AN EXPERT SYSTEM FOR DECISION SUPPORT OF THE

TOTAL ALLOCATION OF AGRICULTURE

NON-POINT SOURCE POLLUTION

A Case of Tiaoxi Watershed

Bo Yu

, Qingyu Zhang, Liangqian Fan, Qi Liu and Weili Tian

Institute of Environmental Engineering, Zhejiang University, 866# Yuhangtang Road, Hangzhou, China

Keywords: Expert system, Non-point source, AHP; Decision support system, Waste load allocation.

Abstract: This paper develops an expert system for waste load allocation of agricultural non-point source (ES-

WLAANS) to assist environment administrators in making decision in optimal solution of allocation. ES-

WLAANS includes a decision model at its core which is built based on analytic hierarchy process (AHP)

and a closely related database. Tiaoxi watershed is the case study area. The results showed that ES-

WLAANS can enhance the efficiency of allocation of agriculture NPS pollution for environmental

administrators.

1 INTRODUCTION

One watershed always covers several administrative

regions. Agriculture non-point source pollution of

each region has different impacts on water quality

due to different pollutants emission. Waste load

allocation of agriculture non-point source of river

basin (WLAANS) is a scientific management

method to determine the optimal pollutant abatement

for each region. Therefore, it is important to decide

how to allocate among these regions of a river basin.

There have been some forms of optimization

models to solve the Waste Load Allocation (WLA)

problem. Mostafavi and Afshar (2011) use non-

dominated archiving multi-colony ant algorithm to

develop WLA model. Saadatpour and Afshar (2007)

presented a fuzzy WLA model. Burn et al. (2001)

explored the capabilities of genetic algorithms.

Analytical Hierarchy Process (AHP) is multi-

objective decision-aiding method for pollutants

allocation (Li et al., 2005). However, the complexity

of WLAANS and the synthesis of AHP have many

problems including a mass of time, lots of human

power and financial resources. It is very difficult for

government to estimate pollutants abatement amount.

This paper developed an expert system for waste

load allocation of agricultural non-point source (ES-

WLAANS). We select COD, total nitrogen (TN),

total phosphorous (TP) as pollutants and Tiaoxi

Watershed as case study. ES-WLAANS is

developed with Visual Basic (VB), AHP, Microsoft

Access and Matrix Laboratory (MATLAB). It takes

economic factor, social factor, technical factor and

environmental factor into account to establish the

structure of AHP for decision support system.

2 METHODS

2.1 Study Area

Tiaoxi Watershed belongs to Tai Lake basin that is

located in the northwest of Zhejiang Province,

China. There are 6 regions including Changxin,

Huzhou, Anji, Deqing, Yuhang and Lin’an in Tiaoxi

valley. Ongley et al. (2010) found that the

proportion of NPS pollution to Tai lake increased

remarkably in recent years and algal bloomed in the

Tai Lake in 2008.

2.2 Framework of ES-WLAANS

ES-WLAANS consists of three basic components,

including database, inference engine (analysis

model) and interface. The basic structure of ES-

WLAANS is shown in Fig.1.

405

Yu B., Zhang Q., Fang L., Liu Q. and Tian W..

AN EXPERT SYSTEM FOR DECISION SUPPORT OF THE TOTAL ALLOCATION OF AGRICULTURE NON-POINT SOURCE POLLUTION - A Case of

Tiaoxi Watershed.

DOI: 10.5220/0003579004050408

In Proceedings of 1st International Conference on Simulation and Modeling Methodologies, Technologies and Applications (SIMULTECH-2011), pages

405-408

ISBN: 978-989-8425-78-2

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

Figure 1: Architecture of the ES-WLAANS.

2.3 Modelling and Methodology

2.3.1 Allocation Index System and the

Hierarchic Structure

The first step is building a proper index system for

AHP. To make the allocation more acceptable for

each region, the principles of fairness and efficiency

were emphasized in building index system.

Combining the former researches and the situation

of Tiaoxi Watershed (Xiong et al., 2007), we select

some representative indices: agriculture GDP,

consumer price index, expenditure of science and

education et al for economic factor; medical staff

proportion, population density, number of personnel

in agriculture et al for social factor; TN emission per

cultivated area, COD emission per cultivated area,

TP emission per cultivated area et al for technical

factor; forest cover rate, area of paddy field, area of

irrigable land, area of arid land, area of forest land et

al for environmental factor.

