.4 Concentration-discharge relationships

The hydrograph at the Maun station is color differentiated based on receding and rising discharge regimes as in Figure 3c. During the hydrograph recession, the TDS concentrations generally increase with decreases in discharge (Fig. 3d). However, we observe significant TDS concentration variability at high discharge between Log

Discharge of 1.8 to 1.5 m3/s and nearly constant TDS concentrations between Log

Discharge of 1.5 to 0.7 m3/s, before TDS concentration continued to increase to the lowest discharge. In contrast, on the rising limb of the hydrograph, TDS concentrations decrease to near peak discharges of Log 1.5 m3/s, after which The TDS concentrations generally increase but continue to fluctuate between Log 1.5 to 1.8 m3/s (Fig. 3d). The general direction of progression for the C-Q plot for Mohembo is counter-clockwise as shown by the plotting of the sequential receding and rising discharge regimes.

Significant positive and negative perturbations are observed at discharges above log 1.4 m3/s, near peak discharge. The C-Q plot is tightly wound and irregularly shaped (Fig 3d).

22 Chapter 5

DISCUSSION

5.1 Processes controlling variations in river solute concentrations TDS concentrations at the inlet and outlet of the Delta displayed seasonal anomalies characterized by distinct periods of (1) increasing, (2) constant or (3) decreasing TDS concentrations (Fig. 2a and f). At shorter intervals, sub-monthly and sub-weekly TDS concentration fluctuations are superimposed on the seasonal scale. The seasonal and sub-seasonal variability of TDS concentrations at the two stations are markedly different (Fig. 2a vs. f). Differences in the timing and duration of TDS fluctuations indicate that the processes controlling solute concentrations of river water entering the Delta and within the Delta are different. Spatiotemporal solute concentration variations can be driven by climate-controlled vegetative processes such as evapotranspiration-induced saturation (e.g., del Pilar Alvarez et al., 2015; Humphries et al., 2011). Water column processes such as mineral precipitation reactions and hydrologic drivers that connect and transfer solutes from transient stores in the watershed to the river can also affect spatiotemporal solute concentration variations (e.g., Akoko et al., 2013; Ozkan et al., 2008).

5.1.1 Evapotranspiration (ET) as a driver for solute concentration variations Evapotranspiration may be responsible for downriver increases in the solute concentrations from the proximal (Fig. 2a) to the distal (Fig. 2f) portions of the Delta due to evapoconcentration of solutes (Dincer et al., 1978; Sawula and Martins, 1991; Akoko et al., 2013). Air temperature is a good proxy for ET because a high amount of solar radiation provides increased energy for vaporization and for plants to transpire. Air temperature

23 at Mohembo (Fig. 2b) and Maun (Fig. 2g) station showed comparable sub-weekly to seasonal fluctuations. The mean air temperatures for the two stations differed by 0.3 °C (Table 1). This difference of <1% in the mean air temperatures indicates that local weather conditions were not significantly different at the inlet and outlet stations, and therefore across the Delta. Water temperature for Mohembo (Fig. 2c) and Maun (Fig. 2h) showed seasonal fluctuations similar to air temperature and are characterized by similar perturbations at the diurnal, weekly and seasonal time scales. The mean water temperatures for the two stations differed by 0.4 °C, a deviation of less than 2% of the mean annual water temperature. The slight difference in air temperature indicates that local weather conditions at the inlet and outlet portions and across the Delta are similar.

We disregard climatic factors which drive ET as controls for temporal perturbations in river solute concentrations. The lack of ET as a primary control for TDS perturbations is especially true for river water at Mohembo where the general increase in TDS concentrations with increases in water temperature between July and January (Fig. 2a) does not correspond to any anomalous increases in water temperature (Fig. 2c). Moreover, the low TDS concentrations between February and September (Fig. 2a) does not correspond to the slowly decreasing water temperature during that period (Fig.

2c). Conversely, at the Maun station, both TDS concentrations and water temperature increase between July and October (Fig. 2f and 2h), but diverge as the TDS concentrations decrease through January, while water temperatures remain relatively high. The behavior of air temperature and water temperature relative to the timing and magnitude of the TDS concentration perturbations indicate that ET does not dominate the temporal variability of the TDS concentrations.

24 5.1.2 Water column processes as drivers for solute concentration variations

Water column processes can affect solute concentrations by (1) mineral weathering which add solutes to the water column or precipitation reactions which remove solutes from the water column (e.g., Drever, 1971; Velbel, 1985). In the upper watershed, limited interaction and shorter residence times of river allow little possibility for solute concentration variation due to low temperature-regulated geochemical processes in the water column (Frings et al., 2014). The long (4-6 months) river transit time from the proximal to the distal ends of the Delta allows river water to interact with and the riparian ecosystems which might affect solute concentrations. However, the hyperoligotrophic status of the Okavango River indicate the lack of weathering fluxes to the or instream processes as dominant mechanisms for solute input into the river (Smith et al., 1997). The major composition of the of the Okavango River is dominated by Ca2+ with an abundance of HCO-3 (Sawula and Martins, 1991; Mackay et al., 2011). The waters are moderately alkaline with high amounts of silica, which is indicated to precipitate onto floodplains and tree islands. Nutrients such as nitrate and phosphate have been recorded at lower concentrations (<0.1 mg/L) at the upper panhandle and near the Boro-Maun outlet (Akoko et al., 2013; Mackay et al., 2011).

