Using hyperspectral remote sensing to estimate chlorophyll‐a and phycocyanin in a mesotrophic reservoir

Abstract

This research estimates phytoplankton pigment concentrations (chlorophyll-a (chl-a) and phycocyanin (PC)) from hyperspectral Airborne Imaging Spectrometer for Applications (AISA) imagery. AISA images were acquired for a meso-eutrophic reservoir in Central Indiana, USA. Concurrent with the airborne image acquisition, in situ water samples and reflectances were collected. The water samples were subsequently analysed for pigment concentrations, and in situ measured reflectance spectra were used for calibrating the AISA images. Spectral indices, derived from the AISA reflectance spectra, were regressed against the measured pigment concentrations to derive algorithms for estimating chl-a and PC. The relationship between the pigment concentrations and the spectral indices were analysed and evaluated. The results indicate that the highest correlation occurred between chl-a and a nearinfrared to red ratio (coefficient of determination R 2 ¼ 0.78) and between PC and the reflectance trough at 628 nm (R 2 ¼ 0.80). The relationship between PC and the reflectance at 628 nm provides an approach to the estimation of cyanobacteria concentration from hyperspectral imagery, which facilitates water-quality authorities or management agencies in making well-informed management decisions.

Hyperspectral retrieval of phycocyanin in potable water sources using genetic algorithm–partial least squares (GA–PLS) modeling

“…The first band ( 1 ) locates around C-PC absorption spectral region (620-640 nm), and second band ( 2 ) locates around chlorophyll-a absorption or fluorescence peak band (690-700 nm), while the third band generally locates around NIR (715-760 nm) to accommodate the backscattering effects caused by suspended sediments, particular of these caused by nonalgal particles in the water column (Gitelson et al, 2008;Xu et al, 2009). Comparatively, the spectral band position for GR is significantly different from MR, ECR or 2006 combined and all combined datasets, which is mainly due to the dredging activities conducted on GR changed the optical properties of the water column (Randolph et al, 2008;Li et al, 2010).…”

“…This airborne sensor has a programmable set-up, allowing the collection of data in up to 512 discrete channels through the spectral range of 400-1000 nm. Detailed information on image acquisition can be found in Li et al (2010). Images with 62 bands in the spectral region of approximately 392-982 nm with a bandwidth of 7-8 nm was acquired with 1 m spatial resolution.…”

“…Each swath of AISA image for each reservoir was georectified to a 2003 aerial photograph with 1 m resolution (Sengpiel, 2007), with final total root mean square errors (RMSEs) within 1.8-3.4 m. All the swaths over a specific reservoir were mosaicked and portable field spectroradiometer (ASD) measurements from the sample locations were used for image calibration through the empirical line calibration technique (Lillesand et al, 2004). By doing this, the AISA image is converted from radiance into reflectance and the effects of the atmosphere between the sensor and the ground are ideally removed, but most likely significantly minimized (Li et al, 2010). From the calibrated AISA image, spectra for each field measurement with a 3 m × 3 m window were extracted for C-PC modeling to avoid the spatial mismatching error caused by geo-registration (Krishna, 2010).…”

“…Sample filtration and analyzing procedures can be found in Li et al (2010). Extracted samples were analyzed for C-PC concentrations flourometrically using a TD-700 Flourometer (Turner Designs, Inc.) equipped with a Cool White Mercury Vapor Lamp and a Phycocyanin Optical Kit (630 nm excitation and 660 nm emission filters) and calibrated using C-PC from Spirulina sp.…”

Determination of absorption coefficients for chlorophyll a, phycocyanin, mineral matter and CDOM from three central Indiana reservoirs

“…While previous methods have shown promising results (e.g. Li et al, 2010;Randolph et al, 2008), suspended sediments and CDOM interfere and reduce the transferability of these models (Vallely, 2008). By developing a set of mass-specific absorption coefficients for these reservoirs, we can apply them to the bio-optical model for estimating cyanobacteria.…”