Keywords

1 Introduction

1.1 Organisms as Sensors

The term “biosensor” began and was popularised in the late’80s by Clark Jr. (1988) [5] and has been gaining interest since. Currently, many methods of using living sensors are being implemented for various purposes ranging from ecological studies, and detection of substances to developing functional robots actuated by living cells [12, 29, 33]. An example is the use of trained dogs for the detection of chemicals at airport security [30]. We take advantage of the dog’s superior sense of smell to detect the olfactory cues emitted by toxic substances. An analogous scenario is the use of canaries in mines for the detection of high carbon monoxide concentration [32].

The need for aquatic monitoring arose with the increasing water and air pollution that came with intense industrialism [7]. This has caused an urgent need to gather extensive data on the limited freshwater supplies. In order to protect these habitats, large-scale monitoring needs to be established and perfected. One of the projects researching this, is project Robocoenosis [25], which aims at developing a biohybrid system that will serve as a complementary element to traditional water monitoring.

In contrast to living organisms, classical sensors, such as oxygen probes or conductivity sensors are a more precise source of information on the environment, due to their narrow specificity. The aforementioned examples of using organisms as sensors (such as dogs and canaries) are performed with the naked eye thanks to the relatively simple observation methods. However, the monitoring of the aquatic environment is more complex due to its inaccessibility. Continuous observation of the lifeforms inhabiting is much more logistically complex with human observers.

To overcome this limitation, the development of so-called biohybrids gained interest. A biohybrid entity (also called “biohybrid”) is a device connecting living organisms or tissues with mechanical and electrical parts [20, 37]. Biohybrids have many functions, for example as actuators in the functioning of muscles or as actuators for motile cells and can operate on a variety of levels [15, 27, 28]. Here, we present a methodology for using this approach to extract data on the surrounding environment through the observation of living organisms by electronic components.

Several companies have used this methodology to create robust water quality monitoring tools. Devices such as Daphtox II ® and MosselMonitor ® [2] have incorporated animals like Daphnia and Dreissena polymorpha for in-house monitoring. Daphnia is used as a living detection tool for ecological, toxicological and pharmaceutical studies [16, 18, 31]. The findings show altered swimming behaviour and physiology depending on the presence/absence and the intensity of toxic substances, such as paracetamol, antidepressants, microcystin, etc. The majority of these studies focus on visual observation or post-experimental video analysis with the use of various object-tracking software. Here, we translate this approach into the field aiming at an autonomous, long-term monitoring device.

Observing the behaviour of living organisms in a biohybrid module developed by Project Robocoenosis provides a different type of data compared to classical sensors. It is a broad-range sensor which focuses on the detection of a wide variety of changes (see Table 1) resulting in a lower precision for individual stressors (for example, the behavioural reaction to a change in temperature is a less precise reading than a temperature probe). The sensitivity of this biohybrid depends on the animal’s tolerance of the stimuli and our ability to interpret its behaviour. To be able to extract data from the organisms we need to: 1) have the background knowledge on the organism’s tolerance to changes and 2) have performed calibration experiments on the changing behaviour to establish the magnitude and possible sources of stress.

1.2 Project Robocoenosis

In this work, we present the most recent results of selected modules of a biohybrid entity, that incorporates multiple living organisms. The biohybrid modules (here, also called “modules”) attract or host different species and communities and observe them in a non-invasive way. This biohybrid includes several features:

Automated Observation of the Organisms: The biohybrid incorporates several species and communities in order to extrapolate more precise data on the environment, due to different lifeforms having different sensitivity to certain pollutants (for details, see Table 1).

Long-Term Data Collection: The biohybrid operates over long runtimes (relative to other autonomous biohybrid approaches such as [21, 23]: 6 days and 7 days respectively), aiming for several months/years, with little to no maintenance required.

Self-sustainable Energy Source: The biohybrid uses Microbial Fuel Cells (MFCs) as an energy harvesting mechanism allowing the entity to forgo external batteries and be entirely self-sustainable, The MFCs also serve as an additional sensor, giving an insight into the bacterial activity in the sediments. For more details, please see [24, 25, 36].

The biohybrid modules designed and investigated by the project are listed in Table 1. In this work, we present the working principles and preliminary results of selected modules in Robocoenosis.

