Cyanobacteria as indicators of water quality in Campania coasts, Italy: a monitoring strategy combining remote/proximal sensing and in situ data

Cyanobacteria are the most ancient group of photosynthetic microorganisms exhibiting the widest range of physiological adaptations for colonizing diverse ecological niches; their ability to develop specific physiological mechanisms allows them to adapt to various environmental stressors. Understanding the factors that regulate cyanobacterial biodiversity, indeed, represents a challenging research topic. Paradoxically, cyanobacteria can be both beneficial and harmful to humans. For instance, they are able to produce bioactive metabolites that may serve as potential leads in drug discovery (Teta et al 2013, Esposito et al 2015), while in some instances, also release potent neurotoxins that are harmful or even lethal to humans (Teta et al 2015).

The suitability of cyanobacteria for assessing water quality in various aquatic environments has been discussed for many years (Soltani et al 2012). Unicellular cyanobacteria such as Synechococcus spp. and Prochlorococcus spp. have been found to dominate open ocean subtropical and tropical oligotrophic ecosystems. In contrast, a high relative abundance of other coastal cyanobacterial species have been reported in response to eutrophic conditions. This phenomenon arising from excess nutrient input is usually of anthropogenic origin, resulting from a myriad of sources including: municipal wastewater discharges, run-off of fertilizers and manure spread from agricultural areas (WHO 1999), organic pollution, and increasing delivery of fecal coliforms to coastal water bodies (Loza et al 2013). Physiological tolerances or ecological ranges of specific cyanobacterial species can differ depending on taxon-specific cellular requirements (Mateo et al 2015). Under specific conditions of light, nutrients and temperature, cyanobacteria can bloom producing a green mat that covers the surface of water bodies. As a result of eutrophication, the occurrence of CBs has increased during the last few decades. CBs produce a strong oxygen depletion in surface waters, that can result in dead zones and fish kills. In addition, during the bloom some cyanobacteria species produce toxic secondary metabolites (cyanotoxins) that may also cause fish mortality and public health risks ranging from mild skin irritations to severe illness. Therefore, monitoring bloom occurrences, composition, frequency and intensity provides important indicators of degraded water quality (Richardson et al 2014). Satellite remote sensing techniques allow the monitoring of algal and cyanobacterial blooms over large spatial and temporal scales, modelling, predicting (Cracknell et al 2001) the formation and progression of bloom events and assessing related environmental parameters. The most commonly used satellite's ocean colour sensors measure the amount of light reflected from the ocean's surface at specific wavelengths. Each satellite contains sensors with a multitude of bands (multispectral) that cover a range of wavelengths. Therefore, the identification of the spectral bands that are most useful in the detection and monitoring of bloom events is required (Heffner 2008). Satellite measurements are especially useful for algal bloom monitoring because of the unique spectral absorbance/reflectance characteristic of photosynthetic pigments. Chlorophyll and carotenoid pigments contained within algal cells are routinely used to assess algal biomass and phylogenetic composition and distribution in aquatic ecosystems. Similarly, phycobiliproteins can be used to detect the presence of cyanobacteria in water samples. Phycobiliproteins are water-soluble photosynthetic accessory pigments that serve as the major light-harvesting pigments in cyanobacteria. Although phycoerythrin is the dominant accessory pigment in marine open ocean cyanobacterial species, phycocyanin (PC) rich cyanobacteria are widespread in both freshwater and coastal marine ecosystems. PC shows a distinctive reflectance profile, which significantly differs from Chl-a. Some of the multispectral sensors like Sea-viewing Wide Field-of-View Sensor (SeaWiFS), Landsat Enhanced Thematic Mapper Plus (ETM+), Moderate Resolution Imaging Spectroradiometer (MODIS), Advanced Land Imager (ALI) do not provide the resolution for distinguishing prokaryotic cyanobacteria from other eukaryotic algal species as these sensors are not able to detect PC (present primarily only in coastal cyanobacterial spp.). Medium Resolution Imaging Spectrometer (MERIS) bands 6 and 7 allow detecting PC features only if the concentration of cyanobacteria is sufficiently high (Kutser et al 2006). Thus, MERIS can be used in identifying cyanobacteria if they are present in relatively high abundance (Shi et al 2015). Therefore, this tool is not applicable at an early stage of a bloom (Vincent et al 2004). The spatial resolution of these sensors may also cause problems. Even the 30 m spatial resolution of Hyperion may be too coarse compared to the spatial distribution of some cyanobacterial blooms (Kutser 2009). In addition, cloud cover often makes it difficult to acquire good satellite images.

