Spatio-temporal pattern analysis of managed forest landscapes: a simulation approach

This study dealt with research problems at thelandscape level. The objectives of this thesis were todevelop tools to study and characterize landscapes and tointerface with a geographic information system (GIS), toevaluate landscape indices, and to examine development offorest cutting patterns under different cutting methods andexplore alternative forest management strategies.
A computer program was developed for simulation andanalysis of landscape patterns. The primary applicationsof the computer program were (1) to quantify spatialpatterns of landscapes, (2) to perform experiments withdifferent silvicultural strategies and forecast theconsequences of management activities, (3) to examine thebehavior of landscape indices without having a large numberof landscape samples, (4) to interface with and tocomplement GIS in terms of ecological analysis, and (5) to serve as a base on which GIS-related landscape models couldbuilt.
Many extant landscape indices were reviewed, and somenew indices proposed. Each was evaluated in terms of itsability to distinguish four test synthetic landscapes withdistinct spatial patterns. Fractal dimension, patchinessindex, dispersal index, and two fragmentation indices(i.e., the forest interior area and the largest forestpatch size) appeared to be most sensitive to spatialvariations among the test landscape mosaics, and may bemost useful to study and quantify the landscape pattern.On the other hand, some commonly-used landscape indices,contagion and dominance, could not distinguish variationsin distinct landscape patterns.
The simulation program and the landscape indices werethen used to study landscape patterns generated bydifferent forest cutting methods. The results indicatedthat different cutting designs may produce landscapes withdistinct characteristics. Landscapes were clearly lessfragmented when larger sizes of cut-units were used. Whena stream system was included in the landscape structure,the behavior of many landscape characteristics changed.The results suggested that simple landscape models (i.e.,the checkerboard model and random model) may lead tomisleading interpretations of landscape patterns.