Monitoring hemlock forest health in New Jersey using Landsat TM data and change detection techniques
1997
Royle, D.D. | Lathrop, R.G.
Defoliation of eastern hemlock (Tsuga canadensis Carriere) forest caused mainly by the hemlock woolly adelgid (Adelges tsugae Annand) was detected, quantified, and mapped for a 1,267 km2 study area in the New Jersey Highlands using anniversary dates of Landsat Thematic Mapper data (1984 and 1994). A model relating estimates of canopy condition to the temporal difference in near infrared/red reflectance (i.e., the vegetative index difference) was developed to predict and map four classes of hemlock condition across the study area. Data from 105 circular ground plots (90 m diameter) were used to develop the regression model, while data from 50 plots were reserved for accuracy assessment. The vegetative index difference was highly correlated to hemlock damage as measured on the ground (R2 = 0.73). Lightly defoliated hemlock canopy did not differ spectrally from healthy hemlock; thus, these two classes were joined together. Accuracy assessment showed that hemlock condition can be predicted within one-half damage class with an overall accuracy of 64% for four damage classes, 70-72% for three classes, and 78-92% for two classes. Of the 7,735 ha of hemlock forest in 1984, 47% remained healthy to lightly defoliated, 44% had experienced moderate to severe defoliation, and 9% were dead by 1994.
显示更多 [+] 显示较少 [-]