Reliability of data collected by volunteers, a nine-year citizen science study in three Red Sea touristic facilities

Citizen‐science projects vary extensively in subject matter, objectives, activities, and scale, but there is always one common goal: to collect reliable data that can be used for scientific and policy purposes. The quality of data collected by non-professional volunteers in citizen science programs is crucial to produce data that are usable by stakeholders to implement environmental management and protection plans. The aim of this study was to assess the quality of data collected by non-specialist volunteers during the citizen science project SCUBA Tourism for the Environment (STE), carried out in three touristic resorts of the Red Sea between 2007 and 2015. The project was based on the Recreational citizen science protocol that differs from the Traditional one because it does not change dives features (i.e. dive place, depth, etc.), do not require to follow a training course and to pass a final exam before participating in the project. For the recreational protocol, scuba instructors and divemasters were briefly trained during public events about project aims, methods and expected results and then they had directly involved volunteers in the project. STE project involved more than 14,000 volunteer recreational divers in data collection on biodiversity. Through a specifically designed questionnaire, volunteers indicated which of the seventy-two target marine taxa were sighted during their recreational dive, giving an estimate of their abundance. To assess the validity of collected data, a reference researcher randomly dove with the volunteers and independently filled in the project questionnaire. Correlation analyses between the records of the reference researcher and those of the volunteers were performed to assess their quality. The study was performed based on 513 sample dives (dives in which was present our reference researcher with at least three volunteers) with a total number of 3138 volunteers tested. Different parameters were used to analyze data reliability 1) Accuracy, the similarity between data collected by volunteers and those collected by the reference researcher, obtained with Spearman’s rank correlation coefficient; 2) Consistency, obtained correlating data collected only by volunteers during the same dive, without the reference researcher, using Spearman’s rank correlation coefficient; 3) Percentage of identified that is the percentage of taxa registered by volunteers compared to the taxa observed by the reference researcher; 4) Correct identification that is the percentage of volunteers that correctly identified individual taxa when the taxon was present; 5) Correctness of abundance ratings (CAR) that is the correctness in abundance ratings made by volunteers compared to those collected by the reference researcher; 6) Reliability obtained with correlation between data collected by volunteers and those collected by the reference researcher, using Cronbach’s alpha correlation; 7) measure of similarity between each volunteer and the reference researcher, using Czekanowski’s proportional similarity index). The influence of independent variables date, depth, dive time, diving certification level and group size of participants on volunteers’ accuracy was also analyzed. The lowest mean score (mean score 51.6%, 95% Confidence Interval CI 44.1-59.2%) was obtained for Consistency, indicating that, depending on personal interests, volunteers could give attention to different taxa and the highest one for the Reliability parameter (mean score 69.8%, 95% CI 62.8 – 76.9) indicating that volunteers are able to collect good quality data. A positive correlation was found between Accuracy and Correctness of Abundance Ratings (CAR) score and date indicating that long-term projects could achieve a higher quality of data collected by volunteers. Diving certification level and dive time resulted positively correlated with all parameters except for CAR indicating that more expert divers could collect better quality data than less expert ones and that spending more time underwater could have benefits on data quality. Overall, data quality in this study was comparable to that obtained in Traditional citizen science projects where strict training activities and protocols were followed. Independent variables revealed that long-term projects could achieve a higher Accuracy and CAR, this could be due to an improvement of the project with time, in terms of public training events, more clear description of tasks requested to instructors and divemasters and their consequent improvement in volunteers involvement. We also found that expert scuba divers (volunteers with higher diving certification level) were more reliable than the less expert, this could be due to their familiarity, not only with the marine environment, but also with the diving equipment, which allows them focus on the surrounding environment rather than on their balance or equipment. Finally, more time volunteers spent underwater the more reliable their data became. This study showed that Recreational citizen science could significantly support conventional research methods in monitoring biodiversity, notwithstanding careful planning for volunteer skills according to each specific project. The use of the Recreational citizen science protocol could enhance massive volunteers participation in citizen science projects because it do not require changes to the recreational activity in order to participate; this could also allow the collection of huge amount of data in a short period of time