Tip:
Highlight text to annotate it
X
Namaste and welcome back to the video course on watershed management. In module number
6, on use of modern techniques in watershed management, in lecture number 26, today, we
will discuss about the applications of knowledge-based models in watershed management.
So, in this lecture, some of the topics covered include Knowledge based modeling, Multi criteria
decision analysis, Fuzzy Logic based modeling, Fuzzy systems, Applications of knowledge based
systems in Watershed Management. Some other important key words - Knowledge based model,
Multi criteria decision analysis, Fuzzy logic based modeling.
So, as we discussed earlier, say the modern techniques like geographical information systems,
computer models, then remote sensing, then decision support system, all these things,
all these models, all these modern models are very useful in the effective management
of many engineering problems, say like watershed management or water resource management, etcetera.
So, we have seen the applications of computer models or numerical models, then the geographical
information systems, remote sensing and decision support systems earlier. So, we can have some
specified system for various specific types of problems like irrigation management or
the land use management or crop management related to watershed. We can combine many
of these modules together, and then, we can have a system called or a model called knowledge
based models. So, these knowledge based models are very
useful to decide or to decide which way we have to do various management practices, which
way we have to go for various plans as far as watershed is concerned.
So, actually these knowledge based models are also one way. They are also decision support
systems or decision support models, but in the knowledge based models, we are using the
artificial intelligence techniques like Fuzzy Logic or the techniques like genetic algorithm
or Artificial Neural Network, etcetera. So, that is why these kinds of models are generally
called as knowledge based models.
So, now, let us look into various aspects of knowledge based models. So, here, in this
slide, I have given some information on knowledge based models. So, Knowledge based systems
are computer systems; that are programmed to imitate the human problem solving ability
by means of artificial intelligence tools. So, actually, we can have a certain artificial
intelligence tools like here Fuzzy Logic or the AI and or Artificial Neural Network or
genetic operators or genetic algorithm very similar to the human intelligence type of
systems. So, that can imitate the human problem very effectively. So, that way, the computer
systems when models are made; that way, we called those kinds of models as knowledge
based systems or knowledge based models. So, human reasoning using natural language
can be reproduced in knowledge based systems through various artificial intelligence tools
like fuzzy logic. So, if this happens, what will be the outcome? So, like that or between
good and bad, how the variation is it is not a very specific either good or bad, but between
that, what can happen. So, like that, with these human reasoning kind of systems, we
can have knowledge based models.
So, in knowledge based models, the knowledge can be derived from basic analysis or expert’s
knowledge collected through surveys and heuristic information from the field. So, these knowledge
based models can be obtained from the basic data analysis or experts’ knowledge or through
heuristic information related to that particular problem.
So, experts’ knowledge and heuristic information related to these specific problems are generally
stored in the form of rule base. So, depending upon the problem, depending upon the watershed
or area which we are dealing, we can generate specific rules and then we can store these
rules in a rule base, and then, using those rule bases, we can generate scenarios or what
can happen like if this particular thing is done, so, like that we can create various
scenarios. So, those types of models are called knowledge based models.
So, here, further on knowledge based modeling, knowledge basis an organized body of knowledge
that provides a formal logical specification for the interpretation of information. So,
for example, if, if the, the watershed, if, if land what is available, whether it is suitable
for paddy cultivation or a millet cultivation or peas cultivation. So, like that, we can
consider various information on the particular area and then we can generate the suitability
of land depending upon the various details available.
So, in the knowledge based modeling approach, say for example, if watershed is concerned,
watershed assessment is a multi criteria evaluation, in which, knowledge of the experts’ is used
to define the factors characterizing the watershed and the logic relations between the factors.
So, we can see when we deal with watershed assessment or watershed management, a number
of criteria like the land used; then water availability; then the present cropping pattern;
then neighborhood to the ponds or water bodies or neighborhoods to the transport systems.
So, like that, we have to consider various criteria so called multi criteria, and then,
based upon this, we can derive certain logical relationships using these factors and then
we can develop a knowledge based model for the particular type of problem.
So, that way, the knowledge base encapsulates the assessment criteria, and the relationships
in an explicit form so that they can be easily examined, modified or updated. So, depending
upon the problems, so, we can created rule base based upon the database, and then, that
rule base can be encapsulated within the systems which defines the various relationships and
criteria, and then, based upon that, we can have a knowledge based model and that can
be used to predict or to say that yes. This particular land can be used for this kinds
of cultivation or what kind of supplementary irrigations to be generated for the cultivation
of particular crop like that. So, that way, a knowledge based modeling can be done.
