Search

In previous studies, empirical models have been widely used to map water quality parameters for inland waters using remotely sensed imagery. Due to the complex optical properties of inland waters, empirical models often have limited performances and they cannot be extended or reused across space and time. To overcome the limitations of traditional empirical models, this article further explores the ensemble method for water quality parameter retrieval. Based on the Sentinel-2 multispectral images and extensive coincident in situ water quality data, we examined different implementation strategies and evaluated the extensibility of a multipredictor ensemble model for retrieving chlorophyll-a (Chl-a) across space and time. Our analysis shows that the multipredictor ensemble model can improve the Chl-a prediction accuracy by 46% compared with the best traditional empirical model. Our experiments suggest that the selection of heterogeneous component models improves the performance of the ensemble model and the new iterative K-fold calibration approach results in a considerably better ensemble model than the global calibration approach. To evaluate the spatial extensibility, our ensemble model calibrated for Harsha Lake was applied to Caesar Creek Lake and Brookville Lake. This ensemble model calibrated with the data sets acquired on October 7, 2016 was also applied to Sentinel-2 images of Harsha Lake acquired on October 2, 2017 and August 23, 2018, to evaluate its extensibility for different years. Our evaluation results show that the ensemble model can be transferred and reused across space and time to provide reliable and accurate Chl-a estimates.

Various empirical algorithms have been developed to retrieve chlorophyll-a (Chl-a) from multispectral and hyperspectral images as a proxy variable for algal blooms in inland waters. In most previous studies, a single empirical model (global model) was calibrated for the entire water body under study. Our analysis shows that the performance of a global model is limited for optically complex inland waters. We discovered that the global model tends to overestimate in some regions and underestimate in other regions, and that the model residuals (errors) display an apparent spatial autocorrelation pattern. To address the inadequacy of the global empirical model, this paper presents regionally or locally adaptive models to better estimate Chl-a concentrations for the first time. We collected two dense sets of Chl-a measurements over Harsha Lake in Ohio during a Sentinel-2A satellite overpass and a dedicated airborne hyperspectral flight. Based on the atmospherically corrected multispectral and hyperspectral images and concurrent in situ measurements, we implemented and evaluated the performance of regionally and locally adaptive models in comparison with the single global model. Among a number of candidate empirical algorithms, the two-band algorithm produces the best global model for Chl-a retrievals for both the multispectral and hyperspectral image sources. By subdividing the water body under investigation into several regions or a set of local areas, we demonstrate that regionally and locally adaptive models can improve Chl-a estimate accuracy by 13%–28% for the multispectral image and by 33%–47% for the hyperspectral image, in comparison with the best global Chl-a model.

This research proposes an ensemble method for synergistically combining multiple empirical algorithms to better estimate chlorophyll-a (Chl-a) concentration. In previous studies, different empirical algorithms have been employed separately and a single algorithm was often identified as the most suitable predictor for Chl-a retrieval. Our ensemble method combines different individual algorithms to form an ensemble predictor that exploits advantages of each individual algorithm to maximize the overall estimation accuracy. We evaluated two ensemble predictors: the optimally weighted ensemble predictor and the spectral space partition guided ensemble predictor. The ensemble method has been successfully applied to a Sentinel-2A multispectral image acquired over Harsha Lake, Ohio in 2016. Based on in situ water reference data and satellite imagery, we constructed two ensemble predictors that consist of three individual empirical algorithms/estimators, including 2BDA (two-band algorithm), 3BDA (three-band algorithm), and NDCI (Normalized Difference Chlorophyll Index). For the optimally weighted ensemble predictor, the weights for individual algorithms are computed by solving an overdetermined linear system with the pseudoinverse technique. For the spectral space partition guided ensemble predictor, the rules for partitioning spectral space into spectral regions were established as a decision-tree using the CART method. The optimal Chl-a estimate for a pixel is obtained by selectively using the empirical algorithm in the ensemble that has the highest expected accuracy in the spectral region where the pixel is located. Our assessments suggest that the spectral space partition guided ensemble method performs significantly better than three individual empirical algorithms and also better than the optimally weighted ensemble method.

Johansen RA, Beck R, Stumpf R, Lekki J, Tokars R, Tolbert C, McGhan C, Black T, Ma O, Xu M, Liu H, Reif M, Emery E. 2019. HABSat-1: assessing the feasibility of using CubeSats for the detec...

