6 The general purpose of a model is used as a guide of management decisions
Shenk and Franklin, 2001. In case of suitable habitat modeling, it could be used to set up conservation priorities Margules and Austin, 1994 in Guisan and
Zimmermann, 2000 and may yield biological insight to explain species distribution and evaluate land activities.
At present, many algorithms have been developed to formulate habitat suitability model. Any of these algorithms is used to determine the response
variable and estimating model coefficient Guisan and Zimmermann, 2000. Most of the available algorithm families in the present used statistical approach, and
some of them used mathematical and artificial intelligence machine learning approach. The complete algorithm families can be seen in the Appendix 2.
2.2. Wildlife-Habitat Mapping
Wildlife-habitat mapping as a process to produce spatially represented wildlife information. Eventually, it has been being a complementary method in
wildlife research and management. There are several reasons why spatial information becomes a major, in contrast to the other type of information. First,
spatial perspective provides simple, but useful, framework for handling large amounts of data Fischer et al., 1996. For example, in a topographic map, we
could extract information on elevation, land cover, settlement distribution, primary infrastructures, and so forth. Therefore, it is suitable for manager and
policy makers that often do not have time to digest all information, communicate and focus to the important subjects Miller, 1994. Secondly, most of wildlife
information holds geographical dimension. It is already a geographic phenomenon since it has been lying on geographic space and interconnected
with the other geographic elements. And the third, by the capability of GIS, it is possible to have almost complete picture of situation in the concerned place by
linking all relevant geographic features in various types into one single frame, known as overlay or superimposition. Hence, GIS is very useful tools for
studying wildlife that inherently influenced by numerous ecological elements in space and time.
7
2.3. Spatial Multi Criteria Decision Analysis
Mostly decision is made under many criterions criteria, rarely depend on single criterion. The term multi criteria decision making analysis is used to
explain the decision methodology which involve incommensurate criteria. Based on Worral 1991 in Malczewski 1999, 80 of data used by
managers and decision makers is related geographically. Furthermore, spatial decision is multi criteria in nature. Consequently, there is a need for utilizing
spatial multi criteria decision analysis that dealt with geographical data, which previously used for aspatial problem.
Malczewski 1999 thought spatial multicriteria decision analysis as a process that combines and transforms geographical data into resultant decision
through decision rules multi criteria decision procedures. At least three components should be considered to utilize multi criteria decision making, i.e.:
decision criteria, criteria’s weight, and decision proceduresrules. The following subsection describes these components.
2.3.1. Criteria
Criteria are also referred as attributes. In spatially based decision making, the attribute is represented by concerned map. This map is called as an evaluation
criteria map, attribute map, thematic map or data layer, which defined as a unique geographical attribute of the alternative decisions that can be used to evaluate the
performance of the alternatives Malczewski, 1999.
2.3.2. Constraints
Constraint is limitation imposed by nature or by human being that do not permit certain actions to be taken Keeney, 1980 in Malczewski, 1999. It is
important in determining the feasibility of the alternatives of decision. If decision is made which failed to take account the constraints, then the decision will be
called infeasible or unacceptable decision.
8
2.3.3. Criteria Weighting
The purpose of criterion weighting is to express the importance of each criterion relative to other criteria Malczewski, 1999. In decision making, it is a
common that each criterion is incommensurate or each criterion has influence level to decision result. There are some procedures to estimate weight, for
example: rating, ranking, pair wise comparison method, trade-off analysis method, and so forth Malczewski, 1999.
2.3.4. Decision Rules
A decision rule is a procedure that allows for ordering alternatives. It integrates the data and information on alternatives and decision maker’s
preferences into an overall assessment of the alternatives Malczewski, 1999. There are numerous decision rules that can be used for solving the multi
criteria decision making problem. Simple additive weighting SAW methods are the most often used techniques for tackling spatial attribute decision making. The
techniques are also referred to as weighted linear combination WLC or scoring methods
Malczewski, 1999. The following formula is the formal model of SAW method Malczewski,
1999:
∑
=
i ij
i i
x w
A 2.1
where x
ij
, is the score of the i
th
alternative with respect to the j
th
attribute, and w is the normalized weight. The most preferred alternative is selected by identifying
the maximum value of A
i
i = 1, 2, 3… m number of attributescriteria.
2.4. The Use of Principal Component Analysis in Ecological Studies
It is a nature that process and interaction of ecosystem involve many factors biotic and abiotic. These interactions sometimes synergic, which can be
strengthening or weakening, balancing or reinforcing, dynamically affect the behavior, distribution, and abundance of an organism. Hence, understanding the
species response solely on the single variable or factor could be misleading. In fact, many scientists already knew the complexity of ecosystem processes.
9 However, it is remarkable of which after the niche theory was reformulated by
Hutchinson 1957, ecological research especially in wildlife-habitat relationship studies concerns many variables Morrison et al., 1992. One approach to
consider many variables simultaneously in explaining organism responses is multivariate statistics.
According to McGarigal et al. 2000, Principal Component Analysis PCA is the most well-understood and widely used ordination technique. They stated
that ordination essentially seeks to uncover a more fundamental set of factors that account for the major patterns across all of the original variables. They realized
the principle behind using ordination in ecology is that much of the variability in a multivariate ecological data set often is concentrated on relatively few dimension,
and that these major gradients are usually highly related to certain ecological or environmental factor. The important characteristic of PCA in the following is
summarized after McGarigal et al. 2000: • PCA assesses relationship within a single set of interdependent variables,
regardless any relationship they may have to variables outside the set. • PCA does not attempt to define the relationship between a set of independent
variables and one or more dependent variables. • Its main purpose is to condense information contained in a large number of
original variables into smaller set of new composite dimensions, with a minimum loss of information, where each dimension is defined by weighted
linear combination of the original variables called principal components. This linear combination represents gradients of maximum variation within the
data set. PCA usually begins with standardize the raw data set mean centering.
From this matrix, the correlation matrix or variance-covariance matrix is calculated. After that, eigenvalues and eigenvector, the most important quantity
in PCA, are obtained through decomposing matrix and QR Algorithm technique Wikipedia, 2005.
According to Morrison et al. 1992, eigenvectors are best combinations of correlated predictor variables that account for most of the variation in the response
variable, whereas the total variance among observations attributed to each
10 component is measured in eigenvalue. The weights of linear equation of each
principal component PC are actually these eigenvectors.
2.5. Javan Gibbon