Table 1. Biohybrid modules developed by the project Robocoenosis and their functions.
Full size table

For more information on the project’s first concepts and goals, the reasoning behind the choice of organisms and the preliminary results of the biohybrid modules, please see publications [24, 25, 36] respectively. Here, we present a further development of several biohybrid modules and the results of various calibration experiments. This work focuses on the presentation of a Hall sensor approach (Sect. 2.5, furthering the “Daphnia module” development (Sect. 2.4) and the first advances in creating a “Plankton community module” (Sect. 2.6).

2 Module Development

2.1 Test Sites

Field tests and calibration experiments take place in two Austrian lakes: Lake Millstatt and Lake Neusiedl. They represent vastly different habitats which allow for more extensive testing of the biohybrid. Lake Millstatt (\(46^\circ \)47\('\)N \(13^\circ \)34\('\)E) located in Carinthia (Austria) is an oligotrophic lake, characterized by small amounts of nutrients and clear waters [38]. Its depth reaches 150 m which allows for testing the biohybrid on different depths and provides optimal conditions for video-based approaches [26].

Lake Neusiedl (\(47^\circ \)50\('\)N \(16^\circ \)45\('\)E), in Burgenland (Austria), provides vastly different conditions, being a highly eutrophic water body [8]. It is characterized by a high nutrient and algae content. It is extremely shallow (\({<}1.8\,\text {m}\)) while being the largest body of freshwater in Central Europe, surface-wise. During multiple workshops per year, both of these lakes are being used for field experiments.

2.2 Proper Setup Design and Its Calibration

The constructed experimental setups and the final modules must meet the following guidelines:

Enable Normal Behaviour: The animals observed by the biohybrid must have the ability to show their natural behaviours, such as feeding and breeding. For this, the modules cannot be overly restricting, must allow for access to food, etc. To test this, long-term tests are compared to laboratory observations and literature.

Observation Unit: The module hosting the organisms must be able to record their movements, store the data and potentially analyze it on board. The setup for collecting the visual data needs to be adjusted taking external factors into account, such as the water turbidity, light conditions and water movement. This is continuously adjusted and improved depending on the changing conditions.

Calibration of the Biohybrid Entity: In the approach of observing organisms in their natural habitat, classical calibration is challenging due to a wide variety of factors influencing their behaviour in the field. We carry out the calibration in two ways:

  • Traditional calibration: a series of experiments are carried out under laboratory conditions in order to be able to distinguish normal behaviour from a disrupted one.

  • Long-term observation within the setup: it is essential to test whether the confinement brings an additional stress factor for the animals. This is then compared against environmental data taken from classical probes.

  • Data extrapolation: finally, two previous methods can be combined to be able to interpret data from the field and compare it against the laboratory- and literature-based studies. This will enable us to classify different behaviours of lifeforms under different conditions.

2.3 “Daphnia Module”

“Daphnia Module” Description

One of the organisms used in the Robocoenosis biohybrid entity is Daphnia sp. It is a small, planktonic crustacean in the suborder Diplostraca, ranging in size from 0.5 to 6 mm [10]. Daphnia is one of the most intensely studied freshwater organisms [22]. Thanks to its high abundance and broad distribution, it is an important part of the aquatic food chain. In recent years, it also gained popularity in toxicological and pharmaceutical studies because of its high sensitivity to even minor changes in the water chemistry or to the presence of pollutants [4, 19, 34].

One of the modules incorporated into the biohybrid entity is the “Daphnia module” (Fig. 1). This module consists of a flow-through chamber, a camera, a Raspberry Pi (a single-board computer), and a power source. The flow-through chamber hosts Daphnia in a layer of water which can be monitored by the camera. The camera used is a wide-angle camera with an OV5647 sensor with an adjustable focus and focal length of 6mm. Recordings were taken at 30 fps. This camera was chosen because it allowed the closest proximity of the camera to the Daphnia cage without having any areas out of view. The design of this module is comparable to the one described in [25]. This setup allows for continuous observation of the behaviour and detects any changes in the swimming pattern or other behavioural characteristics (for example, attraction to light).

Fig. 1.
figure 1

“Daphnia module” before a mission in Lake Neusiedl. A: Daphnia swimming in a flow-through chamber, B: Camera module recording the swimming animals and C: Raspberry Pi plugged into the energy source.

