Projects
Abstract
In ecology, identification of species is necessary in order to determine the composition of habitats, biodiversity, and the health of ecosystems. However, identification can be a difficult process due to the need for frequent field observations and the existence of visually similar species. In this respect, machine learning offers many benefits for the field of ecology through classification. For the identification of birds, audio analysis can be used to simplify the process by using their unique calls as identifiable features. With the use of citizen science recordings, this experiment shows how analyzing spectrograms of audio data can be used to identify birds from field recordings.
Datasets
For this project, I used insect species occurrence data from the US Department of the Interior and cartographic boundary files from the United States Census Bureau.
The insect species occurrence dataset was tabular and contained an identified individual for each row. Each row included the date, coordinates, and scientific name for the individual. The dataset is hosted by the Global Biodiversity Information Facility, and this site was used to filter the occurrences to be within the 50 states before it was exported and used in the project. Once cleaned, a shape file was produced containing only the individuals that were identified at the species level, their year of occurrence, and the coordinates of the occurrence in WGS84.
- A pest reporting and monitoring app used to coordinate community‑based solutions to pest activity.
- Utilizes PostgreSQL, Node.js, and AngularJS to provide a resilient and responsive web application.
- Contributed to the project as the database engineer, and implemented features in the middleware and frontend.
- Designed the database schema and normalized it to provide efficient data storage and reduce data duplication.