Now, let us look into various aspects of knowledge based systems. So, the knowledge base contains
knowledge and experience for the subject domain. So, whatever we are dealing that subject is
like domain knowledge and specifies the local relations among topics of interest to an assessment.
Say for example, if the assessment is whether particular crop is suitable for that particular
land, so then we have to consider the type of soil; then the water availability; then
various other factors which is required for that particular problems and then we can create
the knowledge base. So then, inference engine performs knowledge
based approximate reasoning to draw conclusions about the state of the systems. So, we can
generate the inference engines and these inference engines performs knowledge based approximations,
and then, based upon that, we can generate particular scenarios or we can take particular
decisions based upon these type of systems. So, by integrating knowledge based reasoning
into a GIS environment, provide decision support for watershed management. So, as we discussed
earlier, for example, if a geographical information systems, say when we integrate this kinds
of knowledge based models or knowledge based systems in a GIS environment, then we can
see that what is happening within the systems or what is happening for that particular problem.
All those with the given inputs and then generate output. We can easily visualize the various
aspects and then we can go for particular decision according to the requirement.
So, that way, the knowledge based model process various relationships for those particular
problems and then that gives certain decisions or scenarios. So, that way, as I mentioned,
GIS applications provide database management, spatial analysis, system interface and map
display. So, when we integrate this knowledge base system with GIS, we can have the database
management, spatial analysis; like, a spatial variation, we can easily identify; then system
interface, say GIS itself is system interface and then even we can generate various maps
for display. And then the assessment system allows users
to evaluate the knowledge base for a specific spatial database and view the results. So,
since when we integrate the knowledge base model or knowledge base systems within a GIS
environment, so, we can easily see what is happening with the spatial variation or within
the spatial database, and then, we can see within the GIS environment the results. So,
that way, it is much easier and much useful for decision makers like if this is done,
what will be happening. So, like that various decisions can be easily taken.
So, now, let us look into the knowledge base structure. So, what is the structure of a
particular knowledge base? So, knowledge base structure is a hierarchy of dependency networks.
So, depending upon the problems, say we have to go through various processes; various hierarchy
of the various networks within the system. So, each network evaluates a specific proposition
about the state of, for example, watershed or watershed condition. So, if this particular
watershed or particular area of the watershed, whether it is suitable for paddy cultivation
or any kind of cultivation or that particular crop, then it has to go through a series of
the various networks and then each network is evaluated, and then, we identify whether
that is suitable or not. So, knowledge base structure is design to
address the issues concerned by the watershed managers, say for example, if you consider
watershed and to reflect their opinions on the importance of each issue. So, in the knowledge
base structure, we can consider various relationships, various issues, and then, we, say the, after
effects if that particular thing is, then what will happens. So that if this is done,
then what will happen. So, like that, it will be given. So, that is will be, that will be
very useful for the watershed management or it will be useful for the decision makers.
So, at the top of the hierarchy is the network watershed condition. Say for example, if you
consider watershed, then at the top of hierarchy is the network watershed condition, what is
the condition of the watershed. For the proposition that overall condition of the watershed is
suitable for sustaining healthy populations of the native. If you consider for the living
condition of the particular watershed, then we consider the watershed condition whether
it is deteriorated or it is very good condition or it is bad condition and then we can come
up with various aspects related to particular decisions to be made whether related to cropping
pattern land use or soil erosion or whatever type of problem we are looking for.
So, the watershed condition generally depends on two lower level networks like a upland
or overland conditions. As we discussed earlier, watershed can be considered as an overland
and then the stream or channel conditions. So, whenever we consider the watershed condition,
so, the two lower level networks other than the overall watershed conditions, say two
levels which we have to consider is how the overland condition and the stream conditions
is so accordingly we can consider various aspects of the problem which we consider,
and then, we can come with a solution or come with a scenario.
So, in all these aspects, when we deal with, say for example, watershed management, so,
we have to consider as we discussed various criteria we have to consider in the analysis.
So, that way, we can see that these type of problems are very much multi criteria based.
So, we have to do a multi criteria decision analysis. So, MCDA or multi criteria decision
analysis very important in knowledge based models.
So, simulation models of various hydrologic components of a watershed, say for example,
rainfall runoff or soil moisture or the water sharing, all these kinds of components. So
then, integrated with the artificial intelligence tools like Fuzzy Logic so as to make use of
expert’s knowledge and heuristic information in decision making process; so, this is use
to help the end users to arrive at the best suitable decisions related to irrigation management.
