3.1. Variation Patterns of Nitrates Measurements
Across the reach of the sampling sites in the Medjerda watershed, nitrates levels measurements recorded using test strips and standard laboratory analysis indicated that the water quality is good in 98.5% (n = 135) of samples ([NO
3−] < 50.0 mg/L) and of low quality in 1.5% (n = 2) of samples ([NO
3−] > 50.0 mg/L) according to the national environmental hydrosystems’ water quality standards for nitrates in Tunisia. For the purposes of US-EPA standards, nitrates values of <1.0 mg/L indicate high-quality water. Excess nitrates values at higher concentrations of more than 10.0 mg/L, indicative of lower quality, can cause low levels of dissolved oxygen (hypoxia) and can become eco-toxic for hydrosystems under certain conditions. Variations in nitrates level measurements recorded seem to be related to fluctuations in turbidity (values exceeded 50 NTU and more than 1000 NTU). This is because turbidity values depend on rainfall and therefore on sampling dates and sites. Nitrates may be flushed across the reach of lower water quality sampling sites in Medjerda than other sites of the river by runoff from urban and agricultural surfaces during precipitation events. There are also spatial differences in nitrate concentrations across the sample sites on the Medjerda River (
Table 2). The analysis of variance indicates that the citizen type and the water sampling sites significantly affect the nitrate measurements, with a probability lower than 5%. The hydrologic season does not significantly affect nitrate concentration. The only significant interaction is the citizen type x water source with a probability lower than 5%. Thus, the nitrate level is mainly related to the water source.
3.2. Reliability and Agreement between Nitrates Test Strips and Field Standard
Results show a high correlation between data collected from CS-water quality test strips and those obtained by standard methods (
Figure 3). Nitrate-sensitive water quality test strips seem to be reliable in the 0–125 mg/L concentration range. Agreement between the citizen’s readings and the standard methods is observed by the comparison of the means and standard deviations of nitrate contents. The results show moderate to very good agreement between the results of citizen-based monitoring and the standard methods. Results also show a high correlation between data collected from CS-groundwater compared to CS-surface water quality test strips and those obtained by standard methods. Nitrate surface water quality test strips seem to be sensitive to variations in turbidity value.
Nitrate test strips provide slight reliability in the 0–10 mg/L nitrate concentration range where zero nitrates are indicated while the standard revealed measurable concentrations. However, the agreements are best for intermediate to high nitrates concentrations (i.e., 10 to 25 mg/L nitrates). Previous results revealed the critical importance of the interpretation of the test strip readings, especially in the range of 10 mg/L nitrates, i.e., the upper health advisory limit for nitrates in ground water [
32]. Common interfering colloids agents may affect the reliability of nitrate test strips compared to standard methods. Good agreement was also obtained between nitrate data collected by citizens using test strips’ measurements and those from measurements by high performance liquid chromatography and colorimetric analysis [
32], ion-selective electrodes, the Szechrome reagent method [
33], and molecular absorption spectrometry [
19,
20].
3.3. Compensation of Turbidity Interference in Nitrate Measurements
In view of turbidity interference that leads to inaccuracy in nitrate measurements using test strips and standard laboratory analysis, a bias correction is made to compensate for the turbidity interference. The bias correction was specific for each citizen type and for two ranges of turbidity (
Table 3). Despite potential interferences, standard method and test strips measurements were reliable as shown in a strong consistency across a wide range of turbidity values (R and R
2 > 0.9). Nitrate measurements from the laboratory standard correlated well with nitrate test strips for each citizen types for turbidity values under 50 NTU. For each citizen types and for all citizens, equations slopes are significantly close to one, and intercepts are significantly close to zero at
p < 0.05. Nitrate measurements from laboratory standard and test strips are affected by turbidity interference for values exceeding 50 NTU. For each citizen types and for all citizens, equation slopes are significantly different from zero (0.8 to 0.9), and intercepts are not significantly close to zero (from 0.6 to 6) at
p < 0.05. It seems that suspended particles and other dissolved substances lead to an overestimation of nitrate measurements, which alters the appearance of the linear regression. Turbidity interference caused by many particles and dissolved solids, which soak up light as ultraviolet, highlights the need for checking test strips data to correct values under turbidity variation of samples and to enhance the performance of nitrate test strips measurements for the final data product. The presence of particles and other suspended solids that cause light scattering leads to a general overestimation of nitrate measurements and thus influences the color change in strip bands and absorption over the entire spectrum [
19,
34]. A sedimentation time of more than 15 min is operationally recommended to allow particulates and other suspended substances to become settled before reading the samples.