The concentration of major in the surface waters of the Okavango increases downstream, and correlate with a general increase in electrical conductivity despite changes in the net proportion of each ion as and silica are precipitated out in the form of sepiolites (McCarthy et al., 1986; Mosepele et al., 2016). The low ionic concentrations of the Okavango River reflect the low solubility and weathering of rocks and sediments in the catchment (Sawula and Martins, 1991). However, relatively high concentrations of bicarbonate, indicate the dominance of carbonate weathering in the

25 watershed (Sawula and Martins, 1991; Ramberg and Wolski, 2008). Although we expect dissolution of calcite and dolomites to occur, low abundance of carbonate sediments (McCarthy and Metcalfe, 1990; Huntsman-Mapila et al., 2005) may explain the low ionic concentrations of their weathering products in river water. We expect water column processes in the Okavango River to be slow and impose a minimal effect on variations in solute concentrations in river water. Therefore, the observed perturbations in TDS concentrations in the Okavango River at the inlet and outlet of the Delta are unlikely to be driven by water column processes.

5.1.3 Hydrology as a driver for river solute concentration variations

There are stark differences in the solute behavior between the Delta inlet station in Mohembo and the outlet station in Maun. For example, although the TDS concentrations generally increase with decreasing discharge during hydrograph recessions at both stations, the rates of change and the magnitude in the TDS concentration perturbations are different (see Fig. 4c vs. 4h). During the rising limbs of the hydrograph, the TDS concentrations for the Mohembo station generally increased (e.g. Fig. 4e), while that for Maun station generally decreased (e.g. Fig. 4i). Additionally, near peak discharge, the marked TDS concentration decrease is followed by a nearly constant TDS concentrations for the inlet station at Mohembo, while at the outlet Maun station, the TDS concentrations generally increase, albeit with marked fluctuations.

26

4. Hydrograph for the inlet (a) and outlet stations (f) as well as sub-plots of log TDS concentration (C) on the y-axis in mg/L and log discharge(Q) on the x-axis in m3/s. Sub-plots are categorized as receding or rising discharge regimes and distinguished by different colours. The upper plots (a-e) show the inlet Mohembo station, and the lower plots (f-j) show data for the outlet Maun station.

27

The qualitative C-Q relationships show that river water flowing through the Delta is highly modified by processes occurring in the Delta, which can be further elucidated by quantitative concentration discharge techniques. The relationships exhibited by C-Q (Fig. 4b-4e and Fig 3g-3j) provide a means of making qualitative interpretations of how solute concentrations vary with discharge (Moatar et al., 2017; Lloyd et al., 2016). To conduct a semi-quantitative assessment of solute transport regimes at the inlet and outlet portions of the Delta, we utilize the relationship between

CVC/CVQ and the Slope (β) (Godsey et al., 2009; Rose et al., 2018).

5.2 Solute transport regimes

5.2.1 Solute transport regime of the upper Okavango River Basin The temporal plots of log discharge for the Okavango Delta showing the discharge regime based on colors is shown (Fig. 5a). The CVC/CVQ for the receding limbs at the Mohembo station ranged from 0.3 to 8.3 and β ranged from -6.0 to 3.3 (Fig.

5b and 5c). Between 76-95% of the data have a CVC/CVQ>1.0, and 83-84% of the data lies at β <0, indicating that during flow recession, the Okavango River is chemodynamic and dominated by dilution (Evans and Davies, 1998; Godsey et al., 2009; Lloyd et al.,

2016; Moatar et al., 2017; Rose et al., 2018; Zhi, 2019). During recession, river’s hydrochemical signal is influenced by freshwater input from seasonal rains (Fig. 2e and 2d). Although the hydrograph recession is dominated by dilution, we observe periodic enrichment. In 2010 and 2011, 6% and 8% of β, respectively were >0 (Fig. 5b and c). These enrichment episodes occurred in December during the rainy season (Fig. 2a and 2e) and mark the mobilization of solutes from the watershed into the river. Given that

28 enrichment occur in the early half of the rainy season, we argue that local rains dissolve and wash out solutes from the watershed into the river. This is supported by the presence of tree islands encrusted with calcite and silica precipitate on the fringes and regions of the island devoid of vegetation due to the toxic accumulation of bicarbonate as trona in the soils (Gumbricht and McCarthy, 2012; McCarthy et al., 1993).

The CVC/CVQ during the rising limbs of the hydrograph at Mohembo ranged from 0.3 to 9.0 and slope (β) ranged from -4.0 to 5.0 (Fig. 5d and e). Between 75% and

85% of the data have a CVC/CVQ > 1.0, indicative of a primarily chemodynamic flow regime. For the rising limb in 2010-2011, 66% of data plotted below a β=0, indicating the predominance of dilution over enrichment (Evans and Davies, 1998; Godsey et al., 2009; Lloyd et al., 2016; Moatar et al., 2017; Rose et al., 2018; Zhi, 2019).

