Water Quality and Remote Sensing: a Case Study of Lake Naivasha, Kenya

Nobuhle P. Majozi1, Suhyb M. Salama2, Stewart Bernard1, David M. Harper3 1 CSIR, Natural Resources and Environment, Box 395, Pretoria, South Africa, 0001 2 Faculty of ITC, University of Twente, Box 217, 7500 AA Enschede, The Netherlands 3 Department of Biology, University of Leicester, University Road, Leicester LE1 7RH, England Abstract Freshwater resources are losing their quality posing a threat to human and aquatic life. Remote sensing has shown potential to monitor these resources. This study aimed to retrieve diffuse attenuation coefficient (Kd), to map euphotic depth (Zeu) of Lake Naivasha, Kenya, using Medium Resolution Imaging Spectrometer (MERIS). Radiometric data was collected at Lake Naivasha. The Kd model was calibrated using in situ data and validated using MERIS 490 nm band, giving an RSME 0.43 and MAE 31%. Zeu validation gave an RSME 0.24 and MAE 20%. Atmospheric correction processors, MERIS Neural Network processors, and the ODESA software, were applied to MERIS images. Eutrophic Lakes was the most accurate at 490nm with MAE 43% and RSME 0.49. The Zeu maps show the spatio-temporal variation of Zeu, indicating the suitability of MERIS to monitor water quality of inland waters. Routine production of similar products for South African waters is currently being developed. Key words: diffuse attenuation coefficient, euphotic depth, Lake Naivasha, remote sensing, MERIS INTRODUCTION Freshwater ecosystems are limited renewable resources essential for socio-economic development and environmental sustainability (Pimentel et al., 1997, Koehler, 2008). However, many are deteriorating rapidly due to anthropogenic activities within their catchments exacerbated by climate change (Shiklomanov, 1998, Vrsmarty et al., 2000, Kundzewicz et al., 2007). Satellite remote sensing is a potential tool for monitoring inland water quality and other environmental phenomena because it offers an opportunity for data collection on systematic synoptic and temporal scales, especially in inaccessible areas (Dekker et al., 2002, Schmugge et al., 2002, Jensen, 2009). Satellite data are becoming more freely available from providers such as the European Space Agency (ESA) and National Aeronautics and Space Administration (NASA). However, in situ observation networks need to be developed for calibration and validation of remote sensing data to improve the applicability and usefulness of earth observation technology in developing countries. Satellite remote sensing of water quality has been undertaken for decades, especially for the open ocean and coastal waters, providing estimates of primary production, suspended particulate matter (SPM), non-algal particles (NAP), coloured dissolved matter (CDOM) and chlorophyll a (Chl a) as a proxy for phytoplankton biomass (Marrari et al., 2006, Cui et al., 2010). The European Space Agencys (ESA) Medium Resolution Imaging Spectrometer (MERIS) onboard the ENVISAT satellite, and the follow up OLCI mission on the Sentinel 3 satellite series, facilitates repeated monitoring of inland water at a moderate spatial scale of 300 m (Matthews et al., 2010, Levrini & Delvart, 2011). Light transmitted through the water column interacts with the particulate and dissolved matter and the water molecules (Mobley, 1994). Hence, it is absorbed and/ or scattered by these constituents, resulting in it diminishing exponentially with depth (z). Typical parameters used to quantify light penetration include the Secchi disk depth, and diffuse (Kd) - or beam (c) - attenuation coefficients. Secchi disk transparency is common, easily conducted but approximate measure of light attenuation and is widely used as an important overall indicator of the trophic state of a waterbody (Preisendorfer, 1986, Kirk, 2011), whereas Kd and euphotic depth (Zeu) give a more quantitative measure of water clarity. The euphotic zone (euphotic meaning well lit in Greek) is where there is sufficient Photosynthetically Active Radiation (PAR) to support photosynthesis (Kirk, 2011). Zeu

