The traditional role of hydrologic models has been as a scientific tool used by scientists and engineers for conducting research, supplementing monitoring activities, and assisting in design. However, as water resource planning and conservation shifts from national and state efforts to county and local efforts, hydrologic models are beginning to take on an educational role. Concerned citizens not only want to know what water quantity and quality levels are extreme, but how they became extreme. This research developed an educational tool that utilizes hydrologic modeling to help educate local government officials and concerned citizens. This tool, the Internet Watershed Educational Tool (InterWET), provides Internet-based water resource education, using as a case study the Spring Creek Watershed in central Pennsylvania. InterWET is available through the Penn State Agricultural and Biological Engineering homepage (http://server.age.psu.edu/) under the research web page in the area of Natural Resource Conservation and Management.

InterWET consists of a set of Internet web pages which use simplified, modular hydrologic models ("microworlds") geared toward education. Developed around the water resource issues of the Spring Creek Watershed, InterWET contains 15 microworlds that are combinations of five water resource components (surface runoff, groundwater, sediment erosion, in-stream nutrients, and fish populations) with three perspectives (a researcher, a conservationist, and a local official.) The web pages in InterWET use these microworlds as part of a self-guided learning environment. In addition, these web pages are sequenced to emphasize either the different water resource components or the different perspectives.

Several different Internet computing techniques were used to create the web pages and microworlds in InterWET. All web pages utilized Hypertext Markup Language (HTML), the formatting technique used to develop web pages. The microworlds presented hydrologic information using specific Internet computing techniques. The microworlds from the researcher perspective presented information on hydrologic processes (runoff, groundwater flow, erosion, in-stream nutrients, and fish populations) based on existing hydrologic, erosion, and population theory. These processes are represented by simple calculators modeled using JavaScript, a client-side scripting language. At the conservationist perspective, the microworlds used interactive maps (a geographic information system or GIS) to give a three-dimensional view of hydrologic monitoring data. This was accomplished through Java, a platform independent programming language that can be compiled to run within a web page. The local official perspective used JavaScript calculator microworlds similar to the researcher perspective. However, these calculators predicted the future change of water resource levels, based on a particular set of local policies, using the Generalized Watershed Loading Functions (GWLF) hydrologic model. Look-up tables, based on GWLF predictions, were used to predict the future change of water resource components as affected by a limited number of policies. Artificial neural networks were developed for water resource components affected by many policies.

Artificial neural networks are modeling methods patterned after the neurological system in the brain. These networks can be trained to replicate a data set (supervised) or to find patterns in a data set (unsupervised) and can greatly increase the speed of calculation. InterWET used supervised, Forward-Backward Propagation artificial neural networks. Using methods detailed in the literature, these types of networks were developed to approximate the GWLF predictions of change in level of different water resource components that had been grouped into eight change categories. After preliminary testing to determine the proper network architecture and training parameters, the artificial networks were trained with 1000 sets of data and tested with 100 sets of data. On average, the networks correctly predicted about 80% of the testing data sets. These trained artificial neural networks were then included in InterWET through a JavaScript calculator interface. The speed of calculation using the artificial neural network-based JavaScript calculator was compared to calculations using the actual PC-based GWLF model and a JavaScript version of the GWLF model. The artificial neural network version was 30 times faster than the PC-based version of GWLF and over 1000 times faster than the JavaScript version of GWLF.

InterWET was designed to be a hydrologic model serving in a purely educational role, not a scientific role. While it is based on established hydrologic modeling and monitoring techniques, many simplifying assumptions were made to bring out only the most important underlying concepts for each water resource component. The values predicted by InterWET should not be used for actual design or decision-making. Rather, the concepts brought out in InterWET, such as the effects of land use on runoff and erosion, should be considered when assessing designs or making decisions. Much more detailed analysis should always be performed for new structures or policies which will impact water resources. By portraying hydrology from many perspectives, InterWET allows local officials and others to be more knowledgeable and reflective about the impacts of their decisions on water resources.

InterWET provides a valuable contribution to the Spring Creek Watershed Community, the Educational Community, and the Scientific Community. For the Spring Creek Watershed Community, InterWET can act as an educational resource for decision makers. Local government and conservation groups can incorporate InterWET into their existing programs. In addition, InterWET can act as a foundation for further use of the Internet for education and information delivery.

However, the contribution of InterWET can also apply to a much larger Educational Community. As a stand-alone, self-guided learning environment, InterWET teaches about different water resources and different perspectives. High school and college courses can use different parts of InterWET to supplement lessons on water resources and the environment. InterWET also can serve as a blueprint for the development of other Internet-based educational models.

