,.f.a...\..va. .4 f...” x l...\4’no?4.ofl.-$\‘b< *2! “4&5? .1. 29:91 (5'51... r Ivtlrv 3?? a I"! I... r 0:). 4:1. . :41. art. y’arl: 9v... 375:9... .4 .. r : ... :..; .: 1: .9 or) S.‘nvtl(r(...l> Ii! 11.3.1: . for! l. I)? kilo-(Iii) I‘I(rll.ia.. fvtrlrrfo J . it!!! 4: (5.. .;l (I: . 03:... all u vaf. 2! 1.52.: , (1...? u .. 5.3:. 331.10 .llal 2" _ . . .13.;erqu rt: {19‘ . ...r1.r1tl!§lvfr life!- . fl» '4) 3.9;.) {(61154 (rlrvf I: ll .éyitsltoxtvaiol- .fr 5a.!!! (0) . 3.! 17:! i. .4 .z . . 1!.“7/(1 re rldrabxttlgi. §:,f¢ (1"!!! Ir: I, 4 ..rrl,li(ria ...)21.:A.|115.4)7 I!..«I.r;vl\..sE'a(:. . A . ;. 1., Ii)}f::l:1;l.. .13); (II;I(' I» titrrfplllilrlf:l . . , v. .2 .l o I'II’.»4..r.o.lrr-EIJ.IE>I .rsr/ Ann I‘IV . MICHIGAN STATE U c . uIY LIBRARIES ““593 lllllllllllllllllllllllllllllllllll l|lllllllllllllllllllll 7 7 3 1293 01706 9786 This is to certify that the thesis entitled WATER QUALITY ASSESSMENT OF THE CHILUNG RIVER USING LANDSAT THEMATIC MAPPER AND AIRBORNE MULTI—SPECTRAL SCANNER IMAGES: TAIPEI, TAIWAN presented by Kin Man Ma has been accepted towards fulfillment of the requirements for Masters of Science , Resource Development degree in Major professor Date ”32¢ I ?7 0-7639 MS U is an Affirmative Action/Equal Opportunity Institution LIBRARY Michigan State University PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. DATE DUE . DATE DUE DATE DUE a '3 0 2 "l0 2 El 2000 .9;- \ 20m ‘7 1‘ $3 63T0120t200’l OCT 0 1 2002 1/98 c/CIRC/DaIaDueip65-pj 4 WATER QUALITY ASSESSMENT OF THE CHILUNG RIVER USING LANDSAT THEMATIC MAPPER AND AIRBORNE MULTI- SPECTRAL SCANNER IMAGES: TAIPEI, TAIWAN By Kin Man Ma A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTERS OF SCIENCE Department of Resource Development 1997 ABSTRACT WATER QUALITY ASSESSMENT OF THE CHILUNG RIVER USING LANDSAT THEMATIC MAPPER AND AIRBORNE MULTI-SPECTRAL SCANNER IMAGES: TAIPEI, TAIWAN By Kin Man Ma Due to the rapid, economic growth of Taiwan and the lack of effective wastewater treatment, Taiwan rivers, such as the Chilung River, have become polluted with increases in suspended sediments and biological oxygen demand, and decreases in dissolved oxygen. The Taiwan EPA field teams collected suspended sediments concentration (SSC) water quality samples. Remotely sensed image reflectance values have been correlated with SSC. Three multi-temporal Landsat TM images fi‘om 1993 to 1995 and an April 1996 Airborne MSS image of the Chilung River were obtained. Landsat TM image reflectance values were normalized, transformed into radiance values, and regressed against monthly SSC data. The Airborne MSS image was geometrically resampled using ground control points and reflectance values were regressed against SSC and turbidity values. Regression analyses yielded significant correlations between SSC and Landsat TM Band 4, R2 = 0.44, p = 0.009. Natural log(SSC) and TM Band 4 correlation was more significant, R2 = 0.57, p = 0.002. Band ratios generated significant correlations between ln(SSC) and the (Band 4/Band 2) ratio, R2 = 0.66, p < 0.001. Regression results between Airborne MSS Bands 8 and 9, and SSC levels yielded R2 of 0.38 and 0.37, both p = 0.001. The turbidity and ln(turbidity) regressions against Airborne MSS Band 8 yielded similar results, R2 = 0.38, p = 0.001. This thesis is dedicated to my parents, Ding Fun Ma and Lai Yung Lam Ma, who have continually encouraged and supported me in my life and studies so that I could complete my Masters degree. ayxzzées‘caeaxa assess ass 0 é‘éiiittififiirs paiiniaasaa+ETSREAassas’aasamaa fiifiifig‘fi 0 iii ACKNOWLEDGEMENTS While conducting this international research project for my thesis, there were numerous people, organizations and institutions who have greatly assisted me and provided support for the completion of this work, both in Taiwan and in the Michigan State University (MSU) community. Very significant assistance was received from the following people and groups in Taiwan: I am very indebted to the Taiwan (Republic of China) Environmental Protection Administration (Project No.: EPA-85-L105-03-20) for the opportunity to participate in the foregoing research project from September 1995 to June 1996 and for funding this project conducted by the research team at the Center for Geographic Information Systems Research at Feng Chia University (FCU GIS Center). I thank the Taiwan EPA for their vision to improve the Taiwan environment through research. Dr. Tien-Yin (Jimmy) Chou, Director of the FCU GIS Center, and a former graduate of the Department of Resource Development, was instrumental in fulfilling my desire to conduct research in East Asia. I joined the FCU GIS Center research team to work as a research assistant on the Taiwan EPA project and learned more about Asia and Taiwan as an exchange student through the MSU— FCU research exchange agreement. I am very thankful that he provided the data images for my thesis. I am also grateful to many of the research assistants, fellow iv graduate students and staff for their academic and social support and encourage- ment, since I was the international student adjusting to a new culture. Thanks especially to Hui-Yen (Liz) Chen who worked on the project research team, guided me through many technical aspects of the research, and helped me understand scientific materials written in Chinese. Mei-Ling (Milly) Yeh, a fellow graduate student, was very understanding and supportive of my intellectual curiosity and showed me the joy of learning. Additional thanks to the FCU GIS Center for their financial support while my wife and I lived in Taichung. Dr. Lung-Shih Yang, Dean of the College of Management, Feng Chia University (FCU), was the principal investigator of the Taiwan EPA project and he was supportive of my research efforts and was the first one to teach me about remote sensing. Dr. Shanshin Ton, aSsociate professor of environmental engineering, FCU, was the environmental engineer on our research team. He has been very encouraging and open to sharing his expertise in water related research in person and over E-mail. Mrs. Agnes Chen, Director of the Mandarin Chinese Language Center, FCU, and other teachers challenged and encouraged my study of the Chinese language so that I could be a more effective researcher in Taiwan. The following people and organizations in the MSU community provided significant assistance: Travel to Taiwan was made possible in part through funding from the Department of Resource Development and the Institute for International Agriculture. My Masters thesis committee members, Drs. Eckhart Dersch, Jon Bartholic, and Michael Kamrin, have been a tremendous source of wisdom and support during my thesis, especially my advisor, Dr. Dersch: Dr. Eckhart Dersch, Professor of Resource Development, Michigan State University (MSU), has been overwhelmingly supportive through every step of my research journey from Michigan to Taiwan. He helped me establish connections with Dr. Jimmy Chou at FCU and be challenged me intellectually about develop- ment issues as well as thinking and acting proactively. Dr. Jon Bartholic, Professor of Resource Development, Director of the Institute of Water Research, MSU, has challenged me to think critically about water research. Dr. Michael Kamrin, Professor of Resource Development, Outreach Coordinator, Institute for Environmental Toxicology (IET), in 1994, gave me an opportunity to be a research assistant in IET and has continually endeavored to support my academic pursuits. He has challenged me intellectually and has been a role model for taking complex research ideas and concepts, digesting them and effectively explaining them to the general public. The staff of IET, Carole Abel, Darla Conley, and Carol Chvojka, have been a tremendous encouragement and help throughout the years. IET has been a very stimulating environment for my intellectual growth regarding environmental issues with the intellectual prowess of Dr. William Cooper and a fellow graduate student, Jon MacDonagh-Dumler. Dr. Cynthia Fridgen, Chair of the Department of Resource Development, has been very encouraging and supportive of my research since 1994. Recognition needs to be given to the Geography Department, MSU, for allowing me to conduct my image data analysis on their computers. The following people provided significant assistance: Dr. René Hinojosa, Chair of the Department of Geography, was supportive of multi-disciplinary connections and allowed me to use the UNIX computers for my research. Dr. Daniel Brown, Assistant Professor of Geography, MSU, taught me about GIS. His advice about image reflectance and the use of the UNIX computers in the Spatial Analysis Laboratory for analyzing the Chilung River images have been invaluable. Mr. James Brown, UNIX System Administrator, also assisted in UNIX account maintenance and answered questions regarding the Lab computers. Jiunn-Der (Geoffrey) Duh, a fellow geography Ph.D. student, has taught me much about remote sensing and Taiwan. He was also very helpful to assist me in writing the Chinese dedication to my parents. The following people also provided