Web Based Decision Support System Using Geoinformatics Techniques for Irrigation Water Management in a Near Real Time Environment .

Mahmood Khan

Research output: ThesisDoctoral Thesis

79 Downloads (Pure)

Abstract

A significant increase in irrigated agricultural productivity over the last few decades has protected the world from episodes of hunger and food shortages. Presently however, water resources are under high stress due to rapid population growth, climate change, recent droughts and increased competition among agricultural, industrial, commercial, environmental and domestic sectors. This situation demands for an efficient water management. Due to the above mentioned factors and environmental concerns, there is much pressure on existing water users in Australia to improve their water productivity. In order to do so, there is a need to have access to reliable data on water availability, quantity and quality to provide the necessary foundation for an efficient
management of water resources. Advances in remote sensing (RS), information technology (IT), geographical information systems (GIS) and hydrological modelling offer water resource managers a novel way to accurately obtain spatial data on actual water use, water demand, and allocation and distribution of water in a near real time environment.

In this context, decision support tools can play an important role, therefore the key focus of this research was on the development of a Decision Support System (DSS), which can hold data, functions and models, to make decision-making transparent and interactive through use of a smart water management system. This study developed a DSS called Coleambally Integrated River Information System (IRIS) for the Coleambally Irrigation Area (CIA) which is a web-based information management system with a focus on time series, geospatial, hydrological, climatic and RS data.

The hydrological models developed and validated in this study are:
i) irrigation water demand forecasting (IWDF) based on data mining techniques using spatio-temporal data; and
ii) automation of RS-based processes for estimation of actual
evapotranspiration (ETa).

These models were tested for the 2007-08, 2008-09, 2009-10 and 2010-
11 summer seasons. These two models can both be used across different
spatial scales i.e., from farm level (one farm) to irrigation system level (the whole CIA).

The DSS for the CIA was developed using the scripting language PHP (Hypertext Preprocessor). The DSS uses several open source software components, including a standard Linux environment with an Apache web server, PostgreSQL, PostGIS and MapServer. The Apache web server is used to connect clients with the HTTP (Hypertext Transfer Protocol) web server. Through the Apache web server, the DSS stores various uploaded data in PostgreSQL databases and the file system, and displays the required data to a user with visually attractive graphical plots. It uses the PostgreSQL database as an object relational database
management system (RDBMS). It allows admin users (administrators of the database) to modify information about data types, functions and access methods that are stored in system catalogue tables without changing any hardcoded procedures in the source code. Minnesota Map Server is used in this DSS for visualisation of the maps.

The IWDF for a seven day period has been achieved using data mining techniques based on spatio-temporal data. This forecasting model can be
applied on different scales ranging from subsystem to system level i.e. farm, node (collection of farms) and irrigation area level to gain accurate knowledge of future irrigation demands. The different data mining techniques used in this study were decision tree (DT), artificial neural network (ANN), decision forest (using SysFor), support vector machine (SVM) and logistic regression. In this process firstly data was collected from three different sources, namely water delivery statements, meteorological data and RS satellite data. The collected data was preprocessed using a novel technique based on knowledge of irrigation
engineering and data mining. Models were built on the pre-processed dataset based on the above mentioned five different techniques. The experimental results indicated that as a result of data pre-processing, the quality of our dataset increased significantly. The prediction accuracies obtained by the models on the unprocessed dataset were 51% by SysFor, 44% by DT, 38% by ANN, 35.5% by SVM and 34.5% by logistic regression. However, a higher accuracy percentage was obtained by the models on our proposed pre-processed dataset i.e. 78% by SysFor, 74% by DT, 61% by ANN, 64% by SVM and 56% by logistic regression.
Moreover, among the five different techniques/models, the prediction
performance of SysFor was found to be the best followed by DT and ANN. Applying these models on the dataset from the year 2009/10, the water demand predicted by SysFor closely matched the actual water consumed. The accuracy of closeness produced by SysFor was found to be 97.5% which was followed by DT and ANN whose closenesses were found to be 96% and 95% respectively.

The RS process for estimation of actual evapotranspiration (ETa) was carried out by mapping the study area using Landsat 5 TM satellite images and applying a simplified hybrid classification approach mainly based on a supervised classification algorithm. The Surface Energy Balance System (SEBS) model was used to map spatial daily ETa from 19 Landsat 5 TM satellite images covering various parts of the cropping season in 2009-10 and 2010-11. For summer 2009-10, the daily ETa values ranged from 0.2mm to 10.6mm and for winter 2010, the daily ETa were a minimum of 0.21mm to a maximum of 5.4mm. From this model it was found that the RS-based energy balance algorithm, coupled with ground data, can be an efficient and reliable method for estimation of
evapotranspiration (ET) at different spatial scales.

