Exploring the Use of Decision Tree Methodology in Hydrology Using Crowdsourced Data

Abstract

AbstractTo fill the observations gap on ungauged streams, crowdsourced distributed hydrologic measurements were considered as a potential supplement for observational data networks. However, citizen science data come with uncertainty as they are provided by the general public. In order to investigate this uncertainty, a decision tree methodology was applied to evaluate existing citizen science data of stream stage based on the CrowdHydrology (CH) network. Quality control (QC) flags were developed and applied to CH sites, dividing Level 1 dataset (raw dataset) into Level 2 (flagged dataset) and Level 3 (processed dataset). Error estimates were calculated to determine uncertainty in the citizen science data. The results indicate that the decision tree could provide reliable QC for citizen science data and demonstrate how uncertainty can be quantified in the QC datasets.