International Symposium On Managing Water Supply for Growing Demand
Bangkok, Thailand 16-20 October 2006
Proceedings
Edited by Sacha Sethaputra, Kitchakarn Promma
IHP Technical Documents in Hydrology No.6 UNESCO office , Jakarta 2006 <&. nd X
International Symposium on
Managing Water Supply for Growing Demand
In conjunction with the
14th Regional Steering Committee Meeting
for
UNESCO - IHP Southeast Asia and The Pacific
The Grand Hotel
Bangkok, Thailand
16-20 October 2006
Sponsored by
UNESCO Office, Jakarta anese Ministry of Education, Culture, Sports, Science and Technology (MEXT) Department of Water Resources, Thailand
Organized by
Thai National Committee for IHP-UNESCO Thailand National Commission for UNESCO Ministry of Natural Resources and Environment, Thailand
ii Preface
The IHP Symposium on Managing Water Supply for Growing Demand was held in conjunction with the 3rd APHW Conference on Water Resources Management towards Sustainable Growth and Poverty Reduction during October 16-20, 2006 at the Grand Hotel, Bangkok , Thailand. It was aimed to strengthen the cooperation among the Asia - Pacific countries to solve the problem of limited and fragile water resources in order to respond to the ever-lasting demand of water. It also provided the international arena for all stakeholders in the region to exchange experiences and information on hydrological sciences in order to achieve both solutions for better life as well as to consolidate a regional commitment so as to upgrade knowledge on water resources management and increase the capacity of relevant stakeholders.
As the IHP Symposium was organized in parallel with the 3rd APHW Conference, it provided the opportunities for all counties to establish the solid cooperation and to make the utmost benefit from experiences and information exchange to tackle with the water-related problem in their particular countries under the framework of UNESCO's International Hydrological Programme onwards.
The Proceedings published for the IHP Conference comprise 26 technical document papers, in which various issues and their interesting analysis are included. All papers were presented in the IHP Symposium sessions and were accumulated as the UNESCO technical papers both in forms of hard copy and CD-ROM. The organizing committee do hope mat this technical paper will be of great benefit to the readers and could contribute to the improvement of water resources management and hydrological knowledge in the future.
Finally, the organizing committee would like to render our great appreciation and thanks to the UNESCO Office in Jakarta and the Japanese Ministry of Education, Culture, Sports , Science and Technology ( MEXT) for their financial supports. Without these supports, the organization of the IHP Symposium and the publishing of the Proceedings would not be possible.
in CONTENT
Preface iii
Content v
A Application of a New Flood Stochastic Simulation Model Developed by Russian Hydrologist in the Basin of Qiantang River YUANFANG CHEN, GUOXINCHEN, WENPENG WANG, SHUJIAN LI, AOMI CHEN 1.
*- Statistical Experiment Study of Design Annual Runoff Estimated by Curve-Fitting Method for Pearson Type- III Distribution Chen Yuan-fang, Zhao Li-hong, Xu Sui, Ma Bing-xun, Wang Wen-peng 9-
= Exact and approximate analytical solutions for steady seepage inflows to horizontal wells L. Q. Liongson 15,
4 Effective management of water resources via demand management: Some examples from Southeast N. W. Chan, V. Nittvattananon 23-
G Managing Increased Water Demand in China: A Great Challenge Z.X.XU.J. Y.LI 39-
fc? Water Distribution Network Analysis for DMA Design of Ladprao Branch, Bangkok, Thailand
A. Pornprommin, S. Lipiwattanakarn, S. Chittaladakorn 45
y Water Rights and water allocation in Andean basins
J. Molina, E. Villarroel, J. Alurralde, A. Apaza,F.Soria 51
O Membrane Filtration Removing Salts and Arsenic in Drinking Water for Rural Areas Pikul Wanichapichart, Wiriya Duangsuwan, Darunee Bhongsuwan, Pusadee Mohamad, and Porntip Sridang, 59
Cf Basin water use accounting method with application to the Mekong Basin Mac Kirby, Mohammed Mainuddin, Geoff Podger, Lu Zhang 67,
A 0 A Framework to Assess Model Structural Stability through a Single-Objective Global Optimization Method GihaLEE, Yasuto TACHIKAWA, Kaoru TAKARA 79
A\ Derivation of Rainfall Intensity-Duration-Frequency Relationships for Short-Duration Rainfall from Daily Data LeMinhNHAT, Y. Tachikawa, T. Sayama, K. Takara 89
A ¿_ Stochastic Modeling of Rainfall Maxima Using Neyman-Scott Rainfall Model Carlo Mondoñedo, Yasuto Tachikawa, Kaoru Takara 97
\ 2) Effect of Moving Storm Rainfall on Soil Erosion and Sediment Transport from Watersheds Guillermo Q. Tabios III, PhD 109
v >|£ Status of Disaster Database and Limitations for Planning T. Merabtene, J. Yoshitani, A. Pathirana 125
\ k ICHARM's Research Strategy toward Effective Implementations of Flood Forecasting and Warning System in Asia K. Fukami, H. Inomata, P. Hapuarachchi, R. Oki 133
f 7 Modified Remote Sensing Information Model of Water Erosion on Hillslopes D.Y. Shen, K. Takara 139
\ 8 Capacity Building - Application of Geoinformatics for Disaster Lai Samarakoon, Takashi Moriyama, Chu Ishida 145
A°\ Scope of Flood Hazard Mapping in Developing Countries Shigenobu Tanaka, Rabindra Osti, Toshikazu Tokioka 153
L 0 The Prediction of Flooding Area in the Pasak River Basin by Using Mathematical Model : A Case Study on Land Use N. Hungspreug, A. Penghuaro 159
Combating Reservoir Sedimentation: A Challenge for Sustainability T. Tingsanchali, N.M. Khan 169
/Lt. Impact of 2004-tsunami Natural Disaster on Water District-10 Banda Aceh of Krueng Aceh River Basin Development Masimin, Zouhrawaty A. Ariff. 179
Surface Water Treatment with Microfiltration at The Village of Pranon, Nakhon Sawan Province, Thailand William R. Sellerberg, P.E., Voravuthi RakTae-ngan 187
/L.¿^, Overview of United States Drinking Water Regulations Anthony M. Wachinski, PhD, PE 193
/]_ g Effect of ENSO on Southeast Asian Rainfall Tsing-Chang (Mike) Chen 201
/)(-, Tropical storms and associated flood risk on Grande Terre, New Caledonia Ray A. Kostaschuk, James P. Terry, Geoffroy Wotling 207
vi Application of a New Flood Stochastic Simulation Model Developed by Russian Hydrologist in the Basin of Qiantang River
YUANFANG CHEN, GUOXIN CHEN, WENPENG WANG, SHUJIAN LI, AOMI CHEN Department of Hydrology and Water resources, Hohai University, Nanjing 210098, China E-mail: vfchen@,mail.edu.cn
ABSTRACT
In order to simulate flood process better, the study focus on the application of a new flood stochastic simulation model developed by Russian hydrologist Sambotsky, which could describe the characteristics of rapid rise and slow fall for flood hydrograph, in the basin of Qiantang river in the southern region of China. In the case study, a lot of stochastic tests have been done for the new model, including the test of annual maximum peak flow, annual flood volumes. The results show that flood process simulated may pass all tests. Meanwhile, the new model is also compared with the conventional models, including seasonal AR(1) model and the Disaggregation model by statistical test for simulation flood process. The results show that the new model is better than the other two models at least in this basin. So it's said that the new model is feasible to simulate the flood process for the basin.