additional significant assistance: Dr. David Lusch, Senior Research Specialist, Center for Remote Sensing and Geographic Information Science, MSU, has taught me the details and wonders of satellite remote sensing. He has been very helpful with critical analysis of my research methods and evaluation of my thesis. Dr. Patricia Norris, Assistant Professor of Agricultural Economics and Resource Development, MSU, gave me an opportunity to use my GIS skills on her research project in Barry County. I have learned how to work hard and have been challenged to think about the economic implications of land use. vii I also would not have completed this thesis without the following large network of spiritually and emotionally supportive friends, family, and the Lord God Almighty: Miss Alexandra Saddler and Dr. Jennifer DeLapp have been long-time friends who have encouraged me with prayers and challenged me- to keep my spiritual perspective even when the challenges seemed overwhelming. The people at my church, University Reformed Church, have been instru- mental for keeping me sane during this long process. Pastor Tom Stark has consistently'preached the Word of God challenging me to love and serve others whenever and wherever I am. Dr. Mark Whalon, Professor of Entomology, Interim Director of the Pesticide Research Center, MSU, has been faithful to provide spiritual support and academic encouragement. The Brinkman family, KiSoon Kim, Dr. David and Elaine Prestel, Dr. Leland and Joella Cogan were always concerned about my academic and spiritual welfare. There are many others, though they are too numerous to name here. Thanks to Ralph and Utami Rahardja DiCosty, fellow graduate students, since we have had many opportunities to pray, encourage and rejoice with one another about our mutual challenges and triumphs. I also cannot forget my long-time doctor friends, Drs. David and Sharon (Blanchard) Gunasti, who have been very supportive in every way. The people from the West Taichung Conservative Baptist Church in Taiwan were crucial in helping me spiritually and emotionally during the stressful time of conducting research in Taiwan. Pastor Lin was encouraging as well as the JiaLe small group, consisting of the family of Sam, Annie, Ellen and Tina Yang and viii others. They were extraordinary in welcoming my wife and I into their hearts and challenging us to seek the Lord so we could both survive and flourish during our Taiwan experience. Graduate InterVarsity Christian Fellowship have been a very encouraging and intellectually challenging group of people. Thanks, especially to Randy Gabrielse, Amy Hirshman, Sally Gaff, Tom Johnson, Mike Pasquale, Michelle Mackey, Sandeep Rao, Ying-ying (Sabrina) Wu, JiaNing Lai, and Reji Varghese. I do not have the space to mention all the others. Hyunsook Lee Ma, my wife, has been truly supportive through the entire process of journeying to Taiwan for research and data collection as well as the recent challenges of thesis writing. I am indebted to her for the enduring patience, love, joy and perseverance she has shown me ever since we met. I consider it an honor to call her my wife. My parents, Ding Fun Ma and Lai Yung Lam Ma, and-my brother, Tony Ma, have been supportive of my studies throughout my life. It is now appropriate to save the greatest thanks for last. The Lord God Almighty, has been my ultimate spiritual support because He is my wisdom. He has given me peace, joy and perseverance when circumstances were beyond my control regarding the use of the Taiwan data images. His protection while traveling to Taiwan, Korea, Hong Kong, New York City and back to Michigan has been reassuring. “To God be the Glory” ...... for this thesis. November 1997 East Lansing, Michigan The Lord God Almighty can see remotely... 11 “If I say, ‘Surely the darkness will hide me and the light become night around me,’ 12 even the darkness will not be dark to you; the night will shine like the day, for darkness is as light to you. 13 For you created my inmost being; you knit me together in my mother’s womb. l4I praise you because I am fearfully and wonderfully made; your works are wonderful, I know that full well.” — Psahn 139: 11-14 TABLE OF CONTENTS LIST OF TABLES .................................................................................................... xiv LIST OF FIGURES ................................................................................................... xv LIST OF ABBREVIATIONS ................................................................................. xvii CHAPTER I INTRODUCTION ........................................................................................................ 1 A. BACKGROUND .......................................................................................................... 1 B. RESEARCH AREA ..................................................................................................... 3 1. Channel Modifications and Wastewater Treatment .......................................... 4 2. Suspended Sediments ......................................................................................... 5 3. Turbidity .............................................................................................................. 5 4. Remote Sensing ................................................................................................... 6 C. STATEMENT OF THE PROBLEM .............................................................................. 10 D. RESEARCH QUESTIONS ......................................................................................... 12 E. ORGANIZATION ...................................................................................................... 13 F. BENEFICIARIES OF THE RESEARCH ....................................................................... 13 CHAPTER II ' LITERATURE REVIEW ........................................................................................... 14 A. SUSPENDED SEDIMENTS ....................................................................................... 14 B. REMOTE SENSING RESEARCH .......................... ' ..................................................... 16 CHAPTER III RESEARCH DESIGN AND METHODOLOGY ..................................................... 20 A. WATER QUALITY DATA .......................................................................................... 20 B. LANDSAT-5 THEMATIC MAPPER SATELLITE IMAGES ............................................. 20 1. Data Use and Analysis Limitations .................................................................. 22 2. Test Pixel Selection ........................................................................................... 24 3. Taiwan’s Transverse Mercator 2 ° Coordinate System .................................... 25 4. Normalization of Landsat TM Reflectance Values ................................... - ....... 25 5. Transformation from Reflectance Values to Radiance Values ........................ 25 xi C. AIRBORNE MULTI-SPECTRAL SCANNER IMAGES ........................ h ........................... 26 1. Differential Global Positioning System ............................................................ 27 2. Water Quality Data ........................................................................................... 28 3. Image Correction and Resampling ................................................................... 31 4. Selection of Test Pixel Positions ...................................................................... 32 D. DATA ANALYSIS METHODOLOGY ........................................................................... 33 CHAPTER IV DATA ANALYSIS, RESULTS AND DISCUSSION .............................................. 34 A. LANDSAT TM RELATED DATA ................................................................................ 34 1. Taiwan EPA Water Quality Data ..................................................................... 34 a. Suspended Sediments .................................................................................... 34 b. Turbidity ......................................................................................................... 35 2. Landsat TM Images .......................................................................................... 36 a. Normalization of Landsat TM images ........................................................... 36 b. Transformation from Reflectance Values to Radiance Values ..................... 43 c. Linear Regression Analysis and Results ....................................................... 46 d. Multiple Regression Analysis and Results .................................................... 47 B. AIRBORNE MSS RELATED DATA ........................................................................... 51 1. Water Quality Data ........................................................................................... 51 a. Suspended Sediments .................................................................................... 