Full size image

2.4 Daphnia Calibration Experiments

The calibration of the “Daphnia module” follows the steps described in Sect. 2.2. An observation unit presented in Fig. 1 was tested in the field for 5 days to check its ability to host the lifeforms and offer them conditions suitable for presenting normal behaviour in semi-long-term conditions. Daphnia calibration experiments took place both in the field and in the laboratory. In the field, the “Daphnia module” setup was placed in water for several days, with the animals placed manually in the chamber. The recordings were then extracted when the mission was terminated.

Laboratory experiments focused on classifying the animal’s swimming behaviour in response to a stressor, where increasing salinity (2.5, 3, and 4 ppt, parts per thousand) was chosen for the first tests. Five Daphnia individuals were placed in Petri dishes filled with aquarium water (total number of individuals tested was n = 65, 100 and 100 for salinity levels 2.5, 3 and 4.5 ppt respectively). After a one-hour acclimation period, an adequate salt solution was added to each Petri dish. The animals were then recorded for 24 h and their swimming trajectories were extracted post-experiment.

2.5 “Mussel Module”

Bivalves are another group of organisms that are used in the Robocoenosis biohybrid entity. Bivalves are recognised as suitable indicator organisms for aquatic environmental monitoring since they can be easily handled and respond with different patterns of behaviour to toxicants and pollutants [3]. Currently, zebra mussels (Dreissena polymorpha) are used in a biohybrid entity due to their wide distribution in natural water bodies in Austria and Europe [17]. However, since Dreissena polymorpha is also a highly invasive species the experimental fieldwork is restricted to Lake Millstatt where the species has been previously confirmed in the project.

The “Mussel module” is currently based on an Arduino UNO R3 microcontroller and is able to get data from up to five zebra mussels simultaneously. This setup collects valve movement data from the zebra mussels by attaching a Hall effect sensor (Honeywell SS495A) to one valve and neodymium magnets (HKCM Z04x04Au-N35) to the other valve of the mussel. This Hall effect sensor was chosen since it provides a linear output with high precision. This enables accurate readings of the mussel valve movement.

In previous works [25, 36], camera-based approaches were used. Taking a step away from visual analysis, where possible, allows for quicker data collection from multiple individuals with reduced power consumption, as cameras and software required for image analysis tend to reduce the run-time of the module. The collected raw sensor data is combined with a timestamp obtained from a real-time clock and saved to an SD Card every 200 ms (Fig. 2).

Fig. 2.
figure 2

Field setup of the Mussel organ tested at Lake Millstatt. A: Arduino UNO R3 with a real-time clock and an SD card board. B: Zebra mussel (Dreissena polymorpha) situated on a Hall sensor with a neodymium magnet on the opposite valve.

Full size image

The “Mussel module” has been tested in the field at Lake Millstatt with locally-collected zebra mussels as well as in the laboratory with specimens of a colony of zebra mussels to compare their behaviour in different settings (data not shown). We also tested the reaction of the mussels to artificial stimulation.

In the preliminary experiment at Lake Millstatt, two zebra mussels collected in the field were used and data was obtained over 25 h. The reaction of the zebra mussels to artificial stimuli was tested by dropping stones next to the “Mussel module”.

2.6 “Plankton Community Module”

The composition of the plankton community in limnic water bodies is a very important indicator to estimate the water quality and the ecological status [14]. To include information on the plankton community in the respective lakes, we are currently developing an optical system based on the work by Cowen & Guigand [6]. With this optical system, called “Shadowgraph”, it is possible to observe plankton organisms in a larger volume of water without having to account for focus planes which limit normal camera systems. A “Shadowgraph” projects 3D objects in a volume (in this case plankton organisms) to a 2D plane by having two plano-convex lenses which parallelize the light through the volume. By using a combination of lenses and cameras, the light is collimated and then refocused. This allows for a sharp image of plankton organisms throughout the volume of water. By developing a laboratory setup with off-the-shelve parts, custom 3D-printed parts, and an aluminium frame, it is possible to calibrate the current “Plankton community module” using Daphnia specimens (Fig. 3).

