Implementation of Quality Assurance and Quality Control Measures in the National Phenology Database

Abstract:

The USA National Phenology Network (USA-NPN; www.usanpn.org) serves science and society by promoting a broad understanding of plant and animal phenology and the relationships among phenological patterns and environmental change. The National Phenology Database has over 5.5 million observation records for plants and animals for the period 1954-2015. These data have been used in a number of science, conservation and resource management applications, including national assessments of historical and potential future trends in phenology, regional assessments of spatio-temporal variation in organismal activity, and local monitoring for invasive species detection. Customizable data downloads are freely available, and data are accompanied by FGDC-compliant metadata, data-use and data-attribution policies, and vetted documented methodologies and protocols.

The USA-NPN has implemented a number of measures to ensure both quality assurance and quality control. Here we describe the resources that have been developed so that incoming data submitted by both citizen and professional scientists are reliable; these include training materials, such as a botanical primer and species profiles. We also describe a number of automated quality control processes applied to incoming data streams to optimize data output quality. Existing and planned quality control measures for output of raw and derived data include: (1) Validation of site locations, including latitude, longitude, and elevation; (2) Flagging of records that conflict for a given date for an individual plant; (3) Flagging where species occur outside known ranges; (4) Flagging of records when phenophases occur outside of the plausible order for a species; (5) Flagging of records when intensity measures do not follow a plausible progression for a phenophase; (6) Flagging of records when a phenophase occurs outside of the plausible season, and (7) Quantification of precision and uncertainty for estimation of phenological metrics. Finally, we will describe preliminary work to develop methods for outlier detection that will inform plausibility checks. Ultimately we aim to maximize data quality of USA-NPN data and data products to ensure that this database can continue to be reliably applied for science and decision-making for multiple scales and applications.