Approximately 60% dilution to about 40% enrichment division for the flow regime during increasing discharge remains nearly the same in 2011-2012, as 62% of β plotted at β < 0. These results indicate that the addition of water to the river from seasonal rains and the annual flood pulse arriving from the upper watershed drives dilution (Fig. 5d and e). The differences in the chemodynamism and dilution observed on the rising limbs of hydrograph for 2010-2011 and 2011-2012 indicate interannual variability in the magnitude of river flow and processes. The total discharge and the discharge at peak flow in 2011(934 m3/s) was much greater than that of 2012 (577 m3/s) and thus lower TDS concentrations in 2010-2011 (Fig. 2a) might have been driven by greater dilution from seasonal rains (Fig. 5d).

29 5. Hydrographs for the inlet and outlet stations color differentiated based on receding and rising discharge. Subplots show the ratios of the coefficient of concentration and discharge (CVC/CVQ) vs. the slope of the linear regression of log concentration (C) and log discharge (Q) for different discharge regimes for the Delta inlet station in Mohembo ( b, c, d, and e) and for the for the outlet Maun station (g, h, i, and j). Red diagonal lines in delineate plot regions where there is dilution (slope < 0) and enrichment (slope > 0). The vertical black line at CVC/CVQ = 1 represents the limit of chemostatic region.

30 5.2.2 Solute transport regime of the Okavango Delta

The hydrograph at the outlet Maun station is color differentiated based on receding and receding discharge regimes (Fig. 5f). The CVC/CVQ for the receding limbs of the hydrograph at the Maun station ranged from 0.2 to 5 and β ranged from -3.0 to

0.3 (Fig. 5g and h). Between 46% and 40% of the data had CVC/CVQ >1.0, with 72% and 75% of the data with β of <0, indicating a chemodynamic transport regime characterized by dilution (Evans and Davies, 1998; Godsey et al., 2009; Lloyd et al., 2016; Moatar et al., 2017; Rose et al., 2018; Zhi, 2019). Freshwater input from the annual flooding event causes this dilution (Fig. 2i and 2j) and is consistent with studies that associated reduced soil nutrient levels during high floods to leaching and dilution, as well as anoxic conditions (Cronberg et al., 1996; Mladenov et al., 2005; Tsheboeng et al., 2013) and transfer of mass from the wetlands to the river (Akoko et al., 2013). The limited enrichment episodes where <0.4% of the data have slopes >0 are observed in the latter half of the rainy season indicating that solute addition to the river is likely driven by local seasonal rains (Fig. 2f and j).

The CVC/CVQ for the rising limbs of the hydrograph at the distal Maun station ranged from 0.2 to 5.0 and β ranged from -2.2 to 1.0 (Fig. 5i and j). The rising limb of the hydrograph in 2012 is excluded from the statistical analysis because of its small sample size. About 52% of the 2011 data had a CVC/CVQ>1.0, indicating a transport regime that was nearly equal parts chemostatic and chemodynamic. In 2011, 59% of β were <0 (dilution) and 3% were >0 (enrichment). The CVC/CVQ indicates that during times of increasing discharge (arrival of the flood pulse) at Maun (Fig. 2i), the river starts in a chemostatic flow regime and then transitions into a chemodynamic regime with dilution from the flood pulse. Although the river solute transport regime is dominated by dilution, we recorded discrete periods of enrichment as indicated by the

31 3% of data with β of >0. These enrichment perturbations occurred between April and

May as the discharge began to increase indicating input of waters carrying higher amounts of solutes. These perturbations are explained by the “piston flow” of evaporated wetland water into the river caused by the advancing flood front (e.g., Glover and Johnson, 1974). The “piston flow” phenomenon is a kinematic pressure wave that pushes “old” floodplain water away from the river channel towards the floodplain during flooding (Hewlett and Hibbert, 1967; Jung et al., 2004). During flood recession, a reversal in hydraulic gradient drains water from the floodplains and adjacent wetlands carrying additional solutes back into the river (Anderson and Burt, 1982; Jung et al., 2004; Akoko et al., 2013). As a result, evaporated water in wetland pools and solutes dissolved at the front of the flood pulse recharge the river and cause solute enrichment.

When these solute sources located beyond the river are depleted or diluted by the new flood, then the river’s transport regime shifts into a dilution dominated regime.

5.2.3 Solute transport regime of humid vs. arid rivers

The CVC/CVQ vs. β for the inlet Mohembo station (Fig. 5b-e) and for the outlet Maun Station (Fig. 5g-j) are characteristically different from those in humid river basins.

In the Okavango River, the CVC/CVQ can reach up to 10 and the β range can reach +6, which exceed the CVC/CVQ range of 2 and the β range of +1 measured in humid river basins (Godsey et al., 2009; Rose et al., 2018; Liu et al., 2020). Similar studies have not been conducted in rivers in arid watersheds. High values of CVC/CVQ and range of slope observed in the Okavango River indicate a fundamental difference in the solute transport regimes of rivers in arid vs. humid environments. These differences are controlled by the different proportions of baseflow vs. seasonal/stormflow and the transport of solutes to the river by the different proportions of water. Rivers in arid such as the

32 Okavango River tend to be influent, with no groundwater contribution, while rivers in humid environments are with contribution from groundwater.

