Elsevier

Remote Sensing of Environment

Volume 135, August 2013, Pages 92-106
Remote Sensing of Environment

Remote sensing of chlorophyll-a as a measure of cyanobacterial biomass in Lake Bogoria, a hypertrophic, saline–alkaline, flamingo lake, using Landsat ETM +

https://doi.org/10.1016/j.rse.2013.03.024Get rights and content

Highlights

  • A Landsat chlorophyll-a retrieval algorithm was developed for Lake Bogoria.

  • Dense blooms of cyanobacteria in the lake are a food source for Lesser Flamingos.

  • The NIR band of Landsat is well represented by a linear relationship with Chl-a.

  • The TOA Landsat NIR:Red band ratio was identified as the best Chl-a algorithm.

  • The algorithm will allow quantification of the Lesser Flamingo's food supply.

Abstract

Lake Bogoria is a saline–alkaline lake in the Kenyan Rift Valley, known for supporting dense blooms of cyanobacteria and large flocks of up to 1 million Lesser Flamingos (Phoeniconaias minor). An algorithm for the remote sensing of chlorophyll-a (Chl-a), as an indicator of cyanobacterial biomass, has been developed using a time series of Landsat images and in situ measurements. In situ measured reflectance spectra were resampled to Landsat bands, and the near infrared (NIR) band, R835, was found to be well represented by a linear relationship to Chl-a (R2 = 0.847; Standard error, SE = 55 μg l 1; Samples, N = 14) for concentrations up to 800 μg l 1. The band ratio R835/R660 also showed a strong linear relationship with Chl-a (R2 = 0.811; SE = 61 μg l 1, N = 14). Similar relationships were derived using Landsat satellite imagery and monthly in situ Chl-a data for the period Nov 2003–Feb 2005. The NIR:Red ratio gave a better fit to Chl-a than a single NIR band algorithm when applied to satellite imagery, and the ratio performed best when based on TOA reflectance rather than atmospherically corrected data. Hence an algorithm for Chl-a was derived based on the TOA Landsat reflectance ratio, R835/R660, which showed a strong fit against Chl-a (R2 = 0.801; SE = 69 μg l 1, N = 33) despite the limitations of time coincidence for the available satellite-in situ matches (less than 8 days). Lesser Flamingos feed on cyanobacteria in saline–alkaline lakes, therefore the algorithm can be used to monitor changes in their food supply, providing valuable information for their future conservation. The study also allowed characterisation of the optical properties in Lake Bogoria and provides insight into the changes occurring during cyanobacterial bloom and die-off events.

Introduction

This paper investigates remote sensing as a tool for the ecological research of inland soda (alkaline–saline) lakes. In particular, it aims to develop algorithms for monitoring primary producers in these unique ecosystems, essential to the life cycle of Lesser Flamingos (Phoeniconaias minor). Landsat ETM + data will be used to establish whether moderate spatial resolution satellite imagery can be used for quantifying chlorophyll-a (Chl-a) concentrations in alkaline–saline lakes, as an indicator for cyanobacterial biomass — the primary food source for Lesser Flamingos. The remote sensing techniques developed here will therefore enable the long term stability and distribution of food available to Lesser Flamingos to be determined, thus providing valuable information for their future conservation.

Alkaline–saline lakes have a distinct ecology characterised by dense blooms of cyanobacteria. Such blooms are often considered a hazard in freshwater lakes due to the toxins they may produce (WHO, 1999), but in alkaline–saline lakes the colonial cyanobacterium, Arthrospira fusiformis, plays a vital role in sustaining a population of Lesser Flamingos, which feed by filtering cyanobacteria from the water of a dozen or so soda lakes in the Rift Valley (Childress et al., 2007, Tuite, 2000). Lesser Flamingos are classified as a near-threatened species (IUCN, 2012) due to their decreasing numbers and limited breeding sites. These flamingos are of great economic importance as they attract tourists to the soda lakes of Kenya and Tanzania where they form a globally renowned spectacle. Their numbers can reach one million individuals at a single lake.

Lesser Flamingos are nomadic birds which move from lake to lake in response to food availability and other environmental factors (Kaggwa et al., 2012). Little is known about the spatial and temporal distribution of their food supply. Occasionally a drastic reduction in cyanobacterial biomass — known as a ‘crash’ or ‘die-off event’ — is observed in the lakes (Harper et al., 2003, Oduor and Schagerl, 2007b), resulting in poor feeding conditions for the flamingos. The causes of these events are poorly understood, partly because the remoteness results in a lack of observations. Satellite observations could give insight into how often these crashes occur and how long they last.

The remote sensing of lakes can be particularly problematic since each lake is different in terms of its scale, ecology, water chemistry and surrounding land use; thus each lake presents individual challenges. The main lake considered in this study is Lake Bogoria in Kenya. This site was selected because it is one of the main feeding sites for Lesser Flamingos (Childress et al., 2007) and also due to the availability of in situ Chl-a measurements, which could be used to develop a semi-empirical algorithm for satellite data. Lake Bogoria, like other alkaline–saline lakes, is poorly studied; its remote location makes field observations extremely difficult. Therefore the ability to monitor these environments with satellite data is urgently needed.