51 b. Turbidity ......................................................................................................... 53 2. Regression Analysis and Results ...................................................................... 53 a. Suspended Sediments .................................................................................... 53 b. Turbidity ......................................................................................................... 58 C. DISCUSSION .......................................................................................................... 63 1. Landsat TM image analysis .............................................................................. 63 2. Airborne MSS image analysis ........................................................................... 65 CHAPTER V SUMMARY, CONCLUSIONS AND RECOMMENDATIONS...... ........................ 67 A. SUMMARY ........................................................................................................... 67 B. CONCLUSIONS .................................................................................................. 69 C. RECOMMENDATIONS ...................................................................................... 71 xii APPENDICES APPENDIX A MAP OF EAST ASIA APPENDIX B NORTHERN TAIWAN, TAIPEI REGION AND SURROUNDING CITIES ....... APPENDIX C EASTERN CHILUNG RIVER, TAIPEI, TAIWAN APPENDIX D WESTERN CHILUNG RIVER, TAIPEI, TAIWAN APPENDIX E TAIPEI MAP LEGEND BIBLIOGRAPHY 73 73 74 75 76 77 78 LIST OF TABLES Table 1.1 -- Taiwan Categories of River Pollution for Water Quality Parameters ..... 2 Table 1.2 -- 1986 and 1993 Comparison of Untreated Wastewater Sources Flowing into the Chilung River ................................................................. 2 Table 1.3 — Landsat Thematic Mapper Spectral Bands ............................................... 9 Table 1.4 — 1987 to 1995 Monthly Water Quality Parameter Averages of Chilung River Bridge Test sites (N anHu Bridge to BaiLing Bridge) 1 1 Table 2.1 -- Comparison of Landsat TM and Airborne MSS Spectral Band Sensitivity ................................................................................................ 17 Table 3.1 -- Water Quality Data and Landsat TM Band Reflectance Values ............ 23 Table 3.2 -- Water Quality Data and 4/25/96 Airborne MSS Reflectance Values ...................................................................................................... 29 Table 4.1 -- Landsat TM Bands 1 to 4 values, Normalized from the 1/09/95 image to the 12/11/95 image ................................................................... 42 Table 4.2 -- Landsat TM Bands 1 to 4 values, Normalized from the 6/28/93 image to the 12/11/95 image ................................................................... 42 Table 4.3 -— Landsat TM Normalized and Transformed Band Radiance Values ....................................................................................................... 44 Table 4.4 -- Landsat Thematic Mapper Data Regression Results ............................. 45 Table 4.5 -- Airborne Multi-Spectral Scanner Data Regression Results ................... 56 xiv LIST OF FIGURES Figure 1.1 -- Earth’s Atmospheric Windows ................................................................. 8 Figure 2.1 -- Effects of Suspended Silt upon spectral properties of water. .................................................................................................. 15 Figure 4.1 -- Graph of the SSC levels of Chilung River bridge test sites ......................................................................................................... 35 Figure 4.2 -- Landsat TM images, 1/09/95 to 12/11/95 normalized regression graph (TM Band 1). .............................................................. 38 Figure 4.3 -- Landsat TM images, 1/09/95 to 12/11/95 normalized regression graph (TM Band 2). .............................................................. 39 Figure 4.4 -- Landsat TM images, 6/28/93 to 12/11/95 normalized regression graph (TM Band 1). ............................................................. 40 Figure 4.5 -- Landsat TM images, 6/28/93 to 12/11/95 normalized regression graph (TM Band 2) ............................................................... 41 Figure 4.6 -- Graph of Correlation between Landsat TM Band 4 and SSC. ................................................................................................. 48 Figure 4.7 -- Graph of Correlation between Landsat TM Band 4 and ln(SSC). ........................................................................................... 49 Figure 4.8 -- Graph of Correlation between Landsat TM (Band 4/Band 2) Ratio and ln(SSC). ...................................................... 50 Figure 4.9 -- Graph of the SSC levels of Chilung River test sites, 4/26/96. ................................................................................................... 52 Figure 4.10 -- Graph of the Turbidity levels of Chilung River test sites, 4/26/96. ........................................................................................ 54 XV Figure 4.11 -- Graph of Correlation between Turbidity values and SSC values. Figure 4.12 -- Graph of Correlation between Airborne MSS Band 8 . and SSC. Figure 4.13 —- Graph of Correlation between Airborne MSS Band 8 and In (SSC) Figure 4.14 -- Graph of Correlation between Airborne MSS Band 8 and Turbidity Figure 4.15 -- Graph of Correlation between Airborne MSS Band 8 and In (Turbidity) 55 59 60 61 62 BOD CAMS DGPS DN DO GCPs GPS MSS nm ROC ROC EPA SDD SSC TM TM2° um LIST OF ABBREVIATIONS Biological Oxygen Demand Calibrated Airborne Multi-Spectral Scanner Differential Global Positioning System Digital Number Dissolved Oxygen Ground Control Points Global Positioning System Multi-Spectral Scanner nanometer Republic of China (official name of Taiwan) Republic of China’s Environmental Protection Administration Secchi Disk Depth suspended sediments concentration (mg/L) Thematic Mapper Taiwan’s National Transverse Mercator 2 ° coordinate micrometer CHAPTER I INTRODUCTION A. Background Taiwan is located in eastern Asia about 130 km (80.6 miles) southeast of the southern province of Fujian of the People’s Republic of China. Taiwan’s official name is the Republic of China (ROC) (See Appendix A). Taiwan’s economic development and the growth of the chemical, manufacturing, and electronics industries over the past 30 years have produced many commodities for the world. However, the release of untreated industrial chemicals and heavy metals associated with the production of these commodities and the growth of untreated residential, livestock and municipal landfill leachate wastewater have caused an increase in biological oxygen demand, an increase in the amount of suspended solids concentration, and an increase in other environmental pollutants, such as nitrogen compounds and phosphorous compounds. All of these contributing factors have lowered the water quality of the Taiwan river systems (Republic of China, Environmental Protection Administration (ROC EPA), 1994a). Out of Taiwan’s twenty-one main rivers, 25% of the total surface area of those rivers are either polluted or badly polluted, and out of Taiwan’s 29 secondary rivers, 22.7% of the total surface area of those rivers are either polluted or badly polluted (ROC EPA, 1994a). The different categories of river pollution are shown in Table 1.1 for the following four water quality parameters: dissolved oxygen (DO), 1 2 biological oxygen demand (BOD), suspended sediments concentration (SSC), and ammonia nitrogen (NH3 -N). Table 1.1 — Taiwan Categories of River Pollution for Water Quality Parameters Categories of River Pollution Unpolluted Slightly Polluted Badly Water Quality Polluted Polluted Parameter Dissolved Oxygen (DO), mg/L > 6.5 4.6 ~ 6.5 2.0 ~ 4.5 < 2.0 Biological oxygen demand (BOD), < 3.0 3.0 ~ 4.9 5.0 ~ 15 > 15.0 93m Suspended sediments < 20 20 ~ 49 50 ~ 100 > 100 concentration (SSC), mg/L Ammonia nitrogen (NHs-N), mg/L < 0.5 0.5 ~ 0.9 1.0 ~ 3.0 > 3.0 Source: Adapted from ROC EPA (1994a), p. 115. Table 1.2 — 1986 and 1993 Comparison of Untreated Wastewater Sources Flowing into the Chilung River Wastewater Sources Units: Metric Tons of Water/ Day Year Residential Industrial Livestock Municipal Agricultural Total Landfill Effluents Leachate 54.67 12.62 18.33 8.02 -- N/A -- 93.64 19861 (Metric Tons) 58.38% 13.48% 19.57% 8.57% -- N/A -- 100.0% (percentage of total) 139.78 24.63 2.60 0.26 0.37 167.64 19932 (Metric Tons) 83.38% 14.69% 1.55% 0.16% 0.22% 100.0% (percentage of total) Sources: 1. ROC EPA, December 1989; 2. ROC EPA, June 1994. 