Fig. 3.
figure 3

Laboratory prototype of the “Plankton community module” (PCO). A: Raspberry Pi 4 B: Raspberry Pi HD Camera. C: Image of Daphnia sp. made with the PCO. D: Plano-convex lenses. E: Modified C-Mount with acrylic glass as a light diffuser. Both the Raspberry Pi HD Camera and the light source are combined with a camera lens with a 16 mm focal length. The black scale bar refers to the overall image and the white scale bar refers to the magnified Daphnia image.

Full size image

3 Results

3.1 “Daphnia Module” Results

Field results of the “Daphnia module” addressed two calibration questions: 1) How do the animals react to prolonged confinement and 2) Can the module successfully film and identify Daphnia in the field?

Firstly, long-term experiments performed in both Lake Millstatt and Lake Neusiedl showed that Daphnia not only survives the confinement but also breeds inside it. This shows that the animals are able to preserve their normal behaviours of which extent will be compared against laboratory experiments.

Secondly, the “Daphnia module” allowed us to successfully take underwater recordings of the swimming animals. With post-mission video analysis, it is possible to identify Daphnia and draw their swimming trajectories. Further analysis will be performed.

Further, laboratory-based calibration experiments aimed for investigating in more detail Daphnia’s stress responses to rising salinity. Behaviour was classified as “disrupted” when an individual showed swimming in circles or movement inhibition throughout the duration of the recording. If Daphnia showed at least one spinning movement, it was classified as “spinning”, while if its movement was inhibited for 15 s consecutively, it was classified as “inhibited”. The combined behaviours were classified as “disrupted”. The percentage of Daphnia showing disrupted behaviour is presented for each salinity in Fig. 4. A Kruskal-Wallis test showed that Daphnia present significantly more disrupted behaviour in the highest salinity range compared to the two lower ranges (\(\hbox {p} < 0.05\)) after a 22-hour exposure.

Fig. 4.
figure 4

The percentage of disrupted Daphnia’s individuals (compiled from “spinning” and “inhibited” behaviours) plotted against rising salinity (2.5, 3 and 4.5 ppt). There was a statistically significant (Kruskal-Wallis test, \(\hbox {p}<0.05^{*}\)) difference in Daphnia’s reactions at a 22 h mark between the three salinity levels. The whiskers of the box plots represent Q1 (lower whiskers), Q2 (box with the median line) and Q3 (upper whiskers reaching maximum).

Full size image

3.2 “Mussel Module” Results

With the current “Mussel module”, we can clearly resolve the valve movement of zebra mussels. The mussels showed no abnormal feeding behaviour during the experiment. The preliminary experiment showed that the water disruption performed during the experiment caused both zebra mussels to close their valves simultaneously (Fig. 5).

Fig. 5.
figure 5

Normalised Hall sensor data obtained from two zebra mussels. Top means open shell, and bottom means closed shell. The dotted line represents a stimulus (a dropped stone) to which the mussels reacted by closing their shell. Data are normalized against the minimum and maximum values of the whole data (where 1 is maximally open and 0 is a maximally closed shell).

Full size image

3.3 “Plankton Community Module” Results

The currently developed prototype of the “Plankton community module” can observe plankton organisms, in this case, Daphnia specimens, in the size between \(500~\upmu \text {m}\) (size of newborn Daphnia) and 5 mm (large adult female Daphnia). Various random water samples from the lakes were also added to the preliminary experiments. The combination of off-the-shelve parts with custom 3D printed parts works excellent to test different combinations and qualities of plano-convex lenses, camera sensors and objective lenses and provides insight into future improvements.

4 Discussion

In this work, we present the most recent results obtained from testing selected modules of the biohybrid entity for environmental monitoring developed by the project Robocoenosis. The biohybrid modules host or attract various species and communities in order to monitor their well-being in the environment (Table 1). From this data, we can draw conclusions about the state of the lake. Project Robocoenosis is developing a biohybrid incorporating a combination of those modules to gain insight into both rapid and long-term changes in limnic systems. The most crucial findings of this work are:

  1. 1.

    Daphnia can be incorporated into automated biohybrid systems to be used as a living sensor. The flow-through cage allows them to feed and breed inside it. Laboratory behavioural experiments gave preliminary calibration results that indicate Daphnia’s varying behaviour depending on the presence and the intensity of the stressor.

  2. 2.