Say for example, if the watershed is concerned, if you are dealing with irrigation management,
so, we can consider various hydrologic components of that particular watershed and then we can
come with- Yes, this is the best irrigation management practice or this is the best cropping
pattern or depending upon the requirement of the crops, this is the possible irrigation
management schedule like that. So then, irrigation assessment and management as I mentioned,
so, that way, is a multi criteria problem and then we have to go for multi criteria
decision analysis. So, in this, we have to use the knowledge
based systems. So, knowledge based systems are very useful for multi criteria decision
analysis. So, MCDA or multi criteria decision analysis, in which, the land suitability,
say for example, if land suitability criteria has to be consider, water availability irrigation
requirement and various other criteria to be evaluated.
So, generally, the objective function can be or objective can be maximize the agriculture
production when we are looking for irrigation management for the particular watershed. So,
if you consider MCDA or multi criteria decision analysis, we can say our objective or objective
function can be we have to maximize the agricultural production. So, accordingly, we can go for
the irrigation management or the various scheduling. So, that way, MCDA or multi criteria decision
analysis models are used in irrigation management to identify areas that can be irrigated and
then water release during different time periods and then best suitable cropping pattern for
the considered area. So, like that, when we deal with MCDA, - multi
criteria decision analysis - we can identify the water release for different seasons of
different time period, and then, what is the best suitable cropping pattern, and then,
what is the irrigation schedule. So, like that MCD a is very much part of any kinds
of knowledge base systems which we can develop for particular watershed management problem.
Now, let us look into a particular knowledge based systems for watershed management when
we consider. So, this flowchart shows a typical type of knowledge based model for watershed
management. So, if you consider watershed or water related issues within the watershed,
then we have to go for hydrological modeling. Say for example, we can find out the rainfall
runoff using soil conservation curve number based model, and once the runoff is determined,
say this runoff also depends upon the soil moisture balance and then crop water requirement
and then irrigation requirement. So then, based upon that once the runoff is
predicted, then we can identify how much water is available for that particular area, and
then, if sufficient water is not available, then we can think over how we can go for water
harvesting. So, what is the water harvesting potential for that particular area? And then,
if you use a fuzzy membership approach which we will be discussing in the coming slides,
what is fuzzy logic and all those things will be discussing detail.
So, if we consider fuzzy membership approach, then we can use those approaches to identify
what will be whether the land or particular area is suitable for that particular crop,
and then, we will get a spatial temporal multi criteria decision system or multi criteria
decision model for identifying the most suitable cropping zones for that particular area.
So, based upon this approach, one of my Ph.D. students Rashmi Devi, 2008 developed a model
in Department of Civil Engineering, IIT Bombay - a specified knowledge based systems for
watershed management and these results were published in the journal of Irrigation Drainage
American Society of Civil Engineers. So, this way a typical knowledge based systems
consists various components for that particular problem, we have to, we may have to use sometimes
some specified models, for example, rainfall to runoff model, and then, we have to consider
the for example land suitability and then the water requirement and all those things
we can combined together integrate together within a GIS environment, and so that, that
becomes a knowledge base model, and that can be effectively utilized, say for example,
for the land use, suitability analysis or the most suitable cropping zone identification
for that particular watershed like that. So, that way, as I mentioned these fuzzy logic
systems which is used in many problems for the last two decades can be effectively utilized
in watershed management also. So, let us now look into important aspects of fuzzy logic
and fuzzy base system and then related modeling techniques.
So, these fuzzy logic systems - say it was first presented by Zadeh in the mid-nineteen
sixties at the University of California in Berkeley and he developed the fuzzy logic
as a way of processing data. So, by considering various problems, how to process this data
between various variation within the parameters. So, later on he introduced the idea of partial
set membership. So, say if, within the one, if the variation is, for example, good to
bad. If we cannot identify certain class is totally good or certain class is totally bad,
in between what happens. Then that kind of partial set membership this doctor Zadeh introduced,
and then, he defined the fuzzy systems as the system which is not clear or distinct
or precise. So, a system which is not very clear or we cannot say that this is exactly
this is the fashion or it is not so precise. So, that kind of system, we can call it as
a fuzzy system, and then, he defined the fuzzy logic as a multi-valued logic that allows
intermediate values to be defined between conventional evaluations like a true or false,
yes or no, high or low, etcetera. So, that way, Zadeh defined the fuzzy logic
as an intermediate value or intermediate values between conventional evaluations like a say
which is exact like a true false; between true and false, what is there, or between
high and low, how is the variations. So, like that, so, system which is not so clear or
distinct or precise that is defined within the fuzzy systems or fuzzy logic.