Monitoring harmful algal blooms (HABs) in freshwater over regional scales has been implemented through mapping chlorophyll-a (Chl-a) concentrations using multi-sensor satellite remote sensing data. Cloud-free satellite measurements and a sufficient number of matched-up ground samples are critical for constructing a predictive model for Chl-a concentration. This paper presents a methodological framework for automatically pairing surface reflectance values from multi-sensor satellite observations with ground water quality samples in time and space to form match-up points, using the Google Earth Engine cloud computing platform. A support vector machine model was then trained using the match-up points, and the prediction accuracy of the model was evaluated and compared with traditional image processing results. This research demonstrates that the integration of multi-sensor satellite observations through Google Earth Engine enables accurate and fast Chl-a prediction at a large regional scale over multiple years. The challenges and limitations of using and calibrating multi-sensor satellite image data and current and potential solutions are discussed.

This study evaluated the performances of twenty-nine algorithms that use satellite-based spectral imager data to derive estimates of chlorophyll-a concentrations that, in turn, can be used as an indicator of the general status of algal cell densities and the potential for a harmful algal bloom (HAB). The performance assessment was based on making relative comparisons between two temperate inland lakes: Harsha Lake (7.99 km2) in Southwest Ohio and Taylorsville Lake (11.88 km2) in central Kentucky. Of interest was identifying algorithm-imager combinations that had high correlation with coincident chlorophyll-a surface observations for both lakes, as this suggests portability for regional HAB monitoring. The spectral data utilized to estimate surface water chlorophyll-a concentrations were derived from the airborne Compact Airborne Spectral Imager (CASI) 1500 hyperspectral imager, that was then used to derive synthetic versions of currently operational satellite-based imagers using spatial resampling and spectral binning. The synthetic data mimics the configurations of spectral imagers on current satellites in earth’s orbit including, WorldView-2/3, Sentinel-2, Landsat-8, Moderate-resolution Imaging Spectroradiometer (MODIS), and Medium Resolution Imaging Spectrometer (MERIS). High correlations were found between the direct measurement and the imagery-estimated chlorophyll-a concentrations at both lakes. The results determined that eleven out of the twenty-nine algorithms were considered portable, with r2 values greater than 0.5 for both lakes. Even though the two lakes are different in terms of background water quality, size and shape, with Taylorsville being generally less impaired, larger, but much narrower throughout, the results support the portability of utilizing a suite of certain algorithms across multiple sensors to detect potential algal blooms through the use of chlorophyll-a as a proxy. Furthermore, the strong performance of the Sentinel-2 algorithms is exceptionally promising, due to the recent launch of the second satellite in the constellation, which will provide higher temporal resolution for temperate inland water bodies. Additionally, scripts were written for the open-source statistical software R that automate much of the spectral data processing steps. This allows for the simultaneous consideration of numerous algorithms across multiple imagers over an expedited time frame for the near real-time monitoring required for detecting algal blooms and mitigating their adverse impacts.

We analyzed 37 satellite reflectance algorithms and 321 variants for five satellites for estimating turbidity in a freshwater inland lake in Ohio using coincident real hyperspectral aircraft imagery converted to relative reflectance and dense coincident surface observations. This study is part of an effort to develop simple proxies for turbidity and algal blooms and to evaluate their performance and portability between satellite imagers for regional operational turbidity and algal bloom monitoring. Turbidity algorithms were then applied to synthetic satellite images and compared to in situ measurements of turbidity, chlorophyll-a (Chl-a), total suspended solids (TSS) and phycocyanin as an indicator of cyanobacterial/blue green algal (BGA) abundance. Several turbidity algorithms worked well with real Compact Airborne Spectrographic Imager (CASI) and synthetic WorldView-2, Sentinel-2 and Sentinel-3/MERIS/OLCI imagery. A simple red band algorithm for MODIS imagery and a new fluorescence line height algorithm for Landsat-8 imagery had limited performance with regard to turbidity estimation. Blue-Green Algae/Phycocyanin (BGA/PC) and Chl-a algorithms were the most widely applicable algorithms for turbidity estimation because strong co-variance of turbidity, TSS, Chl-a, and BGA made them mutual proxies in this experiment.

We compared 10 established and 2 new satellite reflectance algorithms for estimating chlorophyll- a (Chl- a ) in a temperate reservoir in southwest Ohio using coincident hyperspectral aircraft imagery and dense coincident surface observations collected within 1 h of image acquisition to develop simple proxies for algal blooms in water bodies sensitive to algal blooms (especially toxic or harmful algal blooms (HABs)) and to facilitate portability between multispectral satellite imagers for regional algal bloom monitoring. All algorithms were compared with narrow band hyperspectral aircraft images. These images were subsequently upscaled spectrally and spatially to simulate 5 current and near future satellite imaging systems. Established and new Chl- a algorithms were then applied to the synthetic satellite images and compared to coincident surface observations of Chl- a collected from 44 sites within 1 h of aircraft acquisition of the imagery. We found several promising algorithm/satellite imager combinations for routine Chl- a estimation in smaller inland water bodies with operational and near-future satellite systems. The CI, MCI, FLH, NDCI, 2BDA and 3 BDA Chl- a algorithms worked well with CASI imagery. The NDCI, 2BDA, and 3BDA Chl- a algorithms worked well with simulated WorldView-2 and 3, Sentinel-2, and MERIS-like imagery. NDCI was the most widely applicable Chl- a algorithm with good performance for CASI, WorldView 2 and 3, Sentinel-2 and MERIS-like imagery and limited performance with MODIS imagery. A new fluorescence line height “greenness” algorithm yielded the best Chl- a estimates with simulated Landsat-8 imagery.