    Zebra mussel D. polymorpha can be incorporated into automated biohybrid systems to be used as a living sensor. The attachment to the underwater setup did not inhibit the mussels’ ability to feed and breathe. Both field and laboratory experiments show results enabling the detection of change in its behaviour. Preliminary results of the calibration experiments show the change in behaviour in case of a disruption in the environment. The data also shows, that the noise of the Hall effect sensors (e.g. due to temperature change) is insignificant in comparison to the behaviour readings.

  3. 3.

    Shadowgraph is a working setup that allows for the identification of planktonic organisms. Its openness to the water column (lack of confinement for the lifeforms) creates a community-based sensor that could be able to detect large-scale changes in the water chemistry. It is expected that through the Shadowgraph, it will be possible to observe the behaviours that indicate changes in the environment, such as the appearance or absence of certain groups. Then, literature data will be used to compare the observations to life cycles, seasonality and other natural periodicity.

4.1 Similar Initiatives

The use of organisms as sensors has been developed for decades in a multitude of fields. An example of a continuous real-time observation of the Daphnia’s swimming behaviour is a piece of laboratory equipment called DaphTox II ® [11, 13]. It hosts several Daphnia individuals and using image analysis tracks the swimming behaviour of the animals and sounds an alarm in case of an abnormality.

A similar approach was used for mussel observation in a MosselMonitor ® [2]. The system evaluates the behaviour of several mussels in a continuous manner and measures the degree of valve opening, activity levels and mortality. It can be used with both freshwater and marine mussels. Here, we present an adaptation of this approach, moved directly into the field which allows the mussels to remain in their natural habitat. By combining their reactions with several other species we are also able to narrow down the possible cause of stress.

In contrast to the mentioned approaches, project Robocoenosis aims to move the monitoring process into the field and provide an autonomous power source. Although other approaches provide high-resolution data, they are limited in terms of power requirements and cost. Reading of the behaviour is done in real-time in the natural environment of the organisms which bypasses the need for laboratory facilities and staff making the process less costly and time-consuming. This work differs from other approaches as it incorporates multiple organisms and interprets their cumulative reactions. Here, we aim for low-price, long-term observations in off-grid locations.

4.2 Challenges

As mentioned previously, organisms provide a wide spectrum sensor. The interpretation of the behavioural data provides insight into the overall state of the environment. The specifications of the changes occurring can only be as good as our knowledge of their reactions and sensitivity to certain compounds. For this reason, we offer the usage of biohybrids as an additional monitoring tool, rather than a substitute. Should the monitoring mission be focused on a specific parameter, such as oxygen, we suggest using an adequate classical probe. However, especially in protected areas exposed to a multitude of anthropogenic pollution, it is of merit to broaden the spectrum of the detection.

These prototypes of biohybrid modules function continuously over the period of 5 days for the “Mussel module” and 2 days for the “Daphnia module” on various power sources. This period will be extended substantially with the use of energy harvesting and low-power electronics.

5 Outlook

This approach to environmental monitoring is novel in the way it hosts multiple lifeforms in their natural habitat and draws conclusions from a non-invasive observation of their natural reactions to the environment. In order to provide optimal readings, several experiments and improvements need to be carried out.

To further the calibration of the “Daphnia module”, long-term confinement tests will be performed both in the field and under laboratory conditions. The calibration of Daphnia’s normal breeding behaviour must be carried out for various animal densities in the flow-through chamber and for changing water quality. The preliminary results showing no mortality or reproduction inhibition are promising as they show that Daphnia is able to continue its natural processes within the setup. Additionally, more tests will be carried out to optimise the robustness of the module.

The mussels showed feeding behaviour during the experiment and reacted to environmental stimuli such as waves, noise and changes in light. Those changes in valve opening are minor and can be clearly distinguished from a strong stress response (see Fig. 5). Valve movements are a normal occurrence, as shown in literature [1, 9, 35], caused by natural behaviours, like feeding and respiration. The responses to toxic substances and other stressors can be clearly identified and what is more, it is possible to distinguish different stressors from the movement patterns [9]. For example, alarm cues (sent by an injured mussel) cause it to close its valves for a significantly longer period than when being exposed to only predator presence. These and other calibration experiments will be carried out and accounted for in further experiments.