3.4. Citizen’s Typology for Nitrates Water Quality Monitoring
Projections of the citizens in the factorial plan are composed of the two main correspondence axes. The citizen types are defined according to the Ward criterion using the Chebyshev measure. The representation of citizens in the factorial plan shows the dispersion of five types of volunteers around the two axes (
Figure 4). Regarding the personal interest of citizens in data collection on WQM in Medjerda, the factor that most motivates them is learning about the water quality analysis (57.6% of interest during training). The correlation matrix for transformed variables related to citizens involved in nitrates’ assessment using a strip test indicates that motivation and commitment levels are correlated with all the variables except for employment, wearing glasses, mode of observation and the relationship with the project team (
Table 4). The factorial correspondence axis, explained variance, and eigenvalues of the citizens involved in nitrates’ assessment are summarized in
Table 5 Two factorial correspondence axes with correlated variables of the citizens involved in nitrates’ assessment are identified:
Axis 1: Relevance of the project to the citizen’s attitude. Participants involved in the water sector showed commitment to the CS-WQM project, an interest during training sessions by addressing questions, and by following the water strip test learning instructions.
Axis 2: Citizen identification. Motivation levels of the participants are linked to their education level, main occupation, and relation with the project team and age.
We studied and categorized five types of citizens based on their education, commitment, motivation, socio-economic background, and how they were recruited into the group. Type 1, being motivated volunteers, have the lowest inertia while Type 3, being students with high and low inertia, considering the second and first axis. Regarding the personal interest of these types in data collection (57.6% of interest during training), previously achieved using the Likert scale survey, learning about water quality test strips analysis as a new easy tool for the WQM of Medjerda is considered the most important motivation factor in the project involvement. Citizens belonging to Type 2, mainly representing agricultural hand labor, have moderate inertia considering both axes. Types 4 and 5, being to researchers and water operators, have the highest values of inertia theses axis.
Typologies of citizen science based on involvement in research steps and goals are crucial. Success factors for CS in environmental monitoring may be considered during project steps (design, start, and during implementation) [
34]. In previous studies, the level of involvement and influence in CS-WQM programs were categorized into three levels based on the Cornell Lab of Ornithology (CLO) (contribution, collaboration, and co-creation) [
35]. A vast majority of the CS-WQM projects in previous case studies are of the contributory type, including most that collect water quality data, where citizens are responsible for gathering data information, while the scientists or experts are the ones who come up with the question and plan for the research. Goal-setting and data analysis should be reached at the citizen participation level [
36,
37]. The participation of citizens in other steps should lead to other typologies of CS than the CLO model. Education, resource management, and data collection are the goals distinguished for knowledge increase and outreach creation [
38]. Participants who took part in the project felt like they belonged to the community and they were more likely to do things such as talking to people they know about the project or going to public meetings [
39]. In this participative CS approach, levels of harmony, trust, and cooperation in society should be enhanced [
40].
Box plots of nitrates collected by citizens using strips and those obtained by conventional methods for each citizen type and for two turbidity ranges across all sites in the Medjerda watershed are presented in
Figure 5. Variations in nitrate level measurements recorded seem to be related to turbidity variations and citizen types. Besides turbidity interference in readings samples, educational, socio-economic background, and recruitment modality of citizens as well as the amount of nitrate had an influence on variations in nitrates’ measurements.
3.5. Futures Implications of CS-WQM
Because most African countries have very limited experience of CS-hydrological monitoring and pollution control, much effort is needed to build data collection capacity in WQM. An overview of CS related to hydrological monitoring and WQM in some African countries is presented in
Table 6.
The African capacity is still often very poor compared to other continents as a component of WQM. Hence, Tunisia, like all African countries, should be concerned with the promotion of efficient use of such tools and a CS-based WQM approach for their socio-economic sustainable development and for tackling CC-related challenges. The CS-based WQM program in the Medjerda hydrosystem provides opportunities and benefits in terms of powerful smart tools for tackling CC challenges and their impact on water management. The CS-WQM approach in Medjerda has the potential to enhance conservation efforts by providing knowledge, motivating the public to take action, supporting scientific research, and involving citizens actively. The CS-WQM approach also provides a cheap way to gain knowledge and feedback on water monitoring and related policies by using non-traditional data sources and analytical skills, and by involving citizens. CS-WQM as a flexible and robust participative approach can induce socio-economic and institutional transformation, reduce water resource uses, increase ecosystem health, and induce functionality, resilience and integration of the ecosystem.
More community initiatives have to be undertaken progressively to monitor water quality across the Medjerda watershed. Besides contribution involvement of volunteers in data collection, participation should lead to collaboration and co-creation levels. We should be more engaged if we want to use CS to improve water governance and get citizens involved in local problems. We need to put more effort into improving communication, training, feedback, motivation, and connection to make this participatory approach in WQM even better. The attitudes of the citizens, their type and complete acceptance of test strips’ tools are the main factors that determine the success of a CS-based nitrates WQM program. This kind of CS-WQM approach has great potential for identifying local sources of nitrate contamination, which could ultimately be used to reduce the community’s impact on the Medjerda watershed.