So, actually, that way, fuzzy we are not actually dealing with probability. So, probability
generally deals with uncertainty and likelihood of various parameters. Say for example, if
rainfall may or may not happens, so, it is uncertainty of that particular parameters,
but in fuzzy logic fuzzy logic generally deals with the ambiguity and vagueness.
So, whether for example, if particular land is there, that land if say some particular
land, we can specifically it is not at all useful for some paddy cultivation, but some
land will be very suitable for paddy cultivation, but in between, if it is there, then how to
identify. So, that kind of problems we can easily deal with the fuzzy logic based systems.
So, now, as I mentioned, this fuzzy logic is a system which we can utilize where vagueness
or where we cannot exactly define what a situation is.
So, that way, the fuzzy logic is based on intuition and judgment. So, we have to see,
the, what is the, say for example, for the when we deal with particular problem, what
kind of intuition we are getting or what kind of judgment we are having.
So, that way, actually it is not based upon specified mathematical models, but we have
to see that a with our intuition judgment, we have to come up with a methodology in fuzzy
logic, and then, fuzzy logic provides a smooth transition between members and nonmembers.
So, if the member between member and non-member means if the what is the decision is - yes
or no. So, what is there in between, so, that is we, that the transition between members
and non-members or the between high and low, so, what is there; so, that kind of transition.
So, that way, it is relatively a symbol; this fuzzy logics is relatively symbol fast and
adoptive and then it is less sensitive to systems fluctuations, and then, according
to the problem, we can defined or design certain rules. So, that way, it is a rule based operation
we can define, and then, it can be implemented for design objectives or like what is difficult
to express mathematically in terms of linguistic or descriptive rules.
So, mathematically, if we cannot have precise type of rules or type of definition for that
particular problem, then but it may be able, we can put it in terms of linguistic or descriptive
rules. So, there fuzzy logic we can directly utilize. So, that way, if we consider for
example conventional or crisp sets are binary, but fuzzy logic is the variation in between
is considered. So, now, an element either belongs to the
set or does not. So, generally, crisp or the conventional set is particular thing is concerned,
whether it belongs or it does not belong, but like a true and false. So, that way, if
we assign true is zero, then we have for false; we can assign one, so, like that. But in fuzzy
logic where for the type of problems, where it is not possible to specifically true or
false but something in between; so, that means it may can vary from zero to one but it may
not be exactly as zero or one depending upon the problem.
So, like that, if we consider the problem, so, now let us see here, you can see that
in this figure, this a b c - these are all subsets of this particular problem. So then,
if it is specifically the things are in this, then it is a or in this subset, it is b, or
in this is c, but in between if then, what happens.
So, that way, fuzzy logic can be considered. So now, let us look what is fuzzy sets. So,
fuzzy sets we have to consider set of details within the, for that particular problem. So,
this allows elements to be partially in a set. As I mentioned here, this particular
set which we considered, so, allows elements to be partially in a set. Each element is
given a degree of membership in a set. Then a membership function is the relationship
between the values of an element and its degree of membership in a set.
For example, if the variation is this particular function mu, so then this is negative and
this side is positive. So then, a negative positive; then or large, medium, small. So,
in between that how the variations are taking place. So, that way, we can consider for the
particular problem which we consider.
So, that way we can consider the fuzzy sets. So now, in these kinds of problems, we have
to consider the membership. Whether it is in which subset or whether between those sets,
so, like that, we have to consider the membership functions.
So, here, the details of the membership functions are given in this slide. So, the membership
generally can be crisp membership functions. So, crisp membership functions, for example,
this mu which we consider either one or zero; so, exactly one or zero. So, for example,
particular number greater than ten so that we can define like this.
But as far as fuzzy membership functions are concerned, the membership value here is not
exactly one or zero but it is varying. So, the degree of truth of a statement can range
between zero and one, and the linguistic variables are used for to describe this fuzzy measure,
what is happening. So, examples of fuzzy measures include the
particular problem is close, say like a water bodies, close to the to the agricultural land
or it is a medium, heavy; it is a heavy, light, big, small, smart, fast, slow, hot, cold,
tall, short, like that. So, on linguistic terms linguistic variables, we can use between
these parameters and then we can specify. So, that way, in the fuzzy membership functions,
these values it is not crisp like a zero or one but it will be varying between, the, these
parameters.
So now, we can see that that way we need to define the fuzzy logic operations. So, if
we consider two sets whether how that sets are behaving, so, accordingly, we can design
the problem. Now in the fuzzy logic operations, for example, we that the union, for example,
if this is subset a subset b, then the union is maximum mu a x mu b x as shown in this
figure.
