Eutrophication in fresh water has become a critical challenge worldwide and chlorophyll-a content is a key water quality parameter that indicates the extent of eutrophication and algae concentration in a body of water. In this paper, a forecasting model for the high accuracy prediction of chlorophyll-a content is proposed to enable aquafarm managers to take remediation actions against the occurrence of toxic algal blooms in the aquaculture industry. The proposed model combines the ensemble empirical mode decomposition (EEMD) technique and a deep learning (DL) long short-term memory (LSTM) neural network (NN). With this hybrid approach, the time-series data are firstly decomposed with the aid of the EEMD algorithm into manifold intrinsic mode functions (IMFs). Secondly, a multi-attribute selection process is employed to select the group of IMFs with strong correlations with the measured real chlorophyll-a dataset and integrate them as inputs for the DL LSTM NN. The model is built on water quality sensor data collected from the Loch Duart salmon aquafarm in Scotland. The performance of the proposed novel hybrid predictive model is validated by comparing the results against the dataset. To measure the overall accuracy of the proposed novel hybrid predictive model, the Mean Absolute Error (MAE), Mean Square Error (MSE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE) were used.

Organic matter (OM) has an essential biogeochemical influence along the hydrological continuum and within aquatic ecosystems. Organic matter derived via microbial processes was investigated within a range of model freshwater samples over a 10-day period. For this, excitation-emission matrix (EEM) fluorescence spectroscopy in combination with parallel factor (PARAFAC) analysis was employed. This research shows the origin and processing of both protein-like and humic-like fluorescence within environmental and synthetic samples over the sampling period. The microbial origin of Peak T fluorescence is demonstrated within both synthetic samples and in environmental samples. Using a range of incubation temperatures provides evidence for the microbial metabolic origin of Peak T fluorescence. From temporally resolved experiments, evidence is provided that Peak T fluorescence is an indication of metabolic activity at the microbial community level and not a proxy for bacterial enumeration. This data also reveals that humic-like fluorescence can be microbially derived in situ and is not solely of terrestrial origin, likely to result from the upregulation of cellular processes prior to cell multiplication. This work provides evidence that freshwater microbes can engineer fluorescent OM, demonstrating that microbial communities not only process, but also transform, fluorescent organic matter.

This research explores the in-situ bacterial production of aquatic fluorescent organic matter (AFOM) under controlled laboratory conditions. Whilst fluorescence techniques have long been used to monitor AFOM distribution, origin and dynamics within aquatic systems, the extent to which AFOM characteristics are defined by microbial processing in surface freshwaters has largely been overlooked. Current convention champions the assumption that humic-like (Peak C) and protein-like (Peak T) fluorescence signatures are exclusively derived from terrestrial (allochthonous) or microbial (autochthonous) origins respectively, with Peak T having been directly correlated with microbial enumeration. Under intensifying anthropogenic perturbations and changing catchment characteristics, the complexities associated with bacterial-organic matter (OM) interactions in freshwater systems are increasing, challenging our understanding as to the origin and fate of aquatic OM. To what extent the observed AFOM in freshwater systems is defined by bacterial processing and how such processing may be influenced by nutrient availability are key knowledge gaps that need to be addressed. Previous research has observed the in-situ bacterial production of humic-like compounds in a laboratory model system with a high-nutrient and high-carbon content synthetic growth medium. This work describes a non-fluorescing, simulated freshwater matrix which is low in both nutrient and organic carbon concentrations. Using this model, growth curve incubation experiments have been undertaken over a 48-hour period with a monoculture laboratory strain of Pseudomonas aeruginosa. Microbiological and fluorescence analyses undertaken at regular time intervals demonstrate the bacterial production of humic-like OM (Peak C) under oligotrophic (after 8hrs) and simulated high-nutrient conditions (after 6hrs). These findings, albeit under laboratory conditions, are important as they show that this fluorescence region, currently viewed as allochthonous in origin, can also represent labile OM generated in-situ by bacteria and, furthermore, that this bacterial production increases as a function of nutrient loading. In addition, the data quantitatively demonstrates that fluorescence intensities increase independently of cell density. These results challenge the assumption that humic-like AFOM is exclusively terrestrial in origin and suggest that bacteria may “engineer” OM in-situ that gives rise to these fluorescence characteristics as a function of metabolism. Importantly, nutrient availability is a key driver of metabolic activity, outlining the potential for the use of fluorescence as a marker for stream metabolism as opposed to a measure of bacterial numbers. Further development of the laboratory model via the utilisation of environmentally-sourced bacterial communities is required. Ultimately, this laboratory model will inform field studies that look to improve our understanding of how microbial communities respond to catchment stressors, and how these responses influence AFOM fluorescence signatures and ultimately the origin and fate of OM in freshwater systems.