Key words: Flood stochastic simulation , Russia, AR(1) model disaggregation model, Qiantang river, Ap plication, Statistical test
1 INTRODUCTION practical regression order. In fact, regression Traditional flood stochastic model is often exponential is nonlinearity andchanged along with based on different kinds of correlativities in time magnitude of discharge. Recently, a new flood sequence and spatial sequence of daily flow or stochastic simulation model was developed by shorter-time interval flow. And it pays much Russian hydrologist, which could describe the attention to statistical characteristics of aggregate characteristics of rapid rise and slow fall for flood and constituent. This kind of model includes hydrograph. Here 32 years(1958-1989) of hydro- seasonal AR(1) model and the Disaggregation logic observations in Quxian station in the basin model and so on. Generally these conventional of Qiantang river in flood season are selected as a models could reflect main characteristics of daily sample, then our study probe into the feasibility, flow in time sequence well. Therefore they are simulation effect, advantage and disadvantage and widely used nowadays when we simulate daily so on of this new model used in China stream flow. However as we know that flood process is simulation. non-reversible in time sequence for daily flow. For example, there is obvious distinction between 2 MODEL DESCRIPTION the growth curve and regression curve of one The stochastic model of the hydrograph of flood daily flow process in mountain river basin in a flood season describes the temporal variance of southern China. Consequently, in order to describe characteristic of daily flow in time water discharge during a flood season as sequence more clearly, we'd better take this non-reversible characteristic into account. realization of a stochastic process. The time is Seasonal AR(1) model and the Disaggregation considerate with a step per one day within an model is difficult to simulate the flood natural process of rapid rise and slow fall. Though interval ' where ~~ is date of the earliest shot noise models may reflect diversity of t = T characteristic between the growth curve and beginning and is a date of the latest season. regression curve, it's still not a good way to describe main characteristics of daily flow. This The model is based on the approximation of die point of view was put forward by O'Connell in 1979. Meanwhile, for the sake of simplifying hydrograph by function. calculation shot noise models assume regression exponential as a constant, which is not conform Q(t) = q0%{t) + Yjqj T - a : a number of the flood peaks of this year; 2. The variables ' and ' are independent and in most there is not statistically reliable depend- 'i h q the dates of their passage; ence between them and ' . For a whole floodo f 9i,-,?t T ' " T : maximum discharge of these floods, every year the variables " ' * are independent and have the unified probability distribution. The independently formed without an influence of same conclusion is true for the variables base flow and previous floods; 3. The sequence of floodpeak s of all investigated : a dimensionless function of a hydro- rivers can satisfactorily described by a model of Poisson process. In particular, the annual number graph of base flow; of peaks have Poisson probability distribution, scribe a stochastic process ß('} . in a case of n As figure 1 shows, where ' is a duration of the nka years of hydrologie observations a main growth of i -flood; j is an intensity of it's exponential regression after ' observations of the element , , can be used for the estimation of their function of distri bution. Fig. 1 Hydrograph of -flood 3 MODEL APPLICATION (3) Draw flood hydrograph and extract five pa- 1. Data Treatment rameter values in model formula, that is , ° , In our study, we choose flood data from Quxian a station at Qujiang river in Qiantang river basin 1, *, , O = !,•••,*) during whole flood season. Sampling procedure listed below: (D Extract annual number of the flood peaks (1) According to general condition of basin and We meet with two kinds of flood process plotted climate factor, flood season total duration T is in figure2 . It is generally agreed that, if the height ascertained as 90 days. of growth curve is greater than or equal to one- (2) In order to reflect the feature of flood process third of the height of peak, we may consider mis sufficiently and efficiently, here we adopt 8 hours hydrograph as a whole flood process. Accord as time interval Ai. Therewith, the amount of ingly, the left figure is regarded as superposition time interval is 90 days * 8 hours = 270, that says of two flood processes and the rightfigur e is re the amount of kerf is 270. garded as an independent flood process. Q Q (rrf/s) (rr?/s) t 0 Fig. 2 Two kinds of flood process ® Extract base flow Because a base flow of a flood season is very point and duration of the main growth is small relative to corresponding flood discharge, reasonable. it's reasonable to take minimum annual dry- weather flow as base flow of the given year. i Extract maximum discharge 1J The maximal discharge of the kerf J subtracts base flow and previous flood recession flow, thus 9j is gained. © Extract duration of growth Model generalizes growth part as a zooming line, which can reflect characteristic of rapid growth. Fig. 3Extraction of duration of the main growth However the real growth curve usually isn't a regular line plotted in figure3 . Since it is a gener alization, therefore so long as the upper and nether segmented area formed by the fittingline , base flow line and original growth curve are equal, then we deem the extraction of the beginning of rise 3 probability-weighted moments method. By © Extract regression exponential oc.J statistical test we found that when the level of Q. i j significance =0.05, rate of pass is 100%, the Real flow '"' of kerf of receding -flood above hypotheses is acceptable. process is composed of base flow <¡0 and (3) Estimation of Q.J previous flood regression discharge at In distributed function test the NULL hypothesis kerf * . is that the sequence under consideration is a.i,j CC can be figured out through formula uniform distribution. On test law, when =0.25, a,,, =(1110,-Inß,,,)/A/ y2(3) according to = 4.108>2.6740, therefore the NULL hypothesis is acceptable. 