Water site selection

This study is situated around the southern part of Lake Albert and its surrounding upland catchments (~750 km²) (Figure 1). The area is characterized by a diverse hydrological setting, covering small upland streams, larger rivers, swamps, stationary pools, artificial spring wells, lacustrine shorelines, and sheltered bays.

During a field exploration in February 2020, 77 water contact sites were selected for the monitoring program, covering topographic and hydrological gradients within the study area (Figure 1). A site was defined as a delineated section of a water body that is used for human activities, such as fishing or bathing, for at least some part of the year. For rivers and streams, a site was specified as a 10-meter section including both banks. Pools and ponds smaller than ca 100 m² were considered in their entirety, while sites at larger water bodies and lakes were marked by a 10-meter section along the shoreline. Care was taken to avoid potential biogeochemical barriers (e.g., culverts, artificial substrate transitions, major hydrological steps) within a single water contact site. Their presence up- or downstream of the site was systematically recorded.

Data collection

Equipment and data collection protocol

The data was collected using Huawei Y7 smartphones. These devices were chosen for their relatively long battery life (3500 mAh), 13-megapixel rear camera, and GPS functionality at a reasonably low price. The data reporting was done through the freely accessible KoBoToolbox platform, which is specifically designed for data collection in remote environments (KoBoToolbox 2014). Questionnaires can be developed on a cloud-based interface, and the mobile application allows participants to fill out the questionnaires offline and submit them to the server whenever they have an internet connection.

All citizen scientists (see below for selection) received the necessary measurement equipment (Figure 2a) and protective gear to safely perform the monitoring. The equipment is listed together with its respective purpose in Table 1.

Figure 2 

(a) Overview of the equipment provided to the citizen scientists: fine-meshed scoop net, ATRAP project t-shirt, water chemistry test strips, stream thermometer, smartphone, diary and Hanna Instruments pocket tester. Latex gloves, rubber boots and the laminated sheet of millimetre paper are not shown in the photograph. (b) In-field sampling exercise during the two-day training course for the citizen scientists in Kagadi.

Table 1

Overview of the sampling equipment of the citizen scientists. In addition to the equipment mentioned here, every citizen scientist received a personal notebook to record any information they considered relevant.

EQUIPMENT	PURPOSE	DETAILS
Protective gloves	Protection: avoid contact with potentially infected water	Disposable latex gloves + heavy-duty long sleeve rubber gloves
Gumboots	Protection: avoid contact with potentially infected water	/
Stream thermometer	Measurement temperature	Accuracy of 1°C
Single-use test strips	Measurement pH, nitrate (NO3) and chlorine (Cl2) concentrations, carbonate hardness (KH), general hardness (GH)	/
Hanna Instruments HI98129 pocket tester	Measurement of water temperature, pH, electrical conductivity (EC), and total dissolved solids (TDS)	Five citizen scientists were equipped with this device, at different locations and hydrological settings to encompass the diversity within the study area.

The citizen scientists are asked to sample all water contact sites under their responsibility once a week. A single questionnaire with a fixed chronological sequence was developed to standardize this sampling across all sites and citizen scientists (https://github.com/ATRAP-UG/KOBO_Questionnaire). Multiple-choice questions are favoured over open questions, and when possible, open input is restricted to predefined ranges to minimise erroneous entries. The use of skip logic allows only those questions applicable for the particular observations to be shown.

At a water site, the citizen scientists start the questionnaire by identifying themselves and the site, and confirming that they are wearing the protective equipment. The date and time are specified, after which the citizen scientist is prompted to take an overview picture of the water site and its immediate surroundings. Next, a GPS point of the citizen scientist’s location at the site is marked, followed by a series of questions about the human activities taking place at the site and the number of people involved. Subsequently, the stream thermometer is placed in the water while a test strip is dipped in the water for one second and allowed to adjust to the water chemistry for one minute. The citizen scientists are asked to place the strip alongside the corresponding colour-chart and take a photograph, eliminating the need for interpretation in the field. Hereafter, the thermometer is taken out of the water and the read-off is entered directly in the questionnaire. At this point, the citizen scientists with a Hanna Instruments pocket tester are prompted to report the measurements for temperature, pH, electrical conductivity (EC), and total dissolved solids (TDS). After this, another question for the current time marks the start of the snail sampling, which is a 30-minute scooping effort. All sampled snails are kept aside, and the number of scoops is counted. Afterwards, the current time is again recorded, and the citizen scientist is asked to enter the number of scoops. A snail-determination guide helps the citizen scientist to distinguish Biomphalaria, Bulinus, and Radix from all other freshwater snail genera (hereafter called the Pool group). The specimens belonging to each genus are counted and placed on a laminated sheet of millimetre paper for scale, after which photographs are taken and uploaded to the questionnaire. Finally, the completed questionnaire is saved and can be submitted whenever a stable internet connection is found. Once uploaded, the submitted questionnaire is referred to as a report.