And then, if we consider only the intersection, so, for example, minimum of mu a x and mu
b x, so, this is the intersection which we consider, and then, the complement, the negation
of the specified membership function, so, we do not consider this area, but on the other
two sides area if we consider when the compliment.
So, like that, fuzzy logic we can define particular fuzzy logic operations and then we can consider
the particular problems which we are dealing. Now, this fuzzy logic, as I mentioned, in
most of the natural problems which we consider like a, if we, if we consider the watershed
management. So then, the particular area is concerned, crop suitability or irrigation
management, so, like that, various problems, so, we cannot a specifically exactly, that
is, this is the way but it can vary between yes or no, or false or right, or like that
between these parameters vary.
So, that way, fuzzy logic has got large number of applications, for example, in watershed
management. Of course this fuzzy logic was developed for various other types of problems.
Some of the applications I have listed here: like a ride smoothness control; then in all
other kinds of engineering like electric engineering, electronics engineering mechanical engineering
like that. Then braking systems, high performance drives,
air conditioning systems, digital image processing, washing machines, pattern recognition remote
sensing, video game artificial intelligence, graphic controllers for automated policy sketchers
like that. So, large number of applications we can see now in literature related to fuzzy
logic. Now, since our main interest is here related to watershed management management
problem, so, watershed related application also large number of applications, we can
see in literature like in modeling rainfall-runoff processes.
Then, erosive soil measurement, then hydro ecological modeling or watershed, then flood
forecasting, then water quality problems cropping and irrigation management. So, like that,
watershed related or watershed management related number of problems also we can utilize
this fuzzy logic. Since many of these natural problems are related to watershed problems
are very much fuzzy in nature. So, that way, we can utilize the this fuzzy logic or fuzzy
sets or fuzzy based model for watershed management related problems.
So now, let us look this fuzzy logic the concepts we have now discussed. So now, let us look
what are the advantages and limitations of fuzzy logic and with respect to applications
what we are seen some there are certain advantages and some limitations.
So here, some of the advantages are like, allow it, the fuzzy logic allows the use of
vague linguistic terms in the rules. So that based upon that, we can come up with certain
decision for the particular problem, and that way, we do not need any, the, excite mathematical
models for that particular problem. So, based upon the linguistic variations, we can make
decisions or make we can make modeling. And these are rule based and descriptive type.
So, most of the fuzzy logic systems are rule based, and then, some of the limitations include
this is difficult to estimate membership function. So, most of the problems, say what kind of
membership is there, accordingly, it is exactly to define, to estimate the membership is difficult.
There are many ways of interpreting fuzzy rules. So, say the time itself is fuzzy, so,
we can interpret also in different ways. So, we have to the correctness of the decision
or the, you have better decisions or better interpretation if we use the best possible
kind of system. And combining the outputs of several fuzzy
rules and defuzzifying the output. So, we have to first see the fuzzification of the
systems, and then, based upon the rules or, the, the, in terms of linguistic systems and
then we have to again come back and defuzzify the output. So, that way, we have to go through
certain procedure. So, these are some of the limitations as far as fuzzy logic types of
systems are concerned.
So now, let us look what are the important components as far as fuzzy logic is concerned.
So, this slide shows the basic components of fuzzy logic systems. Here, first of all
of course, data input is required based upon the available input only, we consider the
particular problems and then come up with a certain decision.
So, first is the input. So, data input, and then, based upon the problem, we consider
certain types of models to fuzzily the systems. So, that is called the fuzzification, and
then, we consider the fuzzy rules, the base rules applicable for that particular problems.
So, based upon that, we will be getting the fuzzy output, and then, since to the, to normalize
and to for the understanding of the problem, again we have to do a defuzzification and
then we will be getting the output. So, this way, in a fuzzy system, the basic,
there are five basic components, and that way, systematic modeling or systematic operation
operations are possible in a fuzzy based modeling.
So now, let us have a look into the, three two, three components the fuzzification fuzzy
based rule and defuzzifcation, some of the important aspects let us look into.
So, fuzzification means it is a fuzzifier; fuzzifier converts a crisp input into fuzzy
variables. Say for example, if we consider the land use for particular crop, so, it is
the possible crisp inputs are not suitable, then suitable. So, in between, we can have
less suitable, moderate suitable, less suitable, so, like that.
So, that way, this fuzzifier converts each piece of input data to various degrees of
membership. So, the membership function is a graphical representation as shown in this
figure representation of the magnitude of participation. So, like the land suitability,
it is suitable or not suitable; then in between, we can have less suitable, moderate suitable
like that. Then the definition of the membership functions must reflects the designers knowledge;
then provides smooth transition between member and nonmembers of a fuzzy set.