The holistic methodology adopted in this study for developing a DSS is very simple and cost effective and is specifically designed and developed for a particular hydrological response unit (the CIA). The DSS consists of the hydrological database, hydrological models and a user friendly interface. The demand forecasting model using data mining techniques is useful for improved irrigation water management in the CIA as well as in other irrigation systems located in arid and semi-arid regions around the globe. The developed tool will help irrigation managers and farmers in a very practical sense to overcome the risks associated with over and under application of irrigation water by matching the demand and supply in a near real time environment. This DSS will ultimately lead to solutions which will allow more production with less use of water. It will provide opportunities for reducing non-beneficial use of water in the irrigation
system. This will be of immense value to the irrigation companies and the farming community as it will act as a complete management information system.
Original languageEnglish
QualificationDoctor of Philosophy
Awarding Institution
  • Charles Sturt University
Supervisors/Advisors
  • Hafeez, Muhammad, Principal Supervisor
  • Islam, Zahid, Co-Supervisor
Award date19 Nov 2015
Place of PublicationAustralia
Publisher
Publication statusPublished - 2015

Fingerprint

decision support system
water management
irrigation
data mining
artificial neural network
remote sensing
water demand
farm
logistics
water
water resource
irrigation system
energy balance
Landsat
information system
productivity
hydrological response
information management
hunger
hydrological modeling

Cite this

@phdthesis{89dc2f7b5a1a467fb135d98aa77df064,
title = "Web Based Decision Support System Using Geoinformatics Techniques for Irrigation Water Management in a Near Real Time Environment .",
abstract = "A significant increase in irrigated agricultural productivity over the last few decades has protected the world from episodes of hunger and food shortages. Presently however, water resources are under high stress due to rapid population growth, climate change, recent droughts and increased competition among agricultural, industrial, commercial, environmental and domestic sectors. This situation demands for an efficient water management. Due to the above mentioned factors and environmental concerns, there is much pressure on existing water users in Australia to improve their water productivity. In order to do so, there is a need to have access to reliable data on water availability, quantity and quality to provide the necessary foundation for an efficientmanagement of water resources. Advances in remote sensing (RS), information technology (IT), geographical information systems (GIS) and hydrological modelling offer water resource managers a novel way to accurately obtain spatial data on actual water use, water demand, and allocation and distribution of water in a near real time environment.In this context, decision support tools can play an important role, therefore the key focus of this research was on the development of a Decision Support System (DSS), which can hold data, functions and models, to make decision-making transparent and interactive through use of a smart water management system. This study developed a DSS called Coleambally Integrated River Information System (IRIS) for the Coleambally Irrigation Area (CIA) which is a web-based information management system with a focus on time series, geospatial, hydrological, climatic and RS data.The hydrological models developed and validated in this study are:i) irrigation water demand forecasting (IWDF) based on data mining techniques using spatio-temporal data; andii) automation of RS-based processes for estimation of actualevapotranspiration (ETa). These models were tested for the 2007-08, 2008-09, 2009-10 and 2010-11 summer seasons. These two models can both be used across differentspatial scales i.e., from farm level (one farm) to irrigation system level (the whole CIA).The DSS for the CIA was developed using the scripting language PHP (Hypertext Preprocessor). The DSS uses several open source software components, including a standard Linux environment with an Apache web server, PostgreSQL, PostGIS and MapServer. The Apache web server is used to connect clients with the HTTP (Hypertext Transfer Protocol) web server. Through the Apache web server, the