A KNOWLEDGE BASED APPROACH TO THE MANAGEMENT OF SOIL EROSION INFORMATION IN
DEVELOPING COUNTRIES.
E.G. Mtalo and E.Derenyi
Department of Surveying Engineering
University of New Brunswick
Fredericton, N.B. CANADA
COMMISSION VII
ABSTRACT
The study and control of soil erosion and the associated environmental degradation is a difficult problem because it involves a very
wide body of knowledge. The use of knowledge based expert systems in this domain may improve the current efforts by providing an
efficient tool for inter-disciplinary technology transfer and application. Domain expert systems may also be used as substitutes for
human experts to provide guidance to less skilled soil erosion domain personnel.
This paper describes the prototype knowledge based system Soil Loss Estimation and Modelling System(SLEMS) and shows how
such a system can be used to solve some soil erosion problems. This system was developed and implemented at UNB on a Sun
workstation for the management and application of remote sensing and geographic information in soil loss estimation and modelling
problems.
Key Words: Knowledge Base, Soil Erosion, Environment, GIS, Remote Sensing, Developing Countries
1. INTRODUCTION.
Soil erosion is defined as the process whereby detachment and
transportation of soil from its natural location takes place usually
with adverse impact on the environment. Factors and causes of
soil erosion are characterized as: climatic and hydrologic agents
such as rain, runoff, antecedent moisture, wind, geographical
position, and altitude; morphological agents including slope,
slope length and surface roughness; geological and soil agents
which include soil type, parent material and soil texture;
vegetative agents such as plant cover type, density and height;
technical agents consisting of conservation management, tillage
systems, farm equipment and construction activities; and social
economic agents such as population density and distribution,
and other human activities(Wischmeier and Smith, 1957,
Goldman et al, 1986). Morgan (1986) considers soil erosion to
be a multi-faceted problem whose solution involves five issues:
policy, measurement and inventory of soil erosion extent and
severity, assessment and evaluation of soil loss conservation
efforts, modelling and estimation of soil loss and social and
economic issues.
The five components of the soil erosion management problem
are strongly interrelated. The various factors influencing soil
erosion are also intricately intertwined. Clearly soil erosion is a
highly complex problem whose solution requires harnessing
and integrating information and technology from a wide range
of disciplines. Moreover a considerable amount of the
information needed to model and estimate soil erosion is fuzzy
or non-precise. At the same time soil erosion evaluation often
relies on fuzzy decision.
Monitoring, management, and control of soil erosion in
developing countries is a critical issue because the economic
base of most developing countries is mainly agricultural. The
dangers posed by increased soil erosion to the economies of
developing countries are therefore very great.
In the developed world it has become fashionable to handle
complex problems which require intensive human expertise by
knowledge based computer systems. This approach extends the
human ability to analyze complex systems in order to provide
accurate and timely answers to difficult problems. The expert
systems technology releases the human expert from tiresome,
repetitive procedures so that s/he may devote more time on
designing efficient problem solving strategies. One area where
the knowledge based systems approach may be beneficial is in
the solution of soil erosion problems. However at present there
have been very few expert systems designed to solve soil
erosion problems.
In developing countries, where lack of human expertise in all
fields is often acute, knowledge based systems in the form of
expert systems could become an important tool for technology
198
transfer and application. Domain expert systems can be used as
substitutes for human experts in providing guidance to less
skilled personnel in developing countries.
Assuming an appropriate framework in which local experts
work together with foreign experts to develop knowledge based
soil erosion applications the benefits which may be realised
include:
(i) Efficient and sustainable transfer of monitoring and
control technology.
(i) On the job instruction on problem solving procedures
by domain expert systems allowing less skilled
technicians to perform tasks requiring expert
knowledge.
Systematic training of local technicians and skilled
labour on the use and maintenance imported
technology.
Reduced dependence on foreign expatriates and
boosting up self-reliance.
(iii)
(vi)
This paper discusses an experimental knowledge based system
(SLEMS), developed as part of an academic research, for
modeling and estimating soil erosion(Mtalo, 1990). Because of
the need for affordable technology in developing countries the
system was developed with a view to making it locally
implementable. The system is therefore designed to provide a
knowledge based management of attributive information and
data required for modelling and estimating soil erosion.
1.1 Knowledge Representation Issues.
Knowledge based systems, differ from conventional software
systems by the manner in which the knowledge required to
solve specific problems is organised. In Knowledge based
systems the facts (data) and knowledge needed for problem
solving are extracted and stored in a knowledge base using
special knowledge representation structures. The knowledge
may be stored as a set of if...then... rules, frames and semantic
networks(Firebaugh, 1988, Barr et al, 1989; Luger and
Stubblefield, 1989). The process of extracting information from
the knowledge base is referred to as knowledge inference. By
using deductive and inductive inference techniques, knowledge
not directly stored in the knowledge base can be inferred.
Systems which have knowledge about a specific and well
defined domain of expertise, are called expert systems. The
introduction of expert systems transformed artificial intelligence
from a purely research domain into an application field. MYCIN
and the INTERNIET were among the earliest practically useful
rule based cxpert systems(Barr and Feigenbaum, 1981,1982).
Among the few expert systems designed for solving soil erosion
problems is PLANTING commissioned by the United States