We analyzed 27 established and new simple and therefore perhaps portable satellite phycocyanin pigment reflectance algorithms for estimating cyanobacterial values in a temperate 8.9 km2 reservoir in southwest Ohio using coincident hyperspectral aircraft imagery and dense coincident water surface observations collected from 44 sites within 1 h of image acquisition. The algorithms were adapted to real Compact Airborne Spectrographic Imager (CASI), synthetic WorldView-2, Sentinel-2, Landsat-8, MODIS and Sentinel-3/MERIS/OLCI imagery resulting in 184 variants and corresponding image products. Image products were compared to the cyanobacterial coincident surface observation measurements to identify groups of promising algorithms for operational algal bloom monitoring. Several of the algorithms were found useful for estimating phycocyanin values with each sensor type except MODIS in this small lake. In situ phycocyanin measurements correlated strongly (r2 = 0.757) with cyanobacterial sum of total biovolume (CSTB) allowing us to estimate both phycocyanin values and CSTB for all of the satellites considered except MODIS in this situation.

The level of sampling effort required to characterize fish assemblage condition in a river for the purposes of bioassessment may be estimated via different approaches. However, the goal with any approach is to determine the minimum level of effort necessary to reach some specific level of confidence in the assessment. In the Ohio River, condition is estimated and reported primarily at the level of pools defined by lock and dam structures. The goal of this study was to determine the minimum level of sampling effort required to adequately characterize pools in the Ohio River for the purpose of bioassessment. We followed two approaches to estimating required sampling effort using fish assemblage data from a long-term intensive survey across a number of Ohio River pools. First, we estimated the number of samples beyond which variation in the multimetric Ohio River Fish Index (ORFIn) leveled off. Then, we determined the number of samples necessary to collect approximately 90% of the fish species observed across all samples collected within the pool. For both approaches, approximately 15 samples were adequate to reduce variation in IBI scores to acceptable levels and to capture 90% of observed species in a pool. The results of this evaluation provide a basis not only for the Ohio River Valley Water Sanitation Commission (ORSANCO) but also states and other basin commissions to develop sampling designs for bioassessment that ensure adequate sampling of all assessment units.

The use of fish communities to assess environmental quality is common for streams, but a standard methodology for large rivers is as yet largely undeveloped. We developed an index to assess the condition of fish assemblages along 1,580 km of the Ohio River. Representative samples of fish assemblages were collected from 709 Ohio River reaches, including 318 “least-impacted” sites, from 1991 to 2001 by means of standardized nighttime boat-electrofishing techniques. We evaluated 55 candidate metrics based on attributes of fish assemblage structure and function to derive a multimetric index of river health. We examined the spatial (by river kilometer) and temporal variability of these metrics and assessed their responsiveness to anthropogenic disturbances, namely, effluents, turbidity, and highly embedded substrates. The resulting Ohio River Fish Index (ORFIn) comprises 13 metrics selected because they responded predictably to measures of human disturbance or reflected desirable features of the Ohio ...

A measure of stream quality, the index of biotic integrity (IBI), was adapted to great rivers (>3226 km2) and calibrated using a variety of spatial scales. Fish fauna was sampled at 60 localities within 15 impoundments of the Ohio River drainage, eastern Ohio, West Virginia and Pennsylvania, with boat electroshocker methods during the summers and autumns of 1990–1993 to provide biological information for the IBI. Significant correlation was not found between ecoregion or differing reservoirs; however, the IBI was sensitive to differences in land use and variable industrial and municipal loadings. Species richness, the percentage large river faunal group, the proportion of round-bodied sucker species, the number of centrarchid species, the number of sensitive taxa and the proportion of simple lithophilous spawning species showed the greatest change between riverine and lacustrine habitats within an impoundment. The percentage large river faunal group metric was not significantly different between riverine, transitional and lacustrine habitats; however, the metric reflected significant differences when evaluated with habitat information. The number of centrarchid species was higher in lacustrine habitats, whereas round-bodied sucker species were highest in transitional habitats. The inherent variation of proportional metrics was significantly reduced with the removal of gizzard shad. This modification of the IBI will enhance assessment sensitivity over the original approach designed for wadable streams and rivers.