defines the water depth where the PAR is 1% of its initial value at the surface (Scheffer, 2004), and has been used to describe the trophic status of a waterbody as well as its primary productivity (Khanna et al., 2009, Haande et al., 2011, Jin et al., 2011,) and Kd describes the exponential vertical decrease of irradiance, thus quantifying the radiometric attenuation of light in water and the depth of euphotic zone (Mobley, 1994). Satellite-derived chlorophyll has been used to estimate Zeu in Case 1 waters where light attenuation is largely due to phytoplankton pigments (Morel & Berthon, 1989). This approach is unsuitable for inland waters like Lake Naivasha, where light attenuation is caused by all the optically significant water components. An alternative approach is to use the physical relationship between diffuse attenuation coefficient in the PAR region (Kd(PAR)) and Zeu. This involves first defining an empirical relationship between Kd(PAR) from Kd(490) using regression analysis. Research in various areas has indicated a very strong linear relationship between the two parameters (Zaneveld et al., 1993, Barnard et al., 1999, Kratzer et al., 2003, Tang et al., 2007). Kd(490) is a standard product for the Moderate Resolution Imaging Spectroradiometer (MODIS), Sea Wide Field-of-view Sensor (SeaWiFS) and MERIS (Mueller, 2000, Kratzer et al., 2008). Also, considering that Kd is an apparent optical property that can be derived from inherent optical properties, various algorithms linking Kd to absorption and backscattering coefficients, have been developed based on radiative transfer approximations (Smith & Baker, 1981, Kirk, 1984, Gordon, 1989, Morel & Loisel, 1998, Loisel & Stramski, 2000, Lee et al., 2005). The main objective of this study is to quantify Zeu from MERIS images in Lake Naivasha and to evaluate associated errors in the technique using in situ measurements. First, three atmospheric correction approaches were evaluated for their suitability in shallow tropical lakes like Lake Naivasha, namely the MERIS Neural Network processors, Case 2 Regional and Eutrophic Lakes, and the Optical Data Processor of the European Space Agency (ODESA) Bright Pixel atmospheric correction (BPAC). In-water Kd and Zeu algorithms were developed using in situ data and their performance and applicability was evaluated using atmospherically corrected MERIS data. METHODS Site Description Lake Naivasha (0045 S; 360 26 E) is a shallow freshwater lake situated approximately 80km North-West of Nairobi, in a dry, water-scarce zone of the Kenyan Rift valley. It is the second largest freshwater lake in Kenya after Lake Victoria, in a region dominated by soda lakes. It has an average area of 145 km2 and has an average depth of 5m. The lake has no surface outlet; its freshness has been explained by the interaction of the lake with the groundwater (Bergner et al., 2003, Becht et al., 2006b, Ayenew & Becht, 2008). Rich in biodiversity, the lake supports a thriving fishing industry, intensive horticulture and floriculture industries, as well as a geothermal power generation station. It shares the water table with groundwater aquifers that provide water to Naivasha town and surrounding population. Hence, it plays a key role in local and national economic development (Harper & Mavuti, 2004), and was declared a wetland of international importance in 1995 under the Ramsar Convention (Becht et al., 2006a). The lake is under enormous anthropogenic pressure from activities within the catchment which have contributed to the deteriorating water quality of the lake, posing a threat to the aquatic life as evident through a report on the death of the fish in the lake (Otian'a-Owiti & Abiya Oswe, 2007, Becht & Chesterton, 2010, Kona & Mwiti, 2010). The increased input of phosphorus and nitrogen and sediment load through Gilgil and Malewa Rivers has caused a rise in the nutrient load of the lake which has fuelled the growth of phytoplankton, and led to increased turbidity levels (Harper et al., 1994, Hubble -3 & Harper, 2002, Kitaka et al., 2002b). Chl a concentration increased from 0.03mgm to -3 0.178mgm between 1982 and 1995, whereas Secchi disk depth reduced to 60cm within the same period. One hundred and sixteen species of algae and cyanobacteria have been identified in the lake, some of which are indicators of increasing nutrient levels in the lake (Ballot et al., 2009).

Figure 1: Map of Lake Naivasha showing the positions of the sampling routes and stations during the field campaign from 17 September to 03 October 2010, diamond symbols are the sampling stations. In situ Data Collection Intensive bio-optical measurements were undertaken on Lake Naivasha from 17 September to 3 October 2010. These were hyperspectral radiometric measurements: above-water radiance (Lsfc()), sky radiance Lsky(), downwelling above-water irradiance Ed(0+,), upwelling underwater radiance (Lu()), and downwelling under-water irradiance at two depths (Ed(z0,) and Ed(z1,) where z0=0.1m and z1=0.6m). Water samples were collected to analyse for SPM concentration, CDOM and Chl a absorption in the laboratory. MERIS satellite overpass times were acquired from the European Space Agency (ESA) during fieldwork planning and preparation stage. During the overpass days, i.e. 17, 20, 23, 26, 29 September 2010, intensive radiometric measurements were collected 1 hour of the scheduled overpass time for ground-truthing. Radiometric measurements were taken using the Trios RAMSES-ARC radiance and RAMSES-ACC-VIS irradiance sensors. Above-water radiometric measurements, i.e. Lsfc() and Ed(0+,) were taken approximately 30cm above the water surface, at solar azimuth angle to avoid shadow effect and zenith angles between 300 and 50 and 135 surface reflectance (Mueller & Fargion, 2003). The above-water radiance Lsfc() were then corrected for skylight reflection effects to give water-leaving radiance (Lw()) (Fougnie et al., 1999, Mobley, 1999): () = () () Bookmark not defined.) (Error!

(=0.02) is a proportionality factor that depends on direction, wavelength, wind speed, sensor field-of-view and sky radiance distribution (Austin, 1974).