Moreover, InterWET can contribute to the Scientific Community. InterWET is one of only a handful of hydrologic models, which have been adapted to the Internet. It provides examples of how new modeling techniques like JavaScript and Java can be employed for computer modeling. In addition, InterWET employs design features like microworlds, which provide a basis for converting hydrologic modeling for the traditional scientific role to an educational role. Finally, InterWET makes use of artificial neural networks to greatly increase the speed of calculation in hydrologic models. The combination of hydrologic modeling with Internet modeling with artificial neural networks shows the Scientific Community a new approach that can have numerous applications in many fields of research.

Looking at the possible uses for the Spring Creek Watershed Community, the Educational Community, and the Scientific Community, InterWET fulfills the goal of providing a hydrologic model with an educational role. Based on sound educational and hydrologic theory and utilizing the newest technologies and computational techniques, InterWET gives a glimpse of future hydrologic model applications. Not just tools for scientists and engineers, InterWET allows more people to be better informed about their impact on water resources.
 

LIST OF FIGURES

LIST OF TABLES

ACKNOWLEDGMENTS

Chapter 1. Introduction

Chapter 2. Development of the Internet Watershed Educational Tool (InterWET)

    2.1. Abstract
    2.2. Problem Statement and Objectives
    2.3. Design Framework for InterWET
        2.3.1. Theory
        2.3.2. Design Details
        2.3.3. Microworld Sequences
    2.4. Applying InterWET for Watershed Education
    2.5. Summary
    2.6. References

Chapter 3. Techniques for Hydrologic Modeling on the Internet

    3.1. Abstract
    3.2. Introduction
    3.3. Background
    3.4. Techniques
        3.4.1. Hypertext Markup Language (HTML)
        3.4.2. CGI (Common Gateway Interface) Scripts
        3.4.3. JavaScripting
        3.4.4. Java
        3.4.5. Third-Part Software: Helper Applications, Plug-Ins and ActiveX, and Server-Side Software
        3.4.6. Techniques Summary
    3.5. Examples
        3.5.1. Internet Watershed Educational Tool (InterWET)
        3.5.2. SWAT On Line
        3.5.3. USGS Water Resources Web Site
        3.5.4. Enviromapper for Watersheds
    3.6. Conclusion
    3.7. References

Chapter 4. Use of Artificial Neural Networks to Facilitate Hydrologic Modeling on the Internet

    4.1. Abstract
    4.2. Introduction
    4.3. Background
        4.3.1. Techniques for Internet-Based Hydrologic Modeling
        4.3.2. Internet Watershed Educational Tool (InterWET)
        4.3.3. Artificial Neural Networks and Hydrologic Applications
    4.4. Methods
        4.4.1. Step One: GWLF for Spring Creek Watershed
        4.4.2. Step Two: Artificial Neural Networks of GWLF Output
        4.4.3. Step Three: JavaScript Artificial Neural Networks
    4.5. Results
        4.5.1. Generation of GWLF Training Data
        4.5.2. Artificial Neural Network Training and Testing
        4.5.3. JavaScript Implementation
    4.6. Conclusions
    4.7. References

Chapter 5. InterWET User's Guide

    5.1. InterWET Site Description
    5.2. InterWET Web Site Use
    5.3. Scope of InterWET

Chapter 6. Summary and Conclusions

    6.1. Structure of Educational Internet-Based Hydrologic Models
    6.2. Techniques for Internet-Based Hydrologic Modeling
    6.3. Using Artificial Neural Networks for Hydrologic Modeling
    6.4. Using InterWET
    6.5. Contributions to Society
    6.6. Future Work Based On InterWET
    6.7. Final Thoughts

Appendix A. DETAILS FOR RESEARCHER PERSPECTIVE CALCULATORS

    A.1. Microworld: Runoff 1 (R1)
    A.2. Microworld: Groundwater 1 (G1)
    A.3. Microworld: Sediment 1 (S1)
    A.4. Microworld: Nutrients 1 (N1)
    A.5. Microworld: Fish 1 (F1)

Appendix B. DETAILS FOR CONSERVATIONIST PERSPECTIVE MAPS

    B.1. Microworld: Runoff 2 (R2)
    B.2. Microworld: Groundwater 2 (G2)
    B.3. Microworld: Sediment 2 (S2)
    B.4. Microworld: Nutrients 2 (N2)
    B.5. Microworld: Fish 2 (F2)

Appendix C. DETAILS FOR LOCAL OFFCIAL PERSPECTIVE CALCULATORS

    C.1. Microworld: Runoff 3 (R3)
    C.2. Microworld: Groundwater 2 (G3)
    C.3. Microworld: Sediment 3 (S3)
    C.4. Microworld: Nutrients 3 (N3)
    C.5. Microworld: Fish 3 (F3)
 

