Loading...
Detecting Spatially Localized Anomalies In Vegetation Health Using A Short-term NDVI Baseline Dataset For Potential Point-source Pollutant Monitoring
Silwal, Abinash
Silwal, Abinash
Citations
Altmetric:
Abstract
Monitoring vegetation health is crucial for assessing the impacts of environmental changes and anthropogenic activities. This research developed methods to detect spatially localized vegetation health anomalies using a short-term (2016-2023) Normalized Difference Vegetation Index (NDVI) baseline dataset from Sentinel-2 imagery. The study focused on a 4x4 mile block of land in central Kansas, USA, slated for future CO2 underground injection, but the methods can be adapted for other sites and remote sensing-based monitoring activities. To prepare a reliable baseline dataset, pixel-level cloud cover impacts were mitigated using both the QA60 band and the Sentinel-2 cloud probability dataset. For anomaly detection, we focused on peak growing season NDVI, which generally is when vegetation development is the most sensitive to environmental conditions. We devised assessment methods appropriate for short-term time series centered around the Jeffries-Matusita (JM) Distance statistic, which characterizes distributional distinction between two samples. To detect anomalies, NDVI values are compared between locations at zone and pixel block levels. In addition, we integrated PRISM temperature and precipitation data to contextualize the NDVI results, in case weather-related factors may help explain the outcomes. A Google Earth Engine (GEE) application was developed for performing real-time analysis and visualization. By establishing a reliable NDVI baseline and devising innovative vegetation health anomaly detection methodologies, this research supports broader environmental sustainability efforts and provides valuable insights for future ecological assessments and interventions.
Description
These are the slides from a presentation given at American Association of Geographers on 03/24/2025.
Date
2025-03-24
Journal Title
Journal ISSN
Volume Title
Publisher
University of Kansas
Files
Loading...
SilwalA_2025.pdf
Adobe PDF, 2.4 MB
Research Projects
Organizational Units
Journal Issue
Keywords
NDVI, Vegetation health, JM Distance, Sentinel 2, Remote sensing