DSS stores various uploaded data in PostgreSQL databases and the file system, and displays the required data to a user with visually attractive graphical plots. It uses the PostgreSQL database as an object relational databasemanagement system (RDBMS). It allows admin users (administrators of the database) to modify information about data types, functions and access methods that are stored in system catalogue tables without changing any hardcoded procedures in the source code. Minnesota Map Server is used in this DSS for visualisation of the maps.The IWDF for a seven day period has been achieved using data mining techniques based on spatio-temporal data. This forecasting model can beapplied on different scales ranging from subsystem to system level i.e. farm, node (collection of farms) and irrigation area level to gain accurate knowledge of future irrigation demands. The different data mining techniques used in this study were decision tree (DT), artificial neural network (ANN), decision forest (using SysFor), support vector machine (SVM) and logistic regression. In this process firstly data was collected from three different sources, namely water delivery statements, meteorological data and RS satellite data. The collected data was preprocessed using a novel technique based on knowledge of irrigationengineering and data mining. Models were built on the pre-processed dataset based on the above mentioned five different techniques. The experimental results indicated that as a result of data pre-processing, the quality of our dataset increased significantly. The prediction accuracies obtained by the models on the unprocessed dataset were 51{\%} by SysFor, 44{\%} by DT, 38{\%} by ANN, 35.5{\%} by SVM and 34.5{\%} by logistic regression. However, a higher accuracy percentage was obtained by the models on our proposed pre-processed dataset i.e. 78{\%} by SysFor, 74{\%} by DT, 61{\%} by ANN, 64{\%} by SVM and 56{\%} by logistic regression.Moreover, among the five different techniques/models, the predictionperformance of SysFor was found to be the best followed by DT and ANN. Applying these models on the dataset from the year 2009/10, the water demand predicted by SysFor closely matched the actual water consumed. The accuracy of closeness produced by SysFor was found to be 97.5{\%} which was followed by DT and ANN whose closenesses were found to be 96{\%} and 95{\%} respectively.The RS process for estimation of actual evapotranspiration (ETa) was carried out by mapping the study area using Landsat 5 TM satellite images and applying a simplified hybrid classification approach mainly based on a supervised classification algorithm. The Surface Energy Balance System (SEBS) model was used to map spatial daily ETa from 19 Landsat 5 TM satellite images covering various parts of the cropping season in 2009-10 and 2010-11. For summer 2009-10, the daily ETa values ranged from 0.2mm to 10.6mm and for winter 2010, the daily ETa were a minimum of 0.21mm to a maximum of 5.4mm. From this model it was found that the RS-based energy balance algorithm, coupled with ground data, can be an efficient and reliable method for estimation ofevapotranspiration (ET) at different spatial scales.The holistic methodology adopted in this study for developing a DSS is very simple and cost effective and is specifically designed and developed for a particular hydrological response unit (the CIA). The DSS consists of the hydrological database, hydrological models and a user friendly interface. The demand forecasting model using data mining techniques is useful for improved irrigation water management in the CIA as well as in other irrigation systems located in arid and semi-arid regions around the globe. The developed tool will help irrigation managers and farmers in a very practical sense to overcome the risks associated with over and under application of irrigation water by matching the demand and supply in a near real time environment. This DSS will ultimately lead to solutions which will allow more production with less use of water. It will provide opportunities for reducing non-beneficial use of water in the irrigationsystem. This will be of immense value to the irrigation companies and the farming community as it will act as a complete management information system.",
author = "Mahmood Khan",
year = "2015",
language = "English",
publisher = "Charles Sturt University",
address = "Australia",
school = "Charles Sturt University",