2 parameters of uniform receding flood flow formula distribution are gained from maximal and minimal value of observed base flow. Q,,j =qJ*exp(-aIJ*At) a. (4) Estimation of and At rj Where: is the duration from to the t a} calculating kerf , =1,2,..., , is the The NULL hypotheses of and are type amount of date of regression curve. Average of x) a P-III distribution. In the case of , when a. a.lj J serves as value of the -flood re ceding process. *a(12) =0.050, =21.03>19.0835, assuming 2. Distribution test and estimation of parameters k *j (1) Estimation of and to be type P-III distribution is acceptable. In Model theory has indicated that the sequence of the annual number of flood peaks k have Poisson a k the case of , the same conclusion is drawn. probability distribution, and with the fixed for a flood period of a concrete years the dates a Ex Cv TJ j j and Cs of and ' are esti mated by fitting a curve method and probability- 1 ' ' ' ' » k weighted moments method respectively. behave like the uniform distribution over interval [0,m], and in our study m=270. 3. Modeltest In order to further judge this model to be (2) Estimation of % good or not, it is necessary to analyze and test applicability for the reproduced series. So in this study, 500 groups of 32 annual flood season series 1j In model theory, the series of are supposed were reproduced as inference population. The as independent random variables. In order to main statistical property tests are how confidence prove the validity of the hypotheses, independence intervals of simulative sample parameters and time interval flow contain those of measured sample. Testing methods include long term series tes* t* off ^> •is •indispensable A- ui . Ex Cv Cs and short term series. The detail results as follow: The testing for measured sample parameters falls of peak flow sequence can be estimated by into certain confidence interval of corresponding simulative ones. Table 1 Testing for measured sample parameters Parameter Measured value Long series Short series Short series mean square deviation Ex 8.1 8.1 8.1 0.5 k Cv 0.29 0.36 0.35 0.05 Cs 0.08 0.35 0.3 0.41 Ex 10.8 10.5 10.5 1.1 9o Cv 0.52 0.58 0.58 0.07 Cs 0.13 0.02 0.01 0.27 Ex 3.93 3.93 3.93 0.1 T 0.48 J Cv 0.48 0.48 0.03 Cs 1.28 1.27 1.24 0.28 Ex 0.3 0.3 0.3 0 a 0.36 0.36 0.36 0.20 j Cv Cs 0.79 0.8 0.78 0.21 Ex 1790 1780 1780 81.0 This model only gives distribution function of the From above table, it shows that only k number of peak , so that the other parameter of doesn't pass through test in case of one distribution only can be identified by experiential mean square deviation while others preserve try. After many time comparisons this research Cs k eventually recognized these distributions, and statistical properties well. Though of from current hypotheses and test results can infer Cs that the selection of distribution is reasonable. Meanwhile we should concern that our study is pass the test, oscillation of is a little big. The just base on single station and sample size is cause exists in the reproduced series incredibly limited, so how to select proper distribution is still preserves second or third moment of real series a question. k for when real series is fittedb y use of mean (2) Testing of time interval flow statistical properties k Qj Tj Because of real data discontinuousness, the for Poisson distribution. Whereas length of time interval is not too long or at most not more than the duration of one flood process. a j According to the measured data 10 time interval and series are all fitted by use of three (length of each interval is 8 hours) were properly parameters for P—IH distribution, their testing selected from Quxian station. result are of course better. Table 2 Testing of time interval flow statistical properties ( short series ) Time interval Parameter 1 2 3 4 5 6 7 8 9 10 ( 8 hours) Observed data 1.07 2.01 2.77 3.42 3.98 4.46 4.91 5.35 5.78 6.15 Ex Simulated data 1.17 2.09 2.87 3.54 4.11 4.60 5.02 5.38 5.70 5.98 Mean square 0.08 0.15 0.20 0.25 0.29 0.33 0.37 0.40 0.42 0.45 deviation Observed data 0.348 0.366 0.376 0.385 0.386 0.392 0.401 0.408 0.419 0.423 Cv Simulated data 0.362 0.366 0.371 0.375 0.380 0.384 0.388 0.393 0.397 0.401 Mean square 0.054 0.055 0.055 0.056 0.057 0.057 0.058 0.058 0.059 0.059 deviation Observed data 0.457 0.514 0.566 0.645 0.609 0.665 0.714 0.694 0.719 0.705 Cs Simulated data 0.681 0.688 0.697 0.704 0.712 0.720 0.726 0.733 0.737 0.742 Mean square 0.501 0.501 0.504 0.511 0.514 0.518 0.519 0.519 0.520 0.520 deviation Table 3 Testing of time interval flow statistical properties (long series) Parame Time interval 1 2 3 4 5 6 7 8 9 10 ter ( 8 hours) Observed data 1.07 2.01 2.77 3.42 3.98 4.46 4.91 5.35 5.78 6.15 Simulated Ex 1.17 2.09 2.87 3.54 4.11 4.60 5.02 5.38 5.70 5.98 data Observed data 0.348 0.366 0.376 0.385 0.386 0.392 0.401 0.408 0.419 0.423 Simulated Cv 0.368 0.373 0.377 0.381 0.386 0.391 0.395 0.400 0.404 0.408 data Observed data 0.457 0.514 0.566 0.645 0.609 0.665 0.714 0.694 0.719 0.705 Simulated Cs 0.844 0.858 0.870 0.886 0.898 0.908 0.915 0.922 0.928 0.934 data a is less than actual value, which induces the Carefully analyzing table 2 and table 3, we simulative flood discharge inclination to increase find two phenomena as follow. at the beginning. In