For CVC/CVQ to be >1, solute concentration has to be greater than discharge variability indicating large variation in solute concentrations compared to flow. This will likely occur if groundwater does not support baseflow as in influent rivers located in arid watersheds. For β to be >0, the solute introduced into the river causes significant increase in concentration, while a β <0, will require significant input of water without the solute of concern, generally during large storms or flooding events causing dilution

(e.g., Musolff et al., 2015; Liu et al., 2020). The results in the Okavango River indicate that (1) the solute regime is not influenced by groundwater because of the much greater range of CVC/CVQ compared to rivers in humid regions and (2) significant solute transport to the river changes the river solute regime on sub-weekly to sub-monthly periods. The results also show that the metrics used to interpret solute transport regime are insufficient for rivers in arid watersheds. The dynamic hydrology of rivers in arid watersheds affecting stream discharge, hydrologic flowpathways and connectivity, while climatic factors controlling solute storage and mobilization.

5.3 Solute load and solute flux into and out of the Delta

5.3.1 Temporal variations in the instantaneous solute load The hydrograph is color differentiated based on receding and rising discharge regimes for the inlet Mohembo station (Fig. 6a) and for the outlet Maun station (Fig.

6b). The accompanying daily solute load plots are also segmented according to the receding and rising discharge regimes for the inlet (Fig. 6c) and the outlet stations (Fig. 6d). Notched box and whisker plots show the spread and data distribution of daily solute

33 load under different discharge regimes at the inlet (Fig. 6d) and the outlet stations (Fig.

6e). The instantaneous solute load measured at Mohembo ranged from 2.1 to 85.6 Mg/h and averaged 15.4 +12 Mg/h (Table 1). On an annual basis, solute load increased with increasing discharge (Fig 6a and c). Decreases in daily solute load began during peak discharge and corresponded with times of freshwater input in the late rainy season. The daily solute load reached low values in August. On the receding limbs we observed a sharp solute load peak in September (Fig 6b). The instantaneous solute load at Maun station where the Okavango River exits the Delta ranged from 0.14 to 12 Mg/h and averaged 3.36 +2.85 Mg/h (Table 1). Temporal variations in the instantaneous hourly discharge is shown in Fig. 6d and the daily solute load is shown in Fig. 6e. On an annual basis, the solute load shows higher values at peak discharge (Fig 6d). The daily solute load distribution shows a symmetry similar to discharge, such that the maximum discharge coincides with the maximum solute load. Instantaneous solute load begins to decrease with the receding discharge and reaches low values in January. From January to June when the solute load is low, we observe solute load increases in February, during the middle of the rainy season. At Mohembo, the box and whisker plots show higher means, medians and ranges in the solute load concentrations during the rising limb of the hydrograph compared to the falling limb (Fig. 6c). We surmise that higher solute influx into the Delta occurs during the pulse flooding and lower solute fluxes occur at other times. The higher interannual solute load in 2011-2012 compared to 2010-2011 may be due to higher relative discharges which transferred higher amounts of solutes from the watershed to the river. The box and whisker plots show similar means and medians, although the

34 range in the solute load concentrations are larger during the rising limb of the discharge

(Fig. 6f). We surmise that there is high solute outflux from the Delta coinciding with the annual flood pulse during the dry season, and that there is a minor increase in the solute load during the late rainy season, initiated by flushing of solutes from the floodplains, wetlands and islands by rains.

6. Solute load and boxplots showing instantaneous solute load variations during different phases of the hydrograph. The boxplots are color differentiated to distinguish the receding and rising limbs of the hydrographs for different years. Notched boxplots show mean, and outliers are shown as red stars lying beyond the lower and upper bounds of the plot.

35 5.3.2 Temporal solute flux into and out of the Delta

The annual surface solute flux into the Delta through Mohembo is estimated at

266,160 Mg/y and the solute outflux measured at the outlet Maun station is 46,981 Mg/y (Table 1). Previous studies have estimated that TDS load at Mohembo ranged between 256,197- 415,104 Mg/y (Gondwe et al., 2016) and 400,000 Mg/y (McCarthy and

Metcalfe, 1990; Gumbricht et al., 2004; Zimmermann et al., 2006). TDS influx at Mohembo in this study falls near the lower ends of previous estimates. Solute fluxes indicate solute amount brought into the Delta by the Okavango River through Mohembo on an annual basis is 4.5 times the amount that leaves the Delta at Maun. Gondwe et al. (2016) estimated an annual solute load of 457,390 Mg/y from surface inflows to 29,832 Mg/y in surface outflow which is a 15 times difference between solute influx and outflux. However, these studies much like ours underestimate surface outflows by only considering the main outlet of the Jao-Boro channel that flows out of Maun. In spite of the fact that 83% of the solute load is sequestered in the Okavango Delta, the Okavango River remains fresh (Wetzel, 1983; Ellery and McCarthy, 1994).

