SPATIO-TEMPORAL PREDICTIVE MODELS OF
MEDITERRANEAN VEGETATION DYNAMICS
Yohay Carmel(1), Ronen Kadmon(1), and Ronit Nirel(2)
(1)Department of Evolution, Systematics and Ecology
(2)Department of Statistics
The Hebrew University of Jerusalem
Jerusalem 91904, Israel
Empirical modeling of vegetation dynamics can be used for
predictive purposes. The goal of the present study is to construct and evaluate possible
approaches for empirical modeling of vegetation dynamics, and investigate
their potential use in planning and management. Specifically we studied three major
issues: 1. Finding an optimal spatial resolution for a vegetation change database, that
would maximize both positional accuracy and resolution; 2. Assessing the performance of
linear models, and the performance of stochastic vs. deterministic realizations of a
logistic model; and 3. Evaluating two different approaches to cope with spatial
autocorrelation, namely the use of a small sample of the database and the introduction of
An empirical model of Mediterranean vegetation dynamics was constructed using a case study
of vegetation change in an area in the Galilee mountains, northern Israel, over 28 years.
Present vegetation in any location was modeled as a function of past vegetation and
environmental factors (e.g. topography, and various disturbances); future vegetation
pattern was then modeled as a function of current vegetation and effects of environmental
factors. In order to assess
model performance, the actual vegetation map was compared with maps
representing model realizations, for the study area and for an external validation area.
Three types of measures were used to compare the predicted and actual vegetation maps:
overall vegetation composition, pattern indices and cell-by-cell match.
Our results indicate that landscape-scale vegetation dynamics can be fairly well modeled,
using few biologically important variables. The logistic and linear models had similar
performance, in spite of the reduced information on which the logistic models were based.
The use of only a 4% sample of the database resulted in a negligible reduction in model
performance. Model performance was reduced - but was still fair - when applied to an
Vegetation Dynamics, Predictive modeling, Empirical models, GIS,
Mediterranean vegetation, Aerial photographs, Succession,
Polychotomous logistic model, Autocovariate.
Department of Forest Sciences
Colorado State University
Fort Collins, CO 80523
Phone: (970) 491-3650
Fax: (970) 491-6754
[ To Publications ]