Inland Water Quality Monitoring Using Remote Sensing: an Australian Perspective
Malthus, Tim1; Hestir, Erin1; Dekker, Arnold1; Anstee, Janet1; Botha, Elizabeth1; Cherukuru, Nagur1; Brando, Vittorio1; Clementson, Lesley2; Oliver, Rod1; Lorenz, Zygmunt1
1CSIRO Land and Water, AUSTRALIA; 2CSIRO Division of Marine and Atmospheric Research, AUSTRALIA
Australia’s inland water quality is ranked among the worst of the developed countries and it is getting poorer (Dekker and Hestir 2012, Emerson et al. 2012). Similar to the challenges facing many countries, existing data are scarce and declining, have poor geographic and temporal coverage, and may be of questionable accuracy (Srebotnjak et al. 2012). Consistent and accurate information on inland water quality over wider areas of the continent are required such that current condition can be assessed and changes in response to other impacts such as changes in land use, fires, flooding and climate change investigated (Salmaso and Mosello 2010, Vorosmarty et al. 2005). Such assessments are needed across a range of scales, from continuous in situ measurements to satellite remote sensing for synoptic supra-regional investigations. Approaches at different scales are complementary; assessment at each scale supports the other, but traditional in situ sampling approaches require repeated travel often to remote areas and costly laboratory analyses. Optical remote sensing offers a method to objectively assess inland water quality over multiple spatial scales. However, while ocean water colour remote sensing is relatively mature, both empirical and semi-analytical algorithms for inland water quality may suffer from several limitations: retrieval of information is often on only a single water quality constituent, specific sensor applicability, the requirement for ongoing coincident in situ data for parameterization, and limited transferability across different inland water optical types, time and concentration ranges (Matthews 2011, Odermatt et al. 2012). The dearth of globally representative in situ water quality data for inland waters, particularly Specific Inherent Optical Properties (SIOPs), prevents understanding the range of variability in inland water types necessary for the development of algorithms for widespread application. Understanding of the spectral shape, variability and relative contributions of phytoplankton, non-algal particles and dissolved matter to total scattering, backscattering and absorption in the water column is required if we are to achieve this goal (GEO 2011). Collation, standardization, and meta-analyses of existing data are needed to understand variability and to inform priority areas for future sampling efforts.
2. INCREASING AUSTRALIAN UNDERSTANDING
Unique algorithms for the detection of several water quality parameters from the reflected light from a water body have been developed by CSIRO's Division of Land and Water, employing an adaptive linear matrix inversion approach. The algorithm is used to remotely detect optically active water quality variables (chlorophyll, cyanobacterial pigments, organic and inorganic sediments and dissolved organic matter) from satellite and in situ spectroradiometric observations. By varying SIOP shape and amplitude parameters through a small group of predetermined conditions, the algorithm creates a set of candidate model parameter sets that correspond to a naturally occurring set of SIOP shape and amplitude parameters estimated from a suite of in situ measurements and samples collected concurrently during a field campaign. The inverse problem is then constrained by only considering a limited number of naturally occurring parameter sets. Thus unnatural or highly unlikely combinations of parameters are avoided. This approach allows more accurate retrieval of water quality parameters than compared to previous 'fixed' parameter approaches (Brando et al. 2012). Its application to satellite data over inland waters may allow regional and continental assessment of many of Australia's inland waters.
This paper presents the results of recent research aimed at collating existing and new optical water quality data for Australian inland waters and testing the adaptive linear matrix inversion algorithm for retrieval in both in situ spectroradiometric methods and remotely sensed data. We present the result of analyses of existing and currently obtained in situ optical data for Australian waters to highlight known optical variability and to target areas for further investigation. The paper also presents the results of applications at regional and local scales to ENVISAT MERIS and DigitalGlobe World- View 2 sensor data and draws conclusions with respect to the potential abilities of Sentinel 2 and Sentinel 3 data for mapping blooms and other optical water quality parameters in Australian and other inland waters.
3. CONCLUSIONS The project provides the basis for a library of optical characteristics of Australian inland waters which allows for algorithm optical parameterization that better matches the optical conditions observed and leads to more accurate retrieval of water quality parameters from satellite observations. Thus a framework is being established for an approach to systematically monitor Australian inland waters that will serve as a blueprint for approaches to monitoring inland water quality at global scales (Malthus et al. 2012). The paper will set out the ongoing challenges toward realizing the goal of accurate inland water quality products over wider spatial areas. These include sensor spatial resolution, cloudiness, atmospheric correction, limited in situ data with which to both provide adequate validation and to adequately parameterize algorithms.