So, that way, we have to do this Fuzzification; so, it should provide smooth transition between
member and nonmembers of the fuzzy set. Then typical shapes of the membership functions
we can have a Gaussian variation, Gaussian variation like this, or we can have a trapezoidal
variation as shown here or we can have a simple triangular variation also. That way, this
kinds of variations we can use, in the, in the fuzzification processes.
Then the second one is the fuzzy base rule. So, this fuzzy base rule is actually the important
component in any of these kinds of modeling. So, these fuzzy base rules include all possible
fuzzy relationships between inputs and outputs. So, we have seen that inputs are there and
then corresponding outputs will be there. So, this fuzzy relation, we can generate relations
based upon this inputs and outputs. So, this include all possible fuzzy relationships
between this inputs and outputs. So, rules are expressed in the if-then format; that
means if this done, if this particular check dam is constructed, then how much will be
the flooding problem or how much is the area can be irrigated. So, like that, if this is
done, then what will happen? So, if-then format, then we can the rules reflect experts decisions.
So, this, the, whatever the rules which we generate, based upon that, the final decisions
made. So, that way, they should reflect the expert’s decisions and then rules are tabulated
as fuzzy words. So, like for example, if a particular person
is there, we can by looking to that particular person, his actions or his conditions, we
can say that weather he is healthy; then we can say that he is whether somewhat healthy,
less healthy or unhealthy.
So, this can be based upon various conditions like a whether he is how much tall or whether
he is a fat or he is thin or he is overall health conditions. So, for example, we can
generate a particular fuzzy base rule, say for example, related to healthy or unhealthy,
healthy is concerned. If height is tall and weight is medium, then
we are healthy. So, like that, if-then relationships we can form, and for example, if height is
small and then weight is more than unhealthy, so, like that, we can have a, we can generate
various fuzzy base rules. So, this is related to the health of a person,
but for example, related to the crop suitability particular area or land suitability, we can
consider various aspects like if water is available, then irrigation can be done so
that this particular crop like a paddy is possible. So, like that we can generate the
fuzzy based rules; so, that will be very useful in in this fuzzy based modeling.
So, that way, I have shown here we can have various conditions like a fuzzy based decision
as shown in this slide. So, the fuzzy based decisions, we can give various weightages
here and then function is shown here; so, this is related to unhealthy, less healthy,
somewhat healthy or healthy, so, like that. So, most of this fuzzy based rules are based
upon, if this is the condition, then what is the situation, so, like that.
So then, third component is the defuzzification. So, once the input is there and then fuzzification
is done, then fuzzy based upon the fuzzy rules, we are now going getting a output which is
in the not understandable for a normal person; so, we have to defuzzify the system.
So, defuzzification, as shown in this here converts the resulting fuzzy outputs from
the fuzzy inference engine to a number. So, generally, in computer terms, it will be represented
in terms of a number and this number will showing how the variations is taking place
whether it can be between zero to one or what way it is.
Then converting the output fuzzy variable into a unique number, so, that unique number
represent, whether that person is healthy, less healthy, somewhat healthy or that particular
area is suitable, less suitable or more suitable, so, like that. So, numbers of defuzzification
methods are available in literature, like weighted average conditions, maximum membership,
average maximum membership or central gravity of that particular the area triangular or
trapezoidal like that that variation. So, these different methods are there; anyway,
we are not going into the details. But the defuzzification is required when we
use the fuzzy logic and that gives the particular output the system which we are looking for
that particular problem. So, that way, defuzzification is also very important. So, now, what we are
discussing is the fuzzy logic and then the structure of fuzzy logic systems and then
how the fuzzy logic is working and then its applications. That is what we are discussing
so far. So now, today is our main topic is knowledge
based model. So, related to watershed management, how we can develop a knowledge based, knowledge
based, knowledge based model related to watershed management? So, that way, now further in the
coming few slides, we will discuss a particular knowledge based model for related to watershed
management.
So, let us look into this particular model developed by PHD student Reshmi Devi and presented
in her thesis knowledge based model for supplementary irrigation assessment in agricultural watershed.
So, here, she developed a fuzzy rule based inference system for land suitability evaluation.
So, for, if a particular land is a particular watershed for particular crops, how effectively
that is suitable, less suitable or more suitable, like that.
And then she developed a spatial temporal multi creative decision analysis model for
s m s MCDA model for identifying the scope for supplementary irrigation. So, based upon
this fuzzy rule based system, she developed a model to identify whether if we consider
the particular watershed and then its cropping pattern, its irrigation availability and so
whether we have to go for supplementary irrigation and then how effectively we can do within
the context of a knowledge based model. So, that way, model has been developed and then
also she developed a graphical user interface for this particular model.