It is known that soda lakes support extremely high cyanobacterial biomass (mean Chl-a: L. Bogoria 388 μg l 1, L. Nakuru 646 μg l 1, L. Elementaita 267 μg l 1; Oduor & Schagerl, 2007b) so cyanobacteria are likely to dominate their optical properties. However, the relative concentrations and contribution to reflectance, of cyanobacteria, sediment and coloured dissolved organic matter (CDOM) are unknown. Hence, the spectral signatures and simultaneous water parameter measurements collected in this study have provided vital information.

Remote sensing is used for the detection of harmful cyanobacterial blooms caused by natural or man-made eutrophication (Kutser, 2004, Kutser et al., 2006, Matthews et al., 2010) and for monitoring water quality parameters, including Chl-a, CDOM, total suspended solids (TSS) and Secchi disc depth (Brezonik et al., 2005, Matthews et al., 2010, Mayo et al., 1995). In oligotrophic, Case I waters, most Chl-a algorithms are based on simple blue to green band ratios but for Case II waters, simple band ratios are often not suitable for recovering non-covarying water parameters (Tyler et al., 2006). However, in coastal and inland waters with extremely high biomass concentrations, phytoplankton dominate the optical properties (Kirk, 1994, Kutser, 2009), so simple band ratio algorithms can be utilised. For these more productive waters, algorithms based on bands in the green to near infrared (NIR) region are more effective for retrieving Chl-a concentration (Gitelson, 1992, Gower et al., 2005, Kutser, 2009). Studies of hypertrophic waters have found water-leaving reflectance in the NIR useful for quantitative monitoring of cyanobacteria (Kutser, 2004, Reinart and Kutser, 2006, Wynne et al., 2008), mapping cyanobacterial surface scums (Hu, 2009, Matthews et al., 2012) and for identifying areas of floating vegetation (Fraser, 1998). The pigment phycocyanin has been used in some studies to distinguish cyanobacterial biomass from that of other forms of phytoplankton (Ruiz-Verdú et al., 2008, Simis, Peters and Gons, 2005, Simis et al., 2007). However, phycocyanin is a difficult pigment to measure and there is no standard measurement technique. Furthermore, since Lake Bogoria is always dominated by cyanobacteria (Harper et al., 2003) there would be no advantage in using phycocyanin as the detection pigment in this study and it is not considered here.

A number of satellite sensors have been used for quantitative monitoring of cyanobacterial blooms, e.g. MODIS, MERIS, SeaWiFS and AVHRR (Kutser, 2009). In this study we require high spatial resolution data due to the small size of the lakes (1–3 km across) and a long timeseries of data. Therefore the Landsat Enhanced Thematic Mapper (ETM +) was selected due to its high spatial resolution (30 m), allowing for detection of small scale spatial variability across the lake surface, and long archive of data, from ETM + (launched 1999) and past sensors (MSS, TM), which will be continued by the Landsat Data Continuity Mission (LDCM). Landsat imagery has been used for the remote sensing of water parameters and cyanobacteria in lakes (Brezonik et al., 2005, Galat et al., 1990, Hamed et al., 2007, Vincent et al., 2004, Yacobi et al., 1995), with empirical relationships typically being derived from combinations of Landsat bands. Its revisit time (16 days) and broad spectral bands are not ideal for monitoring bloom dynamics (Galat and Verdin, 1989, Strong, 1974), but for our purposes these constraints are considered secondary to the need for high spatial resolution data.

Section snippets

Lake Bogoria

Lake Bogoria (0°15′N, 36°06′E) is a highly saline (conductivity approximately 80 mS cm 2) and alkaline (pH 10), endorheic lake in the Eastern Rift Valley, Kenya (Oduor & Schagerl, 2007b). The lake is 16 km long and 1–4 km wide, consisting of 3 basins separated by two necks of land. Rivers enter the lake at the northern and southern ends as well as the west side of the central basin. Significant water inflow also comes from numerous hot-springs along the western and south-eastern shores. The lake is

Method

Two approaches were adopted in this study: i) a field spectroscopy study was carried out at Lake Bogoria in 2010 to improve understanding of the optical properties of soda lakes and to investigate the relationship between water constituents, particularly Chl-a, and the observed reflectance spectra and ii) a timeseries of in situ Chl-a data from an existing study (Oduor & Schagerl, 2007b) was utilised in combination with coincident Landsat ETM + imagery to develop an empirical algorithm for Chl-a

Limnological conditions

During the field study (19th–27th April 2010) the lake appeared a vivid green colour due to high cyanobacterial biomass and mats of cyanobacterial scum. A sediment plume, originating from the Waseges River, was observed which increased in size until it covered most of the north basin. Here the lake appeared milky brown due to the high sediment concentration, however this was only a surface layer — in the wake of the boat vivid green water was visible, showing that the lake was highly

