LBN - a Perl Bayes Network Propagation

NAME

SYNTAX
DESCRIPTION
INPUTS
SUMMARY OF THEORY
FILES

SYNTAX

lbn <Ro> <Sl> <SD> <As> <HI> <choice1> <choice2>

Belief or Bayesian Networks are directed acyclic graphs in which nodes represent multi-valued variables, comprising a collection of mutually exclusive hypotheses and the arcs signify direct dependencies between the linked variables. The strength of these dependencies are quantified by conditional probabilities. lbn is using such a Baysian Networks to assess the danger of desertification, erosion and Regeneration Potential of a burned forest in the Mediterranean region by combining information coming from disparate sources with varying degrees of reliability. Can represent uncertainty by assigning probabilities to each state of the node and therefore represent our current belief in the node being in that state.

INPUTS

Specifies the surface Geology and must be in the range (1 - 100). classes : permeable, semi-permeable, impermeable Specifies the Slope derived from the DEM and must be in the range (1 - 100) classes : gentle , middle, steep Specifies the Soil Depth in the range (1 - 10) classes : bare, shallow, deep Specifies the Aspect derived from the DEM and must be in the range (0 - 360) classes : south, west/east, north Specifies the Human Influence in the range (1 - 10) classes : slightly, moderately, heavily grazed Use measurement -999 to indicate that no data are available. Specifies to output the confidences for 0:Erosion, Regeneration Potential, Desertification or 1:Erosion or 2:Regeneration Potential or 3:Desertification

SUMMARY OF THEORY

FACTORS INFLUENCING DESERTIFICATION A) Soil Erosion B) Regeneration Potential Rules: 1) The higher the risk of erosion the higher the risk of desertification 2) The lower the regeneration potential the higher the risk. A) FACTORS INFLUENCING EROSION A.1) Rock type A.2) Slope A.3) Soil Depth Rules: 1) The more impermeable the rocks that cover the site the higher the risk of erosion. 2) The steeper the slope the higher the risk of erosion. 3) The shallower the soil the higher the risk of erosion. B.) FACTORS INFLUENCING REGENERATION POTENTIAL B.1) Soil Depth B.2) Aspect B.3) Human Influence Rules: 1) The deeper the soil the higher the potential to regenerate. 2) The more the slope is oriented towards the north the higher the potential to regenerate. 3) The less the site is grazed the higher the regeneration potential.

FILES

loopfile : Is the file where the conditional probability matrix is kept. (SHOULD NOT BE CHANGED) Bnd_Var.dat : Is the file where the boundaries of its class and the variance for the measurements is kept. It has to be edited when the boundaries between the classes and the variance for the measurement have to be changed. Example Soil Depth is measures from 1 - 10. We need 3 classes (bare, shallow, deep) We use as boundaries of these classes: 0 - 3 bare 3 - 6 shallow 6 -10 deep and if uncertainty 1 is given the program computes the probabilities of SD to belong to each class. If no uncertainty is known put variance 1/100 of the range of values.