There are 4 layers in the allocation system—

objective layer, factor layer, index layer and decision

layer, showed in Fig.2.

2.3.2 Dimensionless Evaluation Factors

There are different units among these indices, so it is

essential to make all factors non-dimensional-

normalized for better comprehensive analysis

(Xiong et al., 2007). There are negative and positive

factors. Positive factors were advantageous to WLA

such as forest cover rate, and negative factors were

disadvantageous to WLA such as emission of

pollutants per cultivated area. Therefore, the positive

factors should be dimensionless with equation (1)

and negative factors should be dimensionless with

equation (2):

Figure 2: Frame of AHP of waste load allocation.

100'

minmax

min

−

(1)

1001

minmax

min

⎟

⎠

⎞

⎜

⎝

⎛

−

−=

(2)

where

i is index,

x is original value,

maxi

x and

mini

x are the maximum and minimum value.

2.3.3 Indices Weights Estimation

The first layer was broken down into four layers

including economic layer, social layer,

environmental layer and technical layer to establish

the pair-wise comparison matrix. The pair-wise

comparisons are done in terms of which element

dominates the other (Xiong et al., 2007). The matrix

was expressed with A=

(

)

44×

a .

⎥

⎦

⎤

⎢

⎣

⎡

44434241

34333231

24232221

14131211

aaaa

(3)

We can assign the value of element in the matrix

as follows: if social factor is more important than

economic factor, the value of

a can be assigned

greater than 1. At the same time, the value that is

less than 1 means less important than another and if

the value of 1 means that the two factors have equal

importance. This study selects MATLAB to

calculate the eigenvalue and eigenvector of the pair-

wise comparison matrix. Members of the

SIMULTECH 2011 - 1st International Conference on Simulation and Modeling Methodologies, Technologies and

Applications

406

eigenvector represent corresponding weights of the

indices. Weights of indices in the lower lay can also

be obtained with the same method.

2.3.4 Rate of Decision Layer Estimation

Based on the weight of each index, we can calculate

the synthetic evaluation value of each unit by the

following formulas.

4321

VVVVV

+++=

(4)

∑

××=

kkjj

wuWV

(5)

where

i is the unit to be allocated,

and k are

index type in the first layer and lower layer,

V is the

synthetic evaluation value,

is the evaluation value

of the first layer,

W and w are the weight of the

index in the first layer and lower layer,

u is the

value of the index in the lower layer.

Finally, we can calculate the rate of WLA for

each unit based on the synthetic value (equation (6)) :

∑

×−=

LPT

(6)

where

i is the unit to be allocated,

is the amount

of pollutant to be reduced,

is the total amount of

pollutant,

L is the total environmental capacity

added by all the units.

2.4 Database

Microsoft Office Access has two distinguishing

characteristics what are Microsoft Jet Database

Engine and Graphical User Interface which render it

suitable for developing the database.

2.5 Implementation of ES-WLAANS

ES-WLAANS is implemented by VB which is an

advanced programming language characterized by

visualization, object-oriented and event-driven (Liu

and Lu, 2010). Owing to the complexity of model

calculation for AHP, this system selects MATLAB

to support the calculation. Though MATLAB has

the advantage of data process and graph plot, it is

weak in functional interface. Fortunately, Visual

Basic can call dynamic link library compiled by

MATLAB so that the system easily integrates

advantages of AHP and MATLAB (Zhou et al.,

2004).

3 APPLICATION OF ES-WLAANS

3.1 Selecting Index for AHP

According to the present situation of Tiaoxi

Watershed, we select 18 indices in the lower layer

shown in Fig.3.

Figure 3: Frame of AHP of waste load allocation.

3.2 Data Input Interface

As showed from GUI of data input in the Fig.4,

users can input the pair-wise comparison matrix and

kinds of data such as pollutants emission, local

environmental capacity and data of each index.

Figure 4: The GUI of data input.

4 RESULTS AND DISCUSSION

Fig.5 shows the result estimated with ES-WLAANS.