order to achieve a reasonable ® All of the parameters pass through test on a, occasion of confidence interval is one mean simulative result, was corrected by multiply- Cs c c square deviation. However, of simulated ing corrective coefficient ( >1, in this study data fluctuates in a relatively large scope. As we know, the estimation of high-order moment of C =1.2). stochastic variables by use of samples with small (3) Parameter parsimony test capacity is unreliable, therefore it's not strange to During selecting model, it is necessary to take the meet this phenomenon. number of parameters into account in terms of ® The accuracy of extracting information is R = 10 31 influenced greatly by model maker. We find the finite information. In our study, , phenomenon that from interval 0 to 8 real mean value is less than simulation's and following Based on Chinese hydrology data, the range of situation is on the contrary, although simulative D result is satisfying. is ascertained from 5 to 10 or so. In this case As figure 1 show, why leading to this phenome this model parameter number meets requirement. non is that current flood process is influenced not only by former flood but also by next one, as re (4) Comparison with and Disaggregation sults in identifying initial time of flood rising Model (Dis) more difficult. In order to extract receding Because of the model firstly applied to China, it is indispensable to compare simulation result of this exponent easily, effects of die next flood rising are neglected, so mat new model with seasonal model and Dis aggregation model, judging whether superior to these two conventional models or not Based on measured data, time intervals from 1 to ® Disaggregation model generally applies to 30 (each time interval is 8 hours) were determined as modeling sample series, containing a maximal simulate a individual flood process. , , individual flood process of every year season Cs flood. The reason is (D discontinuous real data of maximal 1, 2, 3 total time interval dis AR(X) charge were selected to compare with each other. don't permit and Disaggregation models Comparative results as follows: to simulate each year whole flood season hydro- graph, Table 4 Comparison about time interval flow statistical properties (short series) Time interval 1 2 3 ( 8 hours) Parameter Sea Sea Sea Model New Dis sonal New Dis sonal New Dis sonal AR(1) AR(1) AR(1) Measured data 1.07 1.07 1.07 2.01 2.01 2.01 2.77 2.77 2.77 Simulated data 1.12 0.95 0.95 2.09 1.88 1.85 2.87 2.62 2.55 Ex Mean square 0.08 0.07 0.09 0.15 0.13 0.17 0.20 0.17 0.24 deviation 0.37 0.37 Measured data 0.348 0.348 0.348 0.366 0.366 0.366 0.376 6 6 0.37 0.45 Simulated data 0.362 0.468 0.537 0.366 0.462 0.557 0.564 Cv 1 8 Mean square 0.05 0.07 0.054 0.071 0.140 0.055 0.070 0.142 0.140 deviation 5 0 0.57 0.57 Measured data 0.457 0.457 0.457 0.515 0.515 0.515 0.579 9 9 0.69 0.81 Simulated data 0.681 0.882 1.367 0.688 0.845 1.370 1.335 Cs 7 2 Mean square 0.50 0.62 0.501 0.625 0.956 0.501 0.630 0.951 0.936 deviation 4 1 Table 5 Comparison about time interval flow statistical properties (long series) Time interval 1 2 3 ( 8 hours) Parame ter Sea Sea Sea Model New Dis sonal New Dis sonal New Dis sonal AR(1) AR(1) AR(1) Measured data 1.07 1.07 1.07 2.01 2.01 2.01 2.77 2.77 2.77 EX Simulated data 1.12 0.95 0.95 2.09 1.88 1.85 2.87 2.62 2.55 Measured data 0.348 0.348 0.348 0.366 0.366 0.366 0.376 0.376 0.376 Cv Simulated data 0.362 0.468 0.537 0.366 0.462 0.557 0.371 0.458 0.564 Measured data 0.457 0.457 0.457 0.515 0.515 0.515 0.579 0.579 0.579 Cs Simulated data 0.681 0.882 1.367 0.688 0.845 1.370 0.697 0.812 1.335 4 CONCLUSION After analyzing table 4 and table 5, we find that the simulation effect of new model is better than Through investigation and application to AR(\) the new flood stochastic simulation model and Disaggregation models. developed by Russian hydrologist, the following primary conclusions were drawn: 7 1. In this research, this new flood stochastic model REFERENCES was firstly exploratory application for Quxian station simulating flood process, and the Ding Jing, Deng Yuren, Stochastic Hydrology, simulation results show that it is feasible to Chengdu University of Science and reproduce the flood process for the basin. Technology Publish House. 2. Distinguish characteristic of this model is that Ding Jing, The Progress of Stochastic Hydrology, simulated flood hydrograph describes the Journal of Chengdu University of Science character of rapid rise and slow fall, which well and Technology, No 4, 1986, pl33-134. fits South China mountain rivers flood process. Ding Jing, The Application of Disaggregation 3. By use of this model, assumption population Model in Flood Simulation, Journal of was tested statistically and flood stochastic Chengdu University of Science and simulation was investigated. Results have shown Technology, No 4, 1986, pl41-148. that it can efficiently use the given hydrologie Ding Jing, Deng Yuren, Yang Rongfu, On The data on the basis of the account of the particular Stochastic Approach in Flood Simulation, feature of a flood flow, and preserve various Papers in fifthConferenc e on Hydrology in statistical properties. China, Science and Techology Publish 4. The advantages of this model are that it has House, 1992, pi66-170. clear physical concept, simple structure, proper Chen Yuanfang, The Application of Monte-Carlo parameter and general applicability. Method, Heilongjiang People's Publish House. Hu Kangping, Verification and Validation of Stochastic Streamflow Models, Engineering Journal of Wuhan University, No 1, 1987, pll-17. Huang Zhengping, Hydrologie Statistics, Hohai University Publish House. Statistical Experiment Study of Design Annual Runoff Estimated by Curve-Fitting Method for Pearson Type- III Distribution Chen Yuan-fang Zhao Li-hong