Study on dynamic behavior of water quality in lake victoria, kenya

Part Two describe a chemical analysis which was used for 14 bio-chemical parameters: pH, dissolved oxygen DO, total nitrogen TN, total dissolved nitrogen TDN, nitrite NO2-, nitrate NO3-,

Study on Dynamic Behavior of Water Quality in

Lake Victoria, Kenya

2017 年　2 月　27 日 長崎大学大学院工学研究科 国際水環境工学コース

番号　52115705

(LE THI KIM QUI)

The purpose of this research aims to classify and determine characteristics of a unique polluted water region in Nyanza Gulf, from which to find out the treatments and adaptations to each region The reason for this research is the demand for clean and safe water for the residential areas surrounding Nyanza Gulf, Kenya Therefore, in LAVICORD project, the monitoring programs including frequent water samplings (biweekly and bimonthly) at 22 sites (KISAT, KL1, KL2, D1, D2, D3, D4, KL3, KL4, H1, H2, H3, CG3, CG5, CG6, COL, KIBOS river, KIBOUN river, SONDU river, NYANDO river) were carried out in 1 year (from June 2014 to September 2015) The thesis is composed of five chapters, each of them dealing with different aspects of a whole problem of water quality in trophic lake

Chapter One is an overview of the current state of water quality in Nyanza Gulf, Kenya The urgent need to carry out LAVICORD projects aimed at finding solutions for eutrophication in lakes and addressing the needs of clean water to people around the lake

Chapter Two is a literature review of related issues which has treated in the study Chapter Two consists of 4 parts express four major aspects of the problem of eutrophication lakes Part One defines the phenomenon of eutrophication and its mechanism Part Two shows the current conditions of eutrophication in Nyanza Gulf in recent years Part Three outlines the environmental health issues related to this phenomenon The reference to water resources management policy for Tropic lakes in the world is presented in Part Four

Chapter Three is subdivided into three parts and provides an outline of relevant methodology was used in this research Part One illustrates the approach of field observation Two physical parameters: Water temperature and turbidity, were measured

by a submersible Conductivity-Temperature-Depth profiling system (CTD) Part Two describe a chemical analysis which was used for 14 bio-chemical parameters: pH, dissolved oxygen (DO), total nitrogen (TN), total dissolved nitrogen (TDN), nitrite (NO2-), nitrate (NO3-), ammonium (NH4+), total phosphorus (TP), total dissolved phosphorus (TDP), phosphate (PO43-), Chlorophyll-a (Chl-a), total suspended solid (TSS), total Microcystin (TMC), dissolved Microcystin (DMC) Part Three shows multivariable statistical methods which were applied for complex data matrix such as: Friedman test (ANOVA two-way with blocking for non-parametric method) and ANOVA two-way were used to analyze spatial and temporal variation of 16 water quality parameters; Cluster analysis was used to classify the polluted water region base

on about 16 water quality parameters; Multivariable linear regression analysis was

carried out to make a predicted model of correlation between Chlorophyll-a and nutrient parameter or physical factor.

Chapter Four concentrates on problems resulting in four parts Part One presents a data descriptive and normal distribution of 16 parameters Almost observed data have not followed a normal distribution and strong skewed with many outliers in the left tail

of the distribution Part Two shows significant spatial and temporal variation of 16 water quality parameters Part Three shows the result of the classification of pollution, all 17 parameters (exclude PO43-, include PON and POP) were used As the results, 18 monitoring stations were classified into 3 groups Group 1 named Coastal group includes: Kisat, KL1, D1 and H4 which are located near the shoreline of Nyanza gulf Group 2 named Middle group includes the stations (KL2, KL3, KL4, H1, H2, H3, D2, D3, D4) which are distributed in the middle of Nyanza gulf Group 3 named Mouth group includes the others: CG3, CG5, KL5 and COL which are located at the mouth of the Nyanza Gulf Part Four, basing on 3 pollutant areas, this method was carried out to make a predicted model of correlation between Chlorophyll-a and nutrient parameter or physical factor for each region Three models were established In model no.1, Chlorophyll-a of Coastal area showed positive correlation with TN, TSS and water temperature This result is quite consistent with natural phenomena when both TN and TSS represent for organic pollutants (R2=0.48) In model no.2, the Chlorophyll-a of Middle area showed negative correlation with DO and turbidity (R2=0.28) The model no.3 of Mouth area showed the rather completely relationship between Chlorophyll-a and nutrient parameters (TDN, nitrite and nitrate) as well as physical parameters (Temperature, pH and DO) (R2=0.82) Compare with model of previous research, this study show the more comprehensive relationship between Chlorophyll-a and the other water quality parameters The correlation base on not only nutrient parameter (nitrogen parameters) but also water temperature, pH and dissolved oxygen

Conclusions are drawn in Chapter Five There is significant spatial and temporal variation in water quality in Nyanza Gulf and mechanisms of development of phytoplankton in each polluted water region is different I suggest that either increasing discharge distance of the industrial parks from the coastal zone or take advantage of offshore water (which is not much severely polluted) to serve the needs of clean and safe water for the people around the Nyanza Gulf Besides, raising public awareness about Microcystin in water through the media has to be fostered

My thanks also go to the experts who were involved in the International Water Engineering Course In particular, I would like to acknowledge Prof Itayama, Assoc.Prof Suzuki and Assoc.Prof Fujioka Without their passionate participation and enthusiastic help, my study could not have been successfully done and my dream surely ends at any moment in the past.