3 Taipei City, the capital of the Republic of China, has grown steadily over the past 20 years. The 1993 population in Taipei City’s Chilung River watershed was 1.89 million people and is projected to be 2.08 million people by the year 2001 (ROC EPA, 1994b). With population growth, there has been corresponding growth in the amount of untreated residential wastewater released into the Chilung River which flows through the city. The total amount of untreated residential and industrial wastewater has almost tripled from 1986 to 1993. Residential wastewater as a percentage of the total daily effluent load has increased tremendously from 58.4% to over 83% in only 7 years (see Table 1.2). Nearly all of Taipei City's untreated residential sewage and industrial wastewater flows into the Chilung River causing oxygen depletion and changes in BOD, SSC, and NH3. Field research teams have commented abut the hydrogen sulfide smell and the bubbling of the water’s surface while taking water quality samples (ROC EPA, 1994a). As of 1996, there were no signs of living fish species in the downstream region of the Chilung River. During high tide, some saltwater fish species from the Taiwan Strait have swum into the Chilung River via the saltwater undercurrents (see Appendix B). B. Research Area The Chilung River begins in the Ching Tong mountains in southern Taipei County and flows west through Chilung City and through Taipei City and then it joins with the Tanshui River at Guan Du. The Chilung River is approximately 87 km (53.9 miles) long and has a surface area of 501 sq. km (193.4 sq. mi.), and is 250 meters wide (820 feet) near the mouth of the Chilung River and as narrow as 50 meters (164 feet) near the NanHu bridge. The region of the Chilung River entering 4 Taipei at the N anHu Bridge receives the largest amount of pollutants and therefore the eastern section of the research study area begins from Taipei City's NanHu Bridge and extends westward toward the mouth of the Tanshui River (see Appendices B, C, D and E). 1. Channel Modifications and Wastewater Treatment The slope of the Chilung River is relatively flat with slopes ranging from 1 meter/500 meters or 0.2% at the NanHu bridge on the eastern side of the research area to 1 meter/700 meters or 0.14% at the BaiLing Bridge on the western side of the research area (see Appendices C and D). The average slope of the entire Chilung River is about 1 meter/118 meters or 0.85%. From about November 1991 to late 1993, there were two (2) large channel modifications of the Chilung River for 1) increasing the flow rate of the Chilung River to prevent flooding during the summer monsoon season, and 2) to more effectively transport the increasing amounts of untreated residential wastewater from Taipei City residents toward the Taiwan Strait (see Appendices B, C and D). There is presently one primary wastewater treatment plant near the DaZhi Bridge which is treating approximately 5% of the residential wastewater from Taipei City (ROC EPA, 1996). Beginning in 1994, Taipei City began a plan to build a network of wastewater diversion pipes and another primary wastewater treatment plant near the mouth of the Tanshui River close to the Taiwan Strait. The pipes would extend west along the Chilung River toward the point where it flows into the Tanshui River and then to the mouth of the Tanshui River (see Appendix B). There the sewage wastewater and storm drainage water would undergo primary wastewater treatment and then be emitted about 100 meters offshore into the 5 Taiwan Strait (ROC EPA, 1994b). As of June 1996, the wastewater pipes have yet to be completely built because of engineering problems and miscalculations regarding the total amount of wastewater which would bypass the Chilung River and flow toward the wastewater treatment plant before exiting into the Taiwan Strait. If the Chilung River doesn’t receive any wastewater from the Taipei City residents or storm drainage water, the total water volume of the river would decrease, and ironically the river’s concentration of pollutants may even increase. 2. Suspended Sediments The suspended solids concentration (SSC) in the water is a physical indicator of water quality (Dzurik, 1990). When the SSC in the water is high, there are more suspended solids in the water, whether soil, algae, residential or industrial waste. If the SSC is low, the water can be assumed to have less silt, clay and fragments of organic matter, which indicates in general, better water quality. 3. Turbidity Turbidity levels are another physical measure of water quality. Turbid water decreases the transparency of the water body and this opacity can be measured by a Secchi Disk. A Secchi Disk is a plastic disk approximately 20 cm in diameter with four quadrants. The quadrants are alternately black and white. This disk is lowered gradually into the water body until the white quadrants are no longer visible. This depth is then recorded as the Secchi Disk Depth (SDD). The SDD is related to the sediment content within the water body. Another “measure of turbidity is the use of nephometric turbidity units (NTU), which is measured by the intensity of light which passes through a water sample.”1 1 James B. Campbell, Introduction to Remote Sensing, (The Guilford Press, New York, NY, 6 Water of high turbidity decreases the intensity of the light in a manner that can be related to sediment content. Therefore, as sediment concentration increases the water body ceases to act as a “dark” object, absorbing most of the solar radiation, but slowly becomes more and more of a “bright” object, reflecting increasing amounts of light (Campbell, 1987). 4. Remote Sensing Remote sensing is defined by Campbell (1987) as “the science of deriving information about the earth’s land and water areas from images acquired at a distance. It usually relies upon measurement of electromagnetic energy reflected or emitted from the feature of interest.”2 After the invention of the airplane remotely sensed photographs and images were taken flying over the areas of interest. Moreover, with the 1960’s discovery of space exploration and satellite technology, the United States in 1972 launched the Landsat 1 (Land Satellite), one of the first earth-orbiting satellites designed specifically for observation of the earth’s land areas. Remotely sensed images record the “interaction of electromagnetic radiation with the earth’s surface. Electromagnetic radiation comes from several sources, such as changes in energy levels of electrons, decay of radioactive substances and the thermal motion of atoms and molecules.”3 Nuclear reactions in the sun produce a full range of electromagnetic radiation from ultraviolet rays, wavelengths of 0.30 micrometers (um) to far infrared wavelengths, 7.0 — 15.0 um. However, the earth’s atmosphere of water vapor, ozone, and carbon dioxide gases absorbs certain portions 1987), p. 408. 2 Ibid., p. 2. 3 Ibid., p. 21. 7 of the electromagnetic spectrum preventing some solar radiation from penetrating the earth. There are a series of important atmospheric windows in which solar radiation can be transmitted. Therefore, only certain electromagnetic wavelengths within the earth’s atmospheric windows can be used for measuring solar radiation reflectance from the earth’s landscape. e.g. visible light, and the near-infrared, mid-infrared, and far-infrared wavelengths (see Figure 1.1). Since 1972 many satellites such as the Landsat 2, 3, 4, and 5 were launched during the late 1970’s and early 1980’s. The developers of the Landsat 5 Thematic Mapper (“TM”) satellite system set the sensitivity for each of the bands at specific wavelength intervals for principal applications of water body penetration, delineating water bodies, and for soil moisture discrimination (see Table 1.3). The Landsat 5 satellite orbits the earth at an altitude of 705 km (438 miles) in a near polar orbit at a 982° angle to the equator. The satellite crosses the equator on the north-to-south portion of each orbit at 9:45 AM local sun time. Each orbit takes approximately 99 minutes, with about 14.5 orbits a day. The earth rotates approximately 2,752 km (1,710 miles) during each orbital pass, therefore the satellite sensor passes over the same 185 km (115 miles) swath area every 16 days. All of the Landsat TM band numbers 1 to 7 have pixel resolution sizes of 30 meters (98.4 feet), except the thermal band number 6 which has a pixel resolution of 120 meters (393.6 feet) (see Table 2.1, page 15). .m.N .95me Jim .Q .mefi NZ £5? 3oz .395 30.350 mars .mfimewm 835% S eotosmcotfi fionmawo .m moamw HwQRSOm 95353, omaognmogua.‘ mbuuam I HA wanna c0339: mnm_w_> 023382820 .6 cocqcompm 3:88ro 85 omm Omm I Q Ii‘r] 42 98$ «.2 .2 a an 2me armada . NfibmAN wofivdm Hammwm mm mv wv mm wfiflmm mmEmE 800.3 wmmm.mm mmvodm vommfim vm 9%» 9S» ow cmflquogN mm\wm\® vmmbdfi wfimvdm ooovAm vmmmSm Hm mv bdm. méw ENmQ mmEmB $5.2 NSVBN owemfim Naomdm mm mdm mm mdm 523052 mmEmE www.mNH Nmmm.mm mevflm wflmvdm mdm 95‘ mdm on «mzwcoao mm\wm\© 3%:me muvaam «#molndm afiwamm vfivcmm muwamm «#vnmm Humanism 35 3am wouzaagoz 63:9:qu wouzmfiaoz wosszgoZ SE. SE EB 28 owvmnm omeH owns: mmhdmfi may 3 owmfim meNE 9%. 89¢ 63:58qu .339, v o» H mvsam SE. gamwamq I may @3238 983 a: .2 Em .3 033. 