Xu Sui Ma Bing-xun Wang Wen-peng (Dept. of Water Resources and Hydrology, Hohai University, Jiangsu Nanjing, 210098) email: yfcheniSimail.edu.en ABSTRACT The curve-fitting method is a common way to estimate design annual runoff values under different high probabilities in water resource assessment. However, there's nearly no theoretical analyses done about exactitude of the design values calculated under different curve-fitting criterion. Therefore, it's still a question that how these different criterion influence results. Meanwhile whether we should use all the collected data to do the curve- fitting or not is also a question. In the paper, the above questions are discussed by Monte-Carlo experiments with consideration Pearson-III as population distribution. The criteria for evaluation are the un-biasness and efficiency of the parameter and design value with given probability, Many calculations show that the estimation result is better when the proportion of data for curve-fitting from 50% to 60% than that when the proportion less than 20%, even slightly better than traditional curve-fitting method which we used all the data, and the curve-fitting by absolute criterion is better than that of square criterion. Key words: curve-fitting method; criterion of curve-fitting; partial data curve-fitting; un-biasness; efficiency; annual runoff ; P-III distribution 1 INTRODUCTION exactitude of design values of annual runoff and In the planning and design of all water how to estimate it better. conservancy works, hydrological frequency analysis There are two kind of problems to be solved and calculation should be made in order to promptly in the application of the curve-fitting determine the quantiles or design values of rain method to estimate the design annual runoff. One is storm or flood under different low probabilities that the exactitude of design values estimation and including 1%,0.1% et al. Because the exactitude of the effect of different criterion of the curve-fitting the quantiles is very important to both project on estimation result are not clear. The other one is investment and security, a lot of scholars have been that there is still a disputed question whether we carrying out much research for it and received should use all the collected (or observed) data to do plentiful and satisfactory results. At present, the curve-fitting or not. Nowadays some hydrolo- Pearson-III distribution is adopted as the population gists often use only partial lowest-collected data to distribution for the annual maximum flood peak & do the curve-fitting. Is the result of this treatment flood volume and rainfall in China, and the curve- better than that using all the collected data? If yes, fitting methods (including a curve-fitting by eye what is the best proportion data to be used to do the estimation and a curve-fitting by mathematical curve-fitting? optimal calculation) are also recommended to Based on above view of point, the exactitude estimate their distribution parameters in Chinese difference of the parameters and design value esti Regulation for calculating design flood of water mated by different curve-fitting methods is com [,] resources and hydropower projects . However, puted and analyzed by Monte-Carlo experiments in water resources assessment, it's necessary to esti with consideration Pearson-III as population distri mate hydrological design values, such as design bution. The criteria for evaluation are the un- annual runoff, under some high probabilities (p biasness and efficiency of the parameter and design equals to 75%, 85% or 95%),. In general, it is con value with given probability. Through these, the sidered to use directly above-mentioned method of above-mentioned questions could be solved , and the design flood to estimate these parameters. So it's useful for annual runoff frequency analysis in there are nearly no theoretical researches about the the future. 9 2 DISTRIBUTION FUNCTION AND 3.1 The curve-fitting with all the collected PARAMETERS (observed) data The density function of Pearson-III distribution (1) Square criterion (px [2] is as follows : » , .2 Objective function: A = £ vC - *P°„, ) (4) m =1 a (i flo)(;c a /w=4^(^- or^ " - »>(i) (2) Absolute criterion ç2 v 7 r(«)v 0/ a ß a " i i Where 0 are the parameter of density Objective function: A = ^J*» - *p. I (5) a ß function, and >0, >0. The relationship Where, x°_ is the population design value with between them and three population statistical parameters EX, Cv, Cs is listed as following: frequency P™ 4 Ci While me parameters EX, Cv, and Cs are 2 ß = „„„„ (2) EXCvCs given, the corresponding objective function can 2 Cv EX ! ~ Ci be calculated. The parameters , " , ' Three statistical parameters EX, Cv, Cs can be with which tends to the minimum are the final a ß estimation results of parameters by the curve-fitting with certain criterion. estimated from certain observed data, then , , ° can be obtained by using equation (2), in this 3.2 The curve-fitting with the partial collected data case the design values estimated from the observed The curve-fitting with partial collected data data under different frequency could be finally means that it doesn't use all collected data to do the estimated. curve-fitting, but it uses partial data just in the right side of Haisen frequency curve to do it. 3 DIFFERENT CURVE-FITTING METHODS Let k indicate the percentage of right side data In China, the curve-fitting method is a common used for curve-fitting to all collected data. The parameters estimation way for Pearson-III value of k may be 10%, 20&, 30%, etc. It's obvious distribution. It is recommended by the Ministry of that when k= 100% it is the curve-fitting with all the Water Resources [1] to use the curve-fitting method collected data. The objective function may be with eye estimation