The dissolved load is precipitated as calcretes and silcretes in the soils of the tree islands and on the surfaces of drying floodplains after evapotranspiration losses (Dincer et al., 1978). These findings have led to the conclusion that the hundreds of thousands of islands in the Delta “capture” and transfer salts deep into the subsurface by density- driven flow and fingering (Ramberg and Wolski, 2008; Bauer et al., 2006). We name this the “island desalinization” concept and introduce the “hydrologic desalinization” concept. The hydrologic desalinization is a surficial process that mobilizes and transfers salts from wetland-floodplains and islands back into the Okavango River. As the flood pulse extends laterally into the Delta and eventually subsides, old evaporated wetland water pushed into the river by the arriving floods and return flows rich in dissolved salts

36 from the island and mixed with evaporated saltier water from isolated wetland pools flow back into the river which flush out the salts from the Delta. While the island desalinization process removes dissolved salts from the islands into the subsurface, the hydrologic desalinization process is responsible for the lateral transportation of dissolved salts from wetland, floodplains and islands to the river and out of the Delta.

37 Chapter 6

CONCLUSION

A time-series investigation of solutes in the Okavango River showed distinct differences in the temporal solute concentrations at the inlet and outlet of the Okavango

Delta. These differences are attributed to the time-variant hydrologic connection of the river to solute stores in the floodplains, wetlands and tree islands in the Okavango Delta. Intermittent hydrologic connections between the river and solute stores in the watersheds are initiated by local rains and the flood pulse. The timing, magnitude and persistence of solute perturbations at the inlet versus the outlet of the Delta are starkly different and correspond to the timing of local rains and the flood pulse. The overall solute regime of the Delta is characterized by chemodynamic-dilution patterns caused by local rains and pulse flooding. Modifications of river chemistry by river-wetland- islands interaction show that 82% of the 266,160 Mg/y of dissolved solute load that flow into the Delta is retained, and 46,980 Mg/y is removed from the Delta via the main outlet. The seasonal flood pulse accounts for 87% of the solutes removed from the Delta, while 4% of the load is removed during the rainy season and 9% is removed at other times through the main outlet at Maun. This study complements previous studies that have (1) identified “salt islands” as transient storages of salts which accumulated on the centers and fringes of the islands (e.g. Gumbricht et al., 2004; Bauer et al., 2006), (2) established the Delta as an evapoconcentration system that accumulate 200,000-300,000 Mg/y, of which 80% is deposited on islands (e.g. Bauer et al., 2006), and (3) observed the time variant inputs of solutes initiated by local rains and the flood pulse water across the Delta (e.g. Todd et al., 2003). Identifying the Okavango Delta as a chemodynamic-dilution driven system

38 has highlighted the role of the flood pulse and local rains in desalinizing the Delta by flushing soils at high flow and increasing connectivity between river channel and solute stores in floodplains, islands and isolated wetland pools. By utilizing high-frequency data sensors and spatiotemporal data, we established strong linkages between hydrological changes in the Delta (local rains and the seasonal pulse flooding) and the timing and magnitude of solute perturbations. The flood pulse plays a significant role in the temporal solute behavior by regulating removal of salts from tree islands, floodplains and isolated pools through intermittent connectivity to the river channel. Despite the dilution and solute mobilization effect of the flood pulse and local rains, endorheism and aridity may account for the significant sequestration of solutes within the Delta.

A new understanding of the spatiotemporal impacts of the annual flood pulse in enhancing solute transport and activating the temporary desalination of the Delta indicate the significance of hydrologic perturbations in keeping the Delta fresh. This study could inform solute transport prediction models and water quality monitoring regimes in semi-arid watersheds. The approach and experimental setup are transferable to rivers in arid and semi-arid environments. This involves spatiotemporal analysis of physicochemical parameters applied in conjunction with high-frequency sampling and concentration-discharge metrics to assess and characterize solute transport regimes. Longer-term studies, field experiments and numerical models are needed to explore solute transport mechanisms in semi-arid endorheic basins and forecast hydrochemical responses under different environmental stressors.

39 Chapter 7

FUTURE WORK

This study highlighted the role of local rains and the annual flood pulse in solute transportation in rivers in a semi-arid watershed. Freshwater input from local rains and the annual flood pulse flush solutes from the watershed, help to keep the Okavango Delta fresh and enhance solute movement through increased intermittent connectivity between stores and the river channel. In arid watersheds, high evapotranspiration (ET) rates and transmission losses diminish flow and increase solute concentrations down river. However, the mechanics of solute transport in rivers in arid watersheds remain unassessed in relation to climate variables such as ET and hydrology. Future work investigating similar questions of solute transport regimes in rivers in arid watersheds could benefit from: • Machine learning models trained from the time-series dataset used in this study which might be instructive in identifying specific periods of high solute

mobilization and informing the development of solute prediction models in rivers in semi-arid environments. • Remote sensing and aerial imagery analysis to investigate connectivity between solute stores and the river channel changes during the arrival of the seasonal

flood pulse and at the beginning of the local rainy season. • Advanced soil-vegetation studies to investigate soil-flushing as initiated by the flood pulse. Such research could complement ecological studies that have

already been done in the Delta on the ecological productivity triggered by the flood pulse such as fish spawning (Linhoss et al., 2012; Merron, 1991), and the aquatic food web dynamics (Høberg et al., 2002).

40 REFERENCES

Akoko, E., Atekwana, E.A., Cruse, A.M., Molwalefhe, L. and Masamba, W.R., 2013. River-wetland interaction and carbon cycling in a semi-arid riverine system: the Okavango Delta, Botswana. Biogeochemistry, 114(1-3), pp.359-380.