36de mwmohm wmsmmm mmmmdv ad miwm mdm 93V mfiflwm @990: momm.NH mwmohm Nvmfihm vmmmwm m m.wm vm mv ENwQ mmBo: vwmmz: wvmmfim wmbmmm omgdm 9w m.mm m.mm Q? 525052 $50: vfivnam mnwswm «Mmenwm Saddam Efiwnam mficsam Nuvqwm Hiram 3mm 3am 63:23:02 wouflaagoz uoszagnoz 6329982 29 EB 29 EB omgam omwfi: owes: 32:3 0:» 3 owns: 35¢: 23 88am 60238.82 .33?» v 3 H madam SB ”Rwanda l Haw ozmfi 43 b. Transformation from Reflectance Values to Radiance Values The Landsat TM normalized reflectance values were then transformed into radiance values according to Formula 2 mentioned in Chapter 3. DN — B G DN = digital number value recorded G = slope of response function (channel Gain) L = spectral radiance measured (over the spectral bandwidth of the channel) B = intercept of response function (channel Offset) 2) L: Source: Lillesand & Kiefer (1994), pp. 534-535. The Gain and Offset values for each of the Landsat 5 Thematic Mapper bands for data processed in Taiwan are listed below: Gain B {Offset} Landsat TM #1 1659939989 2523108752 Landsat TM #2 0850992808 2416819498 Landsat TM #3 1241057044 1.452036688 Landsat TM #4 1227673274 1853786631 Source: ROC EPA (1996), p. 102. The results of the transformed Landsat TM radiance values for all three images and their four TM bands are shown in Table 4.3. 44 .NJ. 98 :3. .H.m $33. 89a H6385 6me m2 .2 8M .3 Eng. «Sud $9....de momme mmmeN SHdH H3.HN :de Hmmwm NmHm.m mom wfiflmm meNE vvawd mmmwdH EwNHm wwomdm Hoo._vH www.mm avodm omafim Hbvmd cHN amsmwaoAN «93$ 33$ @8ka 38.3” mHmmHN www.mH www.ma 8me mmmfim mmmfim «E ENmQ meNE wooed oEmdN 536m wooH.NN 2..de HHvSN women momdm 5.3.“. mg 828082 meNE 32.3 $3.3 $3.3 HENNN vaNH mmmdm 85.3 «3.3 9.86 cm SSEGEO meNE amowd $3.3 mfimém $55.3 «8.3 mmoém 3.9mm mmmdv $.me Hm mafidmm 85o: ommmd 95de @2QO vomaHN omm.mH mmodm vadmw mmmdm mmed Hv ENdQ mmBQH mmwmd Hwbde wNHmdm wvmmdm wmmHH www.mu 3.9mm. mHvdm NHmm.m No 885082 mmBQH confia vmvméH mummHm mbdeN n6 de mm mm wovvd m.mH mfidwm mm\HH\NH Sofia $5de $8.8 $8.3 m wH mm vm vomwd wH augmeHoAN mQHQNH ommad mmdeH $323 85.2.. m 95 am pm $.23 05 ENNQ mQHde human wmmm.mH ummodm 33.3 m wH mm on 33d 93 888082 mQHQmH 823 33.3 vame $8.3 ad «a 9mm 9mm $.36 mSm 828880 mQHHRH good“ NNNbHH mmeSN «8me m mH mm mm. «2.9m mdm smamz mm\HH\mH v 685 m 68am a 83m H waam v Haam SE. n vndm SE. a 68am SE. H 33m SE. SE. SE. SE. SE. A€w8v ovum 35G 8.88889 8uomm8aurH. 88%8688 88%838 8.82 .8qu .8qu .8qu ADmmV 8 0mm cwgum owa8H $35» oonwmwam Hawm 6288088988 U8» 60838.32 SE. 8368an l m4. @358 45 Table 4.4 — Landsat Thematic Mapper Data Regression Results Water Thematic Mapper Band and R 2 p Quality Band Combinations F- test *(p < 0.001) Parameter SSC Band 1 0.137 1.905 0.193 SSC Band 2 0.006 0.068 0.799 SSC Band 3 0.094 1.246 0.286 SSC Band 4 ** 0.444 9.599 0.009 10.611 * (Band 4) - 17.844 SSC Band 4/Band 2 ** 0.575 16.238 0.002 SSC Band 4/Band 3 ** 0.560 15.255 0.002 SSC In (Band 4) 0.389 7.633 0.017 SSC In (Band 4/Band 3) 0.406 8.189 0.014 SSC In (Band 4/Band 2) 0.443 9.554 0.009 SSC Band 2, Band 3, Band 4 ** 0.844 18.054 * 0.001 In LSSC) Band 2 0.065 0.761 0.402 In (SSC) Band 3 0.271 4.081 0.068 In (SSC) Band 4 ** 0.574 16.176 0.002 In (SSC) (Band 4/Band 2) ** 0.658 23.055 * 0.001 In (SSC) (Band 4/Band 3) ** 0.629 20.332 0.001 In (SSC) (Band 3/Band 4) 0.268 4.389 0.058 In (SSC) (Band 2/Band 4) 0.335 6.046 0.030 In ASSC) In (Band 1) 0.279 4.641 0.052 In (SSC) In (Band 2) 0.086 1.134 0.308 In (SSC) In (Band 3) 0.268 4.393 0.058 In (SSC) In (Band 4) ** 0.518 12.878 0.004 In (SSC) In (Band 4/Band 3) ** 0.492 11.618 0.005 In (SSC) In (Band 4/Band 2) ** 0.551 14.704 0.002 In (SSC) In (Band 2), In (Band 3), ** 0.851 19.002 * 0.001 In (Band 4) - 14.105 * (1n B#2) + 7.575 * (In B #3) + 0.893 (In B#4) + 30.714 ** yielded relatively high R2 values Table by Kin M. Ma (1997) 45 Table 4.4 — Landsat Thematic Mapper Data Regression Results Water Thematic Mapper Band and R 2 p Quality Band Combinations F- test *(p < 0.001) Band 4 ** 0.444 9.599 10.611 * - 17.844 4/Band 2 ** Band 4/Band ln ln ln (SSC) In (Band 2), In (Band 3), ** 0.851 19.002 * 0.001 In (Band 4) — 14.105 * (ln B#2) + 7.575 * (In B #3) + 0.893 (In B#4) + ** yielded relatively high R2 values Table by Kin M. Ma (1997) 46 0. Linear Regression Analysis and Results Each of the four normalized and transformed radiance band values were regressed against the SSC values. The regression of SSC and TM Band 4, 0.7 6 - 0.90 pm, yielded the only significant correlational relationship (R2 = 0.444, p = 0.009) (see Table 4.4 and Figure 4.6). Campbell (1987) and Kirk (1989) have shown that the relationship between SSC and spectral radiance has a logarithmic relationship at increasingly higher SSC concentrations (see Figure 2.1, page 15). Therefore, additional regression analyses were performed between the natural log (In) of the dependent variable, SSC, and the other independent band variables. As shown in Table 4.4, the regression of the ln(SSC) dependent variable and the Band 4 radiance values yielded the highest significant single band correlational relationship (R2 = 0.574, p = 0.002) with a regression equation: ln(SSC) = O.157*(Band 4) + 2.582. The graph for this correlation is in Figure 4.7. In addition, Braga et a1. (1993) performed regression analyses on band ratio combinations. Since increasing amounts of sediment in the water will shift the amount of reflected energy to the longer visible and near-infrared wavelengths, there may be a relationship with the Infrared/Red, Band 4/Band 2 and Band 4/ Band 3 combinations. The regression of SSC with the Band 4/Band 2 band ratio yielded R2 = 0.575, p = 0.002 and the Band 4/ Band 3 band ratio yielded R2 = 0.560, p = 0.002. When ratio band regression relationships were tested with ln(SSC), the logarithmic relationships also yielded significantly high R2 values. For example, the regression of the ln(SSC) variable and the (Band 4/Band 2) ratio yielded R2 = 0.658, p < 0.001 (see Table 4.4 and Figure 4.8) and the regression of ln(SSC) variable and 47 the (Band 4/Band 3) ratio yielded R2 = 0.629, p = 0.001 with the following regression equations: ln(SSC) = 7.440 * (Band 4/Band 2) + 2.723; ln(SSC) = 4.049 * (Band 4/Band 3) + 2.465. When the log/log function relationship was tested, ln(SSC) and ln(Band 4/Band 2) yielded R2 = 0.551, p = 0.002. d. Multiple Regression Analysis and Results When multiple regression analyses were computed the Log/Log relationship between ln(SSC) and the ln(Band2), ln(Band3), ln(Band4) band combinations yielded the highest R2 = 0.851, p < 0.001 in the following multiple regression equation: ln(SSC) = -- 14.105*(ln Band_2) + 7.575*(1n Band_3) + 0.893*(ln Band_4) + 30.714 When the Landsat TM 2, 3 and 4 band combination were regressed against SSC, it yielded a significant correlation of R2 = 0.844 and p < 0.001 with the following regression equation: SSC = --31.109 *(Band_2) + 23.983 *(Band_3) + 19.423 *(Band_4) + 592.721. 48 250 Landsat TM Band 4] SSC Correlation SSC levels (mg/L) 200.. 150 .. 100.. 50 .. i 4 e e Landsat TM Band 4 Radiance Values 10 Figure 4.6 — Graph of Correlation between Landsat TM Band 4 and SSC. R2 = 0.444, p = 0.009 SSC = 10.611 * (Band_4) -- 17.844 49 Landsat TM Band_4 I ln(SSC) Correlation In (SSC) 0 ; ; r 4 0 2 4 6 8 10 Landsat TM Band 4 Radiance Values R2 = 0.574, p = 0.002 In (SSC) = 0.157 * (TM Band_4) + 2.582 Figure 4.7 — Graph of Correlation between Landsat TM Band 4 and ln(SSC). 5O Landsat TM (Band 4JBand 2) Ratio] ln (SSC) Correlation 0 0.05 01 035 02 025 0:3 0.35 Landsat TM (Band_4/Band 2) Ratio of Radiance Values R2: 0.658, p < 0.001 In (SSC) = 7.440 * (TM Band_4/TM Band_2) + 2.723 Figure 4.8 — Graph of Correlation between Landsat TM (Band 4/Band 2) Ratio and ln(SSC). 51 B. Airborne MSS related data 1. Water Quality Data a. Suspended Sediments All of the water quality data were from the April 26, 1996 field collection day. The SSC values for the test points showed that the water quality of the Chilung River was slightly polluted t0 polluted according to Taiwan’s water quality standards in Table 1.1, page 2, with an average SSC value of 24.9 mg/L, a high value of 94.4 mg/L and a low value of 3.0 mg/L. The average of the BOD values was 9.7 mg/L, with a high value of 13.7 mg/L and a low value of 6.0 mg/L. The average BOD values places it in the “polluted” rivers category, 5.0 — 15.0 mg/L, as shown in Table 1.1, page 2. The SSC readings at the last three sites, Nos. 23, 24, and 25, were especially high, averaging 89.0 mg/L, because the river is very shallow, only 2 meters deep, east of the MinQuan Bridge and it may have been influenced by the ongoing nearby construction on the Taipei highway (see Figure 4.9). 52 Suspended Sediment Cone. Levels of Chilung River test sites, 4I26I96 100 90» ’ 70.. 40«» ISSC levels (mg/L) I 30~~ 20~ 10~ o e 10 1‘5 2'0 25 Test Sites from Chilung River’s Mouth to NanHu Bridge Figure 4.9 — Graph of the SSC levels of Chilung River test sites, 4/26/96. The test sites begin from the mouth of the Chilung River and the site numbers increase going eastwards along the Chilung River toward the NanHu Bridge (see Table 3.2, page 29 for reference). 