Friend, the beautiful friendship is a vital piece in our life and success Dr Kim Sangyeob, Japanese beloved boys in River Engineering laboratory and “Vietnamese popular Fours” who always stand by me, smile to me to let me know I don’t alone at all Thank you so much my friends

Finally, I must express my very profound gratitude to my husband and to my son for providing me with unfailing support and continuous encouragement throughout my years of study They always be here because of me and me too “Dear adorable son, without you, without me stronger than ever today”

Thank you for everything

Table of contents

List of figures vi

List of tables ix

Abbreviations x

Chapter 1: Introduction 1

1.1 Introduction 1

1.2 Objectives of study 2

1.3 Structure of thesis 2

Chapter 2: Literature review 3

2.1 Trophic lakes in the world 3

2.2 Lake Victoria water environment 6

2.3 Nutrient enrichment in Nyanza Gulf 9

2.4 Correlation between Chlorophyll-a and nutrient factors in trophic lakes 11

2.5 Environmental health problem caused by harmful cyanobacteria 13

2.5.1 Environmental health problem caused by harmful cyanobacteria 13

2.5.2 Microcystin in Nyanza Gulf 15

Chapter 3: Materials and methodology 17

3.1 Population of study area 17

3.2 Monitoring sites 21

3.3 Data treatment and multivariate statistical methods 24

3.3.1 Data treatment 24

3.3.2 Two-way analysis of variance (ANOVA) 25

3.3.3 Cluster analysis 25

3.3.4 Multivariable linear regression analysis 26

Chapter 4: Results and discussion 28

4.1 Data descriptive 28

4.2 Temporal variation of water quality in Nyanza Gulf 30

4.3 Spatial variation of water quality in Nyanza Gulf 45

4.3.1 The difference in water quality in vertical direction of water column 45

4.3.2 Spatial variation of 16 water quality parameters 49

4.3.3 Nutrient loading in Nyanza Gulf 55

4.4 Similarity of monitoring sites 58

4.5 Correlation of phytoplankton and nutrient parameters in Nyanza Gulf 68

4.5.1 Correlation of Chlorophyll-a and the other parameters in Coastal area 68

4.5.2 Correlation of Chlorophyll-a and the other parameters in Middle area 70

4.5.3 Correlation of Chlorophyll-a and the other parameters in a Mouth area 71

Chapter 5: Conclusion 76

Reference 79

Appendix 86

List of figures

Figure 2.1: Largest freshwater lakes in the world, including the proportion of the total

freshwater volume contained in them 5

Figure 3.1: Population and density of population of counties locate around Nyanza Gulf (2009) 18

Figure 3.2: Former Nyanza Province 18

Figure 3.3: Siaya county of Nyanza Province, Kenya 19

Figure 3.4: Homa Bay county of Nyanza Province, Kenya 19

Figure 3.5: Kisumu county of Nyanza Province, Kenya 20

Figure 3.6: Density people per km2 of Kisumu 20

Figure 3.7: Sampling map in Nyanza Gulf, Kenya 21

Figure 4.1: Histogram combine with Turkey boxplot of 16 parameters 29

Figure 4.2: Precipitation and temperature in Kisumu station, Kenya 31

Figure 4.3: Temporal variation of TN biweekly observation in four rivers 32

Figure 4.4: Temporal variation of TP biweekly observation in four rivers 32

Figure 4.5: Temporal variation of Chlorophyll-a biweekly observation in four rivers 32

Figure 4.6: Temporal variation of TSS biweekly observation in four rivers 33

Figure 4.7: Temporal variation of TN biweekly observation in northern Nyanza Gulf 34

Figure 4.8: Temporal variation of TP biweekly observation in northern Nyanza Gulf 34

Figure 4.9: Temporal variation of Chlorophyll-a biweekly observation in northern Nyanza Gulf 35

Figure 4.10: Temporal variation of TN bimonthly observation in southern Nyanza Gulf .36

Figure 4.11: Temporal variation of TP bimonthly observation in southern Nyanza Gulf .36

Figure 4.12: Temporal variation of Chlorophyll-a bimonthly observation in southern Nyanza Gulf 37

Figure 4.13: Temporal variation of TSS biweekly observation in northern Nyanza Gulf .37

Figure 4.14: Temporal variation of TSS bimonthly observation in southern Nyanza Gulf .38

Figure 4.15: Temporal variation of Turbidity biweekly observation in northern Nyanza Gulf 38

Figure 4.16: Temporal variation of Turbidity bimonthly observation in southern Nyanza Gulf 39 Figure 4.17: Temporal variation of water temperature biweekly observation in northern

Nyanza Gulf 39

Figure 4.18: Temporal variation of water temperature bimonthly observation in southern Nyanza Gulf 40

Figure 4.19: Temporal variation of pH biweekly observation in northern Nyanza Gulf 40 Figure 4.20: Temporal variation of pH bimonthly observation in southern Nyanza Gulf .41

Figure 4.21: Temporal variation of DO biweekly observation in northern Nyanza Gulf .41

Figure 4.22: Temporal variation of DO bimonthly observation in southern Nyanza Gulf .42

Figure 4.23: Mean of total Microcystin in one year in whole study area 43

Figure 4.24: Total Microcystin biweekly observation in four rivers 43

Figure 4.25: Total Microcystin biweekly observation in northern Nyanza Gulf 44

Figure 4.26: Total Microcystin bimonthly observation in southern Nyanza Gulf 45

Figure 4.27: The difference in Chlorophyll-a between two layers 46

Figure 4.28: The difference in TSS between two layers 46

Figure 4.29: The difference in total Microcystin between two layers 47

Figure 4.30: The difference in Temperature between two layers 47

Figure 4.31: The difference in DO between two layers 48

Figure 4.32: Boxplot of 16 water quality parameters biweekly monitored over space 50

Figure 4.33: Boxplot of 16 water quality parameters bimonthly monitored over space 51 Figure 4.34: Boxplot of 16 water quality parameters bimonthly monitored over time in 4 rivers 52

Figure 4.35: Spatial comparison on 22 sampling sites 57

Figure 4.36: Dendrogram of cluster analysis with Euclidean distance 59

Figure 4.37: Cluster plot against 1st and 2nd principal components 59

Figure 4.38: TN (Mean ± SD) of three regions in dry season and rainy season 62

Figure 4.39: TP (Mean ± SD) of three regions in dry season and rainy season 62

Figure 4.40: TSS (Mean ± SD) of three regions in dry season and rainy season 62

Figure 4.41: Chlorophyll-a (Mean ± SD) of three regions in dry season and rainy season .63