The water samples were collected approximately 15 cm below the water surface at each sampling station and were analysed for SPM concentration, CDOM and Chl a absorption coefficients. The UV/ visible spectrophotometer was used to measure the absorbances of

CDOM, Chl a and Total Suspended Solids, which were then converted to absorption coefficients. SPM concentration was determined using the gravimetric method. 60-100 ml of water samples were filtered through preweighed 0.45 m cellulose nitrate Whatman GF/F filters. The filters were immediately dried in a 75C oven, and reweighed using an electrobalance. To determine CDOM absorption, 30-50 ml water samples were filtered through 0.2 m pore-diameter Whatman GF/F membrane filters, and the filtrate used to measure absorbance by (CDOM) at 400, 412, 440, 490 and 555 nm. Absorbance spectra were then converted to absorption using the following equation: () = 2.303
( ) (750)

where a() is the absorption coefficient, L is the pathlength (=1 cm), and Abs(750) is used to correct for the scattering effect of the cuvette. Chl a absorbance was determined by filtering 150-300 ml sample through 0.45 m Whatman GF/F filters, then by extracting the pigments using 90% acetone following Wright and Jeffrey (1997) and Aminot and Rey (2000). After centrifuging, the samples were allowed to extract for 1 hour. The absorbance spectra were converted to Chl a concentration using: () = 29.62 ((665) (750))

(2)

(3)

where aph(), Ve and Vs denote absorption coefficient of Chl a, absorbance, volume of extraction agent and volume of sample, respectively. MERIS Data Processing MERIS Level 1B products were acquired from ESA via their site Earth Observation LinkStand Alone (EOLiSA). Five images were captured during the field work: 17, 23, 26 and 29 September 2010, however after image screening only 3 were used as matchups. BEAM version 4.9 and ENVI 4.7 software were used for image processing. The Level 1B radiance products were processed to remove the absorption and scattering effects of the atmosphere, using two BEAM plug-in algorithms based on the MERIS Case-2 Core Module, Case-2 Regional (C2R) and Eutrophic Lakes (Doerffer & Schiller, 2008) and the ODESA BPAC processor, based on: The images were also pre-processed for land adjacency effect using the Improved Contrast between Ocean and Land algorithm (ICOL) version 1.1.2. The results of the atmospheric correction with and without ICOL were compared. Results were analysed against the in-situ remote sensing reflectance. Bio-optical Model Development Downwelling profiles of irradiance at the two depths were used to calculate the spectral diffuse attenuation coefficient by applying the exponential coefficient based on the Lambert-Beer law. () =
1

Rrs() was computed using in situ radiometric measurements of water-leaving radiance (Lw()) and downwelling irradiance Ed(0+,) following Mobley (1999): () =
(0+ , ) ( )

where Ed (z0 ; ) and Ed (z1 ; ) and are spectral underwater irradiance measurements at depths z0 and z1.

(0 ; )

(1 ; )

: |1 | > |0 |

(5)

(6)

Kd(PAR) Derivation PAR, defined as the solar radiation integrated from 400nm to 700nm that is used by phytoplankton during photosynthesis, was computed from the underwater irradiances (Lee et al., 2005): () = ; 400
700

(7)

The attenuation coefficient in the PAR range (Kd(PAR)) was then determined from the PAR at z0 and z1 depths by:
1

In order to determine Kd(PAR) from the MERIS-derived Kd(), an empirical relationship was further developed between Kd() and Kd(PAR) from field data, based on various studies (Zaneveld et al., 1993, Barnard et al., 1999, Kratzer et al., 2003) that discovered a strong linear relationship between Kd(PAR) and Kd(490): ( ) = (490) + (9)

() =

(0 ; )

(1 ; )

(8)

where m and j are derived by linear regression between Kd() and Kd(PAR), and represents 490nm. Euphotic depth was finally derived from Kd(PAR) based on the depth where Photosynthetically Active Radiation (PAR) is reduced to 1% of the initial value (Mobley, 1994) at the water surface (PAR(z1 ; )/ PAR(z0 ; ) = 0.01), where z0=0m and z1=Zeu: = = Replacing Kd(PAR), equation 9 then became:
( )+ 4.605 ( ) 4.605

(10)

(1)

MERIS Kd(490) Derivation Taking into account that other methods only estimate a single spectral value of attenuation coefficient at a time, i.e. Kd(490), the proposed approach sought to estimate the whole spectral range Kd all at once. Using the correlation of Kd() and Rrs(), the data were subdivided into 4 distinct classes. The result of the classification was: Spectral range 1 440-600nm Spectral range 2 600-800nm Spectral range 3 800-930nm For this study, a non-linear regression analysis was performed based on the band ratio algorithm modified by Mueller (2000) to derive Kd() from Rrs(): () =
(560) ( )

where , and are model coefficients derived from the regression analysis, Rrs(560) is the reference remote sensing reflectance and Rrs() is the spectral remote sensing reflectance, using the 560 nm channel because MERIS does not have a channel at 555 nm. The above model was then run on each of these spectral ranges, and the accuracy of the results assessed using the mean absolute error (MAE), root mean square error (RMSE) and coefficient of determination (R2).

