Assessing Land Cover Accuracy in the NLCD 2019 Database
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 data for the conterminous United States?
Conclusion
The NLCD 2019 land cover data shows improved accuracy compared to previous versions, particularly in overall accuracy for 2016.
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%.
- Accuracy assessment results are essential for improving future land cover classifications.
Takeaway
This study checks how well the land cover maps match real-life conditions, and it found that the latest maps are better than before.
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 compared to previous assessments, which may affect the precision of accuracy estimates.
Statistical Information
Confidence Interval
± 1%
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website