Quantifying and mapping tree mortality due to mountain pine bark beetles via analyses of remote sensing data in northern Colorado
Date
2024
Journal Title
Journal ISSN
Volume Title
Abstract
In the past two decades, Mountain Pine Bark Beetle (MPBB) infestations have become more pervasive due to increasing temperatures and drought conditions related to climate change causing regional-scale mortality. Insect effects on tree die-off, fuels, and fire behavior can vary widely. A key problem in understanding insect-fire relationships is the lack of empirical maps that show interrelated changes in the distribution of insect infestations and fire zones over space and time. This study demonstrates an approach to tracking and mapping the spread of MPBB by analyses of vegetation indices calculated from Landsat TM data in the study site in northwestern Colorado. These indices were used for calculations in the Random Forest (RF) classifier algorithm and the Support Vector Machine (SVM) classifier algorithm to determine the presence or absence of MPBB and to illustrate the changes in the distribution of infestations with time. A comparison was made between the accuracy of the two classification algorithms (RF and SVM) in tracking and mapping the spread of MPBB. R2 has proved to be a reliable measure of accuracy of regression models. If the statistical accuracy of all the models, (RF vs. SVM and binary vs. regression) are compared, both the regression and binary models based on RF are more accurate. The results of this study can provide a useful tool for forest managers to make decisions about how changing conditions affect potential problems in forest management.
Description
Rights Access
Embargo expires: 08/16/2025.
Subject
Landsat data analysis
random forest classifier
time series
mountain pine park beetles
binary vs. regression analysis
support vector machine classifier