Improving the spectral unmixing algorithm to map water turbidity Distributions

Abstract

In this paper we evaluate the suitability of the spectral unmixing algorithm to map the turbidity in the Curuai floodplain lake and enhance its applicability using autocorrelation modelling. The Spectral Unmixing Model (SMM) was applied to a Moderate Resolution Imaging Spectroradiometer (MODIS) surface reflectance (MOD09) image, taking in-situ measurements close to the acquisition date. Fraction images of inorganic matter-laden water, dissolved organic matter-laden water, and phytoplankton-laden water were generated by SMM, using 4 MODIS spectral bands (blue, green, red, and near infrared). These endmembers were selected based on the dominance of these components, which affect water turbidity. These fraction images allowed assessing the turbidity distribution in the study area but showing only places with high or low turbidity. The kernel estimation algorithm was then used to verify the spatial correlation among the in-situ measurement data. The occurrence of clusters suggests that there are different spatial water regimes. One spatial regression model was then compiled for each water regime, each of which presented a better turbidity estimation as opposed to the one derived from the Ordinary Least Square (OLS). The methodology applied was hence useful to analyze the spatial distribution of turbidity in the Curuai floodplain lake.

Introduction

Water quality modelling is the linkage between the sources of pollution and the in-stream water quality of a given water body (Freni et al., 2009). Different models have already been applied to a variety of studies including the assessment of relocation impact (Dias et al., 2009), trans-boundary conflicts (Zhao, 2009). Many of those models, however, use in-situ water quality data as input (Wu et al., 2009).

Conventional water quality monitoring is expensive and time consuming. This is particularly problematic if the water bodies to be examined are large. Conventional techniques also bring about a high probability of undersampling (Hadjimitsis et al., 2006). Conversely, remote sensing is a powerful tool to assess aquatic systems and is particularly useful in remote areas such as the Amazon lakes (Alcântara et al., 2008).

The remote sensing technique has been used not only for water surface modelling but also for modelling land surface temperature (Sheng et al., 2009), annual variability of nitrate concentrations (Montzka et al., 2008), spatially distributed sediment budgets (Wilkinson et al., 2009), crop yield (Pan et al., 2009).

Data collected using this technique can provide a synoptic overview of such large aquatic environments, which could otherwise not be observed at a glance (Dekker et al., 1995). However, remote sensing is not easily applied to aquatic environment monitoring mainly because of concerns related to the mixture of the optically active substances (OAS) in the water. Several approaches have been proposed to cope with this issue such as derivative analysis (Goodin et al., 1993), the continuum removal (Kruse et al., 1993), and spectral mixture analysis (Novo and Shimabukuro, 1994, Oyama et al., 2009).

The two first approaches are more suitable for hyperspectral images, whereas the spectral mixture analysis can be used for both hyperspectral and multispectral images. The SMM has largely been used for spectral mixture analysis, uncoupling the reflectance of each image pixel (Tyler et al., 2006) into the proportion of each water component contributing to the signal. The result of a spectral mixture analysis is a set of fraction images representing the proportion of each water component per image pixel. This technique has been applied to TM/Landsat images to determine the concentration of suspended particles (Mertes et al., 1993), chlorophyll-a concentration (Novo and Shimabukuro, 1994); as well as to MODIS images, to determine the chlorophyll-a concentration in the Amazon floodplain (Novo et al., 2006), to characterize the composition of optically complex waters in the Amazon (Rudorff et al., 2007), and to study turbidity distribution in the Amazon floodplain (Alcântara et al., 2008).

Remote sensing data has been extensively used to detect and to quantify water quality variables in lakes and reservoirs (Kloiber et al., 2002). One of the most important variables to monitor water quality is turbidity, because it reveals information about the availability of light, as determined by the concentration of inorganic and organic (alive and dead) components in the water column (Goodin et al., 1993). Although turbidity is brought about by organic and inorganic particles, one unresolved issue is to distinguish between them using remote sensing (Wetzel, 2001). The Spectral Unmixing Model (SMM) can, however, be useful to analyze the turbidity caused by inorganic particles and by phytoplankton cell scattering.

A previous study used the SMM to decompose the optical water components and map the turbidity distribution in the Curuai floodplain at high water level (Alcântara et al., 2008). However, it is uncertain if this approach will work when applied to rising water conditions.

We believe that changes in turbidity can be detected in remote sensing images because they bring about changes in the upward irradiance leaving the water surface. This paper describes the experiment carried out to test this hypothesis and to assess the Amazon floodplain lake turbidity using MODIS imagery.