CrowdWater: Motivations of Citizen Scientists, the Accuracy and the Potential of Crowd-Based Data for Hydrological Model Calibration

Abstract

Citizen science is a promising tool for the collection of environmental data because it allows data to be collected at many more locations than individual scientists could cover. The citizen science project CrowdWater aims to collect hydrological data using a smartphone app but does not require any physical installations in the stream or the ground. With the app, citizens can collect water level class data using a virtual staff gauge, submit streamflow estimates, report a qualitative soil moisture class, or report the state of intermittent streams or plastic pollution. This thesis focuses on the water level class and streamflow estimates. I investigated the motivations of the citizen scientists that contributed to CrowdWater and compared it to citizen scientists who contributed to the Naturkalender project using an online questionnaire. Naturkalender is an Austrian citizen science project that uses a similar app as CrowdWater and focuses on the collection of phenological observations of indicator plant and animal species. Citizen scientists who contribute to the projects are mainly driven by their desire to contribute to science, help society and to protect the environment, as well as to learn something new. While most CrowdWater participants agreed that their motivations to engage in the project are also fulfilled by participation, most Naturkalender participants agreed that enjoyment and learning something new were also being fulfilled by their participation. While the enjoyment aspect was not a major reason to join the projects, it was a main reason to continue contributing to both projects. This is encouraging for the further collection of crowd-based water level class observations. The quality of crowd-based streamflow and water level class observations were first assessed in a survey along nine streams in Switzerland. The results showed that water level classes were easier to estimate and had fewer and smaller errors than the streamflow estimates. The quality of the crowd-based water level class observations obtained with the CrowdWater app was also assessed by comparing them to measured water levels. The correlation between the water level class observation and the water level measurements was very good when the staff was gauge well placed. The correlation was better when the observations were made by individual citizen scientists using the app, rather than multiple citizen scientists who were asked to contribute using signs. Some of the dedicated citizen scientists contributed more than one observation per week. A modelling study, using synthetic streamflow time series based on the errors from the survey showed that these data are not useful for calibrating the hydrological model HBV-light because the errors are too large. Model calibration with synthetic water level class time series based on errors from the survey, however, showed that these data are valuable because they led to a significantly better model performance compared to simulations using random parameter sets that represent a situation without any data. The model performance was little affected by errors or the number of water level classes that were used but depended on the number of observations and the timing of the observations throughout the year. This thesis thus shows that citizens are willing to participate in hydrological data collection, that the quality of these data are good and that these data are useful for the calibration of hydrological models. Therefore, crowd-based water level class observations are a promising source of data for catchments where otherwise no information or very little information on streamflow is available. These data could potentially be used for the calibration of models that can be used for flood warning or to predict the effects of droughts.