usually, and the curve-fitting rewritten as follows: method by mathematical optimum calculation could be also used. The plotting position formula of the (1) Square criterion Ç>^ expectation value is also recommended to calculate empirical frequency. Due to difficulty to consider Objective function: A y / (6) the curve-fitting method with eye estimation for m = n- nk +1 Monte-Carlo experiments, so the main study focus on the curve-fitting by mathematical optimum (2) Absolute criterion ç> 2 calculation ,including by absolute criterion and square criterion. Objective function : A _ (7) Suppose the simple hydrological sample m =n-nk +1 x,,x„---,x„ , x'(m = \,2,--,n) , range them to If a certain design value under a given frequency is according to the descending order. Plotting position less than 0, it's put to zero. equation is as following: pm= 7 0»=l,2,-n) (3) n+l 10 4 THE STANDARD OF EVALUATION 5.2 Scheme design of the curve-fitting with different The standard for evaluation are the un-biasness data proportion and efficiency of the parameter and design value n= 30, 50, 100; EX0 = 1.0; Cv0= 0.3, 0.5, 1.0; with given probability'31. Cso/Cvo = 2, 3, 4. And the partial data proportion The un-biasness and efficiency of the takes 0.1, 0.2, 0.3, 0.5, 0.6, 0.7 and 1.0, in fact parameters are expressed by the average and the k= 10,20,30,50,60,70,100 respectively. There are mean square error respectively. If the average is 189 schemes in total. The curve-fitting method with very close to the population mean, it can be thought absolute criterion is only considered for this case. to be un-biased estimation. The smaller the mean For all the above statistical experiments, Ns= 2000 square error is, the better the efficiency is. and P= 0.50, 0.75,0.85, 0.95. The un-biasness and efficiency of the design value are expressed by the average of the relative 6 RESULTS AND ANALYSIS error B^and the relative error of the root mean 6.1 Effect of two different criterion in the square Sxp as follows: curve-fitting to the estimation result of parameters and quantités N s 11 (8) The partial results are listed in table 1. For the I ix' -x°] ÁN X°) *IOO% s results of the parameters, Both estimation of Cv and J i = l\p p) \ p) Cs by two criteria are positively-biased slightly, and the un-biasness and efficiency of the parameters by absolute criterion are better man that of square criterion in most schemes, especially for the latter it B =i^i *100% (9) is positively-biased seriously when the population XP N x0 values of Cv and Cs are large. For die results of s p quantités, both design values by two criteria are less x° Where p is the design guaranteed rate (or man the population value slightly in most schemes. |Bxp| by absolute criterion is less than 5% x°p design probability); is the population design for almost all schemes, so it may be considered as x' un-biased. However, the result by square criterion is value corresponding to p; is the design value unstable, because the design values are negatively- estimated by the i* stochastic sample corresponding biased seriously when Üre population values of Cv to p; Ns is the total number of stochastic samples for and Cs are large, but in some case ,it's positively- statistical experiment. biased seriously. The efficiency of the design values If B,tp >0, it means that the design value is by two criteria are almost equivalent in total, but positively biased, else negatively biased. The larger that of absolute criterion is better slightly and more | Bxp | is, me more serious the bias of the design stable. value is. When |B^,| isn't exceeded 3%, the design values can be considered to be un-biased. The smaller S^ is, the better the efficiency is. 5 MONTE-CARLO EXPERIMENT SCHEMES 5.1 Scheme design of two curve-fitting criteria The curve-fitting methods with absolute criterion and square criterion are considered respectively, and the curve-fitting with all collected data is studied. The other parameters values are as follows :n= 30, 50, 100; EX0 = 1.0; Cv0= 0.3, 0.5, 1.0; CSQ/CVO = 2, 3.There are 36 schemes in total. Table 1 partial result of parameters and quantiles for the curve-fitting methods with two criteria Population parameters Criteria ECv ECs SCv SCs Bxp3 Sxpl N n Cv Cs Bxpi Bxp2 Bxp4 Sxp2 Sxp3 Sxp4 5 5 0.30 0.60 0.32 0.66 0.04 0.43 -0.58 -1.34 -1.66 -206 498 5.75 640 9.73 0 0 S 5 0.30 0.60 0.31 061 0.03 0.39 -0.18 -0.64 -0.93 -1.54 4.82 553 6.35 10.15 0 0 Note: P¡=50%, P2=75%, P3=85%, P4=95%, EX0=1.0, it is same in the table 2. Therefore, the curve-fitting by absolute 6.2 Effect of the curve-fitting with different data criterion is better man that of square criterion proportion to the estimation result of parameters and according to the results of Monte-Carlo quantiles experiments. Table 2 partial result of parameters and quantiles estimated by the curve-fitting method with the different data proportion Population parameters data pro- ECs SCv SCs Bxpl Sxpl porti ECv B*p2 Bxp3 Bxp4 Sxp2 Sxp3 Sxp4 N n Cv Cs on 0 30 0 50 1.00 0 1 0,53 1 18 0.09 0 34 -2 31 -3 13 -2.54 1.47 10 05 13 4 15.94 23 11 30 30 30.50 1,00 02 0,52 1 00 0.08 0-52 -0.66 -1.56 -2 14 -3 94 10 62 13 49 15.51 24 97 30 30 0.50 100 03 0 52 0 97 0.08 0 53 -0 33 -1 00 -1 59 -379 1060 13 21 15 17 25 46 30 30 0 50 1 00 0 5 051 0 96 0 07 0 53 -0.06 -0 57 -1.19 -3 81 1055 12 72 14 78 26.94 30 30 0.50 1.00 06 0 51 094 0 07 0.52 0 17 -0.33 -107 ^130 10 57 12 63 14 92 28 29 30 30 0 50 1.00 0.7 051 0 94 0 07 0 52 0 18 -0 27 -0.99 -4.13 10 54 1258 1504 28 92 30 30 0 50 1 00 10 0.52 099 0 08 0,54 -0 46 -1.62 -2 59 -5 80 1075 13 02 15 33 28 56 30 30 1 00 200 01 102 2 08 0 17 0 39 -2 04 0.61 6 45 37 70 22 64 33.50 42 95 86 19 30 30 100 2.00 02 103 2 06 0 16 044 -1 90 -1 22 1 06 10 71 22 69 32.27 40 46 94.06 30 30 1 00 200 03 102 2 05 0.16 0 43 -1 31 -0 20 2 15 11 18 22.44 31 86 39 