Anderson, M.G. and Burt, T.P., 1982. The contribution of throughflow to storm runoff: an evaluation of a chemical mixing model. Earth Surface Processes and Landforms, 7(6), pp.565-574.

Aubert, A.H., Gascuel-Odoux, C., Gruau, G., Akkal, N., Faucheux, M., Fauvel, Y., Grimaldi, C., Hamon, Y., Jaffrézic, A., Le Coz-Bouhnik, M. and Molenat, J., 2013. Solute transport dynamics in small, shallow groundwater-dominated agricultural catchments: insights from a high-frequency, multisolute 10 yr-long monitoring study.

Basu, N.B., Destouni, G., Jawitz, J.W., Thompson, S.E., Loukinova, N.V., Darracq, A., Zanardo, S., Yaeger, M., Sivapalan, M., Rinaldo, A. and Rao, P.S.C., 2010. Nutrient loads exported from managed catchments reveal emergent biogeochemical stationarity. Geophysical Research Letters, 37(23).

Bauer, P., Supper, R., Zimmermann, S. and Kinzelbach, W., 2006. Geoelectrical imaging of groundwater salinization in the Okavango Delta, Botswana. Journal of Applied Geophysics, 60(2), pp.126-141.

Bufford, K.M., Atekwana, E.A., Abdelsalam, M.G., Shemang, E., Atekwana, E.A., Mickus, K., Moidaki, M., Modisi, M.P. and Molwalefhe, L., 2012. Geometry and faults tectonic activity of the Okavango Rift Zone, Botswana: Evidence from magnetotelluric and electrical resistivity tomography imaging. Journal of African Earth Sciences, 65, pp.61-71.

Costa, A.C., Bronstert, A. and de Araujo, J.C., 2012. A channel transmission losses model for different dryland rivers. Hydrology and Earth System Sciences, 16(4), pp.1111-1135.

Cronberg, G., Gieske, A., Martins, E., Nengu, J.P. and Stenström, I.M., 1996. Major ion chemistry, plankton, and bacterial assemblages of the Jao/Boro River, Okavango Delta, Botswana: the swamps and flood plains. Archiv für Hydrobiologie. Supplementband. Monographische Beiträge, 107(3), pp.335- 407. del Pilar Alvarez, M., Carol, E. and Dapeña, C., 2015. The role of evapotranspiration in the groundwater hydrochemistry of an arid coastal wetland (Península Valdés, Argentina). Science of The Total Environment, 506, pp.299-307.

41

Destouni, G., Persson, K., Prieto, C. and Jarsjö, J., 2010. General quantification of catchment-scale nutrient and pollutant transport through the subsurface to surface and coastal waters. Environmental science & technology, 44(6), pp.2048-2055.

Dincer, T., Hutton, L.G. and Kupee, B.B.J., 1978. Study, using stable isotopes, of flow distribution, surface-groundwater relations and evapotranspiration in the Okavango Swamp, Botswana. In Isotope hydrology.

Dolan, D.M., Yui, A.K. and Geist, R.D., 1981. Evaluation of river load estimation methods for total phosphorus. Journal of Great Lakes Research, 7(3), pp.207- 214.

Drever, J.I., 1971. Chemical weathering in a subtropical igneous , Rio Ameca, Mexico. Journal of Sedimentary Research, 41(4), pp.951-961.

Ellery, W.N. and McCarthy, T.S., 1994. Principles for the sustainable utilization of the Okavango Delta ecosystem, Botswana. Biological Conservation, 70(2), pp.159-168.

Ferguson, R.I., 1987. Accuracy and precision of methods for estimating river loads. Earth surface processes and landforms, 12(1), pp.95-104.

Frings, P.J., De La Rocha, C., Struyf, E., Van Pelt, D., Schoelynck, J., Hudson, M.M., Gondwe, M.J., Wolski, P., Mosimane, K., Gray, W. and Schaller, J., 2014. Tracing silicon cycling in the Okavango Delta, a sub-tropical flood-pulse wetland using silicon isotopes. Geochimica et Cosmochimica Acta, 142, pp.132-148.

Glover, B.J. and Johnson, P., 1974. Variations in the natural chemical concentration of river water during flood flows, and the lag effect. Journal of Hydrology, 22(3- 4), pp.303-316.

Godsey, S.E., Kirchner, J.W. and Clow, D.W., 2009. Concentration–discharge relationships reflect chemostatic characteristics of US catchments. Hydrological Processes: An International Journal, 23(13), pp.1844-1864.

Gondwe, M.J. and Masamba, W.R.L., 2016. Variation of physico-chemical parameters along a river transect through the Okavango Delta, Botswana. African journal of aquatic science, 41(2), pp.205-215.

42 Gumbricht, T. and McCarthy, T.S., 2003. Spatial patterns of islands and salt crusts in the Okavango Delta, Botswana. South African Geographical Journal, 85(2), pp.164-169.

Gumbricht, T., McCarthy, J. and McCarthy, T.S., 2004. Channels, wetlands and islands in the Okavango Delta, Botswana, and their relation to hydrological and sedimentological processes. Earth Surface Processes and Landforms: The Journal of the British Geomorphological Research Group, 29(1), pp.15-29.

