Retrieval of water quality from airborne imaging spectrometry of various lake types in different seasons

Abstract

The suitability of the AISA airborne imaging spectrometer for monitoring lake water quality was tested in four surveys carried out in southern Finland in 1996–1998. Altogether, 11 lakes were surveyed and the total number of stations with concurrent remote sensing and limnological measurements was 127. The ranges of the water quality variables were: the sum of chlorophyll a and phaeophytin a 1–100 μg l⁻¹, turbidity 0.4–26 FNU, total suspended solids 0.7–32 mg l⁻¹, absorption coefficient of aquatic humus at 400 nm 1.2–14 m⁻¹ and secchi disc transparency 0.4–7 m. For the retrieval analyses, 24 AISA channels in the 450–786 nm range with a channel width of 6–14 nm were used. The agreement between estimated and observed water quality variables was generally good and R² for the best algorithms was in the range of 0.72–0.90 over the whole dataset. The channels used for May were, in most cases, the same as those for August, but the empirical parameters of the algorithms were different. After seasonal grouping, R² varied from 0.84 to 0.95. The use of apparent reflectance instead of radiance improved the estimation of water quality in the case of total suspended solids and turbidity. In the most humic lake, the empirical algorithms tested were suitable only for the interpretation of total suspended solids and turbidity.

Introduction

Lakes are optically complex waters, as the concentrations of optically-active substances can vary considerably between lakes and independently of each other. Moreover, the variation in specific inherent optical properties, e.g. due to different origins of substances, makes it even more difficult to monitor lakes by remote sensing. To develop specific interpretation methods for remote sensing which are not specific to region or lake type, the testing data should cover a wide range of lake regions.

Hyperspectral radiances or reflectances have been measured from lakes for water quality retrieval using airborne spectrometers (Dekker et al., 1992, Dekker, 1993; Mille et al., 1992; Dekker, 1993, Hamilton et al., 1993, Jupp et al., 1993, Fraser, 1998), above-water spectrometers aboard vessels (Vertucci and Likens, 1989, Gitelson et al., 1993, Jupp et al., 1993, Yacobi et al., 1995, Kutser et al., 1998a, Schalles et al., 1998b) and underwater spectrometers (Dekker, 1993, Kutser, 1997, Schalles et al., 1998a). The interpretation of water quality in lakes has mainly been based on the use of empirical algorithms. The most common chlorophyll a algorithm has been the reflectance ratio between two channels in the region 670–720 nm. The algorithms for total suspended solids (TSS), turbidity (TURB) and secchi disk transparency (Z_SD) have varied between studies. Another interpretative approach is analytical modelling, in which reflectance spectra are simulated by using the specific inherent optical properties of substances in water. Such models, also referred to as bio-optical models, can be used for sensitivity analyses of the effect of optically active substances on reflectance and to determine the best retrieval algorithms for each case (Dekker, 1993, Hoogenboom et al., 1998a, Kondratyev et al., 1998, Pozdnyakov et al., 1998). Bio-optical models have also been used in the retrieval of water quality variables in lakes from their reflectance spectra using inverse modelling (Kutser, 1997, Kutser et al., 2000) and the matrix inversion technique (Hoogenboom et al., 1998b).

In Finland, monitoring of lakes is a huge task since they account for approximately 10% of the total surface and the number of lakes larger than 0.01 km² is 56 012 (Raatikainen and Kuusisto, 1988). Even the large lakes are usually characterized by the presence of several subbasins and islands, resulting in spatial differences in water quality. Monitoring of lakes in Finland by traditional methods based on water sampling at a few fixed stations cannot produce information on all the lakes or the spatial differences within a lake. Because of the wide spatial coverage, remote sensing is a technique with potential for improving the effectiveness of lake monitoring. One of the main water quality problems in Finnish lakes is eutrophication, which is measured in terms of chlorophyll a concentration in routine monitoring programmes and can therefore be detected by remote sensing.

The main objective of this study was to determine the best empirical algorithms for the retrieval of water quality from airborne spectrometer data in typical Finnish lakes. The algorithms tested are taken from the literature or specially developed for our own data. The main problems with similar studies published in the literature have been the small number of in situ samples and that they have focused on one season or on one lake only. Therefore, special attention here has been directed at finding algorithms that are not specific to lake type and to studying the limitations imposed by differences in water quality and optical properties related to season or lake type.