98 95.18 12 data Population parameters pro- ECv SCv Byp2 Bxpj Oxp4 portí ECs SCs Bxpl Bxp4 Sxpl Sxp2 Sxp3 N n Cv Cs on 30 30 1 00 200 05 1 01 2.00 0 14 0 41 0 03 1 73 364 1041 22 12 32 10 40.48 10169 30 30 1 00 200 06 100 198 0.14 0 45 0 70 2 93 4 42 12 70 22 60 32 25 41 43 116 24 30 30 100 200 07 1 00 1.96 0 14 0,46 106 3 10 3 74 11 04 22 66 32 05 41 69 118.04 30 30 1 00 200 1.0 104 2 03 0 17 0 55 -1 81 -3 70 -6 64 291 23 89 33 10 44.10 12157 50 50 0 50 1 50 01 0 50 143 009 0 43 0 53 0 90 0 76 -0 66 8 85 931 8.51 8 20 50 50 0 50 150 02 0.51 1.49 0 07 0 37 -0.10 -0 10 -0.09 -0 45 8 00 835 802 8 22 50 50 0.50 1 50 0.3 0.50 1.48 0.07 0 35 0.05 009 008 -0.34 7.78 8 08 7 80 8 34 50 50 0 50 150 0.5 0.50 1.48 0.06 0 35 0 16 0 22 0 18 -0 40 7 69 7.89 7.66 940 50 50 0.50 1 50 0.6 0,50 1.47 0 06 0.36 0 28 0 34 0 24 -0.58 7.72 7.83 7.68 10 13 50 50 0 50 1 50 07 0 50 146 0.06 0 37 0.34 0 39 0 24 -0.76 7.73 7 81 7 79 10 82 50 50 0 50 1 50 10 051 150 0.06 0 40 -0 27 -0 68 -0 91 -1.76 788 7 94 7 94 11.60 50 50 100 3.00 01 0.85 2.52 0.26 0 84 14 92 15 60 940 -0.77 26 70 22.62 13 14 168 50 50 1 00 300 0.2 0 97 2.94 0,17 0 49 2 77 3 15 1,79 -0 54 16 46 9 29 4 58 2 26 50 50 1 00 300 0,3 0 99 2.99 0 15 0,45 127 206 142 -0 05 15.05 7 97 4 21 2 57 50 50 1 00 3 00 05 098 2 97 0 13 038 1,45 194 1 32 -0 15 14,47 8.07 4.23 3 11 50 50 1 00 300 06 098 2.96 0.13 0 39 1.75 2.18 1 36 -0 43 14 87 8.22 439 4 56 50 50 1.00 300 0.7 0 98 2.94 0 13 0 39 2 14 2 26 1 18 -0 89 14 94 8 24 4.70 5.82 50 50 1 00 3O0 1 0 102 3.05 0 16 0 53 -0 87 -0 05 -0 65 -2 50 15 98 800 6 65 10 34 100 100 0 30 120 01 030 121 004 0 29 -0 20 -0 30 -0 29 -0 36 3 51 3 72 3 54 3.42 100 100 0 30 1.20 0.2 0.30 1 19 0.03 0.25 -0 02 -0.11 -0 14 -0 33 3 30 3 43 3.31 3 34 100 100 030 1 20 0.3 0 30 1.19 003 0 26 -0 05 -0 12 -0 13 -0 27 3 27 3.32 3.20 3 42 100 100 0 30 120 05 0 30 1.19 0 03 0 27 -0 02 -0 07 -0 07 -0.23 3.25 3.19 3.10 3.78 100 100 0 30 120 06 0.30 1,19 0 03 0.27 -001 -0 05 -0.06 -0 23 3.25 3.17 3.12 3.98 100 100 0.30 120 0.7 0.30 1 19 0.03 0.27 0 01 -0.04 -0.07 -0 27 3 23 3 16 3 16 4.14 100 100 0 30 1 20 1,0 0 30 121 003 0 28 -0 15 -0 31 -0.34 -0 47 3 31 3 22 3.19 4 29 100 100 1 00 400 0.1 084 338 0.27 102 13 73 5.35 1 87 -0.10 20 20 751 2.75 0 18 100 100 1.00 400 0.2 092 3 70 0.19 0 65 6 43 170 0.35 -0.13 10 66 2 24 0.45 0 19 100 100 100 400 03 0 98 3.95 0.16 050 1 83 0 54 006 -0 06 6.87 1.10 0 20 0.18 100 100 1 00 400 05 100 4.03 0 15 0 45 049 0 32 0 08 0 01 6 02 0.95 0.31 0 31 100 100 1 00 400 06 0.99 3 99 0.13 0 35 0,82 034 0 05 -0 03 6 00 0,97 0.31 0 39 100 100 1 00 400 0.7 0.98 3.97 0.14 039 1 14 0 35 -0 05 -0 18 661 098 0,68 0 94 100 100 100 4 00 1 0 101 4 07 0.16 0 53 -0 10 -0 22 -0 58 -0 71 694 1,50 209 244 The partial results are listed in table 2. Form 1. As a whole, the data proportion in the the above results, some conclusions can be drawn as curve-fitting influences the results of the parameters follows: and design values to a certain extent. The effect is 13 small when the data proportion is above 20%. When 3. This research conclusion is not only suitable the proportion is small, such as less than 20%, the to the calculation of design annual runoff under effect will be large; and the results of the design different high probabilities, but also to the values are bad obviously when the population values calculation of design annual rainfall values. of Cv and Cs are large. 4. In fact, there exists usefully auto-correlation 2. The un-biasness of the design values is for both annual runoff series and rainfall series, acceptable with the proportion of above 20%, except which is even large sometimes. However, as of the scheme of the population values of Cv= 1, you know,the curve-fitting method is only used for Cs= 2. It is because that |B^,| in these cases are less the independent homologous samples in theory. than 5%, especially in majority less than 3% ,so the Therefore, whether the auto-correlation affects the un-biasness is considered as good. But in the estimation result of quantiles or not and how great scheme of the population values of Cv= 1, Cs= 2, the influence is, it is necessary to do some further the un-biasness of the curve-fitting with all the research on it. collected data is better than that of the partial collected data, because the values of |Bxp| for the REFERENCES latter are about 10%, so it's positively-biased [1] Ministry of Water Resources & Ministry of seriously. The efficiency of the design values with Energy. Regulation for calculating design the partial collected data is better slightly than that flood of water resources and hydropower of all the collected data. However, the results of the projects SL44-93 [S]. Beijing: Chinese Water design values become bad when the data proportion Resources Publishing House, 1993. is less than 20%, especially when it is less than 10%. [2] Cong shuzheng et al. Statistical Testing Research Generally, the results are relatively stable and have a on the Methods of Parameter Estimation in relative high precision in the data proportion of Hydrological Frequency Analysis[J], Journal 50-60%. of Hydraulic Engineering, 1980(3), PP 1-9. [3] Chen Yuanfang, Hou Yu. Study on the pa 7 CONCLUSIONS rameter estimation for Pearson-Ill distri- Based on the results of lots of Monte-Carlo bution[J], Journal of Hohai University, experiments and the analysis to the questions of 1992,No3,PP 24-31. curve-fitting methods encountered in practical water resource assessment, the final conclusions can be [4] Chen Yuanfang. Monte-Carlo Method and its drawn as follows: Application [M]. Haerbin: Heilongjiang 1. To the two criteria, the curve-fitting by People's Publishing House, 2000. absolute criterion is better than that of square [5] Ding Jing, Deng Yuren. Stochastic Hydrology criterion in the precision and stabilization for [M]. Chengdu: Chengdu University of estimation of parameters and quantiles. So it's Science and Technology Publishing proposed to use the curve-fitting method with House, 1998. absolute criterion in the calculation of design annual runoff under different high probabilities. 