53 b. Turbidity Also, the average of the turbidity values was 61.3 NTU, with a high value of 106 NTU and a low value of 46 NTU. The turbidity readings at the last three sites, Nos. 23, 24, and 25, were especially high, averaging 101.6 NTU, because the river was very shallow, only 2 meters deep, east of the MinQuan Bridge and it may have been influenced by the ongoing nearby construction on the Taipei highway (see Figure 4.10). Turbidity levels are highly significantly correlated with SSC values, R2 = 0.855, F-test 136.11, p g 0.001 (see Figure 4.11). 2. Regression Analysis and Results 3. Suspended Sediments Linear regression analyses of Airborne MSS Bands 1 to 5 and SSC did not yield any significant correlational relationships, at a 95% confidence level. All of the regression analyses of the Airborne MSS Bands 6 to 10 were significant (p .<_ 0.05). The linear regressions of SSC and MSS Bands 8 and 9 yielded the strongest correlations, R2 = 0.377, p = 0.001, and R2 = 0.366, p = 0.001, respectively (see Table 4.5 and Figure 4.12). Logarithmic combinations were also tested and again correlations with MSS Bands 6 to 10 were significant though they were not as highly correlated as the above MSS Band 8 correlation. For example, when ln(SSC) was regressed against MSS Band 8, R2 =0.305, p = 0.004 (see Figure 4.13). Many band ratio regressions were also tested such as the MSS Band 9/Band 7 (Infrared Red/Red) ratio though almost all of the correlation results were not significant at the 95% confidence level, and therefore not included in the results. 54 Turbidity Levels of Chilung River test sites, 4126/96 120 100.- . o 80w 60 ~- Turbidity (NTU) 40-- 20-- O 5 1O 15 20 25 Test Sites from Chilung Mouth to NanHu Bridge Figure 4.10—Graph of the Turbidity levels of Chilung River test sites, 4/26/96. The test sites begin from the mouth of the Chilung River and the site numbers increase going eastwardsalong the Chilung River toward the NanHu Bridge (see Table 3.2, page 29 for reference). 55 Correlation between Turbidity and SSC levels 100 90~ 80i- 60~- 50,, 4o“ SSC levels (mg / L) 20 ‘l" 10«- 0 20 470 60 80 100 120 Turbidity levels (NTU) R2 = 0.855, F-test 136.11, p g 0.001 Figure 4.11 — Graph of Correlation between Turbidity values and SSC values. 56 Table 4.5 — Airborne Multi-Spectral Scanner Data Regression Results Water Airborne MSS Band and R 2 p Quality Band Combinations F- test *(p < 0.001) Parameter SSC Band 1 0.137 3.655 0.068 SSC Band 2 0.143 3.838 0.062 SSC Band 3 0.164 4.499 0.045 SSC Band 4 0.128 3.375 0.079 SSC Band 5 0.170 4.696 0.041 SSC Band 6 0.216 6.332 0.019 SSC Band 7 0.284 9.105 0.006 SSC Band 8 ** 0.377 13.910 0.001 1.495 * (MSS Band_8) — 34.197 SSC Band 9 ** 0.366 13.261 0.001 SSC Band 10 0.269 8.453 0.008 SSC In (Band 6) 0.173 4.828 0.038 SSC In (Band 7) 0.215 6.309 0.019 SSC In (Band 8) 0.274 8.667 0.007 SSC In (Band 9) 0.261 8.103 0.009 SSC In (Band 10) 0.199 5.717 0.025 In (SSC) Band 1 0.164 4.515 0.045 In (SSC) Band 2 0.154 4.183 0.052 ln(SSC) Band 3 0.176 4.910 0.037 In (SSC) Band 4 0.171 4.756 0.040 In (SSC) Band 5 0.193 5.485 0.028 In (SSC) Band 6 0.219 6.438 0.018 . 1n (SSC) Band 7 0.259 8.024 0.009 In (SSC) Band 8 ** 0.305 10.108 0.004 In (SSC) Band 9 0.285 9.146 0.006 In (SSC) Band 10 0.258 7.998 0.016 Turbidity Band 6 0.245 7.468 0.012 Turbidity Band 7 0.303 9.995 0.004 Turbidity Band 8 ** 0.378 13.987 0.001 Turbidity Band 9 ** 0.335 11.610 0.002 Turbidity Band 10 0.218 6.411 0.019 Table by Kin M. Ma (1997) 57 Table 4.5 — (Cont’d) '** 0.400 Turbidity Band 7, Band 8 7.348 0.004 —0.386 * (Band 7) + 1.757 * (Band 8) + 28.176 Turbidity Band 9, Band 10 ** 0.450 8.984 0.001 3.983 * (Band 9) — 3.141 * (Band 10) + 49.398 ln(Turbidity) Band 5 0.209 6.069 0.022 ln(Turbidity) Band 6 0.261 8.130 0.009 ln(Turbidity) Band 7 0.314 10.520 0.004 ln(Turbidity) Band 8 **0.377 13.928 0.001 ln(Turbidity) Band 9 0.322 10.912 0.003 ln(Turbidity) Band 10 0.210 6.124 0.021 ln(Turbidity) Band 7, Band 8, Band 9, **0.488 6.709 0.001 and Band 10 ln(Turbidity) In (Band 6) 0.208 6.036 0.022 Influrbidity) In (Band 7) 0.240 7.256 0.013 ln(Turbidity) In (Band 8) 0.280 8.932 0.007 ln(Turbidity) In (Band 9) 0.234 7.030 0.014 ln(Turbidity) In (Band 10) 0.160 4.371 0.048 ** yielded relatively high R 2 values Table by Kin M. Ma (1997) 58 b. Turbidity Linear regression analyses generated low correlations with R2 ranging from 0.20 to 0.30. However, using a logarithmic function the ln(turbidity) and MSS band 8 regression yielded R2 = 0.377 and p = 0.001 (Figure 4.14). Through multiple regression, the correlation coefficient of turbidity improved slightly. By regressing Bands 7 and 8 with turbidity the results yielded a correlation relationship of R2 = 0.400 and p = 0.004 with a regression equation of: Turbidity = --0.386 * (Band 7) + 1.757 * (Band 8) + 28.176 (see Table 4.5). By testing the logarithmic function, the highest correlation coefficient for ln(Turbidity) and MSS Band 8 was similar to the regression of SSC and MSS Band 8, R2 = 0.377 andp = 0.001. Another regression of Bands 9 and 10 and turbidity was significant, R2 = 0.450 and p = 0.001, with a regression equation of : Turbidity = 3.983 * (Band 9) — 3.141 * (Band 10) + 49.398. An additional multiple regression between ln(Turbidity) and MSS Bands 7, 8, 9, and 10 yielded R2 = 0.488 and p = 0.001, F-test = 6.709. The correlation relationship was significant though there is a low F-test and the individual p values for each band were not significant, with approximate p values of 0.20 and 0.50, signifying that MSS Bands 7 and 10 were forced into the regression equation. This caused the overall multiple regression equation of MSS Bands 7, 8, 9, and 10 and ln(Turbidity) to be less significant then the single band regression of ln(Turbidity) and MSS Band 8 already mentioned above. 59 Airborne MSS Band 8! SSC Correlation 100 90.. 0 30.. O 70.. 60~~ 50»- 40‘» SSC levels (mg/L) 20-— 10- 0 10 20 30 40 50 60 70 Airborne MSS Band 8 Reflectance Values R2 = 0.377, p = 0.001 ssc = 1.495 * (MSS Band_8) -- 34.197 Figure 4.12 — Graph of Correlation between Airborne MSS Band 8 and SSC. 60 Airborne MSS Band 8! ln(SSC) Correlation 5 4 5 ’ ’ O 4 8 .. . o 00 ’ 3.5 A 3 4’ :5: 2.5 5 2 .. 1.5 - 1 . 0.5 0 : . . . . . 0 1o 20 30 40 50 60 70 Airborne MSS Band 8 Reflectance Values R2 =0.305, p = 0.004 In (SSC) = 0.0435 * (MSS Band_8) + 1.136 Figure 4.13 — Graph of Correlation between Airborne MSS Band 8 and ln(SSC). ' 61 Airborne MSS Band 8/ Turbidity Correlation 120 100 Q 80. 60.. Turbidity (NTU) 40. 20. 0 10 20 30 40 50 6'0 70 Airborne MSS Band 8 Reflectance Values R2 = 0.377 andp = 0.001 Turbidity = 0.949 * (MSS Band 8) + 23.790 Figure 4.14 — Graph of Correlation between Airborne MSS Band 8 and Turbidity. 62 Airborne MSS Band 8/ In (Turbidity) Correlation 5 4.5 O o O 4 8 .. . o 00 ’ 3.5 E a » f 2 1.5 1 0.5 O . . . . 0 10 20 30 40 50 60 70 Airborne MSS Band 8 Reflectance Values R2 = 0.377 andp = 0.001 In (Turbidity) = 0.0131 * (MSS Band_8) + 3.573 Figure 4.15 — Graph of Correlation between Airborne MSS Band 8 and In (Turbidity) 63 C. Discussion 1. Landsat TM image analysis The linear regression of ln(SSC) and TM Band 4 yielded a significant correlation of R2 = 0.571, though it was not as highly predictive as Braga et al.’s (1993) study yielding a R2 = 0.74 for the correlation between total suspended solids and TM Band 4. Also Khorram and Cheshire’s (1985) study regressed the Landsat MSS Bands 4, 5, 6, and 7 complex band ratios and SSC and yielded a R2 = 0.64. These significant correlations support Campbell’s (1987) Figure 2.1 showing that high amounts of SSC will reflect more radiant energy in the red and infra-red wavelengths of light, such as Landsat TM Band 4. The differences in R2 values may be due to several factors: a) temporal variability of the SSC, b) inaccuracies in the image normalization process, and c) shallowness of the Chilung River body. a. Temporal Variability in SSC Water quality data should be temporally matched to the Landsat TM images. However, none of the Landsat TM images were temporally matched to the water quality monitoring dates and times. For example, there was usually a day difference though in the December 11, 1995 image, there is a four day difference in time from the December 7, 1995 water quality monitoring date. During the intervening days between the image date and the testing date, there may have been variations in sediment levels. As Table 1.4, page 11, shows, the SSC levels from 1987 to 1995 greatly varied between high and low values throughout the Chilung River. 64 b. Inaccuracies in Image Normalization Since there were an insufficient number of Taiwan EPA SSC water quality test sites from one monthly monitoring sample, three Landsat TM images were normalized to obtain a sufficient number of test data points. During the normalization process, the generated regression equations were very high ranging from R2 = 0.874 to R2 = 0.929, for TM Band 4. There was some loss of reflectance value in the normalization process since the R2 coefficient did not attain the perfect 1.00. This alteration of reflectance values from the original Landsat TM images may have contributed to lowering the Band 4/SSC R2 correlations in this study. c. Shallowness of the River body The Chilung river is a small, relatively shallow river with many sections less than 5 meters deep (ROC EPA, 1996). Landsat TM reflectance may penetrate the shallow water of the river into the riverbed. Moore’s (1978) research showed that solar radiation in the blue-green location, 0.52 pm, can penetrate 2 meters into clear water and all the way up to 20 meters in very, clear water. Therefore, this water penetration may lead to inaccuracies of estimating the amount of SSC from the water quality samples as opposed to estimating the depth of the riverbed. However, this reason may not be very important because the color of the Chilung River was black and very little penetration could occur. 