(2)

 = =1
1

= ) ( ) ]2 =1[(

( ) ( ) ( )

100%

(3) (4)

0.5

The statistical analysis was also performed for the 3 atmospheric correction algorithms and MERIS products of the Zeu. RESULTS AND DISCUSSION In situ Measurements and Derivations Water colour ranged from brownish to green, with clumps of phytoplankton floating on the surface in the main lake. The lake frequently experienced strong winds during the afternoons, which resulted in the turbulent mixing of waters. This could explain the continuous appearance of algal blooms on the water surface on calmer days during the sampling period. Figure 2 illustrates the spectral signature and diffuse attenuation coefficient of the lake water, revealing the presence of Chl a, NAP, and CDOM in the water, characteristic of turbid, productive, inland waters (Gitelson et al., 2007). The first significant dip between 400 and 490 nm (Figure 2 (left)) is attributed to absorption by chlorophyll, dissolved organic matter and suspended sediment. As the spectrum moves across to the green range, absorption by chlorophyll pigments decreases, as well as scattering by all particulate matter, resulting in rising reflectance. The peak around 560 nm is attributed to minimal absorption by phytoplankton pigments and scattering by SPM, thus high reflectance in that region. The subsequent dip within the 620-630 nm range is the result of absorption by Phycocyanin (Gitelson et al., 2007). The reflectance peak in the nearinfrared part of the spectrum, around 705nm, highlights the high concentration of phytoplankton in the water. The attenuation coefficient is a function of absorption and backscattering of the optical components of water as well as the angular distribution of the light field (Siegel & Dickey, 1987). In the visible range, the attenuation coefficient is a combination of both absorption and backscattering coefficients of Chl a, CDOM and SPM play a role (Mobley, 1994). The effect of absorption of water (aw) can be seen in the near-infrared, as shown by the significant increase of the attenuation coefficient after 700 nm. The main optically active component in Lake Naivasha at the time of sampling was Chl a, which explains the high Kd around 440 nm and the dip at 705 nm.

Figure 2: Measured remote sensing reflectance (left) and diffuse attenuation coefficient (right) Table 1 shows a summary of the in situ measurements of the lake. The mean Secchi disk depth (SDD) of the lake was 37.8610.66cm. SPM concentration ranged between 6.8 and 101 mgl-1, with an average of 37.8 mgl-1. Chl a concentration was a

mean of 30.41mgm-3 9.07mgm-3, after multiplying the original Chl a concentration values by the factor 1.64 (Sartory & Grobbelaar, 1984). Various studies show that acetone has low extraction efficiency compared to ethanol and methanol in cyanobacteria dominated waters (Robarts & Zohary, 1984, Sartory & Grobbelaar, 1984, Ppista et al., 2002). Also, for maximum pigment extraction, literature recommends samples to be kept 24 hours before reading (Holm-Hansen & Riemann, 1978). However, these samples were only kept for a maximum of 1 hour for extraction before reading, resulting in underextraction of the Chl a pigments (Buffan-Dubau & Carman, 2000, Cartaxana & Brotas, 2003). Table 1: Statistical summary of the optical properties of the lake showing the minimum, maximum, range, mean and standard deviation Optically active component N Min Max Range Mean Std at(490) (m) Cspm (mgl1) Cchl a (mgm-3) Kd(490) (m) Kd(PAR) (m) Secchi disk depth (m) Zeu (m) aph(440) aCDOM(440) 85 85 95 85 73 85 73 136 136 2 6.8 12.95 2.466 1.729 0.21 0.685 0.034 0.5 11.1 101 71.25 11.93 6.721 0.77 2.663 0.15 1.8 9.1 94.2 57.30 9.464 4.992 0.56 1.978 0.116 1.3 6.37 37.8 30.41 6.59 4.31 0.38 1.15 0.105 1.252 1.474 18.6 9.07 2.748 1.71 0.11 0.4095 0.0224 0.3065

The Kd(490)/Kd(PAR) regression analysis resulted in the equation that was substituted into equation 7 to compute Zeu on the MERIS images: =
0.5601 (490)+0.3029 4.605

(5)