}

Web Based Decision Support System Using Geoinformatics Techniques for Irrigation Water Management in a Near Real Time Environment . / Khan, Mahmood.

Australia : Charles Sturt University, 2015. 260 p.

Research output: ThesisDoctoral Thesis

TY - THES

T1 - Web Based Decision Support System Using Geoinformatics Techniques for Irrigation Water Management in a Near Real Time Environment .

AU - Khan, Mahmood

PY - 2015

Y1 - 2015

N2 - A significant increase in irrigated agricultural productivity over the last few decades has protected the world from episodes of hunger and food shortages. Presently however, water resources are under high stress due to rapid population growth, climate change, recent droughts and increased competition among agricultural, industrial, commercial, environmental and domestic sectors. This situation demands for an efficient water management. Due to the above mentioned factors and environmental concerns, there is much pressure on existing water users in Australia to improve their water productivity. In order to do so, there is a need to have access to reliable data on water availability, quantity and quality to provide the necessary foundation for an efficientmanagement of water resources. Advances in remote sensing (RS), information technology (IT), geographical information systems (GIS) and hydrological modelling offer water resource managers a novel way to accurately obtain spatial data on actual water use, water demand, and allocation and distribution of water in a near real time environment.In this context, decision support tools can play an important role, therefore the key focus of this research was on the development of a Decision Support System (DSS), which can hold data, functions and models, to make decision-making transparent and interactive through use of a smart water management system. This study developed a DSS called Coleambally Integrated River Information System (IRIS) for the Coleambally Irrigation Area (CIA) which is a web-based information management system with a focus on time series, geospatial, hydrological, climatic and RS data.The hydrological models developed and validated in this study are:i) irrigation water demand forecasting (IWDF) based on data mining techniques using spatio-temporal data; andii) automation of RS-based processes for estimation of actualevapotranspiration (ETa). These models were tested for the 2007-08, 2008-09, 2009-10 and 2010-11 summer seasons. These two models can both be used across differentspatial scales i.e., from farm level (one farm) to irrigation system level (the whole CIA).The DSS for the CIA was developed using the scripting language PHP (Hypertext Preprocessor). The DSS uses several open source software components, including a standard Linux environment with an Apache web server, PostgreSQL, PostGIS and MapServer. The Apache web server is used to connect clients with the HTTP (Hypertext Transfer Protocol) web server. Through the Apache web server, the DSS stores various uploaded data in PostgreSQL databases and the file system, and displays the required data to a user with visually attractive graphical plots. It uses the PostgreSQL database as an object relational databasemanagement system (RDBMS). It allows admin users (administrators of the database) to modify information about data types, functions and access methods that are stored in system catalogue tables without changing any hardcoded procedures in the source code. Minnesota Map Server is used in this DSS for visualisation of the maps.The IWDF for a seven day period has been achieved using data mining techniques based on spatio-temporal data. This forecasting model can beapplied on different scales ranging from subsystem to system level i.e. farm, node (collection of farms) and irrigation area level to gain accurate knowledge of future irrigation demands. The different data mining techniques used in this study were decision tree (DT), artificial neural network (ANN), decision forest (using SysFor), support vector machine (SVM) and logistic regression. In this process firstly data was collected from three different sources, namely water delivery statements, meteorological data and RS satellite data. The collected data was preprocessed using a novel technique based on knowledge of irrigationengineering and data mining. Models were built on the pre-processed dataset based on the above mentioned five different techniques. The experimental results indicated that as a result of data pre-processing, the quality of our dataset increased significantly. The prediction accuracies obtained by the models on the unprocessed dataset were 51% by SysFor, 44% by DT, 38% by ANN, 35.5% by SVM and 34.5% by logistic regression. However, a higher accuracy percentage was obtained by the models on our proposed pre-processed dataset i.e. 78% by SysFor, 74% by DT, 61% by ANN, 64% by SVM and 56% by logistic regression.Moreover, among the five different techniques/models, the predictionperformance of SysFor was found to be the best followed by DT and ANN. Applying these models on the dataset from the year 2009/10, the water demand predicted by SysFor closely matched the actual water consumed. The accuracy of closeness produced by SysFor was found to be 97.5% which was followed by DT and ANN whose closenesses were found to be 96% and 95% respectively.The RS process for estimation of actual evapotranspiration (ETa) was carried out by mapping the study area using Landsat 5 TM satellite images and applying a simplified hybrid classification approach mainly based on a supervised classification algorithm. The Surface Energy Balance System (SEBS) model was used to map spatial daily ETa from 19 Landsat 5 TM satellite images covering various parts of the cropping season in 2009-10 and 2010-11. For summer 2009-10, the daily ETa values ranged from 0.2mm to 10.6mm and for winter 2010, the daily ETa were a minimum of 0.21mm to a maximum of 5.4mm. From this model it was found that the RS-based energy balance algorithm, coupled with ground data, can be an efficient and reliable method for estimation ofevapotranspiration (ET) at different spatial scales.The holistic methodology adopted in this study for developing a DSS is very simple and cost effective and is specifically designed and developed for a particular hydrological response unit (the CIA). The DSS consists of the hydrological database, hydrological models and a user friendly interface. The demand forecasting model using data mining techniques is useful for improved irrigation water management in the CIA as well as in other irrigation systems located in arid and semi-arid regions around the globe. The developed tool will help irrigation managers and farmers in a very practical sense to overcome the risks associated with over and under application of irrigation water by matching the demand and supply in a near real time environment. This DSS will ultimately lead to solutions which will allow more production with less use of water. It will provide opportunities for reducing non-beneficial use of water in the irrigationsystem. This will be of immense value to the irrigation companies and the farming community as it will act as a complete management information system.