Figure 4.42: TDN (Mean ± SD) of three regions in dry season and rainy season 63

Figure 4.43: TDP (Mean ± SD) of three regions in dry season and rainy season 63

Figure 4.44: Nitrite (Mean ± SD) of three regions in dry season and rainy season 64

Figure 4.45: Nitrate (Mean ± SD) of three regions in dry season and rainy season 64 Figure 4.46: Ammonium (Mean ± SD) of three regions in dry season and rainy season

Figure 4.47: Temperature (Mean ± SD) of 3 regions in dry season and rainy season 65

Figure 4.48: Turbidity (Mean ± SD) of 3 regions in dry season and rainy season 65

Figure 4.49: pH (Mean ± SD) of three regions in dry season and rainy season 65

Figure 4.50: DO (Mean ± SD) of three regions in dry season and rainy season 66

Figure 4.51: PON (Mean ± SD) of three regions in dry season and rainy season 66

Figure 4.52: POP (Mean ± SD) of three regions in dry season and rainy season 66

Figure 4.53: TMC (Mean ± SD) of three regions in dry season and rainy season 67

Figure 4.54: DMC (Mean ± SD) of three regions in dry season and rainy season 67

Figure 4.55: Distribution of residual against fitted value and Q-Q plot of Chlorophyll-a in coastal area 69

Figure 4.56: Q-Q plot with 95% confidence interval of Chlorophyll-a model in Coastal area 69

Figure 4.57: Distribution of residual against fitted value and Q-Q plot of Chlorophyll-a in Middle area 71

Figure 4.58: Q-Q plot with 95% confidence interval of Chlorophyll-a model in Middle area 71

Figure 4.59: Distribution of residual against fitted value and Q-Q plot of Chlorophyll-a in Mouth area 72

Figure 4.60: Q-Q plot with 95% confidence interval of Chlorophyll-a model in Mouth area 73

List of tables

Table 2.1: Estimates of volumes of water contained in world lakes 5

Table 3.1: Coordination of sampling sites 22

Table 4.1: Summary p – value of Shapiro test 28

Table 4.2: The difference between surface and 5m below surface of water quality 49

Table 4.3: Summary spatial and temporal variation in whole study area 54

Table 4.4: Summary of spatial and seasonal variation of water quality 61

Table 4.5: Summary model of chlorophyll - a in 3 area 73

LAVICORD The Lake Victoria Comprehensive Ecosystem and Aquatic

Environment Research for Development ProjectANOVA Analysis of variance

WHO World Health Organization

UNECE United Nations Economic commission for Europe

Chapter 1: Introduction 1.1 Introduction

The Lake Victoria basin is a source of livehood for millions of riparians in East Africa The Lake supports Africa’s largest inland fishery activities, thus a source of food and income for its riparians are an important issue for the country However, these resources have been overexploited over time and slowly losing their sustainability Besides, availability of clean water has also a remain problem in the region In response, The Lake Victoria Comprehensive Ecosystem and Aquatic Environment Research for Development Project (LAVICORD) was processed This was a multidisciplinary project which integrates modern technologies of water engineering and fisheries In order to address environmental and socia-economic problems of the riparian communities living around Lake Victoria Also, this project enhanced awreness in resource conservation LAVICORD was operated by the Ministry of Environment and Natural Resources of Kenya Government to achieve sustainable development for Vision

2030 (LAVICORD, 2016) LAVICORD was proposed by Maseno University, Kenya and Nagasaki University, Japan

A reasonable estimation of the quality of surface waters will be provided by spatial and temporal variations of water quality parameters Therefore, this monitoring programs including frequent water samplings (biweekly and bimonthly) to collect 14 bio-chemical parameters (pH, dissolved oxygen (DO), total nitrogen (TN), total dissolved nitrogen (TDN), nitrite (NO2-), nitrate (NO3-), ammonium (NH4+), total phosphorus (TP), total dissolved phosphorus (TDP), phosphate (PO43-), Chlorophyll-a (Chl-a), total suspended solid (TSS), total Microcystin (TMC), dissolved Microcystin (DMC)) and 2 physical parameters (water temperature (Temp) and turbidity (Turb)) in

two kind of water sample (surface and 5m below surface) at 22 sites (KISAT, KL1, KL2, D1, D2, D3, D4, KL3, KL4, H1, H2, H3, CG3, CG5, CG6, COL, KIBOS river, KIBOUN river, SONDU river, NYANDO river) were carried out The sampling period was from June 2014 to September 2015

The application of different multivariate techniques such as two-way ANOVA, two-way ANOVA with blocking test, cluster analysis (CA), and multivariable linear regression method for the interpretation of these complex data matrices The findings revealed a better understanding of water quality and ecological status of the studied systems Then, the research offers a valuable tool for responsible management of water resources as well as rapid solutions to pollution problems (Morales, 1999; Reghunath, 2002; Wunderlin, 2001)

1.2 Objectives of study

In the current study, the large dataset obtained during over 1-year monitoring program (407 observations, 16 parameters) subjected to different multivariate statistical approaches to extract information about:

- To identify spatial and temporal trends of water quality;

- To classify the polluted water region, then figure out the influence of possible sources (natural and anthropogenic) on the water quality in Nyanza Gulf;

- To determine the relationships between Chlorophyll-a and nutrient concentrations in Nyanza Gulf for identifying essential and effective nutrients on Chlorophyll-a concentration

1.3 Structure of thesis

The thesis is composed of five chapters: (1) Introduction, (2) Literature review, (3) Materials and methodology, (4) Results and discussion, (5) Conclusion

