Author

Suming Jin

Date of Award

8-2005

Level of Access

Campus-Only Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Forest Resources

Advisor

Steven A. Sader

Second Committee Member

Thomas B. Brann

Third Committee Member

William Halteman

Abstract

The research applied new methods that integrated remote sensing and other spatial data bases to answer questions about changes in the forests of northern Maine during the past three decades. Normalized Difference Moisture Index (NDMI) and Tasseled Cap Wetness (TCW) omission and commission errors were compared in detecting forest type disturbances and harvest intensity at five, two, and one year Landsat acquisition intervals. The NDMI and TCW were highly correlated (>0.95 r2) for all five image dates. No significant difference existed between NDMI and TCW for detecting forest disturbances. However, the lesser known NDMI performed slightly better when images were collected less than two years apart. Forest ownership maps from 1994, 2000, and 2004 were analyzed with forest cover (early 1990s) and NDMI forest change detection maps derived from Landsat imagery (1991, 2000, and 2004) to quantify the relationship between forest harvest pattern, ownership type, and ownership change. Industrial landowners harvested the highest percentage of forest in the 1980s. In the 1990s and early 2000s, Timber Investment Management Organizations and Logger/Short-term Investors harvested a higher percentage of forest than other owners. The estimated average disturbance rotation on the Industrial ownerships was 51 years, compared to 70 years for Non-Industrial Private Forest lands. Landowners showed a general preference to harvest softwood and softwood-hardwood in the 1980s, softwood-hardwood and hardwood-softwood stands in the 1990s, and nearly a balanced proportion of four forest types available in the study area between 2000 and 2004. Finally, the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor was evaluated to quantify detection accuracy associated with disturbed forest patch size. The MODIS data had good overall classification accuracy (agreement around 85%) with the Landsat change detection database used for reference. MODIS change detection maps had approximately 70% and 90% detection accuracy when the disturbed patch size was larger than 20 ha and 50 ha, respectively. The MODIS forest area change estimates precision compared favorably to change estimates derived from U.S. Forest Service, Forest Inventory Analysis data. MODIS data showed potential for statewide forest change detection applications.

Share