Quantifying vegetation structure and fire fuels in montane pine forests impacted by mountain pine beetle using remotely piloted aircraft system multi-spectral, photogrammetric and lidar technologies
No Thumbnail Available
Date
2025
Authors
Parsian, Saeid
University of Lethbridge. Faculty of Arts and Science
Journal Title
Journal ISSN
Volume Title
Publisher
Lethbridge, Alta. : University of Lethbridge, Dept. of Geography and Environment
Abstract
Jasper National Park, one of the oldest Canadian national parks, has a long history of fire suppression. This has increased the homogeneity of mature pine forests within the park over time, and in recent years has been severely affected by mountain pine beetle (Dendroctonus ponderosae) outbreak. Tree mortality associated with Mountain Pine Beetle (MPB), dry fuels, and the potential for fire in the Athabasca and Miette Valleys is of significant concern to the park. Measuring forest fire fuels in the field is difficult and time-consuming, and field plots do not measure all the spatial variability of fuel distribution (though field measurements are effective for measuring surface organic layers (e.g., duff layers), which are difficult to measure using remote sensing or other techniques). Remote sensing platforms, such as remotely piloted aircraft systems (RPAS), can be used to quantify 3D structures and fuels, however, RPAS platforms and sensors have not been fully explored as a tool for identifying fuels in MPB-affected stands. Therefore, the objectives of this thesis are to: 1) quantify and compare fuel attributes using RPAS-photogrammetric, lidar point clouds and 2D multispectral imagery; and 2) examine the distribution of fire fuels based on the proportion of proportion of tree mortality, likely associated with MPB outbreak phases. To answer these objectives, geographic object-based image analysis (GEOBIA) was employed to identify vegetation species, MPB phases and map coarse woody debris. Then, 3D fuel attributes were quantified using RPAS-photogrammetric and -lidar point clouds (a single wavelength collected using the Zenmuse L1 sensor (as a comparison). The results indicated that GEOBIA effectively identified tree species and achieved an overall accuracy of approximately 90% compared to field-based validation. Photogrammetric point clouds were accurate for quantifying tree structures, including tree height (R2= 0.96, Root Mean Square Error (RMSE)= 1.22 m, Normal Root Mean Square Error (NRMSE)= 6% and Bias=-0.34 m). Crown base height estimated using a novel region-based approach was also identified when compared with field-based validation (R2= 0.76, RMSE= 2.29 m, NRMSE=17%, Bias= 0.73 m). RPAS-lidar point clouds demonstrate higher accuracy in measuring tree height (R2= 0.99, RMSE= 0.59 m, NRMSE=3% and Bias= -0.23) and crown base height (R2= 0.91, RMSE= 1.32 m, NRMSE= 8% and Bias= 0.16 m) compared with measured, illustrating the value of using lidar, especially in dense canopies. RPAS point clouds demonstrated moderate accuracy for estimating crown fuel load (R2 = 0.38, RMSE = 1.92 kg. tree-1, NRMSE=16% and Bias= 0.30 kg. tree-1) Lastly, fuel distribution was assessed by comparing foliage volume and canopy fuel load and bulk density across 16 plots, and a downward shift was observed in canopy fuels over the progression of MPB outbreaks from green to gray phases. Unaffected plots (e.g., plot 15 (V = 2600 m3) had greater foliage volume and canopy fuel load than affected plots (e.g., plot 11 (V = 515 m3)).
Description
Keywords
forest fire fuels measuring , remote sensing platforms , mountain pine beetle affected pine forests , tree mortality