MERIS Processing

Atmospheric Correction Figure 3 shows the measured and the C2R and Eutrophic Lakes, and BPAC derived remote sensing reflectances for four sites on three different days in the lake. It can be seen that the C2R estimated spectra successfully reproduced the in situ spectral shape throughout the spectrum in all the sampling stations on the various days. However, in the main lake, it overestimated the reflectance right across the spectrum, with the largest bias in the green. Eutrophic Lakes, on the other hand, gave more accurate results, although it failed to simulate the reflectance peak at 709 nm reflectance throughout the lake, as well as underestimating the reflectance in the red spectral region. Our results show that C2R gave an RSME of ~67% and MAE of 47%, whereas Kratzer and Vinterhav (2010) recorded an RSME of 17.4% and mean absolute bias of 25.6% at 490nm in their study at the Baltic Sea. Eutrophic Lakes gave the most accurate correction at 490nm with an RSME of 49% and MAE of 43%, whereas ICOL+C2R had an RSME of 102% and of 69%.but failed to simulate the 709nm peak. It could be attributed to the high concentration of Chl a in the water and that the processor is beyond its application range despite the fact that it was optimized for high eutrophic conditions (Doerffer & Schiller, 2008). ICOL-preprocessed C2R gave higher reflectances compared to the C2R correction alone, a result highlighted by Kallio (2012) and Zhu et al., (2012). BPAC on the other hand, failed completely in the blue region of the spectrum, as indicated below, giving negative reflectance values between 400 and around 530nm. This algorithm was developed for SPM-dominated Case 2 waters, whereas Lake Naivasha is phytoplankton dominant. Kratzer and Vinterhav (2010) also showed that the BPAC algorithm failed in coastal waters, especially without the ICOL-preprocessing. These results reveal the challenges of accurately performing atmospheric correction of inland waters.

Figure 3: Comparison of different atmospheric correction algorithms for the 3 MERIS matchup days to the ground measurements BPAC gave negative reflectances in the blue region of the spectrum, a result attributed to the fact that the algorithm was developed for high SPM concentration waters, whereas Lake Naivasha is Chl a dominant lake (Aiken & Moore, 2000, Lavender et al., 2005). Model Validation Table 2 shows the results of the Kd model calibration and validation. The 440-600 nm spectral range, which we are most interested in gave quite robust results 23% MAE and R2~0.98. Table 2: Statistical analysis of the band ratio algorithm developed to derive Kd() showing the calibration and validation results Calibration Validation Spectral RMSE R RMSE MAE (%) R Range 3.752 1.245 -0.16 0.275 0.974 2.805 23.436 0.976 440-600 1.92 0.737 1.079 0.115 0.77 0.82 32.031 0.797 600-800 1.846 0.803 0.01 0.078 0.925 0.262 17.698 0.98 800-930 To model Kd(490) from Rrs(), the parameters derived for the 400-600 nm spectral range were used: (490) = 3.753
(560) 1.245 (490)

Derivation of Kd(490) and Zeu from MERIS images using the developed models was successful as shown in Table 3. Table 3: Statistical analysis of Kd(490) and Zeu derived from MERIS matchup images

0.16

(65)

Kd(490) Zeu

RSME MAE (%) 0.426 31.336 0.236143 20.31099

Mapping Zeu Kd(490) and Zeu derived from MERIS FR Eutrophic Lakes processed data of 20, 23 and 26 September 2010, using Equations 15 and 14 respectively, are displayed in Fig 6. A spatiotemporal variation of Zeu is seen across the 3 days. This shows how dynamic the euphotic depth and other water quality properties are, and the effect the environmental conditions have on them. On 20 and 23 September, the northern part of the lake, where the river inlets are found, low Zeu (~0.8 m) was recorded. On 26 September, a cloudy day characterized by strong winds and clouds, the Zeu was very low and uniform throughout the lake at around 0.8 m, as indicated on the map. Environmental conditions, i.e. strong winds and rains that were experienced on the previous day caused mixing of waters across the lake, resulting in the low, uniform Zeu. The water on this part of the lake is clear green and the Secchi depth was almost double of that of the main lake. The extreme depth of up to 1.5 m could be a result of floating vegetation that resulted in high reflectance.

Figure 6: Maps showing distribution of Zeu in the lake on three days, 20, 23 and 26 September 2010 Research on Zeu undertaken on other shallow lakes includes a study by Zhang et al., (2006), which reveals that the Zeu of Lake Taihu, which is defined as a typical shallow lake, ranges from 1.04 to 1.95m. These concur with the results reported by Phlips et al., (1995) in Okeechobee Lake, a large shallow lake, in Florida, United States of America. CONCLUSION Although earth observation techniques have been used for research on the lake, they have not been employed to monitor its water quality. The purpose of this study was, thus, to investigate the possibility of using earth observation techniques to map the euphotic zone depth of Lake Naivasha. The results show that MERIS FR data successfully mapped the spatiotemporal variability of Kd(490) and Zeu in the lake. The diffuse attenuation coefficient and euphotic depth models both accurately retrieved the parameters from MERIS imagery. They illustrate that MERIS FR can be used to effectively map the optical water parameters in shallow eutrophic inland waters like Lake Naivasha. They further show the opportunity of using satellite remote sensing data like MERIS and other sensors to continually monitor water quality of highly variable inland waters. The atmospheric correction results also indicate the challenges that are faced when it comes to accurately performing atmospheric correction in optically complex inland waters. Also, a gap is still realized in African waters when it comes to using remote sensing techniques to monitor environmental conditions. With this in mind, strong in situ networks need to be established to develop more robust algorithms suitable for African waters. Acknowledgements This study was part of the Faculty of Geo-Information Science and Earth Observation (ITC), University of Twentes research project at Lake Naivasha. My colleagues from ITC, Semhar Gebrehiwot, Girma Adera Kebede, Mussie Ghimai Habte and Bashana Wondimu Daksa, for their participation in the field campaign and data analysis. The Council for

