Supporting tools for the integrated management of drinking water reservoirs contaminated by cyanobacteria and cyanotoxins

The main objective in the BLOOWATER project was to develop a decision support system for public water supplies and agencies to prepare and respond to the risk of blooms of toxin-producing cyanobacteria in lakes used for drinking water production. Cyanobacteria, also known as blue-green algae, can quickly multiply under favourable environmental conditions, and may produce and release toxins that are harmful to humans. Occurrences of cyanobacterial blooms are increasingly detected in lakes due to emissions of nutrients originating from human activities as well as due to climate change. Though, there may be years between each outbreak, the effects may be devastating for those affected. Therefore, it is important to be able to predict when, where and at which scale such toxic blooms will occur as soon as possible to be ready to implement adequate emergency measures. Hence, the project aimed to develop an on-lake and in-air surveillance system coupled with a forecasting model that can help to give early warnings of potential upcoming cyanobacterial blooms to both the operators at the local waterworks and the local authorities. The chlorophyll content of the cyanobacteria and other phytoplankton can tell a lot about how many there are, and, with certain limitations, what kind of algae are present. It was therefore an chosen to analyse the chlorophyll content in more detail with the various analysis techniques the project had available. A relatively strong correlation was found between the chlorophyll analyses made directly in water samples and those based on multispectral analysis of images taken by satellites. Although there was a weaker correlation with the measurement data from the drone-based sensors, they made it possible to identify cyanobacteria specifically on the surface of Lake Albano in Italy. Both process modelling (PB) models and models based on machine learning (ML) were assessed to see if they were suitable for predicting cyanobacterial blooms. Historical data series from lakes in Italy, Sweden and Norway from previous blooms were used as input data in the models. Unfortunately, none of the model approaches were able to simulate the timing of algal blooms with sufficient accuracy to serve as a warning to water utilities of the potential for an upcoming bloom. Since the decision support system that were going to be developed in the project was completely dependent on sufficient accuracy in the predictions provided the models, the further development of the system, after the framework for the content and architecture had been determined, unfortunately had to be halted. However, a negative result is still of value, and this was one of the first comprehensive evaluations of both PB and ML models for the use in operational forecasting and will no doubt benefit others. The modelling methods that were developed and tested in BLOOWATER can in any case be useful for assessing seasonal changes in total algal biomass, and it turned out that ML modelling was well suited to simulating the development of the concentration of dissolved oxygen, which is closely linked to the activity of the algae, and with direct consequences for the lake ecology. It is also vital that the measures are ready to be used when needed, and that they will be effective against the actual threat as long as it is still present. The project therefore also tested and compared in lab scale tests the effectiveness of conventional technologies and newer technologies such as polymer-enhanced ultrafiltration (PEUF) used in the production of drinking water as barriers against toxins produced by cyanobacteria. Several laboratory-scale tests were run to see the effect of membranes with different nominal pore sizes (NF and low UF) on the retention of toxins of different sizes (microcystin, MC>cylindrospermopsin, CYN>saxitoxin, STX>anatoxin-a, ATX) under different operating conditions. Not unexpectedly, the membrane with a nominal pore size smaller than the smallest of the toxins was found to be most effective at removing all of the toxins, but STX and ATX were not effectively removed by either of the membranes. Experiments were also carried out with chemical precipitation with iron-based coagulants and chitosan ahead of the membrane filtration to see if it was possible to create larger aggregates of toxins and other material in the water that could be more easily removed by the membrane. Only MC and CYN were flocculated, so the removal of STX and ATX was equally weak. There has been little focus on analysing STX and ATX in Norwegian drinking water sources until now, but the results indicate that it may be important to get a better overview of these in Norwegian water bodies. Norwegian Institute for Water Research (NIVA) was the only Norwegian partner. The other partners were the Italian National Agency for New Technologies, Energy and Sustainable Economic Development, the Italian Marche Polytechnic University and the Swedish Uppsala University.