In the Italian province of Campania, coastal waters are optically complex (Morel and Prieur 1977) due to increased optical interference from atmospheric aerosols and the backscattering from optically active constituents, including phytoplankton, suspended solids and coloured dissolved organic matter (CDOM). In addition, microbial biomass may not be solely dominated by cyanobacteria. A versatile approach that can distinguish factors related to the local ecology, watershed hydrology, land use and anthropogenic impact is clearly warranted. In this study, we propose a hierarchical monitoring strategy combining remote and proximal sensing and traditional in situ aquatic monitoring. The remote sensing approach has been integrated with the use of specific indices for extracting information related to the cyanobacteria bloom. For in situ monitoring, analytical spectrophotometric analyses for determination of cyanobacterial pigment content have been integrated with the metagenomic analysis of the DNA isolated from the water samples to estimate biomass and taxonomically characterize the cyanobacterial population. Four different areas along the Campania coast, differing by location and salinity, but all extremely impacted by anthropogenic pressure, were selected as a test bed for this research.

It's noteworthy that only a few reports have documented the occurrence of cyanobacterial blooms in Campania region (Ferranti et al 2008) and none of them in the coastal waters; moreover, to our knowledge, this is the first time that the blooms are used as diagnostic bioindicators of degraded water quality.

In this study we used a hierarchical monitoring approach ranging from remote/proximal sensing to analytical in situ methods for the detection of cyanobacteria (figure 1).

Zoom In Zoom Out Reset image size

2.1. Remote and proximal sensing

Algal and cyanobacterial blooms are often noticeable within the visible and non-visible areas of the electromagnetic spectrum (EM). Consequently, by using remote and proximal sensing data, it is possible to detect and track bloom development in coastal, fresh and marine waters.

In this study the remote/proximal sensing approach includes data from satellite and aerial platforms. Specific objectives of our study were prioritized to include accomplishing the following tasks. First, our major objective was to maximize the sensitivity of detecting certain surface properties. Ideally, such responses should be linearly related to allow both ease of scaling and use over a wide range of surface conditions. Second, to facilitate making consistent spatial and temporal comparisons from different environments we normalized or reduced various potential biases including, effects due to sun angle, viewing angle, the atmosphere, topography, and instrument signal to noise ratios. Lastly, we endeavored to ensure that the measurements were linked to surface processes or specific parameters (e.g. a biophysical parameter such as LAI, biomass, or APAR; Jensen 2007).

Remote sensing research is often integrated with the development of algorithms able to perform functions such as calculating phototrophic biomass (Mishra and Mishra 2014).

As in previous studies (Lega and Endreny 2016), the satellite remote sensing data were obtained from the Landsat 8 operational land imager sensor (OLI) to create Normalized Difference Vegetation Index (NDVI) images for the area. NDVI index, first applied by Rouse et al (1973), identifies the photosynthetic affinity or 'greenness' of the vegetation through the reflective properties of the chlorophyll and mesophyll layers within the plants in the NIR and red part of the EM spectrum.

The NDVI index is quantitatively defined as:

$$\begin{equation} \text{NDVI} = \text{(}R_{\text{NIR}} - R_{\text{RED}} \text{)}/\text{(}R_{\text{NIR}} \text{ + }R_{\text{RED}} \text{)}\text{.} \end{equation}$$

The principle is that phototrophs display a sharp increase in the red reflectance spectra near 700 nm (due to chlorophyll a) so the difference between R_NIR and R_RED is a proxy for the vegetation density. Normalization relative to (R_NIR + R_RED) can partially remove the atmospheric effects from different measurements (Hu 2009). In the case of photosynthetically active vegetation, low red reflectance is observed along with very high NIR reflectance producing a NDVI, approaching 1. Soil is observed to have NDVI values close to zero, whilst water is normally associated with negative values.