Discussion

Field measurements confirmed that the optical properties of Lake Bogoria are dominated by high cyanobacterial biomass, particularly at NIR wavelengths. However, the high CDOM concentrations will contribute significantly to the absorption at shorter wavelengths. Although the highest CDOM concentration was observed in the sediment plume, this input is infrequent and the majority of CDOM in Lake Bogoria will originate from the cyanobacteria. Hence, Lake Bogoria could be considered a Case I water

Conclusion

This work has shown that satellites can provide ecologically valuable information for the study of alkaline–saline lakes. It has illustrated that there is still a role for simple empirical algorithms, particularly for small water bodies where data is limited to multispectral sensors. Field spectroscopy identified the clear signal of high biomass cyanobacterial blooms as elevated reflectance beyond 700 nm. In situ reflectance spectra, re-sampled to Landsat ETM + bands, showed a strong linear

Acknowledgements

The authors would like to thank the following people and organisations: The NERC Field Spectroscopy Facility for providing equipment and training; The Centre for Interdisciplinary Science, University of Leicester; S. O. Oduor and M. Schagerl for providing in situ chlorophyll data; Steve Ison and the rest of the Biology support staff; Camp staff including camp manager Velia Carn, cooks Willy and Patrick and drivers James Njoroge, Reuben Ndolo and John Kaba; Roland Leigh and Rosie Graves for

References (58)

  • A. Strong

    Remote sensing of algal blooms by aircraft and satellite in Lake Erie and Utah Lake

    Remote Sensing of Environment

    (1974)
  • R.K. Vincent et al.

    Phycocyanin detection from LANDSAT TM data for mapping cyanobacterial blooms in Lake Erie

    Remote Sensing of Environment

    (2004)
  • Y. Zhang et al.

    The contribution of phytoplankton degradation to chromophoric dissolved organic matter (CDOM) in eutrophic shallow lakes: Field and experimental evidence

    Water Research

    (2009)
  • A. Abeliovich et al.

    Photooxidative death in blue green algae

    Journal of Bacteriology

    (1972)
  • P. Brezonik et al.

    Landsat-based remote sensing of lake water quality characteristics, including chlorophyll and colored dissolved organic matter (CDOM)

    Lake and Reservoir Management

    (2005)
  • A. Bricaud et al.

    Absorption by dissolved organic matter in the sea (yellow substance) in the UV and visible domains

    Limnology and Oceanography

    (1981)
  • B. Childress et al.

    East african flyway and key site network of the Lesser Flamingo (Phoenicopterus minor) documented through satellite tracking

    Ostrich: Journal of African Ornithology

    (2007)
  • A. Eaton et al.

    Standard methods for the examination of water and wastewater

    (2005)
  • R.N. Fraser

    Hyperspectral remote sensing of turbidity and chlorophyll a among Nebraska Sand Hills lakes

    International Journal of Remote Sensing

    (1998)
  • D. Galat et al.

    Patchiness, collapse and succession of a cyanobacterial bloom evaluated by synoptic sampling and remote sensing

    Journal of Plankton Research

    (1989)
  • D. Galat et al.

    Large-scale patterns of Nodularia spumigena blooms in Pyramid Lake, Nevada, determined from Landsat imagery: 1972–1986

    Hydrobiologia

    (1990)
  • A. Gitelson

    The peak near 700 nm on radiance spectra of algae and water: Relationships of its magnitude and position with chlorophyll concentration

    International Journal of Remote Sensing

    (1992)
  • A.A. Gitelson et al.

    Optical properties of dense algal cultures outdoors and their application to remote sensing of biomass and pigment concentration in Spirulina Platensis (Cyanobacteria)

    Journal of Phycology

    (1995)
  • J. Gower et al.

    Detection of intense plankton blooms using the 709 nm band of the MERIS imaging spectrometer

    International Journal of Remote Sensing

    (2005)
  • G. Groom et al.

    Using object-based analysis of image data to count birds: Mapping of Lesser Flamingos at Kamfers Dam, Northern Cape, South Africa

    International Journal of Remote Sensing

    (2011)
  • A.F. Hamed et al.

    Floristic survey of blue-green algae/cyanobacteria in saline–alkaline lakes of Wadi El-Natrun (Egypt) by remote sensing application

    Journal of Applied Sciences Research

    (2007)
  • D.M. Harper et al.

    Aquatic biodiversity in saline lakes: Lake Bogoria National Reserve, Kenya

    Hydrobiologia

    (2003)
  • P. Hunter et al.

    The spatial dynamics of vertical migration by Microcystis aeruginosa in a eutrophic shallow lake: A case study using high spatial resolution time-series airborne remote sensing

    Limnology and Oceanography

    (2008)
  • IOCCG

    Remote sensing of ocean colour in coastal, and other optically-complex waters

  • Cited by (140)

    View all citing articles on Scopus
    View full text