Chapter 2: Literature review

2.1 Trophic lakes in the world

The largest lakes are the most important reserves of surface fresh water on our planet However, the eutrophication in lake has been a relevant challenge for more than

a century If such enrichment occurs because of natural processes in its catchment, the process is known as natural eutrophication If caused by anthropogenic factors, however, it is called anthropogenic eutrophication Natural eutrophication can continue for several 1,000 years, while anthropogenic eutrophication can disrupt the natural balance of a lake system and produce the same result much more rapidly, over a decadal cycle in some cases According to Ullmann’s Encyclopedia, ‘the primary limiting factor for eutrophication is phosphate’ (Werner, 2002) Human activities have increased the rate of phosphorus cycling on the Earth by approximately four times, due mainly to production and application of agricultural fertilizer Between 1950 and 1995, an estimated 600,000,000 t of phosphorus were applied to the Earth’s surface, primarily on croplands (Carpenter, Caraco, & Smith, 1998) Others sources of excess phosphate are detergents, and surface run-off from pastures, industry and domestic sources With the phasing out of phosphate-containing detergents in the 1970-s, industrial, domestic and agricultural run-off have emerged as the dominant nutrient contributors to eutrophication (Werner, 2002) In addition to phosphorus, another important limiting factor for eutrophication is nitrogen Nitrogen transport is correlated with various indices of human activity in watersheds (Arbuckle & Downing, 2001), although agricultural ploughing activities are one activity contributing most to nutrient loading Elevated levels of atmospheric compounds of nitrogen also can increase nitrogen availability in lake waters (Paerl, 1997)

Anthropogenic eutrophication processes have presently affected not only sized and middle-sized lakes, but also most of the largest lakes in the world According

small-to Wetzel (1983) the small-total volume of water contained in the world’s freshwater lakes is about 125,000 km3 Most subsequent publications provide estimated lake values that comprise both the freshwater and saltwater lakes (Table 2.1) Although Wetzel and Shiklomanov provide separate estimates of such values, their accuracy is questionable According to Shiklomanov & Rodda (2003), the volume of water in freshwater lakes is only 91,000 km3, whereas only the 100 largest lakes contain about 95,000 km3 of water (most of the volumes of these lakes are accurately defined) Even if the volume of newly discovered Antarctic Lake Vostok is subtracted from this total estimate, about 88,500 km3 of water remains, which is closer to Shiklomanov’s estimate for all freshwater lakes However, the estimate of Wetzel may be exaggerated because of a lower estimate of the total water volume in lakes reported in later publications (Ryanzhin, 2005) New estimates of the global abundance of surface water bodies were obtained, utilizing new data sources, enhanced spatial resolution, and new analytical approaches (Downing et al., 2006; Verpoorter et al., 2014)

The distribution of the water volume contained in the 93 largest lakes in the world

is presented in Figure 2.1 All freshwater lakes with surface areas exceeding 10,000 km2

are presented in the circular chart, whereas the volume of the 79 lakes with surface areas ranging from 1,000 to 10,000 km2 is shown aggregately, as a share of each of them is comparably less As evidenced in Figure 2.1, slightly <95% of all the water volume is contained in the 14 largest freshwater lakes with surface areas exceeding 10,000 km2 The three largest lakes among them contain about 58% of this volume Water bodies located in rift zones of the Earth typically contain the greatest water volumes, as well as

those occupying the large tectonic cavities that underwent glacial processing during the period of the last Quaternary glaciation.

Table 2.1: Estimates of volumes of water contained in world lakes

contained in all lakes of Earth,

km 3

Volume of water

in the freshwater lakes, km 3

Volume of saline lake waters, km 3

(Source: Izmailova & Rumyantsev, 2016)

Figure 2.1: Largest freshwater lakes in the world, including the proportion of the total

freshwater volume contained in them(Source: Izmailova & Rumyantsev, 2016)

Prior to the beginning of anthropogenic eutrophication (i.e during the period of

‘natural’ lake conditions), most of the largest lakes were oligotrophic, with about 30% being mesotrophic, and <10% being eutrophic Analysis of changes in the trophic status

of all the largest lakes indicates the fastest processes of anthropogenic eutrophication occurred in shallow lakes, both in tropic latitudes and in the temperate zone These processes were even more amplified in tropical lakes, with constantly high temperatures promoting accelerated biological processes (including phytoplankton growths) Among the largest tropical lakes whose tropic status was affected by anthropogenic eutrophication by the beginning of the 21st century was the hypertrophic condition exhibited by Lake Chapala (average depth of 7.2m) Lake Managua (average depth of 7.8m), and the eutrophic and hypertrophic status exhibited by Lake Taihu (average depth about 2m) A hypertrophic status also was exhibited by some deeper lakes under considerable anthropogenic pressures, including Lake Albert (average depth of 25m) and Lake Victoria (average depth of 40m) A considerable quantity of polluted water is contained in the Lake Victoria, its share being 8.6% of the total volume of lake waters

in the tropical zone (Izmailova & Rumyantsev, 2016), one of the most polluted largest

lakes in the world today

In summary, the anthropogenic eutrophication issues for water bodies located in various regions of the world, the process has the most serious consequences for water bodies in tropical countries Because the freshwater lake is one of the most important sources of freshwater on the Earth Therefore, the idea of preventing and reducing the impact of anthropogenic eutrophication deserve urgent attention

2.2 Lake Victoria water environment

Lake Victoria is one of the world’s largest freshwater lake (area = 68,800 km2,

mean depth = 40 m) and it supports one of the world’s largest inland fisheries, yielding almost one million tons per annum (Lewis Sitoki et al., 2010) It has attracted much interest in recent years because of the extraordinary ecological changes that occurred in

it over the last three decades, comparable in their magnitude to those in newly created man-made lakes These changes were triggered in the mid-1980s by an introduced

predatory fish, the Nile perch Lates niloticus (L.), and its subsequent destruction of the

endemic haplochromine cichlids These fishes, which formed a species flock with some