We analysed the range of NDVI values only for the data related to the water surface, using data extraction techniques based on the NDWI (Normalized Difference Water Index, Gao 1996) defined as:

$$\begin{equation} \text{NDWI} = (R_{\text{NIR}} - R_{\text{GREEN}} )/(R_{\text{NIR}} + R_{\text{GREEN}} ). \end{equation}$$

Typically, remote sensing studies only use satellite data. However, only other types of aerial platforms (e.g. aircraft, rotorcraft and UAV/DRONE) permit enhanced resolution and flexibility on various spatial and temporal scales. In addition, with this type of focused monitoring, algae and cyanobacteria can be detected at lower concentrations due to the increased resolution.

In this study, to perform proximal sensing activities, we used an aircraft (Rockwell Aero Commander 685) equipped with a camera containing four bands (Intergraph DMC) and housed on an inertial platform (Applanix POS/AV 510-DG) and linked to a GPS system. The flight plan of the aircraft was programmed for a specific altitude to guarantee a resolution of 10–15 cm/pixel.

To overcome problems involving lower spatial resolution and a large acquisition bandwidth associated with panchromatic and multispectral images respectively, we utilized a recently developed method of pan-sharpening that combines panchromatic and multispectral data to create new multispectral images with a higher geometric resolution (Belfiore et al 2016). The Landsat multi-spectral data were pre-processed with specific routines (Basile Giannini et al 2015). These data were then used to compute NDVI indices and, finally, classified. The same procedure, without an atmospheric correction routine, was used for the aircraft data.

Several classes (3 for satellite data and 5 for aircraft data) of NDVI were generated for each image to illustrate different intensities of possible photosynthetic activity at each of the sites and to highlight the bloom fronts and near field regions, as well as delineating the bloom geometry. The classes were defined by a software-based analysis that factors in the relative concentration. Specifically, the lowest concentrations fall into Class 1 and the highest concentrations fall in classes 3 and 5. The analysis of the NDVI satellite data was constrained to pixels bounded by at least 1 other pixel that was fully contained within the aquatic region. This boundary condition was set to reduce the bias associated with analyzing pixels with spectral radiation signatures derived from terrestrial regions.

2.2. Description of sites and sampling protocols

In situ data were collected in a similar fashion as reported in our previous coastal water sampling campaign (Lega et al 2010, Lega et al 2012). We selected four anthropogenically-impacted areas along the Campania coast for this study (figure 2), partly covering the territory well known as the 'Land of Fires'. The name comes from the fire and/or the burial of toxic wastes by the Ecomafia that in addition to the illicit discharges of untreated urban wastewaters has caused a real environmental crisis. This region has been plagued with an increased incidence rate of diseases such as various cancers as well as malformations and other birth defects.

Zoom In Zoom Out Reset image size

A short description of each selected area with some related key points is reported in the Supplementary Materials section.

Water samples from sites A-D were collected in the region where an anomaly was detected via remote/proximal sensing. Visual observation of the sites was also performed in order to pinpoint precisely the areas in which foam, high turbidity or an altered colour was visible.

At each site, hydrographic measurements of surface water temperature, salinity and pH were performed (see table 1).

2.3. Determination of the Cyanobacterial community composition

3.1. Remote sensing and proximal sensing

Remote and proximal data were analysed in order to highlight potential alterations of the water quality.

Visual and digital processing were carried out to detect and track bloom development in coastal, fresh and marine waters and to assess the bloom extent.

In the visual analysis, different bands of the electromagnetic spectrum in Landsat 8 and aircraft images were combined. The normalized difference vegetation index (NDVI) was obtained from all platforms that generated images and data. The results were processed andseveral classes were generated. Due to the poor spatial resolution (30 m/pixel), the satellite analysis only permitted a limited identification of the area within the bloom region. Conversely, the high resolution of the aircraft data (10–15 cm/pixel) permitted a clear identification of the bloom. The definition of specific ground-truth areas (labelled in the pictures GT) enabled the calibration of remote-sensing data and aided in the interpretation and in the analysis. Specifically, it was possible to define three characteristic targets within each site including:

• GT1, with the highest NDVI value (class 4 and 5) chosen as the sampling site;
• GT2, showing an absence of an algal bloom and therefore used as a negative control;
• GT3, used as reference value for the specific area, located in the middle of the water body with no interferences.

