Digital aerial photogrammetry for assessing cumulative spruce budworm defoliation and enhancing forest inventories at a landscape-level

作者：Goodbody Tristan R H^*; Coops Nicholas C; Hermosilla Txomin; Tompalski Piotr; McCartney Grant; MacLean David A

来源：ISPRS Journal of Photogrammetry and Remote Sensing, 2018, 142: 1-11.

DOI：10.1016/j.isprsjprs.2018.05.012

• Digital aerial photogrammetry
• Forest monitoring
• Spruce budworm
• Cumulative defoliation
• Enhanced forest inventory
• Partial least squares

Spruce budworm (Choristoneura fumiferana [Clem.], Lepidoptera: Tortricidae) is a native defoliating insect with an important disturbance role in the eastern boreal forests of North America. With an extensive history of outbreaks and associated impacts on forest structural changes and timber supply, the mapping of spruce budworm defoliation has been of major management importance. In this study we assessed the ability of high spatial resolution digital aerial photogrammetric (DAP) data to predict cumulative defoliation as well as basal area and merchantable volume in spruce budworm host stands in the Gordon Cosens Forest south of Kapuskasing, Ontario, Canada. To do so, DAP derived structural and spectral metrics were incorporated to implement a stratified sampling design to improve the efficiency and cost-effectiveness of field surveying. Standard forest inventory measurements including diameter and height, as well as ocular and branch level defoliation assessments were undertaken on thirty 400 m(2) radius plots. A partial least squares analysis was performed to determine whether structural metrics from a DAP point cloud could be influenced by spruce budworm defoliation, as well as determine the relative effectiveness of spectral (e.g. mean NDVI) vs. structural (e.g. 90th percentile of height) metrics, or their combination, for predicting cumulative defoliation. Results indicated that spectral metrics were the most effective for predicting cumulative defoliation (R-2 = 0.79), while structural metrics were the least effective (R-2 = 0.49). Metrics characterizing variance of the spectral values were found to be the most important predictors. Structural metrics and linear regression were also used to estimate landscape-level volume and basal area per hectare yielding R-2 = 0.80 and R-2 = 0.90, respectively. Outcomes of this analysis indicate that DAP-derived spectral metrics were more capable of modeling cumulative defoliation, while structural metrics were effective for landscape-level estimations of standard forest inventory attributes. This analysis indicated that the provision of both spectral and structural metrics from a single aerial imagery survey has potential to enhance defoliation monitoring and forest attribute modeling at a landscape-level.

• 出版日期2018-8