, 2012 
ange in Mo- 
pp. 42-64. 
hniques for 
lied Optics, 
lectance and 
aircraft and 
2 
ig combined 
. J. Remote 
Multispectral 
rithm. [EEE 
2251-2259. 
metrics, 15, 
lquist, 2004. 
ectral data. 
ession with 
Shallow sea 
ech. Chron. 
summary) 
ipping from 
alized ratio 
nth passive 
-1578. 
lence on the 
vol.8, no.3, 
r the Social 
etermination 
nagery over 
7-556. 
erivation for 
ite imagery 
esy, 31, pp. 
eanography, 
Engineering, 
, Greece (in 
ric use of 
e®, Inc. 
poral water 
ote Sensing, 
REMOTE SENSING OF WATER QUALITY IN OPTICALLY COMPLEX LAKES 
T. Kutser *°*, B. Paavel*, C. Verpoorter 5 T. Kauer?, E. Vahtmàáe* 
? Estonian Marine Institute, University of Tartu, Máealuse 14, 12618, Tallinn, Estonia — Tiit.Kutser(a)sea.ee 
b Limnology / Department of Ecology and Genetics, University of Uppsala, Norbyvágen 18D, Uppsala, 75236, Sweden 
Commission VIII/4 
KEY WORDS: Lakes, remote sensing, water quality, aquatic optics, chlorophyll-a, CDOM, suspended matter 
ABSTRACT: 
Solving of several global and regional problems requires adequate data about lake water quality parameters like the amount and type 
of phytoplankton dominating in the lakes, the amount of dissolved and coloured dissolved organic matter and/or concentration of 
suspended sediment. Remote sensing is the only practical way to study many lakes provided it can produce sufficiently accurate 
estimates of the water characteristics. We studied optically very variable lakes in order to test both physics based methods and 
conventional band-ratio type algorithms in retrieval of water parameters. The modelled spectral library used in the physics based 
approach provided very good results for chlorophyll-a retrieval. The number of different concentrations of CDOM and suspended 
matter used in the simulations was too low to provide good estimates of these parameters. Extending the spectral library is currently 
in progress. Band-ratio type algorithms worked well in chlorophyll-a and CDOM retrieval. None of the algorithms tested for total 
suspended matter, organic suspended matter and inorganic suspended matter retrieval performed well enough and there is need in 
further testing. 
1. INTRODUCTION 
Monitoring of lake water quality in regional to global scales is 
important from several aspects. For example recent studies 
(Cole et al., 2007; Battin et al., 2009; Tranvik et al., 2009) 
indicate that lakes play very important role in the global carbon 
cycle. These estimates are mainly based on upscaling regional 
in situ data to global scale. The true role of lakes in the global 
carbon cycle can be determined only by remote sensing as it is 
not possible to study sufficient amount of lakes by means of in 
situ sampling. However, reliable remote sensing algorithms for 
retrieval of parameters like coloured dissolved organic matter, 
CDOM, will be needed as CDOM can be used as a proxy to 
estimate also DOC (dissolved organic carbon) concentrations 
and CO, saturation in lakes in global scale (Kutser et al. 
2005a,b, 2009a, Sobek et al. 2003). Monitoring of lake DOC is 
important also from drinking water quality point of view as 
chlorination produces cancerogenic compounds and increasing 
DOC in lakes used as a drinking water source impacts human 
health (McDonalnd and Komulainen 2005). 
Frequency and extent of intense phytoplankton blooms has 
increased in inland and coastal waters around the world 
(Hallegraeff 2003, Sellner et al. 2003, Glibert et al. 2005) 
Potentially harmful effects of the blooms (Edler et al. 1985, 
Horner et al. 1997, Landsberg 2002, Hallegraeff 2003, Sellner 
et al. 2003, Glibert et al. 2005, Backer and McGillicuddy 2006) 
on human and animal health, drinking water quality and 
recreational use of water bodies have caused awareness of 
general public, environmental agencies and water authorities. 
Conventional monitoring networks, that are based on infrequent 
sampling in a few fixed monitoring stations, cannot provide the 
information needed as the blooms are very heterogeneous both 
spatially and temporally. Only remote sensing can provide the 
spatial and temporal coverage needed. Many authors (see 
  
* Corresponding author. 
references in review papers by Kutser 2009b and Matthews 
2011) have worked on recognition and quantitative mapping of 
potentially harmful blooms. However, the quantitative mapping 
problem is still unsolved as there is no sufficient information on 
optical properties of different blooms and vertical structure of 
biomass in the blooms (for these species that can migrate in the 
water column). 
Physics based methods using modelled spectral libraries (look- 
up tables) in interpretation of remote sensing data have gained 
popularity in both mapping of water depth and shallow water 
benthic habitats (Kutser et al. 2002, 2006; Mobley et al. 2005; 
Lesser and Mobley 2007; Brando et al. 2009) but also in 
retrieving optically active constituents of the water column 
(Kutser et al. 2001, Brando et al. 2009) or estimating 
phytoplankton biomass in heavy blooms (Kutser 2004). We 
used Hydrolight radiative transfer model to simulate remote 
sensing reflectance spectra and then used procedure called 
Spectral Angle Mapper, SAM, to find the modelled spectrum 
most similar to the measured one. It was assumed that the 
concentrations of chlorophyll-a, CDOM and mineral particles in 
the water match the concentrations used in modelling if the 
measured and modelled reflectance spectra match. This would 
allow to retrieve the three concentrations simultaneously. 
Suitability of several band-ratio type algorithms for retrieving 
the optically active water constituents were also tested. 
2. STUDY SITES AND METHODS 
2.1 Study sites 
The selected lakes are variable in their optical water properties 
and large enough for satellite remote sensing with sensors like 
MERIS with it's 300 m spatial resolution. Lake Peipsi in 
Estonia is the fourth largest lake in Europe (3555 km”). It's