The GT1s, where NDVI anomalies were registered at all the sites were found in specific locations harboring CB blooms.

Zoom In Zoom Out Reset image size

The first case study focused on the Agnena River (site A, figure 3), particularly on a small lake that linked with the river delta. This area suffers considerable human impact effects related to illegal discharge of wastewaters derived from a small residential area. In GT1 the presence of an unauthorized wastewater pipeline was clearly observed.

Zoom In Zoom Out Reset image size

Patria (site B) is a coastal lake, linked to the sea by with a small channel that shows permanent signs of water contamination including: fish mortality, pungent odors and the presence of a white foam along the perimeter of the lake. On the lake, the GT1 area corresponded to a pipeline linked to a restaurant and was adjacent to an artificial channel that passed through a residential area on the east coast. Notably on the east side, the lake shows a semi-permanent presence of an algal mat.

Zoom In Zoom Out Reset image size

In the past, the second coastal lake Lucrino (site C) was used for the breeding of mussels and even today networks, poles and other infrastructures, that reflect the stagnation of specific areas, are present. The presence of cyanobacteria was detected on the west coast (figure 5); in detail, this area is the furthest away from the connection with the sea and is in close proximity to a spa resort.

The fourth case study focused on the Caracciolo waterfront (site D); it is the principal touristic waterfront of the urban region of Naples.

The presence of cyanobacteria was detected in an area behind the breakwater that protects the waterfront road and part of a pedestrian area. That specific area appeared as a tide pool, but the presence of a fig tree was noticeably (clearly, not a salt water tree) growing in the sea on a breakwater rock just beyond the walkway. The salinity value in this area was measured to be close to zero. The explanation for this seeming paradox (zero salinity in the sea) lies in the presence of an unmapped fresh water spring that flows directly behind the breakwater. In addition, it was also observed that the presence of a nearby pipe coming from a spillway that drains the public sewer from storm water overflow, also contained freshwater.

3.2. Water sampling

Water samples A1, B1, C1, and D1 were collected in the GT1 area of sites A, B, C, and D, respectively. Additionally, sample D2 was collected in the GT2 area of site D (Caracciolo waterfront, figure 6). Sample D2 was chosen as a negative control for our monitoring strategy validation, as both microscopic observation and aircraft NDVI values did not reveal the presence of cyanobacteria. Various chemical and physical properties of the aquatic bodies where samples were taken are summarized in table 1. There was no clear cut statistical correlation among the hydrographical variables measured. The presence of cyanobacterial species was observed in warm (>19 °C) brackish waters. The measurement of ancillary variables such as nutrient concentrations and DOC levels might have been revealing and should be made in all future studies.

Zoom In Zoom Out Reset image size

Table 1. Summary of samples characterization.

Code	Sample type	Site	Salinity	Temp (°C)	Microscopic observations
A1	pond water	Agnena river	10‰	29.7	spiral cyanobacteria
B1	foam	Patria lake		34.5	filamentous cyanobacteria
C1	green mat	Lucrino lake	2‰	35	filamentous cyanobacteria
D1	green mat close to the cliff	Caracciolo waterfront	3‰	28.3	filamentous cyanobacteria
D2	water close to the cliff	Caracciolo waterfront	3‰	19	—

In addition, light microscopic observations of the analysed samples revealed the presence of filamentous and spiral shaped cyanobacteria (figure 7).

Zoom In Zoom Out Reset image size

3.3. Spectrophotometric analysis for determination of PC content

The presence of PC in all the field samples was investigated. Pigments from all the samples were extracted and the crude extracts were analyzed using a UV/Vis spectrophotometer. They all showed the characteristic absorption peak at 620 nm (figure 8), corresponding to the absorption peak of PC from Spirulina sp. In the same spectra, peaks attributable to Chl-a were also detected, by comparison with the UV/Vis spectra of Chl-a from Spirulina.