500 described forms (Witte et al, 2007), accounted for over 80% of the fish biomass in the lake (Kudhongania & Cordone, 1974) but they were small fish thought to be of little commercial value Nile perch were therefore introduced with the specific intention of converting haplochromines into a more valuable fish product (Anderson, 1961), which they did so effectively that it was feared that most haplochromines were facing extinction (Witte et al., 1992) As a result, food chains in the lake were greatly simplified with only three fish species dominating the biomass; Nile perch and Nile

tilapia Oreochromis niloticus (L.) which were both introduced, and the native cyprinid

Rastrineobola argentea (PELLEGRIN)

In addition, the lake became eutrophic following the extremely rapid growth of the predominantly rural human population, which is still increasing at more than double the global average (PRB, 2009) Land-based activities in the drainage basin, primarily deforestation and increasingly intensive agriculture, but with some contribution from urban runoff have increased nutrient loading to the lake and led to its progressive enrichment (Scheren et al, 2000) There is considerable evidence to show that eutrophication began as early as the 1950s in Kenyan waters, but rather later in Uganda and Tanzania, with the phytoplankton changing from a diatom flora to one dominated

by cyanobacteria Fossil assemblages of chironomids also provide evidence that deep water anoxia became more prevalent around that time (Stager, Hecky, Grzesik, Cumming, & Kling, 2009) The relationship between Nile perch and eutrophication is controversial, but the almost complete destruction of the endemic fishes, especially the phyto- and zooplanktivorous haplochromines, must have had significant ecological impacts and probably accelerated the process of eutrophication through cascading effects caused by the loss of phytoplanktivorous species (Goldschmidt et al., 1993).The effects of eutrophication included a change in the composition of the phytoplankton with a shift from diatoms to cyanobacteria (Gophen et al., 1995) along with an increase in algal biomass, which resulted in fish kills and more severe deoxygenation of deeper (Hecky et al., 1994; Ochumba, 1990; Ochumba & Kibaara, 1989) The transparency of the water also declined, partly because of the algal blooms and partly because of silt carried in by the rivers, and this has been implicated in the decline of the haplochromines by disrupting their mating behavior, which relies heavily

on visual cues (Seehausen, Alphen, & Witte, 1997) A further consequence of

eutrophication was an explosion of the water hyacinth Eichhornia crassipes (MART.)

SOLMS This plant was first reported from Lake Victoria in 1989 and it began to increase in the early 1990s until it infested some 200 km2 by 1998 Its spread was almost certainly a consequence of nutrient enrichment but the infestation was brought under control by weevils introduced for that purpose and the plant is now no more than

a local nuisance (Wilson et al., 2007) These environmental changes were reflected by changes in the fishery which, prior to 1980, amounted to about 100,000t per annum and consisted primarily of haplochromines, tilapias and other native species Catches increased dramatically from 1980 onwards to reach about 500,000t by 1989 with Nile

perch making up 70% of the total (Reynolds et al., 1995) By 2007 the total catch had risen to around 1,000,000 t with the Nile perch catch being around 240,000t (25% of the

total) with Rastrineobola, Oreochromis and haplochromines making up the balance

The decreased proportion of Nile perch in the catch led to concerns that the stock was becoming overfished (Matsuishi et al., 2006) and the biomass of Nile perch in the lake has indeed declined significantly over the last decade (Lewis Sitoki et al., 2010) Simonit & Perrings (2005) asserted that landings of all fish had declined since 1994 and, while recognizing the contribution of overfishing, suggested that eutrophication was an important factor This was taken further by Kolding et al., (2008) who claimed that the Nile perch stock has not been affected by fishing but that environmental conditions in the lake may now limit its productivity The increased intensity of deoxygenation during the stratified period, when >50% of the water column was said to become anoxic, was seen as a particular threat since Nile perch is sensitive to low oxygen concentrations (Schofield & Chapman, 2000)

2.3 Nutrient enrichment in Nyanza Gulf

The Nyanza Gulf is a semi-closed bay with a limited water exchange with the main basin (Calamari, Akech, & P B O Ochumba, 1995) The isolated situation distinguishes the Nyanza Gulf from the numerous more open bays where the impacts of eutrophication are probably diminished by dilution effects resulting from surface seiches (Haande et al., 2011)

The characteristics of the waters indicate nutrient enrichment According to Hecky (1993), concentration of nitrogen has become higher particularly in inshore waters On the other hand, concentration of silicon in the epilimnion of offshore regions has reduced by a factor of ten (Kilham & Kilham, 1990) The phytoplankton community

that showed clear seasonal successions of diatoms, blue-green algae (cyanobacteria) and green algae and several other taxa (Talling, 1965, 1987) is now persistently predominated by cyanobacteria (Lung’ayia, M’harzi, Tackx, Gichuki, & Symoens, 2000) Overall, algal biomass has increased and blooms dominated by cyanobacteria are common (Ochumba & Kibaara, 1989)

In 2008, Gikuma-Njuru reported that the average TP concentration almost doubled from 68µg/l (2000–2002, Gikuma-Njuru & Hecky, 2005) to 107µg/l (2005–2006) However, the increase in TP is not matched with a similar increase in TN, which has changed from an average of 983 µg/l in 2000 – 2002 (Gikuma-Njuru & Hecky,

2005) to 1,356 µg/l (L Sitoki, Kurmayer, & Rott, 2012) A large portion of this nutrient

increase can be attributed to the allochthonous fluvial nutrient inputs from agricultural and urban areas in the catchment (Gikuma-Njuru, 2008; Hecky et al., 2010) The rapid increase in nutrient concentrations seems to be further enhanced by heavy rainfall (Hecky et al., 2010) Based on the average TP concentrations, the trophic classification system for tropical lakes (Salas & Martino, 1991) assigns the Nyanza Gulf to a hypereutrophic state On the other hand, the dissolved inorganic nitrogen (DIN) mainly

NO3–N concentration is relatively low when compared with TN This discrepancy between DIN and TN seems to be a common feature of many tropical water bodies For example, in Lake Maracaibo, Venezuela nitrogen availability was found to be regulated

by efficient diurnal nutrient (nitrogen) recycling based largely on dissolved organic nitrogen in spite of low DIN (Gardner et al., 1998)

Water hyacinth Eichhornia crassipes, a notorious waterweed, recently invaded the

lake starting in 1989 (L Sitoki et al., 2012) It has spread to all parts of the lake particularly in bays and inshore shallow waters These areas are commonly nursery and

foraging grounds for many fish species including rarities Oxygen depletions due to decomposition of the dead plant matter may affect life cycles, especially recruitment of fish and other aquatic organisms Physical cover by water hyacinth mats reduces penetration of light going into the water, reduces mixing of the water column thereby affecting nutrient dynamics, creates associated habitats and leads to emergence of new organisms The plant is a threat to ecosystem health and productivity.

