Rule-based reasoning

In this network, each circle represents an influencing input factor. Each box represents an attribute that might be influenced by factors or other attributes. The arrows show the dependence from causes to effects.

What the state of each attribute is, depends on the states of the factors or attributes that  influence it. Certain combinations of states of the influencing factors and/or attributes may lead to one or another state. For example, we may say: "If the risk of erosion is low and there is no limitation to regeneration potential, the risk to desertification is low."

We can see immediately that the factors and attributes that appear in this diagram have to be endowed with lists of possible states:

States of Slope: gentle, average, steep

States of Rock Type: permeable, semi-permeable, impermeable

States of Soil Depth: bare, shallow, deep

States of Aspect: south, west/east, north

States of Animal Grazing: slightly, moderately, heavily grazed

States of Erosion:  low, medium, high

States of Regeneration Potential : low, medium, high

States of Desertification:  no/slight, low, medium, high, very high

An expert then can give us a list of rules that express which combinations of states of causes lead to which state of an effect. An example rule is the one mentioned earlier about the dependence of risk to desertification on the risk to erosion and on regeneration potential. Another example rule is:"If the state of Rock Permeability is `permeable' and the state of Soil Depth is `shallow' and the state of Slope is `average' then the state of erosion is `medium'."

The expert may supply us with lots of such rules covering all possible combinations of states of causes leading to all possible states of the effects. Using these rules to draw conclusions constitutes a "rule-based reasoning system". Most Geographic Information Systems (GIS) are rule-based.               

  However...