Search

A hyperspectral inversion algorithm was used to distinguish between cyanobacteria and algal blooms in optically complex inland waters. A framework for the algorithm is presented that incorporates a bio-optical model, a solution for the radiative transfer equation using the EcoLight-S radiative transfer model, and a non-linear optimization procedure. The natural variability in the size of phytoplankton populations was simulated using a two-layered sphere model that generated size-specific inherent optical properties (IOPs). The algorithm effectively determined the type of high-biomass blooms in terms of the relative percentage species composition of cyanobacteria. It also provided statistically significant estimates of population size (as estimated by the effective diameter), chlorophyll-a (chl-a) and phycocyanin pigment concentrations, the phytoplankton absorption coefficient, and the non-algal absorption coefficient. The algorithm framework presented here can in principle be adapted for distinguishing between phytoplankton groups using satellite and in situ remotely sensed reflectance.

This short communication describes several previously undocumented processing steps, pixel flagging procedures, and improvements made to the maximum peak height (MPH) algorithm, which is aimed at the operational detection of chlorophyll a (chl-a) and cyanobacteria blooms in inland and near-coastal waters. The improvements reduce false-positive cyanobacteria detection in oligotrophic waters; enable pixels in clear waters affected by environmental radiances (stray light) to be identified; and enhance detection of floating aquatic vegetation. Several case studies from around the globe are used to illustrate these improvements, and demonstrate new pixel flagging procedures and algorithm operation. The algorithm was validated using a large in situ dataset for chl-a from a wide range of inland water types. The MPH chl-a estimates were more stable than the fluorescence line height (FLH) and maximum chlorophyll index (MCI) values. The results demonstrate the potential for using the MPH algorithm, now released as a plug-in for the Basic ENVISAT A(ATSR) and MERIS (BEAM) processing toolbox, as a standard product for estimating chl-a (and trophic status) in inland and near-coastal waters using MERIS and the forthcoming Sentinel-3 Ocean and Land Colour Instrument.

Eutrophication and increasing prevalence of potentially toxic cyanobacterial blooms among global inland water bodies is becoming a major concern and requires direct attention. The European Space Agency recently launched the Ocean and Land Color Instrument (OLCI) aboard the Sentinel 3 satellite. The success of the mission will depend on extensive validation efforts for the development of accurate and robust in-water algorithms. In this study, four full atmospheric correction methods are assessed over four inland water reservoirs in South Africa, along with a suite of red/NIR based semi-analytic and band difference models for chl-a estimation which were applied to both full and partial atmospherically corrected data. In addition, we tested a novel duplicate pixel correction method to account for duplicate pixels induced by high observation zenith angles. Radiometric errors associated with OLCI Top of Atmosphere (TOA) radiances over small water targets were also investigated by modeling in situ reflectance measurements to at-sensor radiances using MODTRAN. Of the four atmospheric corrections, the 6SV1 radiative transfer code showed the most promise for producing reasonable reflectances when compared to in-situ measurements. Empirically derived band difference models outperformed all other chl-a retrieval methods on both partially and fully corrected reflectances. The Maximum Peak Height (MPH) algorithm applied to Bottom of Rayleigh Reflectance (BRR) performed best overall (R2 = 0.55, RMSE(%) = 99), while the Maximum Chlorophyll Index (MCI) performed best on atmospherically corrected data using 6SV1 (R2 = 0.35, RMSE(%) = 107). Semi-analytic chl-a retrieval methods proved very successful when applied to in situ Rrs, however, are not reliable when applied to low quality reflectance data. The SIMilarity Environment Correction (SIMEC), an adjacency correction applied in conjunction with the image correction for atmospheric effects (iCOR) processor, did not improve retrieval results for these small water targets.

The medium resolution imaging spectrometer full resolution (MERIS FR) archive (2002 to 2012) over South Africa has been processed with the maximum peak height (MPH) algorithm for the 50 largest standing water bodies in South Africa. Time series of chlorophyll a (chl-a), cyanobacteria and surface scum area coverage were used to establish the status, seasonality and trends for each of the water bodies. The majority (62%) of the 50 water bodies were hypertrophic (mean chl-a > 30 mg m− 3), while 23 water bodies had intermediate to extensive cyanobacteria coverage. Cyanobacterial surface scum events posing a high health risk occurred in 26 of the water bodies in varying extents. Analysis of significant trends showed that water bodies both worsened and improved with regard to eutrophication, cyanobacteria and surface scum coverage between 2005 and 2012. Validation of the MPH algorithm using an independent in situ dataset demonstrated that gross trophic status through chl-a can be accurately determined in both eukaryote and cyanobacteria-dominant waters. Chl-a estimation in oligo/mesotrophic waters remains challenging due to a wide range of potential sources of error. This study is the first of its kind providing quantitative chl-a and phytoplankton species information in lakes from a time series of satellite remotely sensed data on a sub-continental scale, demonstrating how global analyses of changes in lake biogeochemistry might be performed in the future. It demonstrates the pivotal role that satellite remote sensing can play in supplementing in situ monitoring efforts, particularly in the developing world. The study concludes that eutrophication and cyanobacterial blooms are widespread in South African water bodies and remain issues of critical concern for water quality.

