Tracking Advance of the Tree Line on the New Hampshire Presidential Ridge Introduction Question Results Remote sensing Has the tree line advanced on The NDVI data show that the change detection tech- the Presidential Ridge? amount of heavy vegetation is in- nology was can track creasing, while the areas of light the effects of climate Which classification method vegetation and no vegetation are change on alpine eco- best tracks tree line migration? decreasing. The MLC data show systems. This study that areas of both heavy and light focuses on changes in Legend 1986 Image 2004 Image vegetation are increasing, while the tree line of the areas of no vegetation are decreas- Heavy Vegetation All analyses performed in this project were based on the Presidential ridge in above images. These images were taken by Landsat 4-5 ing. The accuracy of the MLC the White Mountains Light Vegetation TM in the month of August. classification was greater than the of New Hampshire. No Vegetation accuracy of the NDVI classifica- What is the tree line? tion. The tree line is the Study Area point at which trees Accuracy of Classification Method can no longer grow, Classification due to inhospitable climate. At a certain elevation, Method 1986 2004 which varies by geographical region, trees cannot sur- NDVI 77% 71% NDVI Binary 93% 96% vive the conditions and rocky soil is exposed. This MLC 98% 95% rocky zone above the tree line is known as the alpine MLC Binary 100% 100% tundra. As global climate increase, the tree line ad- vances to areas of higher elevation. 1986 NDVI 2004 NDVI 1986 MLC 2004 MLC Why is tree line important? Tree line shift is an effective way to measure cli- Conclusion NDVI Change Detection MLC Change Detection mate change. Further, tree line has great biological % Sq. Mi. Sq. Mi. significance. Classification Classification % Change Regardless of classification Change Change Change 1986 classification overlaid with 2004 alpine tundra Many species of method, the tree line is advancing. Heavy Vegetation 42.1 8 Heavy Vegetation 3.5 1.1 plant and animal Legend More work is needed to determine Light Vegetation -37 -6.4 Light Vegetation 17.7 1.29 can only survive in if this shift is caused by climate No Vegetation -21.1 -1.3 No Vegetation -37.6 -2.39 2004 No Vegetation specialized envi- change or normal climate fluctua- ronmental niches, NDVI Binary Change Detection MLC Binary Change Detection 1986 No Vegetation tions. By comparing the NDVI % Sq. Mi. % Sq. Mi. Classification Classification and the MLC methods, we see im- such as the alpine Change Change Change Change 1986 Light Vegetation proved classification accuracy tundra. A large- Vegetation 4.4 1.6 Vegetation 6.1 2.39 1986 Heavy Vegetation scale advance of No Vegetation -21.1 -1.3 No Vegetation -37.4 -2.39 with the supervised technique. the tree line will likely result in the severe endanger- ment and possible extinction of many species. Kathryn Bond How can the tree line be studied? Intro to Remote Sensing Most commonly, a Normalized Difference Vegetative Methods gories of land cover: heavy vegetation, light vegetation, and Post Classification Spring 2010 Index (NDVI) is used to analyze changes in the tree To track the advance of the tree line, data from Au- no vegetation. For each category, training sites consisting Once the images were classified, the data were compared gust 1986 was compared with data from August 2004. Data: Landsat 4-5 TM, from line. However, a supervised classification may be able of 8 polygons and over 200 points were created. The classi- using the change detection function of ENVI. First, com- Data Processing to provide more accurate results. In this analysis, both fication was run using ENVI. parison of the two dates was conducted using all three land USGS The Landsat TM 4-5 data were obtained from the 2. NDVI classification. This unsupervised classification classes. Next, the light and heavy vegetation classes were of these methods of land classification will be used to USGS. After being downloaded, the layers were compares the reflectance of the red band and the near infra- combined and a further binary change detection analysis track changes in vegetation. By using two methods, stacked and then clipped to the size of the study area. red band. After NDVI classification, three classes based was done to compare areas with vegetation to areas without For the NDVI analysis, data were converted into re- the overall reliability of the results may be assessed. on the NDVI value of each pixel were created: vegetation. In this way, the results of the MLC and the flectance. -1 to 0.49: No vegetation NDVI classification methods may be directly compared de- Further, the two methods will be compared on their ef- Classification 0.49 to 0.65: Light vegetation spite differences in the classification criteria. fectiveness to evaluate changes in the tree line. 1. Maximum Likelihood classification. This super- 0.65 to 1.00: Heavy Vegetation vised classification was done by creating three cate-