Social and economic problems that are caused by the weed include choking of water pumps, blocking of piers and landing beaches and impeding fishing and boat navigation Measures to control the weed include introductions of weevil species that destroy the plant and mechanical harvesting

2.4 Correlation between Chlorophyll-a and nutrient factors in trophic lakes

Chlorophyll a is the major photosynthetic pigment in a lot of phytoplankton and a trophy index in aquatic ecosystems (Dillon, 1975) Chlorophyll a (Chl-a) is often used

as an estimate of algal biomass, with blooms being estimated to happen when Chl-a concentrations go above 40 μg/l

The importance of phytoplankton in tropical reservoir ecosystems include its use

in estimating potential fish yield, productivity, water quality, energy flow, trophic status and management (Beyruth & Tanka, 2000) These reservoirs are increasingly threatened

by human activities In Lake Victoria, phytoplankton species composition, numerical abundance, spatial distribution and total biomass are in a direct relation with the environmental factors (Lung’ayia et al., 2000) Actually, environmental and temporal changes determine the community present in a lake (Levandowsky, 1972)

So a methods for the estimation of the growth and development of the phytoplankton community is to perform an analysis of photosynthetic pigments, even

though the content of chlorophyll in the cells changes with the availability of light and thus with depth and trophic gradient (Kasprzak et al., 2008)

Because eutrophication is defined as an aquatic ecosystems response to nutrient loading, the ability to identify important factors and predict subsequent algal blooms with the use of a Chl-a equation could be a key lake water management tool (Dillon & Rigler, 1974) Both chemical and physical controls can be used to prevent or remove algae or algae byproducts from the water In particular, information about the form of Chlorophyll-a:nutrient relationships (Prairie, Duarte, & Kalff, 1989) has allowed lake managers to establish nutrient concentration and loading aims Nitrogen and phosphorus are often identified limiting nutrients to algal biomass and silicon is necessary for diatom growth (Hecky & Kilham, 1988) Nitrogen occurs in fresh water in numerous forms: dissolved nitrogen, amino acids, amines, urea, ammonium (NH4+), nitrite (NO2-), and nitrate (NO3-) In aquatic ecosystems, phosphorus (P) can be found either in particulate matter or as soluble inorganic phosphorus, orthophosphate (PO43-) (Knud-Hansen, 1997)

A review of the 1995 to 1997 biological abstracts about significant factor for algal blooms, illustrates that, of 596 articles on estuaries and nutrients, 52% consider only nitrogen, 32 % refer to both nitrogen and phosphorus, and 16% consider only phosphorus, although the preponderance of study on N, the evidence for general N limitation of coastal systems is feeble compared to the data for general P limitation of freshwater systems There have been a small number of comprehensive analyses of the form of phosphorus-chlorophyll relationships The phosphorus-chlorophyll relationship most probably outcomes from the dependence of algal growth rates on phosphorus availability Nitrogen limitation of algal biomass seems to be more general in

subtropical and tropical lakes (Hecky et al., 1993), while phosphorus appears to be the primary limiting nutrient in temperate lakes (Smith, 1990) Other nutrients, for example iron and silicate, have been reported to be limiting in some other regions (Johnson, Chavez, & Friederich, 1999) The limiting nutrient is decided mainly by the mass equilibrium between elements such as C, N, P, and Si, and their relationship to the growth requirements of the phytoplankton (Wu & Chou, 2003).

Evaluation of phytoplankton community structure is essential and useful as an indicator of the water quality Because of Chl-a identifed as a major photosynthetic pigment in a lot of phytoplankton and a trophy index in aquatic ecosystems and the other hand the chlorophyll a concentration in the phytoplankton cells changes with nutrients and environmental factors so know about the e�ective factors on chlorophyll a concentration is very important for ecosystem management

2.5 Environmental health problem caused by harmful cyanobacteria

2.5.1 Environmental health problem caused by harmful cyanobacteria 2.5.1.1 The effect for animals

An example of the inadequate recognition of cyanobacterial poisoning hazards has been presented several times over recent years by mat-forming cyanobacteria

Dog deaths in Scotland, Ireland and New Zealand have occurred after eating fragments of anatoxin-a-producing cyanobacterial mats and drinking the water nearby

(Hamill, 2001) The neurotoxic material, including benthic benthic Oscillatoria and

Phormidium, can be overlooked during waterbody inspection since it is brown-black,

rather than blue-green This is due to the presence of cyanobacterial phycoerythrin, in addition to phycocyanin Shoreline accumulations of the mats are, however, prone to scavenging by dogs, which appear to be attracted to the odorous material Further

benthic cyanobacterial neurotoxicoses involving three dogs, of which two died, occurred at a Dunfermline (Scotland) freshwater lake in May 2003 (Codd, Lindsay, Young, Morrison, & Metcalf, 2005) Anatoxin-a- and Microcystin-containing cyanobacterial mats have accounted for numerous cattle deaths at small lakes in alpine summer pastures in Switzerland (Mez et al., 1997)

Cattle died after drinking water containing a Cylindrospermopsis raciborskii

bloom at a farm pond in Queensland, Australia in 1997

The subsequent findings of Saker et al (1999), including liver histopathology and cylindrospermopsin in the bloom and in a derived laboratory isolate, appear to be the

first report of animal poisonings attributable to C raciborskii and cylindrospermopsin

