Monitoring sub-arctic wetland vegetation using nested scales of spectrometry to inform multiple endmember spectral unmixing of Sentinel-2A imagery
The vegetation of the vast circumboreal subarctic wetlands plays an important role in moderating or exacerbating ongoing climate impacts, making the monitoring of change in vegetation foundational to understanding and predicting the carbon balance at high latitudes. We use nested scales of intersect...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IOP Publishing
2025-01-01
|
| Series: | Environmental Research: Ecology |
| Subjects: | |
| Online Access: | https://doi.org/10.1088/2752-664X/adf448 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The vegetation of the vast circumboreal subarctic wetlands plays an important role in moderating or exacerbating ongoing climate impacts, making the monitoring of change in vegetation foundational to understanding and predicting the carbon balance at high latitudes. We use nested scales of intersecting spectral data to estimate and map fractional vegetation composition of three sub-arctic peat accumulating wetlands using multiple endmember spectral mixture analysis (MESMA). We develop a bottom–up reference library for unmixing based on nested scales of data beginning with the highest resolution of a ground collected handheld spectral measurements, progressing to 1 m ^2 resolution using fused hyperspectral-LiDAR data, and then subsequently map predictively at the spatial resolution of the 10 m ^2 multi-spectral imagery of the European Space Agency’s Sentinel-2A. We assess the accuracy of the MESMA unmixing with a confusion matrix between field sampling plots and satellite (Sentinel-2A) MESMA pixel-plots, and visual assessment. We perform MESMA on imagery four years apart, to estimate the vegetation compositional turnover, at three separate sites representing three different types of wetlands. The spectral libraries we develop return kappa statistics between 0.79 and 0.95, and unmix between 92.4 and 99.1 percent of the wetland imagery. The confusion matrix used to evaluate the model’s classification of vegetation results in misclassification rates ranging from 0.07 to 0.45. Our results demonstrate that MESMA can provide important information about vegetation dynamics at a high resolution in these highly heterogeneous wetland systems. These findings and examples highlight the future potential for extracting meaningful ecological information about expansive, heterogeneous subarctic wetlands. |
|---|---|
| ISSN: | 2752-664X |