Figure 2.1. Location of Spring Creek Watershed

Figure 2.2. Runoff Calculator Display from Researcher Perspective

Figure 2.3. Runoff Map Display

Figure 2.4. Two Microworld Sequences for InterWET

Figure 3.1. InterWET's JavaScript Erosion Calculator

Figure 3.2. InterWET's Java Groundwater ActiveMaps GIS

Figure 3.3. Input Module for SWAT On Line

Figure 3.4. Real-Time Stream Data

Figure 3.5. Enviromapper for Watersheds

Figure 4.1. Typical Artificial Neural Network Setup

Figure 4.2. Location of Spring Creek Watershed

Figure 4.3. Subwatersheds of the Spring Creek Watershed

Figure 4.4 Municipalities of the Spring Creek Watershed

Figure 4.5. JavaScript Calculator for Delivered Sediment using Artificial Neural Networks

Figure A.1. Runoff JavaScript Calculator

Figure A.2. Groundwater JavaScript Calculator

Figure A.3. Sediment JavaScript Calculator

Figure A.4. Nutrient JavaScript Calculator: Wet Year

Figure A.5. Brown Trout Habitat Suitability JavaScript Calculator

Figure B.1. ActiveMaps Java Runoff Map

Figure B.2. ActiveMaps Java Groundwater Map

Figure B.3. ActiveMaps Java Sediment Map

Figure B.4. ActiveMaps Java Nutrient Map

Figure B.5. ActiveMaps Java Fish Map

Figure C.1. JavaScript Calculator for Predicting Surface Runoff Change

Figure C.2. JavaScript Calculator for Delivered Sediment Change

Figure C.3. JavaScript Calculator for Predicting Brown Trout Population Change

Table 2.1. Specific Issues and General Categories of Concern for Spring Creek Watershed

Table 2.2. Microworlds within InterWET

Table 3.1. Summary of Techniques for Hydrologic Modeling over the Internet

Table 4.1. Inputs to the GWLF Water Transport File for the Spring Creek Watershed

Table 4.2. Input to the GWLF Nutrient File for the Spring Creek Watershed

Table 4.3. Policy Choices Modeled with GWLF for the Spring Creek Watershed

Table 4.4. GWLF Outputs Affected by Policy Choices

Table 4.5. Percent Change Categories

Table 4.6. Policy Scenarios, Output Values, and Modeling Techniques for GWLF Outputs

Table 4.7. Training and Testing Error for Artificial Neural Networks for Predicting Change Categories for GWLF Outputs

Table C.1. Effect of Temperature and Sediment Change on Brown Trout Populations
 

I first would like to thank several organizations for funding and support for this research. Thank you to Drs. Dennis Buffington, Harvey Manbeck, and Roy Young for providing financial support from the Penn State Agricultural and Biological Engineering Department. All of the staff in the department, especially Wanda, Sue, Mike, and Donna, have helped countless times with all of the administrative and paper work that went into this research. Thanks also goes to Dean Reischman and his staff at the College of Engineering Dean's Office for support with the GAANN Fellowship and Dean's Fellowship. Finally, I would like to acknowledge the U.S. Department of Education for their outstanding support for this research with the GAANN Fellowship in Environmental Engineering.

I would like to personally thank each of my committee members. Dr. Paul Robillard, thanks for acting as stand-in advisor for Jim during his sabbatical. Dr. Peggy Johnson, thanks for your insight during our discussions and use of your monitoring equipment. Dr. Mirna Urquidi-MacDonald thanks for your help and guidance through our meetings and emails. A special thanks goes to my advisor, Dr. Jim Hamlett. Thank you for patience during the initial stages of this research, reading endless drafts of proposals until the research came together. Also thanks for all of the meetings throughout the years, leading and guiding, while being a good friend and a sympathetic ear to all of the woes of being a graduate student. In the future, when I am in the position to advise others, I have a great role model.

Next, I want to thank my wife, Rachel. Your interest in environmental education has helped get me away from the computer more often and made me remember why I went to Ag. Engineering. Thanks for listening to all of my ideas, letting me hog the computer, and occasionally enduring the reading of a first draft for some paper.

My final thanks to my Lord and Savior, Jesus Christ. Jesus answered, "Everyone who drinks this water will be thirsty again, but whoever drinks the water I give him will never thirst. Indeed, the water I give him will become in him a spring of water welling up to eternal life." John 4:13-14 NIV

Back to Table of Contents

Next to Chapter 1