Selection and training of the citizen scientists

The citizen scientist network was set up under the project Action Towards Reducing Aquatic Snail Borne Parasitic Diseases (ATRAP) (https://www.citizenscienceuganda.info/atrap-project-description.html) through a collaborative effort between the Mbarara University of Science and Technology (MUST) in Uganda, and the Royal Museum for Central Africa (RMCA) and KU Leuven in Belgium. In February 2020, the participants were selected based on a list of criteria that were formulated by the entire project team, which consisted of social scientists, biologists, epidemiologists, and geographers. These criteria include administrative requirements (e.g., minimum 18 years old) and less stringent favourable qualifications, such as gender balance, area of residence, motivation, and volunteer experience in the community (see Supplemental File 1).

The first stage of selection was to organize meetings with leaders from the administrative and political levels, and with health officers and representatives of religious or cultural groups. These leaders know their communities thoroughly and are therefore well placed to indicate who might be interested, motivated, able to participate, and trusted by the community. By devoting ample time for mutual exchange, the views of the community leaders on the project could be probed and integrated in the subsequent steps of setting up the network. The community leaders were asked to jointly nominate and mobilize potential volunteers from within their sub-county based on the predefined set of criteria. Through face-to-face interviews conducted by the research team, a final group of 25 citizen scientists was formed. The citizen scientists were predominantly male (68%), between 23 and 49 years of age, with an average age of 37. The most common occupations were farming, fishing, and shop keeping. A vast majority (84%) had already been involved in voluntary work through, e.g., participation in the Village Health Teams (VHT) and Red Cross projects.

Next, the 25 citizen scientists were invited to a two-day training course in Kagadi. The schedule and content for this training was jointly designed by the entire ATRAP team. The first day consisted of an in-depth explanation of schistosomiasis and the relevance of the citizen scientist’s involvement. Furthermore, an introduction on how to use the smartphone was given. Finally, the use of the thermometer, water test strips and pocket tester were demonstrated, and an entire protocol was jointly run through with mock data. The second day started with a training on snail identification. Snail specimens preserved in ethanol were used to illustrate some characteristic traits of the snail hosts of interest, and to teach participants how to distinguish the specimens of interest from other freshwater snail genera. This was followed by two in-field monitoring sessions to put everything into practice (Figure 2b), and feedback was given on the spot based on incoming reports. During the two weeks following the two-day training course, each citizen scientist was accompanied to their water contact sites by a snail expert, and at least one full questionnaire was jointly completed. This way, additional feedback could be tailored to the experience and provided on the spot.

Citizen scientists are compensated 75,000 UGX (ca. 17 euros) on a monthly basis for transportation costs and mobile data to submit the reports. This sum is thus intended as compensation rather than as a salary and was not specified to the candidates during the selection process. This monthly reimbursement is of a similar magnitude as that of another citizen scientist network reporting on natural hazards in the Rwenzori region (Jacobs et al. 2019).

The citizen scientist network became fully operational on 04 March 2020. In the initial three weeks between operationalisation and a temporary suspension due to the Covid-19 pandemic on 28 March 2020, a total of 224 reports were submitted. These initial reports were validated as described earlier, and based on the results, tailored feedback was provided to the citizen scientists in person. The network restarted the monitoring activities on 24 June 2020, leading to an additional 346 reports submitted by 05 August 2020.