Zoom In Zoom Out Reset image size

3.4. 16S rRNA metagenomic analysis

The cyanobacterial community inhabiting samples A1–D1 and the negative control D2 was probed using a cultivation-independent approach. Metagenomic DNA was isolated from ∼20 mg sediment of each sample and used as a template for PCR amplification with 16S cyanobacterial specific primers CYA106F and CYA781R(a). A 16S rRNA amplicon library was prepared for each sample with the TOPO TA cloning kit (Invitrogen, CA, USA) and plasmids from randomly chosen clones (52 clones overall) were extracted and single-read sequenced.

Eighteen unique partial 16S rRNA gene sequences were found at a dissimilarity threshold ≤2%. The rRNA sequences were analyzed using the RDP (Ribosomal Database Project) Seq Match tool and BLASTn searches, and 12 sequences were shown to belong to the Cyanobacteria group. As expected, all amplicons from the metagenome of the negative control D2 were non-cyanobacterial rRNA sequences.

The cyanobacterial rRNA sequences were assigned to the genera Leptolyngbya (6 strains), Geitlerinema (1 strain), Pseudoscillatoria (1 strain) and Spirulina (1 strain) based on the top matching sequences from BLASTn searches (figure 9). Three sequences could not be assigned because they lacked significant similarities with any known cyanobacterial species.

Zoom In Zoom Out Reset image size

The taxonomic assignment of cyanobacterial sequences retraces the same clade formation observed in 16S rRNA neighbour joining and minimum evolution trees, inferred using either maximum composite likelihood or p-distance methods. The neighbour joining tree of the 16S cyanobacterial rRNA fragments, built using the maximum composite likelihood approach, is reported in figure S1 in supplementary material.

Our study focused on part of the Campania region coastline, in the province of Naples and Caserta, the so-called 'Land of Fires'. Verde et al (2013) analysed the environmental degradation of this heavily-impacted region. The malfunction or absence of wastewater treatment systems, the intensive livestock activities and the unlawful waste disposal methods at this site resulted in significant eutrophication and presumably the presence of microbiological bioindicators.

The multidisciplinary strategy proposed here is based on the hierarchical monitoring of cyanobacterial blooms by a variety of complementary approaches. Each of the applied technologies and techniques used here for environmental monitoring is by itself unable to resolve the problems connected to the early detection of cyanobacterial blooms. We propose that only through the combination and integration of applying different methodological approaches can we be truly effective in detecting early blooms of cyanobacteria. The use of aerial imagery combined with specific indices has revealed a powerful tool to relate anomalies in coastal development to environmental aquatic disturbances, and as a consequence to occurrence of cyanobacterial blooms.

Although satellites may be the instrument of choice for observing large areas, they are somewhat limited in their resolving power on finer temporal and spatial scales. Cyanobacterial blooms are indeed dynamic, spreading rapidly over large areas where satellite spatial resolution is often too low to detect, especially during the early stages of CB blooms. Proximal sensing data, obtained in this study by the use of aircraft, highlighted the presence of anomalies in the NDVI values in all the analysed sites. The presence of cyanobacteria was confirmed by in situ analyses (spectrophotometric analysis, molecular techniques and the traditional microscopic observations). In situ sampling and spectrophotometric and molecular analyses are an essential part of the proposed procedure, in that they allow the recognition of the presence of cyanobacteria directly in field samples and provide decisive identification of developing cyanobacterial bloom species. However, the sampling activity is dramatically reduced, because sampling needs to be performed only in the sites identified by remote and proximal sensing. Further research to study the correlation between blooms of specific cyanobacterial species and specific pollution parameters (e.g. nitrogen to phosphorus ratios) of effluent discharge waters is clearly warranted. The specific taxa of cyanobacteria found in the examined areas confirmed the assumption that cyanobacteria could be used as bioindicators of anthropogenic pressure in aquatic systems, specifically denoting habitats that are indeed extremely polluted sites.