AB - A significant increase in irrigated agricultural productivity over the last few decades has protected the world from episodes of hunger and food shortages. Presently however, water resources are under high stress due to rapid population growth, climate change, recent droughts and increased competition among agricultural, industrial, commercial, environmental and domestic sectors. This situation demands for an efficient water management. Due to the above mentioned factors and environmental concerns, there is much pressure on existing water users in Australia to improve their water productivity. In order to do so, there is a need to have access to reliable data on water availability, quantity and quality to provide the necessary foundation for an efficientmanagement of water resources. Advances in remote sensing (RS), information technology (IT), geographical information systems (GIS) and hydrological modelling offer water resource managers a novel way to accurately obtain spatial data on actual water use, water demand, and allocation and distribution of water in a near real time environment.In this context, decision support tools can play an important role, therefore the key focus of this research was on the development of a Decision Support System (DSS), which can hold data, functions and models, to make decision-making transparent and interactive through use of a smart water management system. This study developed a DSS called Coleambally Integrated River Information System (IRIS) for the Coleambally Irrigation Area (CIA) which is a web-based information management system with a focus on time series, geospatial, hydrological, climatic and RS data.The hydrological models developed and validated in this study are:i) irrigation water demand forecasting (IWDF) based on data mining techniques using spatio-temporal data; andii) automation of RS-based processes for estimation of actualevapotranspiration (ETa). These models were tested for the 2007-08, 2008-09, 2009-10 and 2010-11 summer seasons. These two models can both be used across differentspatial scales i.e., from farm level (one farm) to irrigation system level (the whole CIA).The DSS for the CIA was developed using the scripting language PHP (Hypertext Preprocessor). The DSS uses several open source software components, including a standard Linux environment with an Apache web server, PostgreSQL, PostGIS and MapServer. The Apache web server is used to connect clients with the HTTP (Hypertext Transfer Protocol) web server. Through the Apache web server, the DSS stores various uploaded data in PostgreSQL databases and the file system, and displays the required data to a user with visually attractive graphical plots. It uses the PostgreSQL database as an object relational databasemanagement system (RDBMS). It allows admin users (administrators of the database) to modify information about data types, functions and access methods that are stored in system catalogue tables without changing any hardcoded procedures in the source code. Minnesota Map Server is used in this DSS for visualisation of the maps.The IWDF for a seven day period has been achieved using data mining techniques based on spatio-temporal data. This forecasting model can beapplied on different scales ranging from subsystem to system level i.e. farm, node (collection of farms) and irrigation area level to gain accurate knowledge of future irrigation demands. The different data mining techniques used in this study were decision tree (DT), artificial neural network (ANN), decision forest (using SysFor), support vector machine (SVM) and logistic regression. In this process firstly data was collected from three different sources, namely water delivery statements, meteorological data and RS satellite data. The collected data was preprocessed using a novel technique based on knowledge of irrigationengineering and data mining. Models were built on the pre-processed dataset based on the above mentioned five different techniques. The experimental results indicated that as a result of data pre-processing, the quality of our dataset increased significantly. The prediction accuracies obtained by the models on the unprocessed dataset were 51% by SysFor, 44% by DT, 38% by ANN, 35.5% by SVM and 34.5% by logistic regression. However, a higher accuracy percentage was obtained by the models on our proposed pre-processed dataset i.e. 78% by SysFor, 74% by DT, 61% by ANN, 64% by SVM and 56% by logistic regression.Moreover, among the five different techniques/models, the predictionperformance of SysFor was found to be the best followed by DT and ANN. Applying these models on the dataset from the year 2009/10, the water demand predicted by SysFor closely matched the actual water consumed. The accuracy of closeness produced by SysFor was found to be 97.5% which was followed by DT and ANN whose closenesses were found to be 96% and 95% respectively.The RS process for estimation of actual evapotranspiration (ETa) was carried out by mapping the study area using Landsat 5 TM satellite images and applying a simplified hybrid classification approach mainly based on a supervised classification algorithm. The Surface Energy Balance System (SEBS) model was used to map spatial daily ETa from 19 Landsat 5 TM satellite images covering various parts of the cropping season in 2009-10 and 2010-11. For summer 2009-10, the daily ETa values ranged from 0.2mm to 10.6mm and for winter 2010, the daily ETa were a minimum of 0.21mm to a maximum of 5.4mm. From this model it was found that the RS-based energy balance algorithm, coupled with ground data, can be an efficient and reliable method for estimation ofevapotranspiration (ET) at different spatial scales.The holistic methodology adopted in this study for developing a DSS is very simple and cost effective and is specifically designed and developed for a particular hydrological response unit (the CIA). The DSS consists of the hydrological database, hydrological models and a user friendly interface. The demand forecasting model using data mining techniques is useful for improved irrigation water management in the CIA as well as in other irrigation systems located in arid and semi-arid regions around the globe. The developed tool will help irrigation managers and farmers in a very practical sense to overcome the risks associated with over and under application of irrigation water by matching the demand and supply in a near real time environment. This DSS will ultimately lead to solutions which will allow more production with less use of water. It will provide opportunities for reducing non-beneficial use of water in the irrigationsystem. This will be of immense value to the irrigation companies and the farming community as it will act as a complete management information system.

M3 - Doctoral Thesis

PB - Charles Sturt University

CY - Australia

ER -