Scientific and Industrial Research, Natural Resources and Environment Earth Observation team for funding time and their useful comments and suggestions when preparing this manuscript. The European Space Agency for providing MERIS images. Sarah Higgins, for allowing us to use her premises to access the lake, and various local stakeholders for logistical help and information on the lake. References Aiken, J., & Moore, G. (2000). ATBD Case 2 (S) Bright Pixel Atmospheric Correction. , 2, 1-14 Aminot, A. (2000). Standard procedure for the determination of chlorophyll a by spectroscopic methods. Austin, R.W. (1974). Inherent spectral radiance signatures of the ocean surface. Ocean color analysis, 7410, 1-20 Austin, R.W., & Petzold, T.J. (1981). The determination of the diffuse attenuation coefficient of sea water using the Coastal Zone Color Scanner. Oceanography from Space, 239-256 Ayenew, T., & Becht, R. (2008). Comparative assessment of the water balance and hydrology of selected Ethiopian and Kenyan Rift Lakes. Lakes & Reservoirs: Research & Management, 13, 181-196 Ballot, A., Kotut, K., Novelo, E., & Krienitz, L. (2009). Changes of phytoplankton communities in Lakes Naivasha and Oloiden, examples of degradation and salinization of lakes in the Kenyan Rift Valley. Hydrobiologia, 632, 359-363 Barnard, A.H., Zaneveld, J.R.V., & Pegau, W.S. (1999). Determination of the Remotely Sensed Reflectance and the Absorption Coefficient: Closure and Inversion. Applied Optics, 38, 5108-5117 Becht, R., Odada, E.O., & Higgins, S. (2006a). Lake Naivasha. Experience and lessons learned brief Becht, R., Mwango, F., & Muno, F.A. (2006b). Groundwater links between Kenyan Rift Valley lakes. Becht, R. & Chesterton, M. (2010). Flower farming from a hydrologist's perspective : are Marnie's roses, and the other millions of romantic bouquets, really responsible for draining the water of Lake Naivasha and polluting it with pesticides and fertilizer? Bergner, A., Trauth, M., & Bookhagen, B. (2003). Paleoprecipitation estimates for the Lake Naivasha basin (Kenya) during the last 175 ky using a lake-balance model. Global and Planetary Change, 36, 117-136 Buffan-Dubau, E., & Carman, K.R. (2000). Extraction of benthic microalgal pigments for HPLC analyses. Marine Ecology Progress Series, 204, 293-297 Cartaxana, P., & Brotas, V. (2003). Effects of extraction on HPLC quantification of major pigments from benthic microalgae. Hydrobiologie, 157, 339-349 Cui, T., Zhang, J., Groom, S., Sun, L., Smyth, T., & Sathyendranath, S. (2010). Validation of MERIS ocean-color products in the Bohai Sea: A case study for turbid coastal waters. Remote Sensing of Environment, 114, 2326-2336 Dekker, A., Vos, R., & Peters, S. (2002). Analytical algorithms for lake water TSM estimation for retrospective analyses of TM and SPOT sensor data. International Journal of Remote Sensing, 23, 15-35 Doerffer, R., & Schiller, H. (2008). MERIS regional coastal and lake case 2 water project atmospheric correction ATBD. Inst.Coastal Res., GKSS Res.Center, Geesthacht, Germany, Rep.GKSS-KOF-MERIS-ATBD01, 1 Fougnie, B., Frouin, R., Lecomte, P., & Deschamps, P.Y. (1999). Reduction of skylight reflection effects in the above-water measurement of diffuse marine reflectance. Applied Optics, 38, 3844-3856 Gitelson, A.A., Schalles, J.F., & Hladik, C.M. (2007). Remote chlorophyll-a retrieval in turbid, productive estuaries: Chesapeake Bay case study. Remote Sensing of Environment, 109, 464-472 Gordon, H.R. (1989). Dependence of the diffuse reflectance of natural waters on the sun angle. Limnology and Oceanography, 1484-1489 Haande, S., Rohrlack, T., Semyalo, R.P., Brettum, P., Edvardsen, B., Lyche-Solheim, A., et al. (2011). Phytoplankton dynamics and cyanobacterial dominance in Murchison Bay of

