Estimating Tundra Vegetation Cover Using Near-surface Repeat Field Photographs
PURPOSE: Quick, robust methods to measure vegetation are preferred in polar environments. The success and widespread adoption of plot-level photography in the Arctic is contingent upon the accuracy of image analysis. Geographic Object-Based Image Analysis (GEOBIA or OBIA) may effectively classify im...
Main Author: | |
---|---|
Format: | Text |
Language: | unknown |
Published: |
ScholarWorks@GVSU
2021
|
Subjects: | |
Online Access: | https://scholarworks.gvsu.edu/gradshowcase/2021/biology/2 |
id |
ftgvstateuniv:oai:scholarworks.gvsu.edu:gradshowcase-1338 |
---|---|
record_format |
openpolar |
spelling |
ftgvstateuniv:oai:scholarworks.gvsu.edu:gradshowcase-1338 2023-05-15T14:54:44+02:00 Estimating Tundra Vegetation Cover Using Near-surface Repeat Field Photographs Christoffersen, Hana 2021-04-01T07:00:00Z https://scholarworks.gvsu.edu/gradshowcase/2021/biology/2 unknown ScholarWorks@GVSU https://scholarworks.gvsu.edu/gradshowcase/2021/biology/2 Graduate Showcase: Education for the Future text 2021 ftgvstateuniv 2022-12-09T08:05:14Z PURPOSE: Quick, robust methods to measure vegetation are preferred in polar environments. The success and widespread adoption of plot-level photography in the Arctic is contingent upon the accuracy of image analysis. Geographic Object-Based Image Analysis (GEOBIA or OBIA) may effectively classify images of tundra vegetation, but it has not been thoroughly tested at fine-scales in the Arctic. This study investigates the accuracy of an object-based approach in quantifying vegetation cover from near-surface digital images at Utqiaġvik, Alaska. SUBJECTS, METHODS AND MATERIALS: We tested our approach on twelve images with three bands (red, green, blue). We estimated the cover of six plant functional groups using nearest neighbor classification. ANALYSES: We compared our digital estimates to field-based estimates using paired t-tests. We also generated confusion matrices to assess the performance of our classification. RESULTS: We detected significant differences in estimates of bryophyte (p<0.005) and graminoid cover (p<0.005). There were no significant differences in estimates of dead plant material, deciduous shrub, forb or lichen cover. Between 0.004% to 14.2% of the images were indecipherable due to shadow. Kappa values ranged from 41 – 48%. Overall accuracy ranged from 49 – 55%. CONCLUSIONS: We can use OBIA to partially automate classification of tundra vegetation from digital images. This technique is limited to the estimation of dead plant material, deciduous shrub, forb and lichen cover. We recommend a larger sample size and an investigation of different classifiers for future studies. Plot-level photography maximizes our time, funding and technology in order to monitor terrestrial change in the Arctic. Text Arctic Tundra Alaska Grand Valley State University: Scholar Works @ GVSU Arctic |
institution |
Open Polar |
collection |
Grand Valley State University: Scholar Works @ GVSU |
op_collection_id |
ftgvstateuniv |
language |
unknown |
description |
PURPOSE: Quick, robust methods to measure vegetation are preferred in polar environments. The success and widespread adoption of plot-level photography in the Arctic is contingent upon the accuracy of image analysis. Geographic Object-Based Image Analysis (GEOBIA or OBIA) may effectively classify images of tundra vegetation, but it has not been thoroughly tested at fine-scales in the Arctic. This study investigates the accuracy of an object-based approach in quantifying vegetation cover from near-surface digital images at Utqiaġvik, Alaska. SUBJECTS, METHODS AND MATERIALS: We tested our approach on twelve images with three bands (red, green, blue). We estimated the cover of six plant functional groups using nearest neighbor classification. ANALYSES: We compared our digital estimates to field-based estimates using paired t-tests. We also generated confusion matrices to assess the performance of our classification. RESULTS: We detected significant differences in estimates of bryophyte (p<0.005) and graminoid cover (p<0.005). There were no significant differences in estimates of dead plant material, deciduous shrub, forb or lichen cover. Between 0.004% to 14.2% of the images were indecipherable due to shadow. Kappa values ranged from 41 – 48%. Overall accuracy ranged from 49 – 55%. CONCLUSIONS: We can use OBIA to partially automate classification of tundra vegetation from digital images. This technique is limited to the estimation of dead plant material, deciduous shrub, forb and lichen cover. We recommend a larger sample size and an investigation of different classifiers for future studies. Plot-level photography maximizes our time, funding and technology in order to monitor terrestrial change in the Arctic. |
format |
Text |
author |
Christoffersen, Hana |
spellingShingle |
Christoffersen, Hana Estimating Tundra Vegetation Cover Using Near-surface Repeat Field Photographs |
author_facet |
Christoffersen, Hana |
author_sort |
Christoffersen, Hana |
title |
Estimating Tundra Vegetation Cover Using Near-surface Repeat Field Photographs |
title_short |
Estimating Tundra Vegetation Cover Using Near-surface Repeat Field Photographs |
title_full |
Estimating Tundra Vegetation Cover Using Near-surface Repeat Field Photographs |
title_fullStr |
Estimating Tundra Vegetation Cover Using Near-surface Repeat Field Photographs |
title_full_unstemmed |
Estimating Tundra Vegetation Cover Using Near-surface Repeat Field Photographs |
title_sort |
estimating tundra vegetation cover using near-surface repeat field photographs |
publisher |
ScholarWorks@GVSU |
publishDate |
2021 |
url |
https://scholarworks.gvsu.edu/gradshowcase/2021/biology/2 |
geographic |
Arctic |
geographic_facet |
Arctic |
genre |
Arctic Tundra Alaska |
genre_facet |
Arctic Tundra Alaska |
op_source |
Graduate Showcase: Education for the Future |
op_relation |
https://scholarworks.gvsu.edu/gradshowcase/2021/biology/2 |
_version_ |
1766326500856954880 |