65 2. Airborne MSS image analysis The linear and logarithmic relationship generated correlations of R2 = 0.37 7 for the regression of SSC and MSS Band 8 and R2 = 0.284 for the regression of SSC and MSS Band 7 (see Table 4.5). Curran et al.’s (1987) study yielded significant correlations at similar spectral wavelengths with R2 = 0.46. Also Miller et al.’s (1994) SSC research in Puerto Rico yielded a very significant R2 = 0.85 for regressing low concentrations of SSC at the 0.63—0.69 pm wavelength sensitivity which would be equivalent to this study’s MSS Bands 6 and 7. The large differences in R2 values may be due to several factors: a) temporal variability of the SSC, or b) the small 5 meter pixel resolution size combined with Airborne MSS image resampling inaccuracies and errors. a. Temporal Variability of the SSC Since the water quality data should temporally match the Airborne MSS image in time and space, there was possible variability of the SSC and turbidity between the day the Airborne MSS image was flown, 4/25/96, and the day that the water quality samples were taken, 4/26/96. As Table 1.4, page 11, shows, the SSC levels from 1987 to 1995 greatly varied between high and low values throughout the Chilung River. b. Small Pixel Resolution and Resampling Errors According to Gao and O’Leary (1997), the spatial resolution of the image has a significant role in the accuracy of quantifying SSC in water. Their study showed that a 10 meter pixel resolution produced the best correlation with the SSC data 66 (R2 = 0.97) and that as the pixel resolution diminished below 10 meters, the correlation of SSC and band reflectance decreased from R2 = 0.81 for a 5 meter pixel resolution down to R2 = 0.75 for a 1 meter pixel resolution image (Gao and O’Leary, 1997). The reasons are because even with the increased accuracy of using the Differential Global Positioning System (DGPS), there may be some variations and errors in rectifying and resampling the aerial image. “Geometric errors can still be present in the rectified image due to the lack of geometric controls over the water surface where no distinct features can be used as GCPs” (Gao and O’Leary, 1997). Also with the small 5 meter pixel resolution, the uncertainty of determining the exact location of the water quality reading is greater, especially if flowing river currents shift the field boat several meters to the right or left during the time that the DGPS reading is being taken. e CHAPTER V SUMMARY, CONCLUSIONS AND RECOMMENDATIONS A. SUMMARY Due to the rapid industrial development within Taiwan over the past 30 years, the accumulated releases of phosphates, nitrates, particulates and heavy metals, associated with industrial development have degraded the water quality of Taiwan’s rivers. This caused increases in biological oxygen demand, increases in the suspended sediments concentration (SSC), and decreases in dissolved oxygen. All of these contributing factors have lowered the water quality of the Taiwan river systems. The Chilungr River flowing through Taipei City is no exception. In order to combat the environmental degradation, Taiwan citizens rallied to pressure the Taiwan government to establish the Taiwan Environmental Protection Administration (Taiwan EPA) in 1987. Since 1987, the Taiwan EPA has been sending out monthly field teams to gather water quality monitoring data from the Chilung River and adjoining Tanshui River. However, this monitoring effort is costly and can only provide specific point samples of water quality monitoring data. Extensive research has been done using remotely sensed images of various water bodies, such' as lakes, coastal zones, and large rivers such as the Amazon River (Shimoda et al., 1986; Curran et al., 1987; Mertes, et al., 1994). . 67 5 68 This study utilized several Landsat TM images and an Airborne MSS image to determine whether there were significant correlations between the reflectance values and SSC and turbidity levels. Suspended sediments concentration data for the Chilung River were obtained from the monthly Taiwan EPA field data and were temporally matched with three Landsat TM images. The three images were normalized and the reflectance values were transformed into radiance values for increased accuracy of statistical regression analysis. The Airborne MSS image was flown by the Taiwan Provincial Government, Department of Forestry, Agroforestry Aerial Survey Division, and the field research team collected SSC and turbidity water quality data for over twenty-five sites. Multi-linear regression analyses between the Landsat TM radiance values and SSC showed significant correlations between TM Band 4, the near-infrared band, and suspended sediment concentrations (SSC) with R2 = 0.444, p = 0.009 (see Table 4.4). Utilizing the logarithmic function, the regression of the ln(SSC) dependent variable and the Band 4 radiance values yielded the highest significant single band correlational relationship, R2 = 0.574, p = 0.002, with a regression equation: ln(SSC) = 0.157*(Band 4) + 2.582. Also, the (Band 4/Band 2) ratio regression against ln(SSC) yielded a high correlation, R2 = 0.66, p = 0.001 (see Figure 4.8). Multi-linear regression analyses of the Airborne MSS reflectance values and SSC yielded significant correlations between MSS Band 8 and 9, the near infrared bands (see Table 2.1). Utilizing logarithmic function analysis, the regression of ln(turbidity) and MSS Band 8 was significantly correlated, R2 = 0.377, p = 0.001 (see Figure 4.12). 69 B. CONCLUSIONS The conclusions of this study are presented as responses to the following research questions: Research Questions 1) Can the water quality of the entire Chilung River be efficiently assessed using remotely sensed images and not just areas from which water quality point samples were collected? It is efficient to use Landsat TM images for monitoring the suspended sediments of the Chilung River because the Landsat-5 satellite continues to pass over the Taipei area at regular 16-day intervals, and significant correlations between SSC and TM Band 4 have been found. However, since the Chilung River is relatively narrow at the NanHu bridge, some locations may only have several pixels which span the width of the river and not be very accurate in estimating SSC levels of the Chilung River, since these 900 sq. meter edge pixels of the Chilung River may be mixed with the adjoining land features, such as the eastern region of the research area. The finer resolution of the Airborne MSS image, 5 meter pixel resolution, may help the water quality assessment problem. However, the use of Airborne MSS would be too expensive for use in periodic monitoring since flying and processing one Airborne MSS image costs over US$10,000 in Taiwan. Also, there are inaccuracies of using an image with very fine resolution for water quality assessment, and of ground truthing in a aqueous environment where there are no good landmarks for resampling the MSS image (Gao and O’Leary, 1997). 70 2) Is there a significant correlation between the reflectance values of the remotely sensed images with the collected water quality data of the Chilung River? As stated above, there are significant correlations between Landsat TM radiance values and TM Band 4, the infrared band. The Airborne MSS reflectance bands 9 and 10, the infrared bands, also have been significantly correlated with SSC. These remotely sensed images can be used to periodically monitor the SSC and turbidity levels of the Chilung River, though periodic field water quality testing is still needed, since the Landsat TM data showed a large variability of SSC values at several sites in which the reflectance values were the same. e.g. SSC values of 39.3, 19.5 and 18.0 mg/L in the December 1995 TM image yielded the same reflectance value of “5” (see Table 3.1, page 23). 3) Which type of remotely sensed images, Landsat TM or Airborne MSS, yield better correlations between Chilung River water quality values and their respective reflectance values? As the correlation results show, Landsat TM Band 4 and ln(SSC) yielded the highest R2 = 0.571, while the equivalent Airborne MSS spectral wavelengths, Bands 8 and 9, yielded R2 of 0.38 and 0.37. Landsat TM holds promise to be a better predictor of SSC in this study for periodical water quality monitoring. Airborne MSS reflectance values yielded significant correlations, though there may be some problems of positional accuracy with its small 5 meter pixel resolution. These problems were explained in detail in the Chapter 4 discussion section. 