Previously, the only information concerning the occurrence of cyanobacteria in the Campania region denoted the presence of Aphanizomenon ovalisporum, Cylindrospermopsis raciborskii, Planktothrix rubescens in Lake Averno (Rapporto Istisan 2013, Ferranti et al 2011). In addition, a series of algal species were identified in the brackish environments of Lake Patria (Margalef 1983). To our knowledge, this study is the first that documents the presence of the aforementioned cyanobacterial species at these sites and proposes the use of them as bioindicators of a degraded and eutrophic coastal ecosystem. Leptolyngbia sp. is the most widespread species in all the analysed samples. Some species of Leptolyngbia have been associated with nutrient enriched waters and it has been reported to prefer strongly eutrophic conditions (Mateo et al 2015) and very extreme environments (Komarek 2007). The presence of the cyanobacterial species Oscillatoriales is also indicative of eutrophic conditions.

The index NDVI was successfully used to detect cyanobacterial blooms from aerial images. In fact, NDVI takes advantage of the absorption/reflectance features of Chl-a, and therefore is not expected to be able to distinguish between cyanobacterial and algal blooms. In any case, the high spatial resolution of the images from the aircraft allows us to identify with relatively high accuracy the patterns associated with the blooms. The method proposed here can therefore distinguish the typical baseline distribution of algal biomass along the coast from the episodic bloom-forming presence (in our case associated to cyanobacteria) related to wastewater discharge and related freshwater plumes. However, the characteristic spectral signature of cyanobacterial PC can also potentially be used in the future to define specific absorption/reflectance bands for the sensors carried onboard the aerial platforms, with the aim of improving the specificity of the method. Beyond a generic visualization of coastal areas, this approach can identify early signs of degrading water quality by using bioindicators and various indices of developing bloom formation. In general, the combined monitoring of bloom occurrence, composition, frequency along with physical and chemical hydrographic measurements can be useful as important indicators of degraded water quality. These measurements can support environmental impact studies by governmental officials regarding aquatic pollution problems such as determining the effectiveness of wastewater plants in preventing eutrophication impacts within specific coastal areas (Teta et al 2016).

Finally, we emphasize that the use of these specific remote sensing data products can be used in a wide range of environmental monitoring applications, including fusion of optical data with synthetic aperture radar data to detect cattle ranching (Errico et al 2015) and use of thermal imagery to monitor landfills (Lega and Napoli 2008) and detect illegal dumping (Persechino et al 2013) and other illegal activities (Lega et al 2014). In addition, remote sensing data can be strategically combined with other data layers in geographic information systems to monitor the vulnerability of cultural sites and to anticipate or predict potential environmental violations (Persechino et al 2010).

We have proposed a multidisciplinary strategy for detecting potentially harmful algal blooms and for assessing degraded water quality along coastal regions. Specifically, we revealed that by utilizing a multidisciplinary approach we can target the early detection and the rapid monitoring of cyanobacterial blooms that may become important in degrading water quality in coastal ecosystems. The rapid detection of such blooms may be especially important in not only preventing the degradation of water quality, but also the deterioration of ecosystem health by the potential disruption of natural foodwebs. Early detection of these algal blooms may also prevent adverse human health effects from the release of toxins by harmful algal species. These harmful blooms may begin to proliferate under eutrophic conditions especially in a warming climate.

Key elements of this strategy are the combination and integration of applying different methodological approaches such as the parallel and combined use of satellite, aerial and in-situ data, the simplified multispectral image indexing and classification for a truly efficient method for detecting early blooms of cyanobacteria. The effectiveness of the strategy has been validated also by the specific taxa of cyanobacteria found in the anthropogenically-impacted areas along the Campania coast in southwestern Italy. To our knowledge, this is the first report regarding the presence of cyanobacteria in water bodies along the Campania coast. The results confirmed the assumption that cyanobacterial blooms may serve as useful bioindicators of degraded water quality in coastal ecosystems.

This work has been funded by Università degli Studi di Napoli Federico II under the STAR project SeaLEADS.

This work has been partially funded by Università degli Studi di Napoli Parthenope under 'Bando di sostegno alla ricerca individuale per il triennio 2015–2017'

The authors are grateful for the support from the United States and Italy Fulbright Commission, who have facilitated this collaboration.

Finally, the authors are also grateful for the support from the Agency for Environmental Protection (ARPA) of the Campania Region-Italy and the Nuova Avioriprese company, for providing additional imagery used in this larger research project.