Haddon, I.G. and McCarthy, T.S., 2005. The Mesozoic–Cenozoic interior sag basins of Central Africa: the late-cretaceous–Cenozoic Kalahari and Okavango basins. Journal of African Earth Sciences, 43(1-3), pp.316-333.

Hewlett, J.D. and Hibbert, A.R., 1967. Factors affecting the response of small watersheds to precipitation in humid areas. Forest hydrology, 1, pp.275-290.

Høberg, P., Lindholm, M., Ramberg, L. and Hessen, D.O., 2002. Aquatic food web dynamics on a floodplain in the Okavango Delta, Botswana. Hydrobiologia, 470(1-3), pp.23-30.

Humphries, M.S., Kindness, A., Ellery, W.N. and Hughes, J.C., 2011. Water chemistry and effect of evapotranspiration on chemical sedimentation on the Mkuze River floodplain, South Africa. Journal of arid environments, 75(6), pp.555-565.

Hutchins, D.G., 1976. A summary of the geology, seismicity, geomorphology, and hydrogeology of the Okavango Delta (Vol. 7). Geological Survey Dept.

Jung, M., Burt, T.P. and Bates, P.D., 2004. Toward a conceptual model of floodplain water table response. Water Resources Research, 40(12).

Kinabo, B.D., Atekwana, E.A., Hogan, J.P., Modisi, M.P., Wheaton, D.D. and Kampunzu, A.B., 2007. Early structural development of the Okavango rift zone, NW Botswana. Journal of African Earth Sciences, 48(2-3), pp.125-136.

Kinabo, B.D., Hogan, J.P., Atekwana, E.A., Abdelsalam, M.G. and Modisi, M.P., 2008. Fault growth and propagation during incipient continental rifting: Insights from a combined aeromagnetic and Shuttle Radar Topography Mission digital elevation model investigation of the Okavango Rift Zone, northwest Botswana. Tectonics, 27(3).

King, J. and Chonguiça, E., 2016. Integrated management of the Cubango-Okavango River basin. Ecohydrology & Hydrobiology, 16(4), pp.263-271.

43 Leseane, K., Atekwana, E.A., Mickus, K.L., Abdelsalam, M.G., Shemang, E.M. and Atekwana, E.A., 2015. Thermal perturbations beneath the incipient Okavango Rift Zone, northwest Botswana. Journal of Geophysical Research: Solid Earth, 120(2), pp.1210-1228.

Linhoss, A.C., Muñoz-Carpena, R., Allen, M.S., Kiker, G. and Mosepele, K., 2012. A flood pulse driven fish population model for the Okavango Delta, Botswana. Ecological Modelling, 228, pp.27-38.

Lloyd, C.E.M., Freer, J.E., Johnes, P.J. and Collins, A.L., 2016. Using hysteresis analysis of high-resolution water quality monitoring data, including uncertainty, to infer controls on nutrient and transfer in catchments. Science of the Total Environment, 543, pp.388-404.

Mackay, A.W., Davidson, T., Wolski, P., Mazebedi, R., Masamba, W.R., Huntsman- Mapila, P. and Todd, M., 2011. Spatial and seasonal variability in surface water chemistry in the Okavango Delta, Botswana: a multivariate approach. Wetlands, 31(5), p.815.

McCarthy, T.S. and Ellery, W.N., 1998. The Okavango delta. Transactions of the Royal Society of South Africa, 53(2), pp.157-182.

McCarthy, T.S., McIver, J.R. and Verhagen, B.T., 1991. Groundwater evolution, chemical sedimentation and carbonate brine formation on an island in the Okavango Delta swamp, Botswana. Applied Geochemistry, 6(6), pp.577-595.

McCarthy, T.S. and Metcalfe, J., 1990. Chemical sedimentation in the semi-arid environment of the Okavango Delta, Botswana. Chemical geology, 89(1-2), pp.157-178.

Mendelsohn, J. and el Obeid, S., 2004. Okavango River: the flow of a lifeline. Struik.

Merron, G.S., 1991. The ecology and management of the fishes of the Okavango delta, Botswana, with particular reference to the role of the seasonal floods (Doctoral dissertation, Rhodes University).

Mladenov, N., McKnight, D.M., Wolski, P. and Ramberg, L., 2005. Effects of annual flooding on dissolved organic carbon dynamics within a pristine wetland, the Okavango Delta, Botswana. Wetlands, 25(3), pp.622-638.

Moatar, F., Abbott, B.W., Minaudo, C., Curie, F. and Pinay, G., 2017. Elemental properties, hydrology, and biology interact to shape concentration‐discharge curves for carbon, nutrients, sediment, and major ions. Water Resources Research, 53(2), pp.1270-1287.

44 Modisi, M.P., Atekwana, E.A., Kampunzu, A.B. and Ngwisanyi, T.H., 2000. Rift kinematics during the incipient stages of continental extension: Evidence from the nascent Okavango rift basin, northwest Botswana. Geology, 28(10), pp.939-942.

Musolff, A., Schmidt, C., Selle, B. and Fleckenstein, J.H., 2015. Catchment controls on solute export. Advances in Water Resources, 86, pp.133-146.

