Accuracy Assessment of NLCD 2019 Land Cover
Author Information
Author(s): Wickham James, Stehman Stephen V., Sorenson Daniel G., Gass Leila, Dewitz Jon A.
Primary Institution: Environmental Protection Agency
Hypothesis
How accurate is the NLCD 2019 land cover classification compared to previous versions?
Conclusion
The NLCD 2019 land cover classification shows improved accuracy compared to NLCD 2016, particularly in overall accuracy metrics.
Supporting Evidence
- Overall accuracy for NLCD2019 2016 was 77.5% when using primary reference labels.
- User's accuracies for forest loss and grass gain were greater than 70%.
- Overall accuracy improved by 5% compared to NLCD2016 for the same agreement definition.
Takeaway
This study checks how well the land cover maps of the U.S. match what is actually on the ground, and it found that the latest maps are better than the older ones.
Methodology
A stratified random sampling method was used to collect land cover reference data for accuracy assessment.
Potential Biases
Potential biases may arise from the reliance on Google Earth imagery for reference data.
Limitations
The sample size was reduced from previous assessments, which may affect the precision of accuracy estimates.
Statistical Information
Confidence Interval
± 1% for overall accuracy estimates
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website