From figure 5, we can see that Changxing has the

worst result which requires reduction of 9575.2 ton

of COD, 4056 ton of TN and 223.45 ton of TP

followed by Anji which needs reduction of 6005.2

ton of COD, 2406.2 ton of TN and 85 ton of TP. As

for Deqing, the required reduction of TP is 87.81 ton

just less than Changxing, but COD, 251.5 ton, and

TN, 623 ton, are the second least except Lin’an

which needn’t be reduced anymore. Besides,

AN EXPERT SYSTEM FOR DECISION SUPPORT OF THE TOTAL ALLOCATION OF AGRICULTURE

NON-POINT SOURCE POLLUTION - A Case of Tiaoxi Watershed

407

Allo catio n of Agric ulture NP S Pollution F or Each Area

2000

4000

6000

8000

10000

12000

Wuxing Deqing Changxing Anji Yuhang Linan

Survey Regions

Amount of allocation of COD and T

(ton/a)

100

150

200

250

Amount of allocation of TP (ton/a

)

COD

Figure 5: Result of allocation for each area.

Wuxing and Yuhang have the similar condition that

reduction of TN for Wuxing is 767.6 ton and

reduction of TN for Yuhang is 926.9 ton. COD and

TP in both Wuxing and Yuhang needn’t be reduced

like Lin’an.

5 CONCLUSIONS

ES-WLAANS is a modelling expert decision

support system that assists decision-makers in local

environmental administration selecting an optimal

solution for reduction proportion of pollutants from

agriculture NPS among survey regions. Compared

with other model or system, it has own feature

which are showed in the table.1.

Table 1: Comparison with other decision support systems.

System/

Model

Manual/

computerized

Index

number

Expert

ES-WLAANS C Flexible Not

Delphi-AHP M 18 Need

Input-output

analysis

(Ni et al., 2001)

M 21 Need

Linear

programming

(Deng et al.,

2010)

M 6 Need

The characteristic of ES-WLAANS is listed as

following:

1. It can provide user-friendly interface for

operators without special training such as mastering

MATLAB or AHP.

2. When ES-WLAANS is applied in other

watershed, users can select any index they need.

3. ES-WLAANS can give pollutants abatement

scenarios of agriculture non-point source in each

administrative region.

ACKNOWLEDGEMENTS

This work was supported by grant number

2008ZX07101-006. The special thanks are sent to

the viewers for my paper.

REFERENCES

AL-Harbi, K. A. (2001). Application of the AHP in

project management. International Journal of Project

Management, 19, 19-27.

Burn, D. H. et al. (2001). Waste-Load Allocation Using

Genetic Algorithms. Journal of Water Resources

Planning and Management, 121-129.

Deng, Y. X. et al. (2010). Study on the total water

pollutant load allocation in the Changjiang (Yangtze

River) Estuary and adjacent seawater area. Estuarine,

Coastal and Shelf Science, 86, 331-336.

Li, R. Z. et al. (2005). Delphi-AHP method for allocation

of waste loads in a region. Journal of Harbin Institute

of Technology, 37, 37-39.

Liu, J. M. and Lu, X. L. (2010). Design and realization of

the software for computer room management based on

VB and batch processing. Journal of Henan Institute

of Engineering, 22, 43-45.

Mostafavi, S. A. and Afshar, A. (2011). Waste load

allocation using non-dominated archiving multi-

colony ant algorithm. Procedia Computer Science, 3,

64-69.

Ni, J. R. et al. (2001). Total waste-load control and

allocation based on input-output analysis for Shenzhen,

South China. Journal of Environmental Management,

61, 37-49.

Ongley, E. D. et al. (2010). Current status of agricultural

and rural non-point source pollution assessment in

China.Environmental Pollution, 158, 1159-1168.

Xiong, Y. et al. (2007). Combining AHP with GIS in

synthetic evaluation of eco-environment quality-A

case study of Hunan Province, China. Ecological

Modelling, 209, 97-109.

Zhou, Z.S. et al. (2004). The method to use MATLAB in

VB software. Computer Development and Application,

17, 21-23.

SIMULTECH 2011 - 1st International Conference on Simulation and Modeling Methodologies, Technologies and

Applications

408