Nanson, G.C., Tooth, S.T.., Knighton, A.D., Bull, L.J. and Kirkby, M.J., 2002. A global perspective on dryland rivers: perceptions, misconceptions and distinctions. Dryland rivers: hydrology and geomorphology of semi-arid channels, pp.17-54.

Ozkan, E.Y., Kocatas, A. and Buyukisik, B., 2008. Nutrient dynamics between sediment and overlying water in the inner part of Izmir Bay, Eastern Aegean. Environmental monitoring and assessment, 143(1-3), pp.313-325.

The Permanent Okavango River Basin Water Commission (OKACOM), 2020. Thematic Area 2: Water Resources Management. https://www.okacom.org/thematic-area-2-water-resources-management (accessed June 01, 2020).

Parsons, A.J., Wainwright, J., Stone, P.M. and Abrahams, A.D., 1999. Transmission losses in on dryland hillslopes. Hydrological Processes, 13(17), pp.2897- 2905.

Ramberg, L. and Wolski, P., 2008. Growing islands and sinking solutes: processes maintaining the endorheic Okavango Delta as a freshwater system. Plant Ecology, 196(2), pp.215-231.

Ramberg, L., Hancock, P., Lindholm, M., Meyer, T., Ringrose, S., Sliva, J., Van As, J. and Vander Post, C., 2006. Species diversity of the Okavango delta, Botswana. Aquatic sciences, 68(3), pp.310-337.

Reeves, C.V., 1972. Rifting in the Kalahari? Nature, 237(5350), pp.95-96.

Rose, L.A., Karwan, D.L. and Godsey, S.E., 2018. Concentration–discharge relationships describe solute and sediment mobilization, reaction, and transport at event and longer timescales. Hydrological processes, 32(18), pp.2829-2844.

Runkel, R.L., 1998. One-dimensional transport with inflow and storage (OTIS): A solute transport model for and rivers (Vol. 98, No. 4018). US Department of the Interior, US Geological Survey.

45 Sawula, G. and Martins, E., 1991. Major ion chemistry of the lower Boro River, Okavango Delta, Botswana. Freshwater Biology, 26(3), pp.481-493.

Thomas, D., Thomas, D.S. and Shaw, P.A., 1991. The Kalahari Environment. Cambridge University Press.

Thompson, S.E., Basu, N.B., Lascurain, J., Aubeneau, A. and Rao, P.S.C., 2011. Relative dominance of hydrologic versus biogeochemical factors on solute export across impact gradients. Water Resources Research, 47(10).

Todd, M., Hughes, D., Savenije, H., Gumbricht, T., Wolski, P., Wilk, J., Ringrose, S., Kniveton, D. and Andersson, L., 2003. Water flow dynamics in the Okavango River Basin and Delta-a prerequisite for the ecosystems of the Delta.

Tooth, S., 2000. Process, form and change in dryland rivers: a review of recent research. Earth-Science Reviews, 51(1-4), pp.67-107.

Tsheboeng, G., Bonyongo, M. and Murray-Hudson, M., 2014. Flood variation and soil nutrient content in floodplain vegetation communities in the Okavango Delta. South African Journal of Science, 110(3-4), pp.01-05.

Velbel, M.A., 1985. Hydrogeochemical constraints on mass balances in forested watersheds of the southern Appalachians. In The Chemistry of Weathering (pp. 231-247). Springer, Dordrecht.

Wetzel, R.G., 1983. . –Saunders College Publishing. Philadephia, New York, Chicago.

Zhi, W., Li, L., Dong, W., Brown, W., Kaye, J., Steefel, C. and Williams, K.H., 2019. Distinct source water chemistry shapes contrasting concentration‐discharge patterns. Water Resources Research, 55(5), pp.4233-4251.

Zimmermann, S., Bauer, P., Held, R., Kinzelbach, W. and Walther, J.H., 2006. Salt transport on islands in the Okavango Delta: numerical investigations. Advances in Water Resources, 29(1), pp.11-29.

46 SUPPLEMENTAL MATERIAL

A.1 Levelogger and barologger data from Maun and Mohembo

Table A1. Data from the barologgers and leveloggers collected at the proximal Mohembo station and the distal Maun station. The data is available online at Google Drive-Thesis Appendix Data

A.2 Discharge-water level rating curves

Mohembo rating curve 3.0

2.8

/s) 3 2.6

2.4 Log discharge (m discharge Log y = 1.5574x + 1.9071 2.2 R² = 0.9811

2.0 0 0.2 0.4 0.6 0.8 Log water level (m) B1. Log discharge-log water level rating curve for the proximal Mohembo station. A linear rating equation is shown with the R2 as a measure of fitness. The equation was consequently used to calculate instantaneous discharge for corresponding water level data from the logger.

47 Maun rating curve 2.2

1.8

/s) 3

1.4 Log discharge (m discharge Log 1.0 y = 2.7623x + 0.4572 R² = 0.9294

0.6 0 0.15 0.3 0.45 0.6 Log water level (m) B2. Log discharge-log water level rating curve for the distal Maun station. A linear rating equation is shown with the R2 as a measure of fitness. The equation was consequently used to calculate instantaneous discharge for corresponding water level data from the logger.

48