71 C. RECOMMENDATIONS Recommendations for Scholars More efforts need to be made to coordinate simultaneous flying of the image and conducting the field water quality testing since this will eliminate the potential variability in the water quality parameters between the date and time the remotely sensed image is flown and the suspended sediment or turbidity water quality readings. Since Gao and O’Leary (1997) suggested that a pixel resolution of 10 meters would be the ideal resolution for obtaining the best R2 correlation between SSC and airborne remotely sensed data, further research should seek to find a remote sensor having a 10 meter pixel resolution which matches the band spectral sensitivity of the Landsat TM sensor, or fly the Airborne MSS image at a higher altitude to increase the pixel resolution to 10 meters. Recommendations for the Taiwan EPA Since there were significant correlations between suspended sediment concentrations and Landsat TM radiance values, Landsat TM images can be used to supplement the Taiwan EPA’s water quality monthly monitoring program since images can be obtained at 16-day periodic intervals. If the number of monthly water quality monitoring data points were increased, there would be a more representative distribution of the SSC within the Chilung River and increased likelihood of more accurately predicting the SSC levels 72 in the Chilung River when regressed against the radiance values of the Landsat TM images. Also, several monthly water quality monitoring dates could be planned to coincide with the Landsat TM flight date(s) over the Taipei region to test whether more highly significant correlations will be generated when the variability between the remotely sensed image flight date and the test monitoring date is eliminated. The Taiwan EPA may consider using this method of using Landsat TM images to help monitor SSC and turbidity of other less accessible Taiwan rivers and coastline areas since the Landsat TM sensor can easily fly over any region of Taiwan island. Recommendations for other governments Since significant correlations could be established between SSCS and Landsat TM data for a small, shallow Chilung River in Taipei, Taiwan, other governments can also attempt to obtain Landsat TM images to help monitor the suspended sediments concentration in medium to small rivers in areas of difficult access. APPENDICES 73 APPENDIX A MAP OF EAST ASIA Beijing ° Tokyo . (A 3??“ CHINA Shanghai , PACIFIC OCEAN "x" 6% Fukien Q ‘27 s [\Q ' Carlton Hong Kong m PHILIPPINES The Region of East Asia 400 miles al. Source: James Reardon-Anderson, Pollution, Politics and Foreign Investment in Taiwan, (ME. Sharpe, Inc., Armonk, NY, 1992), p. xii. 74 .82 55583 53m 55.5.,qu 38m 55322 .385er me .568 32 55.me 6% 33 $3.85 5.18m sees 535889 Rm use .330 .o 65868.5 .55 .useék .E. Uessem .Hofim M3330 5.56 5.. e . wflEéCEJ—ZZEmr—ECP l’l .CaficaOm 3::CU 35:51:53.2 55.8..» END. _ _ NM IIIllIi\ nnflOm L232 UCN >4>>WWLL i .552 8. 25.5 I mumuw mus/o . cum 2 Se .2 I WZON i_<_~_.—.WDQZ_ leIhu—OZ QZ< 7.0—Um“ Z II. can «26$ 83 .30me I v53 UmfimE. I 533 v53 tofimEEOZ I ZC_._.OU GZS .Il|\\ Ewe: EM 1% QZm—GHA mg 57:42. W EQZHQQ< BIBLIOGRAPHY BIBLIOGRAPHY Alfoldi, T.T. (1982). Remote sensing for water quality monitoring. In Remote Sensing for Resource Management. (Johannsen, C.J. and J .L. Sanders, Eds.), Soil Conservation Society of America, Ankeny, Iowa, pp. 317-327. Braga, C.Z.F., A.W. Setzer, and L.D. de Lacerda. (1993). Water quality assessment with simultaneous Landsat-5 TM data at Guanabara Bay, Rio de J aneiro, Brazil. Remote Sensing of Environment, 45(1): 95-106. Campbell, J .B. (1987). Introduction to Remote Sensing. The Guilford Press, New York, NY, pp. 283-287, 404-433. Clark, W.A.V. and PL. Hosking. (1986). Statistical Methods for Geographers. John Wiley & Sons, Inc., New York, NY, pp. 334-349. Curran, P.J., J .D. Hansom, S.E. Plummer, and M.I. Pedley. (1987). Multispectral remote sensing of nearshore suspended sediments: a pilot study. International Journal of Remote Sensing, 8(1): 103-112. Dzurik, AA. (1990). Water Resources Planning. Rowman & Littlefield Publishers, Inc., Savage, Maryland, 318 pp. Gao, J. and SM. O’Leary. (1997). The role of spatial resolution in quantifying SSC from Airborne remotely sensed data. Photogrammetric Engineering and Remote Sensing, 63(3): 267-271. Han, L. and DC. Rundquist. (1994). The response of both surface reflectance and the underwater light field to various levels of suspended sediments: Preliminary results. Photogrammetric Engineering and Remote Sensing, 60(12): 1463- 147 1. Jensen, J .R., K. Rutchey, M.S. Koch, and S. Narumalani. (1995). Inland wetland change detection in the Everglades water conservation area 2A using a time series of normalized remotely sensed data. Photogrammetric Engineering and Remote Sensing, 61(2): 199-209. Khorram, S. and H. Cheshire. (1985). Remote sensing of water quality in the Neuse River Estuary, North Carolina. Photogrammetric Engineering and Remote Sensing, 51(3): 329-341. 78 79 Kirk, J .T.O. (1989). The upwelling light stream in natural waters. Limnology and Oceanography, 34: 1410-1425. Lavery, P., C. Pattiaratchi, A. Wyllie, and P. Hick (1993). Water quality monitoring in estuarine waters using the Landsat Thematic Mapper. Remote Sensing of Environment, 46(3): 268-280. Liedtke, J ., A. Roberts, and J. Luternauer. (1995). Practical remote sensing of suspended sediment concentration. Photogrammetric Engineering and Remote Sensing, 61(2): 167-175. Lillesand, TM. and R.W. Kiefer. (1987). Remote sensing and image interpretation. Second edition. John Wiley & Sons, Inc., New York, NY, pp. 450-466. Lillesand, TM. and R.W. Kiefer. (1994). Remote sensing and image interpretation. Third edition. John Wiley & Sons, Inc., New York, NY, pp. 202-214, 462-480, 524-541. ‘ Lusch, DP. (1996). Introduction to GPS: The Global Positioning System. Manual published for the Center for Remote Sensing, Dept. of Geography, and Institute for Water Research, Michigan State University, East Lansing, MI, 28 pp. Mertes, L.A.K., M.O. Smith, and J .B. Adams. (1993). Estimating suspended sediment concentrations in surface waters of the Amazon River wetlands from Landsat images. Remote Sensing of Environment, 43(3): 281-301. Miller, R.L., J .F. Cruise, E. Otero, J .M. Lopez. (1994). Monitoring suspended particulate matter in Puerto Rico: Field measurements and remote sensing. Water Resources Bulletin, 30(2): 271-282. Moore, GK. (1978). Satellite surveillance of physical water quality characteristics. In Proceedings of the Twelfth International Symposium on Remote Sensing of Environmen, Volume 1, April 20-26', 1978. Environmental Research Institute of Michigan, Ann Arbor, MI, pp. 445-462. Pattiaratchi, C.B., P. Lavery, A. Wyllie, and P. Hick. (1994). Estimates of water quality in coastal waters using multi-date Landsat Thematic Mapper data. International Journal of Remote Sensing, 15(8): 1571-1584. Reardon-Anderson, J. (1982). Pollution, Politics and Foreign Investment in Taiwan. ME. Sharpe, Inc., Armonk, NY, p. xii. Republic of China, Environmental Protection Administration. (1989). Report on the Plan to Treat the flow of Wastewater from the Chilung River. December 1989, Taipei, Taiwan (in Chinese). 80 Republic of China, Environmental Protection Administration. (1994a). 1993 State of the Environment, Taiwan, Republic of China. (Lung-Sheng Chang, Ed.), Taipei, Taiwan, 473 pp. Republic of China, Environmental Protection Administration. (1994b). Case Report on the Plan to Treat the Pollution of the Tanshui River. June, 1994, Taipei, Taiwan (in Chinese). Republic of China, Environmental Protection Administration. (1996). The Application of Remote Sensing Technique on the Study of Environmental Pollution Monitoring. Prepared by the Center for Geographic Information Systems Research, Feng Chia University, June, 1996, Taichung, Taiwan, 163 pp., (in Chinese). Republic of China, National Weather Bureau. (1988- 1996). Monthly weather reports from January 1987 to December 1995 (in Chinese). Rimmer, J .C., M.B. Collins, and CB. Pattiaratchi. (1987). Mapping of water quality in coastal waters using airborne thematic mapper data. International Journal of Remote Sensing, 8(1): 85-102. Ritchie, J .C., C.M. Cooper, and ER. Schiebe. (1990). The relationship of MSS and TM data with suspended sediments, chlorophyll, and temperature in Moon Lake, Mississippi. Remote Sensing of Environment, 33: 137 -148. Shimoda, H., M. Etaya, T. Sakata, L. Goda, and K. Stelczer. (1986). Water quality monitoring of Lake Balaton using Landsat MSS data. Seventh International Symposium on Remote Sensing for Resources Development and Environmental Management. (Damen, M.C.J., G.S. Smit, and HT Verstappen, Eds.), August 1986, Enschede, Netherlands, pp. 765-770. Welch, R., M.M. Remillard, and RB. Slack. (1988). Remote sensing and geographic information system techniques for aquatic resource evaluation. Photogrammetric Engineering and Remote Sensing, 54(2): 17 7 -185. "lllllllllllllllll