So, in this particular model for this knowledge based model, five steps are there in the model
development - first one is the fuzzification of the attributes, then estimation of the
intermediate land suitability index, then generation of the fuzzy rule based, then aggregation
of the rules like a fuzzy output in terms of five suitability classes like a less suitable,
suitable, more suitable, like that and then defuzzification. So, the basic steps - fuzzification
estimation, intermediate land suitability index, then generation of fuzzy rules, then
aggregation of the rules and defuzzification.
So, these were the essential steps for this particular model. Then fuzzy rule based inference
system, so, here, the problem with large number of attributes. For example, if a particular
watershed or particular area whether we want to identify whether that particular area is
suitable for particular cropping, so, we have to consider various aspects like the average
for the particular average number of dry days, proximity of water body. So, this is related
to water related issues; then elevation with respect to the nearest water body; then land
related issues like the land use, soil texture, terrain slope, soil depth, drainage density,
lineament density. Then soil related issues like pH, then electrical conductivity, salinity,
etcetera. Then climate related like rainfall temperature,
proximity to road. So then, based upon the fuzzy operations related to water related
issues which is used water potential, and then, from this, the land use related issues;
then the weighted aggregation related to soil; then related to terrain and weighted aggregation
related to the soil use the fertility; then various other attributes also can be considered.
So now, based upon this fuzzy rule based, inference system has been developed and so
this is the overall structure of the model and then defuzzification, and then, from that,
we can get the particular the decision or particular suitability. So, the, here, she
considered a hierarchical classification and then she considered both land potential and
water potential for that particular problem. So, the main issues are related to land potential,
water potential so that a suitable crop index or crop suitability or land suitability for
that particular crop can be identified. So, this system was based upon the fuzzy rule
based inference system. So, as far as fuzzification is concerned as
we discussed earlier, the attribute values are mapped between zero and one, and then,
two types of attributes - like thematic attributes for land potential, unique membership value
for each class, then continuously expressed attributes for land potential semantic import
membership function, then asymmetric left or asymmetric right or optimal range.
So, more details of this, you can see in the journal of irrigation drainage a paper published.
So, the reference will be given later by us. Then the next step is the intermediate land
suitability index. So, here, the weighted aggregation of the attribute membership values
are used, and attributes like a here, she used Saaty’s relative importance scale to
identify the intermediate land suitability index.
And then based upon that, relative importance is assumed based on literature, field observation
and heuristic information, and then, this gives intermediate land suitability index
in three suitability classes like good, moderate and not suitable based on the land and the
water potential.
So this, these are the details as far as the intermediate land suitability index in this
particular knowledge based model. Then the fuzzy rule based and aggregation of the rules
are generated. So, suitability criteria is based upon in the form of if then rules in
terms of intermediate suitability indices, like if land use is good and water potential
is good and the terrain is good and chemical character is good and then other parameters
are good, then the area is excellent. So, like that, the system is made.
And then another scenario, if land use is good and water potential is moderate and terrain
is moderate, then and physical chemical characteristics is moderate and other parameters are moderate;
then the area is moderate, and third one if land use is not suitable and water potential
is not suitable and terrain is suitable and physical chemical characteristics is not suitable
and parameters are not suitable, then area is not suitable.
So like that, if then rules, fuzzy rule base we have generated. So, this we generated the
fuzzy output in terms of five suitability classes for this model like excellent, whether
the area is excellent, good, moderate, less suitable and a not suitable, and then, next
step is defuzzification as I discussed earlier. So, this convert the fuzzy output into a single
value land suitability index. So, here, maximum centroid method is considered so as shown
in this figure. So now, based upon this, using this model,
we can generate the best suitable crop map. So, relative importance and land suitability
index is given. So, three cases: case one is land suitability index of existing crop;
so, that is less then another crop of higher priority. So then, higher priority crop is
selected, so, this one case. Then case two is land suitability index of the existing
crop is less then another crop of lesser priority. Then change in the cropping pattern if less
suitability or not suitable for the existing crop.
Then third case is land suitability index is same for more than one potential crop if
less suitable or not suitable for the existing crop, and if relative importance of the existing
crop, then the than the other crop. Then a change in the cropping pattern is proposed
and then it replaces the existing crop with high priority one. So, like that various systems
we are made in this particular models.
