Print

Weverka, Aiko S.

Natural Resources

2012

MS

Remote sensing can provide a relatively low-cost and low-impact approach to large scale assessment of forest condition and productivity over time. However, the connection between canopy spectral signatures and scalable field metrics is not well understood. To explore this relationship, we compared annual basal area increment (BAl) at 47 sites throughout northern Vermont and New Hampshire to a suite of vegetation indices derived from annual growing season Landsat 5 TM imagery. Correlation analysis was used to evaluate the relationship between annual BAI and these indices at each site from 1984-2010, and a stepwise multiple linear regression model was created to predict BAI growth using a combination ofmultiple indices. Results showed weak significant relationships between BAI and several vegetation indices (mean I p I=0.104 ± 0.032) and that relationships between BAI and vegetation indices do not hold within most sites (<38%). The linear regression model created to predict BAI growth used a combination of four vegetation indices (r²=0.120,p<0.0001), although average residuals were high (mean standard error =24.34) and significantly varied by species type (p <0.001, F =58.07).
These results indicate that while tracking relative changes in productivity is possible and more likely to be successful when species-specific relationships are examined, using remote sensing teclmiques for precise growth monitoring and accurate carbon accounting may be of limited value. The relationship between BAI, canopy characteristics and remotely sensed metrics at the plot level is likely nuanced, and complicated by heterogeneous species composition, variability in tree response to abiotic and biotic stressors, and the inability of single date of imagery to characterize the quality of an, entire growing season. While many have utilized remote sensing to quantify landscape scale productivity, the resulting models should be viewed conservatively.