Lake Victoria (Uganda) in relation to environmental conditions. Limnologica-Ecology and Management of Inland Waters, 41, 20-29 Harper, D., & Mavuti, K. (2004). Lake Naivasha, Kenya: Ecohydrology to guide the management of a tropical protected area. International Journal of Ecohydrology & Hydrobiology, 4, 287-305 Harper, D.M., Phillips, G., Chilvers, A., Kitaka, N., & Mavuti, K. (1994). Eutrophication prognosis for Lake Naivasha, Kenya. Proceedings: International Association of Theoretical and Applied Limnology, 25 Holm-Hansen, O., & Riemann, B. (1978). Chlorophyll a Determination: Improvements in Methodology. Oikos, 30, pp. 438-447 Hubble, D.S., & Harper, D.M. (2002). Phytoplankton community structure and succession in the water column of Lake Naivasha, Kenya: a shallow tropical lake. Hydrobiologia, 488, 89-98 Jensen, J.R. (2009). Remote sensing of the environment: An Earth Resource Perspective. India: Pearson Education Jin, X., Li, Y.M., Wang, Q., Liu, Z.H., Wang, Y.F., Zhang, H., et al. (2011). Primary productivity evaluation based on underwater light distribution in Chaohu Lake. Huan Jing Ke Xue, 32, 1257-1265 Kallio, K. (2012). Water quality estimation by optical remote sensing in Boreal Lakes. Monographs of the Boreal Environment Research Khanna, D.R., Bhutiani, R., & Chandra, K.S. (2009). Effect of the Euphotic Depth and Mixing Depth on Phytoplanktonic Growth Mechanism. International Journal of Environmental Research, 3, 223-228 Kirk, J.T.O. (1984). Dependence of relationship between inherent and apparent optical properties of water on solar altitude. Limnology and Oceanography, 29, 350-356 Kirk, J.T.O. (2011). Light and photosynthesis in aquatic ecosystems. New York, USA: Cambridge University Press Kitaka, N., Harper, D.M., Mavuti, K.M., & Pacini, N. (2002). Chemical characteristics, with particular reference to phosphorus, of the rivers draining into Lake Naivasha, Kenya. Hydrobiologia, 488, 57-71 Koehler, A. (2008). Water use in LCA: managing the planets freshwater resources. The International Journal of Life Cycle Assessment, 13, 451-455 Kona, E. & Mwiti, C. (2010). Mysterious disease kills fish in L. Naivasha. Kenya Broadcasting Cooperation Kratzer, S., & Vinterhav, C. (2010). Improvement of MERIS level 2 products in Baltic Sea coastal areas by applying the Improved Contrast between Ocean and Land processor (ICOL)-data analysis and validation. Oceanologia, 52, 211-236 Kratzer, S., Brockmann, C., & Moore, G. (2008). Using MERIS full resolution data to monitor coastal waters--A case study from Himmerfjrden, a fjord-like bay in the northwestern Baltic Sea. Remote Sensing of Environment, 112, 2284-2300 Kratzer, S., Hkansson, B., & Sahlin, C. (2003). Assessing Secchi and photic zone depth in the Baltic Sea from satellite data. AMBIO: A Journal of the Human Environment, 32, 577-585 Kundzewicz, Z.W., Mata, L.J., Arnell, N.W., Dll, P., Kabat, P., Jimnez, B., et al. (2007). Freshwater resources and their management: Climate Change 2007: Impacts, Adaptation and Vulnerability. Contribution of Working Group II to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. In M.L. Parry, O.F. Canziani, J.P. Palutikof, P.J. van der Linden, & C.E. Hanson (Eds.), (pp. 73-210). Cambridge, UK: Cambridge University Press Lavender, S.J., Pinkerton, M.H., Moore, G.F., Aiken, J., & Blondeau-Patissier, D. (2005). Modification to the atmospheric correction of SeaWiFS ocean colour images over turbid waters. Continental Shelf Research, 25, 539-555 Lee, Z.P., Darecki, M., Carder, K.L., Davis, C.O., Stramski, D., & Rhea, W.J. (2005). Diffuse attenuation coefficient of downwelling irradiance: An evaluation of remote sensing methods. Journal of Geophysical Research-Oceans, 110 Levrini, G., & Delvart, S. (2011). MERIS Product Handbook. Loisel, H., & Stramski, D. (2000). Estimation of the inherent optical properties of natural waters from the irradiance attenuation coefficient and reflectance in the presence of Raman scattering. Applied Optics, 39, 3001-3011