So, this model, the details are given in a flowchart. This is spatial temporal multi
creative decision analysis model for irrigation feasibility analysis. So, here, first we have
to assess the irrigation requirement and runoff availability, then runoff versus irrigation
requirement; so, this gives the water deficit periods. Then land suitability - it is coming
from the fuzzy logic. So, runoff versus priority based irrigation requirement. The irrigable
areas, we can identify, and then, outsource water requirement for different suitability
classes whether we have to go for the further the supplementary irrigation like that we
can decide using this model.
So, these details of these models are available by in this, you know, in the paper published
in 2010 titled knowledge based model for supplementary irrigation assessment in agricultural watersheds
journal of irrigation drainage ASCE 2010 volume 136 page is 376 to 82.
So, let us have a brief look into one case study related to this model which is done
by Rashmi Devi in her thesis. So, location is the Harzul watershed and the area is 10.9
square kilometer and principle crops in this area are paddy and finger millet. So, this
is the watershed area.
So, here based upon the various input data, various data based were generated like heuristic
information and field observation related to attributes; attribute suitability for different
crops, crop priority and agriculture practices, land suitability criteria. So, this is done
in the heuristic information, and then, map layers related to drainage map, contour map,
then soil map, pH map showing spatial variation electrical conducted salinity etcetera were
generated.
And then also land use map, drainage density map, then proximity to water body, proximity
to settlement, so, all these details were generated in the data base so that this fuzzy
based system modeling can be done. And then hydro meteorological data related
to rainfall, stream flow temperature, relative humidity, sunshine duration, wind speed, all
these were collected, and then, using the earlier described knowledge based model, either
modeling has been done, and so, the, for the land suitability related, so, for this watershed,
for example the s c s c n soil moisture simulation model has been used for the runoff generation
for the given condition; so, this shows the output.
And then for example year 2002 which is a dry year we can identify how much is the water
available, and then, for paddy field or finger millet, how is water requirement, and then,
this is the rainfall hyetograph and then we can identify how much is the irrigation requirement
and then accumulation like that. So, that is these all generated using the particular
model.
And then irrigation requirement harzul watershed, for example, four years – 2001, 2002, 2003,
2004 were generated. So, irrigation requirement, non-agricultural area, then finger millet
irrigation, then paddy field, how with 50m requirement, 500m requirement, 150m requirement.
And then based upon that, best suitable cropping zones in harsul watershed has been generated.
So, this yellow shows the area suitable for paddy field and then this red shows non-agricultural
area and this violet this shows the finger millet suitable and not suitable for paddy
or finger millet. So, this way, we can generate the best suitable cropping zones.
So, this shows the land suitability for paddy, land suitability for finger millet. The suitability
classes and range of land suitability index is given here percent area wise related to
paddy and finger millet. So, that way, we can generate the, using this knowledge based
model, we can genera identify the land suitability for particular crop and then which of the
area are most suitable, less suitable; so, like that, we can identify.
So, now, to, finally to conclude many decision making problem solving tasks are too easy
to solve. So, this fuzzy logic can be used for this purpose and knowledge based models
shows the irrigation requirement for the predicted rainfall and predicted rainfall helps to choose
adopt appropriate crops and irrigation management plans for the given area.
So, these are some of the important references used in a for today’s lecture.
So then, before closing some tutorial questions, critically study the applications of knowledge
based systems for various water resource management problems. Study various case studies available
in literature. So, these details we can obtained from the internet. Study the role of knowledge
based modeling in integrated water resource management. So, how we can effectively utilize
knowledge based model?
Then some self evaluation questions - describe the features of typical knowledge based models.
Illustrate the requirements of knowledge based systems. Describe typical knowledge based
systems for watershed management. Illustrate the fuzzy logic operators used in typical
fuzzy logic. What are the important components of fuzzy logic systems? So, these all these
questions you can answer by going through today’s lecture.
So, few assignment questions - describe the structure of a knowledge based system. What
are the important features Multi Criteria Decision Analysis? Illustrate the features
of fuzzy logic based on systems. Describe applications advantages and limitations of
fuzzy logic? Illustrate a typical knowledge based model for watershed management. So,
all these questions we can answer by going through today’s lecture.
And one unsolved problem - critically study a typical knowledge based model for a for
the water and land management in the watershed. For your watershed area, study the scope of
development of knowledge based model considering rainfall, various crops, land use land suitability,
water requirement etcetera. So, today we considered the knowledge based
systems for watershed management. We discuss the fuzzy logic systems, and then, connecting
to that, how we can generate knowledge based models. So, that way, we can see that this
knowledge based models are very useful in watershed management.
So, with this lecture, the particular module on the modern systems for watershed management
module number 6 is over. Now, we will discuss various other aspects of watershed management
issues. Thank you.