Search Filter
- Publications
- Abstract of Theses and Dissertations
- Database
- Integrating information technology and local knowledge to detect and...
Integrating information technology and local knowledge to detect and measure forest degradation
Dissertation Abstract:
Forest degradation, in the context of carbon emission, is difficult to detect and measure. Forest degradation can be defined as “persistent reduction of carbon stocks due to human activities but which does not result in the conversion of forest to nonforest land”. Changes within forest may not easily be detectable via usual remote sensing (RS) approaches. While remote sensing data remains to be the primary material for forest degradation assessment, other sources of data must be explored to enhance detection and estimates. In this study, remote sensing, high-resolution Google Earth imagery (HRGEI), GIS ancillary data, field measurements and observations, and local knowledge were integrated to assess forest degradation. Local knowledge refers to information about the forest structure and the changes it has undergone over a specified time period.
The case study area was a landscape, which land cover has been transformed from predominantly closed canopy forest to fragmented forest due to agriculture. RS (Landsat TM and ETM+), HRGEI, and GIS ancillary data were designated as spatial information technology-based datasets. Local knowledge was collected from individuals whose familiarly with the area of interest was established on the basis of land ownership, length of stay, and actual use of the land. Use of local knowledge was centered on reference land cover polygons, which have great value in accurately assessing land cover classification.
GIS served as tool and medium for integrating the various datasets. A new approach called vecter approach was explored. Vecter was coined from vector and raster—the two data formats in GIS. It requires transformation of all spatial data into a grid cell format.
Aboveground biomass (AGB) was used as proxy to carbon stock. In vecter technique, a land cover classification scheme is needed to show change in carbon stock over a specified time period, where a change from low-carbon category to high-carbon category represents forest degradation.
Results from the vecter models were compared with those of regression models. A total of 59 spectral variables were used to create AGB models. The resulting 15 models had widely varying estimates and most of them performed poorly against ground data and expected map output. The model whose independent variable was only the fourth band of the tasseled cap transformation had the closest landscape AGB estimate with that of the vecter model.
Over a 20-year period, the landscape has lost approximately 27 percent of its AGB (125,000 tons), which translates to around 56,250 tons of carbon. Credibility and integrity of this estimate is vouched for by high overall accuracy ratings of land cover classification for all datasets, robust field measurements, and spatial information based on local knowledge, and agreement of land cover information between Landsat data-based analysis and HRGEI-based analysis.