Report validation: most common errors and the effect of feedback

On average, the citizen scientists spent 42 minutes sampling a single site. An overview of the validation results is given in Table 2. Apart from the location error, all errors showed a decrease after the personalized feedback provided by 24 June 2020. It was found that incorrect location settings on the smartphones of four citizen scientists led to highly inaccurate GPS readings in 30 of the 346 reports submitted after the suspension, thereby artificially raising the locations flagged as invalid. Hence, these reports were not considered in the following analysis. The issue has since been resolved.

Regarding the snail identifications, for all genera, false positive reports were more common than false negatives (Table 2). In other words, there were more reports in which snails were misclassified as Biomphalaria, Bulinus or Radix than reports in which snails belonging to these respective genera were found in the Pool class. The Bulinus identification resulted in the most mistakes, with up to 80% false positive reports being identified after the initial start-up.

To explore the effect of feedback on the level of individual citizen scientists, the average number of errors per ten submissions for each of the 25 volunteers, both before and after the feedback, is shown (Figure 3). First, it should be noted that in each report, a maximum of eight potential errors can be identified. Prior to the feedback, most of the citizen scientists (64%) made on average fewer than five errors per ten submissions. Another 28% made between five and ten errors per ten submissions, with a final 8% making on average more than ten errors per ten submissions. These results improved considerably after the feedback was given, with the most notable changes amongst the citizen scientists with initially the highest error rates. The fraction of citizen scientists making on average less than five errors per ten submissions increased to 92%, with only 8% averaging between five and ten errors. For all snail genera, classifications improved after feedback. Bulinus remained the most difficult genus to identify, but the number of false positive reports reduced considerably to 32% after feedback. Contrastingly, Biomphalaria classifications were remarkably accurate, both in terms of false positive (2%) as well as false negative reports (1%). The citizen scientist with ID 24 misinterpreted the feedback on snail identification and therefore made more errors on this aspect than before. This issue has since been addressed and resolved.

Exemplifying results: sampled snail host abundances

To further explore the potential of this approach, some exemplifying results from the citizen-driven dataset are shown, such as a map of the average sampled snail host abundances in March 2020 (Figure 4). At that time, sites along the shores of Lake Albert were reported to have far lower snail abundances than inland river sites, or even no snails at all. Meanwhile, some inland river sites had total snail abundances of more than 150 specimens per 30 minutes of scooping. Most of the water contact sites had populations dominated by Biomphalaria sp., followed by Radix sp., with only a few sites reporting Bulinus sp.

In addition to assessing spatial distributions at a certain time, the weekly sampling allows us to monitor population dynamics at a high temporal resolution. For example, Figure 5 demonstrates variations in abundances of Biomphalaria sp. at four sites sampled by a single citizen scientist, with some sites showing stable abundances and others characterized by larger fluctuations.

Data quality and research potential of the pilot study

Concerns surrounding citizen science datasets often revolve around the possibility of inaccurate or incomplete measurements, as well as the lack of validation to control for potential biases (Flanagin and Metzger 2008; Pocock et al. 2019). In our research design, we distinguished two levels of abstraction between the actual in situ situation in the field and the information received from the citizen scientists. The first can be defined as an observation bias, which is the difference between the actual in situ conditions and the sampled observation made by the citizen scientists, e.g., due to incomplete observations. The second can be termed a reporting bias—the difference between that sampled observation and the information reported through the questionnaire, e.g., due to errors in the interpretation of the observations or to mistakes when filling in the questionnaire. Ideally, both biases are minimized, or at least accounted for, when interpreting the reported information as a sample of the real-world situation.

Reporting biases were minimized by the layout of the Kobo questionnaire and the validation protocol, for example by restricting entries to realistic ranges. Furthermore, the design of the protocol and the semi-automatic validation allowed us to provide tailored, individual feedback to the citizen scientists where necessary. Similar to other case studies (e.g., van der Wal et al. 2016), the results of the feedback showed clear improvements in the average reporting quality, with the most notable changes among the initially least-performing volunteers (Figure 3). Despite these advancements, the identification of Bulinus sp. remained rather poor with 32% false positive reports. This is likely the result of the higher plasticity in morphology compared with the other genera of interest (Brown 1994). Conversely, the identification of Biomphalaria species was by far the best, with only 2% false positive reports and 1% false negative reports (Table 2). This is very promising because intestinal schistosomiasis, which is transmitted by Biomphalaria snails, is highly endemic in the study area, whereas urinary schistosomiasis, transmitted by Bulinus species, is considered of minimal public health significance (Kabatereine et al. 2007). The moderately positive results for Radix open the possibility for future fasciolosis research.

