Estimation of Insect Infestation Dynamics Using a Temporal Sequence of LANDSAT Data

The current outbreak of mountain pine beetle (Dendroctonus ponderosae) in western Canada has been
increasing over the past decade and is currently estimated to be impacting 9.2 million hectares, with varying
levels of severity. Large area insect monitoring is typically undertaken using manual aerial overview sketch
mapping, whereby an interpreter depicts areas of homogenous insect attack conditions onto 1:250,000 or
1:100,000 scale maps. These surveys provide valuable strategic data for management at the provincial scale.
The coarse spatial and attribute resolution of these data however, make them inappropriate for fine-scale
monitoring and operational planning. For instance, it is not possible to estimate the initial timing of attack
and year of stand death. In this study, we utilise eight Landsat scenes collected over a 14 year period in northcentral
British Columbia, Canada, where the infestation has gradually developed both spatially and
temporally. After pre-processing and normalising the eight scenes using a relative normalisation procedure,
decision tree analysis was applied to classify spectral trajectories of the Normalised Difference Moisture
Index (NDMI). From the classified temporal sequence of images, key parameters were extracted including the
presence of beetle disturbance and timing of stand decline. The accuracy of discriminating beetle attack from
healthy forest stands was assessed both spatially and temporally using three years of aerial survey data
(1996, 2003, and 2004) with results indicating overall classification accuracies varying between 71 and 86%.
As expected, the earliest and least severe attack year (1996), recorded the lowest overall accuracy. The
relationship between the timing of stand attack (i.e. moderate to severe beetle infestation) and NDMI (initial
year of detected disturbance) was also explored. The results suggest that there is potential for deriving
regional estimates of the year of stand death using Landsat data and decision tree analysis however, a higher
temporal frequency of images is required to quantify the timing of mountain pine beetle attack.