No abstract

Retrieval of aquatic biogeochemical variables, such as the near-surface concentration of chlorophyll-a (Chla) in inland and coastal waters via remote observations, has long been regarded as a challenging task. This manuscript applies Mixture Density Networks (MDN) that use the visible spectral bands available by the Operational Land Imager (OLI) aboard Landsat-8 to estimate Chla. We utilize a database of co-located in situ radiometric and Chla measurements (N = 4,354), referred to as Type A data, to train and test an MDN model (MDNA). This algorithm’s performance, having been proven for other satellite missions, is further evaluated against other widely used machine learning models (e.g., support vector machines), as well as other domain-specific solutions (OC3), and shown to offer significant advancements in the field. Our performance assessment using a held-out test data set suggests that a 49% (median) accuracy with near-zero bias can be achieved via the MDNA model, offering improvements of 20 to 100% in retrievals with respect to other models. The sensitivity of the MDNA model and benchmarking methods to uncertainties from atmospheric correction (AC) methods, is further quantified through a semi-global matchup dataset (N = 3,337), referred to as Type B data. To tackle the increased uncertainties, alternative MDN models (MDNB) are developed through various features of the Type B data (e.g., Rayleigh-corrected reflectance spectra ρs). Using held-out data, along with spatial and temporal analyses, we demonstrate that these alternative models show promise in enhancing the retrieval accuracy adversely influenced by the AC process. Results lend support for the adoption of MDNB models for regional and potentially global processing of OLI imagery, until a more robust AC method is developed. Index Terms—Chlorophyll-a, coastal water, inland water, Landsat-8, machine learning, ocean color, aquatic remote sensing.

We used hyperspectral remote sensing with the aim of establishing a monitoring program for cyanobacteria in a South American reservoir. We sampled at a wide temporal (2012–16; 10 seasons) and spatial (30km) gradient, and retrieved 111 field hyperspectral signatures, chlorophyll-a, cyanobacteria densities and total suspended solids. The hyperspectral signatures for cyanobacteria-dominated situations (n=75) were used to select the most suitable spectral bands in seven high- and medium-spatial resolution satellites (Sentinel 2, Landsat 5, 7 and 8, SPOT-4/5 and -6/7, WorldView 2), and for the development of chlorophyll and cyanobacteria cell abundance algorithms (λ550 – λ650+λ800) ÷ (λ550+λ650+λ800). The best-performing chlorophyll algorithm was Sentinel 2 ((λ560 – λ660+λ703) ÷ (λ560+λ660+λ703); R2=0.80), followed by WorldView 2 ((λ550 – λ660+λ720) ÷ (λ550+λ660+λ720); R2=0.78), Landsat and the SPOT series ((λ550 – λ650+λ800) ÷ (λ550+λ650+λ800); R2=0.67–0.74). When these models were run for cyanobacteria abundance, the coefficient of determination remained similar, but the root mean square error increased. This could affect the estimate of cyanobacteria cell abundance by ~20%, yet it still enable assessment of the alert level categories for risk assessment. The results of this study highlight the importance of the red and near-infrared region for identifying cyanobacteria in hypereutrophic waters, demonstrating coherence with field cyanobacteria abundance and enabling assessment of bloom distribution in this ecosystem.