Marrari, M., Hu, C., & Daly, K. (2006). Validation of SeaWiFS chlorophyll a concentrations in the Southern Ocean: A revisit. Remote Sensing of Environment, 105, 367-375 Matthews, M.W., Bernard, S., & Winter, K. (2010). Remote sensing of cyanobacteriadominant algal blooms and water quality parameters in Zeekoevlei, a small hypertrophic lake, using MERIS. Remote Sensing of Environment, 114, 2070-2087 Mobley, C.D. (1999). Estimation of the remote-sensing reflectance from above-surface measurements. Applied Optics, 38, 7442-7455 Mobley, C.D. (1994). Light and Water: Radiative transfer in natural waters. : Academic Press San Diego, CA Morel, A., & Loisel, H. (1998). Apparent optical properties of oceanic water: dependence on the molecular scattering contribution. Applied Optics, 37, 4765-4776 Morel, A., & Berthon, J.F. (1989). Surface pigments, algal biomass profiles, and potential production of the euphotic layer: Relationships reinvestigated in view of remote-sensing applications. Limnology and Oceanography, 1545-1562 Mueller, J.L. (2000). SeaWiFS algorithm for the diffuse attenuation coefficient, K (490), using water-leaving radiances at 490 and 555 nm. , 2427 Mueller, J.L., & Fargion, G.S. (2003). Ocean Optics Protocols for Satellite Ocean Color Sensor Validation, Revision 4, Volume III: Radiometric Measurements and Data Analysis Protocols. Otian'a-Owiti, G. E. & Abiya Oswe, I. (2007). Human impact on lake ecosystems : the case of Lake Naivasha, Kenya. African journal of aquatic science, 32, 79-88 Ppista, ., Acs, E., & Bddi, B. (2002). Chlorophyll-a determination with ethanola critical test. Hydrobiologia, 485, 191-198 Phlips, J., Lynch, T. C. & Badylak, S. (1995). Chlorophyll a, tripton, color, and light availability in a shallow tropical inner-shelf la-goon, Florida Bay, USA. Marine Ecology Progress Series, 127, 223-223-234 Pimentel, D., Houser, J., Preiss, E., White, O., Fang, H., Mesnick, L., et al. (1997). Water Resources: Agriculture, the Environment, and Society. Bioscience, 47, 97-106 Preisendorfer, R.W. (1986). Secchi disk science: Visual optics of natural waters. Limnology and Oceanography, 909-926 Robarts, R.D., & Zohary, T. (1984). Microcystis Aeruginosa and Underwater Light Attenuation in a Hypertrophic Lake (Hartbeespoort Dam, South Africa). Journal of Ecology, 72, pp. 1001-1017 Sartory, D.P., & Grobbelaar, J.U. (1984). Extraction of chlorophyll a from freshwater phytoplankton for spectrophotometric analysis. Hydrobiologia, 114, 177-187 Scheffer, M. (2004). Ecology of shallow lakes. : Springer Schmugge, T.J., Kustas, W.P., Ritchie, J.C., Jackson, T.J., & Rango, A. (2002). Remote sensing in hydrology. Advances in Water Resources, 25, 1367-1385 Shiklomanov, I.A. (Ed.), (1998). World water resources. Paris: United Nations Educational, Scientific and Cultural Organization Siegel, D.A., & Dickey, T.D. (1987). Observations of the vertical structure of the diffuse attenuation coefficient spectrum. Deep Sea Research Part A.Oceanographic Research Papers, 34, 547-563 Smith, R.C., & Baker, K.S. (1981). Optical properties of the clearest natural waters (200 800 nm). Applied Optics, 20, 177-184 Tang, S., Chen, C., Zhan, H., Xu, D., & Liu, D. (2007). Retrieval of euphotic layer depth of South China Sea by remote sensing. Journal of Tropical Oceanography, 1 Vrsmarty, C.J., Green, P., Salisbury, J., & Lammers, R.B. (2000). Global water resources: vulnerability from climate change and population growth. Science, 289, 284 Wright, S., & Jeffrey, S. (1997). High-resolution HPLC system for chlorophylls and carotenoids of marine phytoplankton. Monographs on Oceanographic Methodology, 10 Zaneveld, J.R.V., Kitchen, J.C., & Mueller, J.L. (1993). Vertical structure of productivity and its vertical integration as derived from remotely sensed observations. Limnology and Oceanography, 1384-1393 Zhang, Y., Qin, B., Hu, W., Wang, S., Chen, Y., & Chen, W. (2006). Temporal-spatial variations of euphotic depth of typical lake regions in Lake Taihu and its ecological environmental significance. Science in China Series D: Earth Sciences, 49, 431-442 Zhu, J., Chen, J., Matsushita, B., Yang, W., & Fukushima, T. (2012). Atmospheric correction of ENVISAT/MERIS data over case II waters: the use of black pixel

assumption in oxygen and water vapour absorption bands. International Journal of Remote Sensing, 33, 3713-3732