Assessing—and subsequently minimizing—the observation bias is less straightforward because independent validation on the ground is necessary. Similar to other studies contrasting citizen-driven with professionally collected datasets (e.g., Krabbenhoft and Kashian 2020), we intend to compare the citizen science dataset with a monthly sampling campaign at the same water contact sites by an experienced malacologist. This comparison will provide more conclusive evidence whether citizen-driven data can accurately reflect the spatiotemporal population dynamics, i.e., abundances of snail hosts, and test the reliability of the water chemistry measurements. That way, we will be able to attribute spatiotemporal trends in sampled abundances to either actual variations in the snails’ dynamics or to differences in the citizen scientists’ observational biases. For example, contrary to our expectations, the abundance of Biomphalaria sp. in March 2020 was very high in upland areas where schistosomiasis is generally absent, but very low at the lakeside (Figure 4), where the prevalence of schistosomiasis is highest (Kabatereine et al. 2004). Correctly interpreting and attributing these differences in abundance (e.g., to seasonal or spatial drivers) requires assessing any effects of observational bias on the reported abundances.

Given the large improvements after feedback and the highly accurate identification of Biomphalaria snails, we can cautiously conclude that the current citizen scientist network can generate relevant information on putative transmission sites of intestinal schistosomiasis through a continuously growing dataset, unprecedented in terms of spatiotemporal resolution. The next step will be to quantify the observation biases, necessary to assess the completeness of the observations and to validate the abundance data. If successful, the dataset can form the basis for research on snail population dynamics and species distribution modelling.

Beyond the pilot study: broader perspectives for Citizen Science and vector monitoring in resource-limited settings

Similar to earlier studies on the potential of citizen science in monitoring vectors for human pathogens (e.g., Porter et al. 2019; Braz Sousa et al. 2020), this pilot study was incentivized by the observation that the labour- and resource-intensive nature of vector monitoring by experts hampers multiyear large-scale observations and the setup of effective control programs. An additional lack of (inter)national standards further complicates the monitoring of freshwater snail hosts (Stothard et al. 2017; Perez-Saez et al. 2019; Maes et al. 2021). Given the need for spatiotemporally dense observations, we designed a systematic sampling and classification strategy that can be executed by a selected group of trained volunteers.

Although still in its exploratory stages, the proposed citizen science approach could be a sound candidate for the further development of a standardized snail host surveillance design at larger scales for several reasons. First, the relatively low amount of resources required for the setup and functioning of the citizen scientist network makes the approach applicable to a variety of settings and sustainable for multiyear surveying. Second, the approach can be implemented across different regions provided that the same cloud-based surveying protocol and a consistent training layout is adopted. This implies that datasets from different networks operating across various regions could be identically formatted and stored centrally for easy access and analysis.

Setting up a citizen scientist network should, however, not be regarded as a quick fix. Ample consideration should be given to the research design to maximize the quality of the resulting dataset. Generally, one should aim to minimize the likelihood of misinterpretation during the data collection/reporting by anticipating and accounting for possible mistakes in the sampling protocol, and designing an in-depth training for the citizen scientists. Additionally, the implementation of a standardized validation protocol allows early identification of recurring mistakes and provides the necessary information for targeted feedback to the citizen scientists.

A citizen-driven approach also requires considerations that span beyond data-collection design and implementation. First, participant motivation is key to ensuring sustainable, long-term participant engagement (Pocock et al. 2019; Ashepet et al. 2021; West et al. 2021). Second, Pocock et al. (2019) identified the reluctance of uptake of citizen science by policy makers as an important institutional barrier for their implementation in East Africa. Aside from these two challenges, a citizen science approach does offer the added possibility for the volunteers to act as communicators between the professional researchers and the wider community audience (Jacobs et al. 2019; Pocock et al. 2019; Ashepet et al. 2021). Hence, ongoing research within the ATRAP project will also focus on these three aspects to pursue scientifically, societally, and policy-relevant citizen science.