Mapping Mountain Pine Beetle Mortality through Growth Trend Analysis of Time-Series Landsat Data
Disturbances are key processes in the carbon cycle of forests and other ecosystems. In recent decades, mountain pine beetle (MPB; Dendroctonus ponderosae) outbreaks have become more frequent and extensive in western North America. Remote sensing has the ability to fill the data gaps of long-term inf...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2014-06-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | http://www.mdpi.com/2072-4292/6/6/5696 |
| _version_ | 1828137686380052480 |
|---|---|
| author | Lu Liang Yanlei Chen Todd J. Hawbaker Zhiliang Zhu Peng Gong |
| author_facet | Lu Liang Yanlei Chen Todd J. Hawbaker Zhiliang Zhu Peng Gong |
| author_sort | Lu Liang |
| collection | DOAJ |
| description | Disturbances are key processes in the carbon cycle of forests and other ecosystems. In recent decades, mountain pine beetle (MPB; Dendroctonus ponderosae) outbreaks have become more frequent and extensive in western North America. Remote sensing has the ability to fill the data gaps of long-term infestation monitoring, but the elimination of observational noise and attributing changes quantitatively are two main challenges in its effective application. Here, we present a forest growth trend analysis method that integrates Landsat temporal trajectories and decision tree techniques to derive annual forest disturbance maps over an 11-year period. The temporal trajectory component successfully captures the disturbance events as represented by spectral segments, whereas decision tree modeling efficiently recognizes and attributes events based upon the characteristics of the segments. Validated against a point set sampled across a gradient of MPB mortality, 86.74% to 94.00% overall accuracy was achieved with small variability in accuracy among years. In contrast, the overall accuracies of single-date classifications ranged from 37.20% to 75.20% and only become comparable with our approach when the training sample size was increased at least four-fold. This demonstrates that the advantages of this time series work flow exist in its small training sample size requirement. The easily understandable, interpretable and modifiable characteristics of our approach suggest that it could be applicable to other ecoregions. |
| first_indexed | 2024-04-11T18:23:48Z |
| format | Article |
| id | doaj.art-00d554a7beae41a48b8aacb96dca996e |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-04-11T18:23:48Z |
| publishDate | 2014-06-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-00d554a7beae41a48b8aacb96dca996e2022-12-22T04:09:40ZengMDPI AGRemote Sensing2072-42922014-06-01665696571610.3390/rs6065696rs6065696Mapping Mountain Pine Beetle Mortality through Growth Trend Analysis of Time-Series Landsat DataLu Liang0Yanlei Chen1Todd J. Hawbaker2Zhiliang Zhu3Peng Gong4Department of Environmental Science, Policy and Management, University of California, Berkeley, CA 94720, USADepartment of Environmental Science, Policy and Management, University of California, Berkeley, CA 94720, USAGeological Survey, P.O. Box 25046, DFC, MS 980, Denver, CO 80225, USAGeological Survey, 12201 Sunrise Valley Drive, Reston, VA 20192, USADepartment of Environmental Science, Policy and Management, University of California, Berkeley, CA 94720, USADisturbances are key processes in the carbon cycle of forests and other ecosystems. In recent decades, mountain pine beetle (MPB; Dendroctonus ponderosae) outbreaks have become more frequent and extensive in western North America. Remote sensing has the ability to fill the data gaps of long-term infestation monitoring, but the elimination of observational noise and attributing changes quantitatively are two main challenges in its effective application. Here, we present a forest growth trend analysis method that integrates Landsat temporal trajectories and decision tree techniques to derive annual forest disturbance maps over an 11-year period. The temporal trajectory component successfully captures the disturbance events as represented by spectral segments, whereas decision tree modeling efficiently recognizes and attributes events based upon the characteristics of the segments. Validated against a point set sampled across a gradient of MPB mortality, 86.74% to 94.00% overall accuracy was achieved with small variability in accuracy among years. In contrast, the overall accuracies of single-date classifications ranged from 37.20% to 75.20% and only become comparable with our approach when the training sample size was increased at least four-fold. This demonstrates that the advantages of this time series work flow exist in its small training sample size requirement. The easily understandable, interpretable and modifiable characteristics of our approach suggest that it could be applicable to other ecoregions.http://www.mdpi.com/2072-4292/6/6/5696Landsatmountain pine beetletime-series classificationtemporal segmentationdecision treesample size |
| spellingShingle | Lu Liang Yanlei Chen Todd J. Hawbaker Zhiliang Zhu Peng Gong Mapping Mountain Pine Beetle Mortality through Growth Trend Analysis of Time-Series Landsat Data Remote Sensing Landsat mountain pine beetle time-series classification temporal segmentation decision tree sample size |
| title | Mapping Mountain Pine Beetle Mortality through Growth Trend Analysis of Time-Series Landsat Data |
| title_full | Mapping Mountain Pine Beetle Mortality through Growth Trend Analysis of Time-Series Landsat Data |
| title_fullStr | Mapping Mountain Pine Beetle Mortality through Growth Trend Analysis of Time-Series Landsat Data |
| title_full_unstemmed | Mapping Mountain Pine Beetle Mortality through Growth Trend Analysis of Time-Series Landsat Data |
| title_short | Mapping Mountain Pine Beetle Mortality through Growth Trend Analysis of Time-Series Landsat Data |
| title_sort | mapping mountain pine beetle mortality through growth trend analysis of time series landsat data |
| topic | Landsat mountain pine beetle time-series classification temporal segmentation decision tree sample size |
| url | http://www.mdpi.com/2072-4292/6/6/5696 |
| work_keys_str_mv | AT luliang mappingmountainpinebeetlemortalitythroughgrowthtrendanalysisoftimeserieslandsatdata AT yanleichen mappingmountainpinebeetlemortalitythroughgrowthtrendanalysisoftimeserieslandsatdata AT toddjhawbaker mappingmountainpinebeetlemortalitythroughgrowthtrendanalysisoftimeserieslandsatdata AT zhiliangzhu mappingmountainpinebeetlemortalitythroughgrowthtrendanalysisoftimeserieslandsatdata AT penggong mappingmountainpinebeetlemortalitythroughgrowthtrendanalysisoftimeserieslandsatdata |