2.5.1.2 The effect for humans

There is a small number of publications on human illnesses and deaths associated with exposure to cyanobacterial cells and toxins in during the past 40 years (Carmichael

et al., 2001) These and additional episodes, including skin and respiratory irritations among swimmers and fishermen on the Queensland coast from 1996-1998 (Codd et al., 2005)

The serious health incidents, most notably the deaths of 52 Haemo-dialysis patients up to Dec 1996, out of over 100 affected at a clinic in one episode (Codd et al., 2005), have increased the attention and co-operation of clinicians, epidemiologists, public health professionals, water scientists and cyanobacteriologists, to begin to identify needs, formulate and implement policies to reduce health risks presented by cyanobacterial toxins Indications of the extent of health problems due to recreational exposure are still limited: partly because the reporting of small episodes of relatively mild illness (skin irritation, gastrointestinal upsets) associated with recreational contact

with cyanobacterial blooms, if recognized at all, may only be to local and national health and environmental authorities, rather than to primary journals.

2.5.2 Microcystin in Nyanza Gulf

While the spatial–temporal variability of the phytoplankton composition has been studied previously (Gikuma-Njuru & Hecky, 2005), the concentration of MC has never been recorded in a systematic manner Several genera of cyanobacteria commonly occurring in Lake Victoria (Anabaena, Microcystis) have the potential to produce MC, which are the most widespread cyanotoxins with over 80 structural variants (Krienitz et al., 2002) Krienitz et al (2002) reported MC production from surface water in Kisumu Bay, which was probably the first report on cyanotoxin occurrence from Lake Victoria

In the meantime, MC has been reported from the Mwanza Gulf, Tanzania (Sekadende et al., 2005), and several times from Uganda (Semyalo, Rohrlack, Naggawa, & Nyakairu, 2010) In the study of Krienitz et al., (2002), MC concentrations (<1 µg/l) were reported during an Anabaena bloom (Anabaena spp >90% of the phytoplankton biovolume with

8 x 105 cells/ml) The same authors stated that it was unclear as to whether the MCs that were detected could be attributed to all the cyanobacterial species found and the identification of the responsible MC producing species requires clarification In contrast, the results from several Ugandan waters including Lake Victoria showed that Microcystis was the only MC producer (Okello et al., 2010)

According to Reynolds (2006), even higher phytoplankton biomass (biovolume

>10 mm3/l) could be predicted from the presently observed nutrient concentrations in the Nyanza Gulf in the near future Strategies of dealing with MCs from lake water used for the drinking water supply should involve a regular monitoring of the cell numbers of toxigenic cyanobacteria in the raw water When Microcystis cells pass a certain

threshold (>2x103 cells/ml, Chorus & Bartram, 1999), additional treatment steps would

be necessary, including flocculation and ozonation to remove the particles, followed by sand filtration or activated carbon filtration for removing dissolved MCs and other toxic compounds Particularly sand filtration has been shown to be efficient in removing MCs from raw water (Chorus & Bartram, 1999)

Nevertheless, raising public awareness through the media has to be fostered Since, during dry periods, alternative water sources are scarce, the installation of in-situ/household biosand filtration units that can be used for water purification should be considered

Chapter 3: Materials and methodology

3.1 Population of study area

Surrounding residential areas of Nyanza Gulf are the Kisumu County, Siaya County and Homa Bay County Both of Homa bay County and Siaya County are the region where own the lower density of population compare to Kisumu County Furthermore, their main economic activities are the fishery rely on the mouth of the Nyanza Gulf Meanwhile, Kisumu County with the highest density of population is the largest bustling city in the region Kisumu is not only the center of industrial (textiles, molasses, fish processing plants and agricultural produce processors, etc.) but also the agriculture and fishing Because here is the basin of some large rivers (Nyando and Kibos), most of the water for irrigation comes from River Nyando

Based on the characteristics of the population distribution and geography (Figure 3.1, Figure 3.2, Figure 3.6, Figure 3.7), the northern Nyanza Gulf is generally considered as the area causing the key impact on the water quality here Thus, the sampling network of LAVICORD program has distributed 7 monitoring stations in northern Nyanza Gulf and 4 stations on the 4 rivers with a high sampling frequency (biweekly) Besides, in the southern Nyanza Gulf (the area of the Homa bay County and Siaya County), 11 monitoring stations was set with low sampling frequency (bimonthly) Such monitoring program has been set to expect the assessment of water quality in the region as reasonable as possible

Figure 3.1: Population and density of population of counties locate around Nyanza Gulf

(2009)

Figure 3.2: Former Nyanza Province

Homa Bay County is a county in the former Nyanza Province of Kenya Its capital and largest town is Homa Bay The county has a population of 963,794 (2009 census) and

an area of 3,154.7 km²

Figure 3.4: Homa Bay county of Nyanza Province, Kenya

(Source: Wikipedia)

Kisumu County

Kisumu County is one of the new devolved counties of Kenya Its borders follow those

of the original Kisumu District, one of the former administrative districts of the former Nyanza Province in western Kenya Its headquarters is Kisumu City It has a population

of 968,909 (according to the 2009 National Census) The land area of Kisumu County totals 2085.9 km²

Figure 3.5: Kisumu county of Nyanza Province, Kenya

gulf; 4 sites, that were named KIBOS, KIBOUN, SONDU, NYANDO, locate on 4 rivers with the same name to site.

