This story highlights a current research project at the Hubbard Brook Experimental Forest. To read more about research projects at HBEF, visit Current Research page. Check back regularly to learn about new research projects.
Assessing ice storm damage using Landsat TM imagery


  Contact Info:
  James S. Burnett, PhD
Vermont Forestry Associates, Inc.
Fairlee, VT 05045
phone: (802) 333-4129

SINCE 1967 we have conducted isolated collections of wood fern (Dryopteris spinulosa) fronds to determine plant nutrient concentrations. Wood fern is the most abundant herbaceous species at the Hubbard Brook Experimental Forest, comprising 40 to 50% of the biomass of this stratum. (Lycopodium lucidulum, an evergreen clubmoss, makes up another 40 to 50%, but only 10% of its biomass consists of new growth each year.) Most of our early efforts at Hubbard Brook were focused on nutrient cycling through the tree species. However, occasional samples of wood fern were collected and analyzed to get a rough idea of how the herbaceous layer contributed to nutrient cycling in the ecosystem. In 1985 we started a more consistent annual collection of wood fern from an area just west of the Watershed 6 weir. The annual collection of wood fern fronds was initiated on Watershed 1 in 1996. These quick and simple collections were not part of a larger study, but were conducted simply to monitor gross long-term patterns in plant chemistry. As we discovered in the summer of 2000, it is important to maintain these types of long-term datasets to get an idea of the magnitude of annual variability.

  Wood fern  
  Wood fern Dryopteris spinulosa


Methods of collection and analysis
Features of vegetation and forest canopies can be explored using spectral signatures derived from spaceborne remote sensing instruments, such as Landsat Thematic Mapper (TM). These signatures can in turn be used to measure vegetation health and condition. Images displayed using the Landsat TM 5/4 (1.65/0.83 mm) ratio, for example, have been found to clearly distinguish between high and low levels of canopy damage areas to hardwood forests (Vogelmann and Rock, 1989). In this study the ability of Landsat TM imagery, using the TM 5/4 index, to map three levels of ice storm damage was tested. Nine target areas (1 km x 1 km) selected in the White Mountain National Forest represented the forest canopy conditions prior to and following the January 1998 ice storm. Three of these target areas were in the Hubbard Brook Experimental Forest. Sixty-six, 90m x 90m sites representing light, moderate, and severe ice damage were selected from the 9 target areas. TM 5/4 (TM5 = 1,550-1,750 nm; TM4 = 760-900 nm) was used as a moisture stress index to differentiate levels of damage. The working hypothesis was that the foliage of healthy canopies contained more moisture per unit area and could be distinguished from gaps in the canopy caused by ice storm damage that contained less moisture.

Change detection was applied to pre- and post-event TM scenes to detect levels of change in the moisture content of forest canopies. Pixel-by-pixel subtraction revealed the difference in the canopy condition between scenes. Figure 1 shows pre- and post storm scenes with pink areas indicating low moisture content. In 1998, after the ice storm, new pink areas are associated with ice storm damage.

Ice damaged sites representing three levels of damage were measured on the ground. The spectral characteristics of these sites, derived from the TM 5/4 ratio, were used to classify other areas within the TM imagery. The Tukey test results (Figure 2) demonstrate the significant difference (clear separation between circles) between damage classes when change detection is applied to pre- and post-storm scenes. Figure 3 shows an error matrix analysis for the classification map of the change detection data. The overall accuracy of the classification map was 78%. These results demonstrate that Landsat TM imagery can be used to map levels of ice damage.

Vogelmann, J.E. and B.N. Rock, 1989. Use of Thematic Mapper data for the detection of forest damage caused by the pear thrips. Remote Sensing of Environment, 30:217-225.


  Landsat image pre- ice storm Landsat image pre- ice storm  
  Landsat 5 TM - 1996
• Path 13, Row 29
• Acquired 8/30/96
• Radiometrically Corrected
• Geometrically Corrected
• 10% Cloud Cover
• 30-m pixels
R – TM5
G – TM4
B – TM5
• Path 13, Row 29
• Acquired 8/20/98
• Radiometrically Corrected
• Geometrically Corrected
• 0% Cloud Cover
• 30-m pixels
R – TM5
G – TM4
B – TM5at 5 TM - 199
  Figure 1. Landsat TM imagery for pre- and post-ice storm event in the Hubbard Brook Experimental Forest, Thornton, NH. Three features are labeled on the second image: 1 is the main channel of Hubbard Brook, 2 is Mirror Lake, and 3 is Watershed 5, whole-tree harvested in 1983.    
  Graph of statistical test results  
  Figure 2. One-way ANOVA and Tukey test results for TM 5/4 1998-1996 change detection by damage, 90 x 90 m sites, n=66.  
  Table of error matrix analysis  
  Figure 3. Error matrix analysis for classification map of change detection data.  

Date Prepared: January 2003