Semester

Fall

Date of Graduation

2003

Document Type

Dissertation

Degree Type

PhD

College

Eberly College of Arts and Sciences

Department

Biology

Committee Chair

James B. McGraw.

Abstract

A population of eastern hemlock (Tsuga canadensis L.) was censused from the ground using traditional field methods and from the air using large scale, high-resolution, aerial imagery in the early spring of 1997, 1998 and 1999. A manual crown survey map of the population, prepared from aerial imagery, was compared to a traditional field census. Over 60% of the individuals measured on the ground were not detected in the aerial census. Tree size, crown density and crown position all played roles in determining a crown's visibility from the air. Nearly all large, upper canopy hemlocks were visible in the aerial census. An important minority of small, lower canopy hemlocks were also visible in the aerial census. An automated spatial segmentation procedure was developed to identify and measure individual population units, or blobs, within the forest population. A blob was defined as a distinct portion of crown segmented from its neighbors on the basis of size, shape, and connectivity. To ensure the comparability of multi-year segmentation maps, an automated blob reconciliation procedure was also developed to make certain that no hemlock pixels were assigned to different blobs in different years. Following spatial segmentation and reconciliation, a large majority of hemlock blobs (∼64--72%) were found to be closely associated with ground referenced, manually delineated, individual hemlock crowns. The remaining blobs consisted of spatially distinct parts of a crown or closely clumped multiple crowns. Matrix population models were constructed from the ground-derived and aerial-derived population data. Matrix analysis produced a number of useful population characteristics including overall population growth rate (lambda), stable stage distributions, reproductive values, and sensitivity values. lambda's calculated from the aerial and ground-derived matrices were compared using randomization tests. While providing a different perspective and description of a population than traditional ground studies, demographic studies using remote sensing provide some promising advantages. The spatially explicit nature of the data permits more biologically realistic modeling of the population and the investigation of potential environmental influences on population dynamics. Automated extraction of demographic or megademographic data from remotely sensed images represents an important first step toward scaling population analysis to the landscape and regional levels.

Share

COinS