Figure 3.7: Sampling map in Nyanza Gulf, Kenya

(Source: Google map)The station which located on the head of gulf and rivers were observed biweekly for one and half of year (June 2014 to September 2015) The station which located on and near mount of gulf were observed bimonthly for one year (June 2014 to July 2015)

By considering the physicochemical properties, nutrient constituents and toxic characteristic of water quality, 16 representative variables were collected for testing These variables were water temperature (Temp), pH, dissolved oxygen (DO), turbidity (Turb), total nitrogen (TN), total dissolved nitrogen (TDN), nitrite (NO2-), nitrate (NO3-), ammonium (NH4+), total phosphorus (TP), total dissolved phosphorus (TDP), phosphate (PO43-), Chlorophyll-a (Chl-a), total suspended solid (TSS), total Microcystin (TMC), dissolved Microcystin (DMC)

Table 3.1: Coordination of sampling sites

Order Station Latitude Longtitude Depth (m)

3.3 Data treatment and multivariate statistical methods

Method structure:

3.3.2 ANOVA two-way To determine a different between surface

and from surface 4 – 5m below of data In order to understand the effect of stratification on water quality distribution

To determine a spatial and temporal variation of each parameter to find out the reason of water quality variation in study area

This is the first step to understand complex data matrix

3.3.3 Cluster analysis To recognize the similarity of sampling

sites in the large area from a variety of water quality parameters

3.3.4 Linear regression analysis To find out a predict model of correlation

between Chlorophyll-a and nutrient parameter or physical factor in that areaMain objective: this study aims to find out the pollutant source and the model of phytoplankton and the other factors which has been used to prevent the eutrophication

in this area

3.3.1 Data treatment

In regarding to raw data, it was considered in terms of the distribution of the data Histogram and Turkey boxplot were used to show visually distribution of data Shapiro – test in R was used to quantitatively examine the distribution of the data (p-value > 0.05 represents for the normal distribution)

In the case, the data do not follow the normal distribution, log-transformation was applied for data standardization Input data for some of analysis such as CA, z-transformation was used to avoid misclassifications arising from the different orders of magnitude of both numerical values and variance of the parameters analyzed (Liu,

C.W., 2003; Simeonov et al., 2003) All mathematical and statistical computations were made using Microsoft Office Excel 2016 and R.

3.3.2 Two-way analysis of variance (ANOVA)

Two-way ANOVA with blocking is known as Friedman test (Friedman, 1937; Hollander & Wolfe, 1973) for data which do not follow normal distribution In this method, the only assumption is that the samples are drawn from continuous distributions And the null hypothesis is now tested:

Null hypothesis: the MEDIAN values of the underlying distributions of the TN

concentrations in the two layers are equal

Alternative hypothesis: the MEDIAN values of the underlying distributions of the

TN concentrations in the two layers are different.

In the result, the p-value is almost zero (p = 10-16), and the Friedman test thus confirms the results of the classical parametric ANOVA with blocking The Friedman test is the best choice for testing this hypothesis

In addition to, two-way analysis of variance (ANOVA) was also carried to estimate the temporal and spatial variabilities of water quality in the study area (Mendiguchı´a, Moreno, & Garcı´a-Vargas, 2007)

3.3.3 Cluster analysis

Cluster analysis is a group of multivariate techniques whose primary purpose is to assemble objects based on the characteristics they possess Cluster analysis classifies objects, so that each object is similar to the others in the cluster with respect to a predetermined selection criterion The resulting clusters of objects should then exhibit high internal (within-cluster) homogeneity and high external (between cluster) heterogeneity Hierarchical agglomerative clustering is the most common approach,

which provides intuitive similarity relationships between any one sample and the entire data set, and is typically illustrated by a dendrogram (tree diagram) (McKenna Jr., 2003) The dendrogram provides a visual summary of the clustering processes, presenting a picture of the groups and their proximity, with a dramatic reduction in dimensionality of the original data The Euclidean distance usually gives the similarity between two samples and a distance can be represented by the difference between analytical values from the samples (Otto, 1998) In this study, hierarchical agglomerative CA was performed on the normalized data set by means of the Ward’s method, using squared Euclidean distances as a measure of similarity.

The Ward’s method uses an analysis of variance approach to evaluate the distances between clusters in an attempt to minimize the sum of squares (SS) of any two clusters that can be formed at each step The spatial variability of water quality in the whole river basin was determined from CA, using the linkage distance, reported as

Dlink/Dmax, which represents the quotient between the linkage distances for a particular case divided by the maximal linkage distance The quotient is then multiplied by 100 as

a way to standardize the linkage distance represented on the y-axis (Pesce et al., 2001; Simeonov et al., 2003; Singh, Malik, & Sinha, 2005)

3.3.4 Multivariable linear regression analysis

Regression analysis is used to predict values of a dependent or response variable from the values of one or more independent or explanatory variables (Draper & Smith,

1998; Fox, 1997) The dependent variable y is the variable that will be predicted using the information of the independent, explanatory, variable(s) x1,…, x p with p ≥ 1 When

dealing with only one x and one y variable for correlation analysis, which is somewhat related to regression analysis, the choice of the x and y variable has no influence on the

resulting correlation coefficient For regression analysis, however, this choice is of paramount importance In general, the higher the Spearman rank-order correlations

coefficient between the x and y variable, the more similar regression lines will result.

Regression analysis could, for example, be interesting if one variable is very difficult or very expensive to analyze or the determinations were of poor quality It could thus be tempting to predict values of this variable using analytical results of the

other variables In practice a number of measurements for the x and y variable are

needed to build a model for prediction

Chapter 4: Results and discussion

4.1 Data descriptive

As the LAVICORD project progressed, chemical analysis for 12 bio-chemical parameters and 4 physical parameters in two kind of water sample (surface and 5m below surface) collected at about 22 sample sites were received These needed to be summarized, compared and mapped, so now we enter the field of statistical data analysis

Histogram and Turkey boxplot were used to display a distribution of each parameter in whole study area

From Figure 4.1, almost observed data strong skewed with many outliers in the left tail of distribution In addition to, the result from Shapiro test showed in Table 4.1 lead to a conclusion that the data in this study have not comply with normal distribution Therefore, in data analysis, non-parametric method (Friedman test) will be preferred to choose instead of parametric method to compare among water quality parameters In the other hand, normalization is also extremely important step to prepare input data for other analysis (linear regression analysis)

Table 4.1: Summary p – value of Shapiro test