A novel algorithm is presented for detecting trophic status (chlorophyll-a), cyanobacterial blooms, surface scum and floating vegetation in coastal and inland waters using top-of-atmosphere data from the Medium Resolution Imaging Spectrometer (MERIS). The maximum peak-height algorithm (MPH) uses a baseline subtraction procedure to calculate the height of the dominant peak across the red and near-infrared MERIS bands between 664 and 885 nm caused by sun-induced chlorophyll fluorescence (SICF) and particulate backscatter. Atmospheric correction of the MERIS TOA reflectance data for gaseous absorption and Rayleigh scattering proved adequate given the spectral proximity of the relevant bands and the sufficiently large differential spectral signal. This avoided the need to correct for atmospheric aerosols, a procedure which is typically prone to large errors in turbid and high-biomass waters. A combination of switching algorithms for estimating chl-a were derived from coincident in situ chl-a and MERIS bottom-of-Rayleigh reflectance measurements. These algorithms are designed to cover a wide trophic range, from oligotrophic/mesotrophic waters (chl-a 20 mg m− 3) and surface scums or dry floating algae or vegetation. Cyanobacteria-dominant waters were differentiated from those dominated by eukaryote algal species (dinoflagellates/diatoms) on the basis of the magnitude of the MPH variable. This is supported by evidence that vacuolate cyanobacteria (Microcystis aeruginosa) possess enhanced chl-a specific backscatter which is an important bio-optical distinguishing feature. This enables these phytoplankton groups to be distinguished from space. An algorithm derived from cyanobacteria-dominant waters had a r2 value of 0.58 for chl-a between 33 and 362 mg m− 3 and an error of 33.7% (N = 17). The operational algorithm for eukaryote-dominant algal assemblages gave a coefficient of determination of 0.71 and a mean absolute percentage error of 60% for chl-a in the range 0.5–350 mg m− 3 (N = 48). A flag based on cyanobacteria-specific spectral pigmentation and fluorescence features was used to identify cyanobacterial-dominance in eutrophic waters. Global applications demonstrate how the MPH algorithm can offer rapid and effective assessment of trophic status, cyanobacterial-dominance, surface scums and floating vegetation in inland and coastal waters.

Eutrophication and cyanobacterial algal blooms present an increasing threat to the health of freshwater ecosystems and to humans who use these resources for drinking and recreation. Remote sensing is being used increasingly as a tool for monitoring these phenomena in inland and near-coastal waters. This study uses the Medium Resolution Imaging Spectrometer (MERIS) to view Zeekoevlei, a small hypertrophic freshwater lake situated on the Cape Flats in Cape Town, South Africa, dominated by Microcystis cyanobacteria. The lake's small size, highly turbid water, and covariant water constituents present a challenging case for both algorithm development and atmospheric correction. The objectives of the study are to assess the optical properties of the lake, to evaluate various atmospheric correction procedures, and to compare the performance of empirical and semi-analytical algorithms in hypertrophic water. In situ water quality parameter and radiometric measurements were made simultaneous to MERIS overpasses. Upwelling radiance measurements at depth 0.66 m were corrected for instrument self-shading and processed to water-leaving reflectance using downwelling irradiance measurements and estimates of the vertical attenuation coefficient for upward radiance, K u , generated from a simple bio-optical model estimating the total absorption, a ( λ ), and backscattering coefficients, b b ( λ ). The normalised water-leaving reflectance was used for assessing the accuracy of image-based Dark Object Subtraction and 6S Radiative Transfer Code atmospheric correction procedures applied to MERIS. Empirical algorithms for estimating chlorophyll a (Chl a ), Total Suspended Solids (TSS), Secchi Disk depth ( z SD ) and absorption by CDOM ( a CDOM ) were derived from simultaneously collected in situ and MERIS measurements. The empirical algorithms gave high correlation coefficient values, although they have a limited ability to separate between signals from covariant water constituents. The MERIS Neural Network algorithms utilised in the standard Level 2 Case 2 waters product and Eutrophic Lakes processor were also used to derive water constituent concentrations. However, these failed to produce reasonable comparisons with in situ measurements owing to the failure of atmospheric correction and divergence between the optical properties and ranges used to train the algorithms and those of Zeekoevlei. Maps produced using the empirical algorithms effectively show the spatial and temporal variability of the water quality parameters during April 2008. On the basis of the results it is argued that MERIS is the current optimal sensor for frequent change detection applications in inland waters. This study also demonstrates the considerable potential value for simple TOA algorithms for hypertrophic systems. It is recommended that regional algorithm development be prioritized in southern Africa and that remote sensing be integrated into future operational water quality monitoring systems.