At prosjektet ikke kom i mål med hovedmålsettingen om et funksjonelt verktøy som skulle kunne identifisere algeoppblomstringer så tidlig at det ville gitt mulighet å sette inn tiltak raskt nok, gjør at også den potensielle betydningen av det gjennomførte arbeidet har blitt vesentlig svekket. Men prosjektet har uansett bidratt til å styrke kunnskapsnivået på flere områder, kanskje spesielt innen utvikling av metoder og arbeidsflyt ved bruk av maskinlæring til å predikere oppblomstringer. Også det at verken filtrering med NF- eller lav-UF-membraner alene, kjemisk felling alene eller i kombinasjon med UF eller MF-filtrering ser ut til å være gode barrierer mot de minste cyanotoksinene; saxitoxin (STX) og anatoxin-a (ATX) er viktige bidrag. Hvordan mikrocystein påvirkes i ulike behandlingsprosesser er godt dokumentert, men det er gjort betydelig færre studier med CYN, STX og ATX. Funnene vil således kunne få betydning ved risikovurdering av råvannskilder hvor man kan forvente oppblomstringer av cyanobakterier som produserer denne typen toksiner, j.fr. det nye EU-direktivet 2020/2184 «Drinking water — essential quality standards», som er ment å understøtte risikovurderingen av vannforsyningssystemene. Det har også vært noe begrenset med feltforsøk med fokus på fjerning av cyanotoksiner i pilotskala, og spesielt med miljøvennlige alternative teknologier som PEUF. De regionale miljømyndighetene i området der pilotskalaanlegget ble testet ut (Castreccioni-sjøen) har vist stor interesse for den nye teknologien og resultatene som ble fremskaffet. Selv om det har vært relativt få hendelser med oppblomstringer med giftproduserende cyanobakterier i Norge, er det en økende oppmerksomhet rundt denne utfordringen, j.fr. VKM-rapporten 2021:13 «Cyanobakterier og cyanotoksiner i norske drikkevannskilder». I all hovedsak er det mikrocystein som det blir analysert for, men det er grunn til å tro/håpe at det blir mer fokus på å analysere også for saxitoksin og anatoxin-a på bakgrunn av dokumentasjonen gjort i dette prosjektet.

The BLOOWATER project develops a decision support system (DSS) for public water supplies and agencies to prepare and respond to the risk of cyanotoxins in drinking water. The project aims to develop a support tool based on a multiple barrier-approach that integrate innovative monitoring techniques, responsive surveillance strategies and bloom/toxin-specific treatment measures at the waterworks. The project intends to create forecasting models and systems of surveillance and early warning of toxic blooms. Accordingly, source water monitoring and system observations can inform a water system’s decisions about if and when to start cyanotoxin monitoring in raw water, when and how to adjust treatment plant operations, and when to communicate with external stakeholders and the citizens. This will be useful for various stakeholders and in particular for waterworks operations. Concerning the treatment processes, polymer-enhanced ultrafiltration (PEUF) is proposed to remove both cyanobacteria and cyanotoxins from the raw water. The biopolymer chitosan is used to complex the cyanotoxins together with other natural organic matter prior to the ultrafiltration unit thereby improving their removal. NIVA provides historic data from previous cyanobacterial blooms in Norwegian freshwaters for the mechanistic modelling as a part of the development of forecast systems. NIVA coordinate the work related to the treatment technologies targeting cyanotoxins, review and test appropriate reference treatment systems among conventional technologies (e.g. nanofiltration). For the treatment testing, selected cyanobacterial cultures are provided from NIVA’s extensive Culture Collection of Algae. NIVA contributes to the development and optimization of PEUF treatment systems by investigating membrane biofouling over the full length of a UF membrane to predict the fouling caused by algae and algal organic matter to reduce maintenance costs. NIVA contributes also with national data for the DSS.