2. In using the curve-fitting method to calculate design annual runoff values under different high probabilities, the data proportions affect the estimation results to a certain extent. When the data proportion is large, the effect is not obvious. However the precision of the design values becomes low and unstable when the data proportion used for the curve-fitting is small, such as less than 20%,especially less than 10%. Generally, the results are relatively stable and have a high precision in the data proportion of 50-60%. But, it should be mentioned that the curve-fitting with all the collected data be considered as un-biased in all schemes, which is better than that with partial collected data. EXACT AND APPROXIMATE ANALYTICAL SOLUTIONS FOR STEADY SEEPAGE INFLOWS TO HORIZONTAL WELLS L. Q. Liongson Department of Civil Engineering and National Hydraulic Research Center, College of Engineering, University of the Philippines, Diliman, Quezon City, Philippines ABSTRACT Horizontal infiltration gallery systems as water-supply sources have become an attractive alternative to vertical groundwater wells and open-channel flow diversions in the Philippines and other countries. One advantage of horizontal infiltration galleries over vertical wells is the avoidance of excessive drawdown of the piezometric head, thereby avoiding too the upconing of saline groundwater in coastal or island aquifers. Another advantage is that the water from river sources undergoes natural filtration by seepage through the stream bed deposit and riverbank soil and attains much reduced level of suspended solids inside the wells. This paper presents the method and examples for the estimation of the inflow yield of horizontal wells based on analytical models of steady groundwater flow in a fully-saturated isotropic homogeneous aquifer, using potential flow theory and source-sink pair elements. Both exact and approximate analytical solutions are de rived for seepage inflow rates, taking into account the effects of well radius and its centerline distance from the horizontal recharge boundary and the vertical or horizontal impermeable boundaries of the aquifer. KEYWORDS: horizontal wells, infiltration gallery, aquifer, groundwater flow 1. INTRODUCTION Where *P = - K h = potential function; The use of horizontal infiltration gallery K = permeability or hydraulic conductivity of the systems (horizontal wells or radial collector wells) aquifer medium; as water-supply sources has become an attractive h = hydraulic or piezometric head = pressure alternative to vertical groundwater wells and head + elevation head; open-surface flow diversions in the Philippines V = bulk seepage velocity, equal to the gradient and other countries. One advantage of horizontal of the potential function. infiltration galleries over vertical wells is the In general, a complex potential is defined by avoidance of excessive drawdown of the piezo analytic functions: metric head localized near few vertical well w(z) = A fast method of preliminary estimation of Both ç and \|/ satisfy Laplace Equation: the inflow yield and sizing of an infiltration gal V- /5y = u- i v (9) 15 2. PREVIOUS RELATED WORKS r ^ 7t(y + r/2)' (11) c = -ln 4 sin m sin 2B An early application of complex analysis with conformai mapping using image wells in = the effective flow resistance coefficient; for two-dimensional flow theory was done by List r<0.1B (1964) for the problem of the steady flow of Ha = hydraulic head around or outside the precipitation (percolating recharge) to an infinite horizontal well; series of horizontal tile drains above an Hs = hydraulic head inside the horizontal well; impervious layer. The free surface boundary r = radius of the horizontal well (assumed small conditions are exactly satisfied, but the shape of compared to die aquifer thickness); the impervious layer is approximated. Involving y = centerline height of the horizontal well above computations with trigonometric and hyperbolic die impervious aquifer bottom; functions, the analytical results can be applied to a B = Üiickness of the confined aquifer. series of horizontal wells which lie above an impervious layer and receive inflows from vertical In the same paper, Haitjema derived by recharge. Ilyinsky and Kacimov (1992) made a conformai mapping with two source-sink pairs die study of seepage to a system of empty horizontal effective resistance to converging flow to a drains for one of three possible patterns of water horizontal well underneath a recharging stream tables and flow regimes: (1) an empty drain (an and above an impermeable bottom: isobar), (2) a drain partially filled with water, and (3) a drain completely filled with water (an f 7tr Ï ^(y + r/2)^ c = -ln tan — equipotential line). tan — - ) V 2B J (12) In a more recent work by Kolymbas and Wagner (2006) an analytical expression for the to be applied to Equation (10) in which Ha is estimation of the steady inflow into a drained defined as the constant hydraulic head along the tunnel of circular cross section under a constant horizontal recharge boundary located above the head line is derived on the basis of conformai horizontal well, and die other symbols retain the mapping for a source-sink pair. Their results are same meaning. This result is also derived in this identical with those in this paper but are obtained present paper witii an equivalent conformai here with a different method. mapping but is expressed in a different manner. Haitjema (2006), who is an noted 3. THE SOURCE-SINK PAIR MODEL AND proponent of me analytic element method in ITS VARIATIONS groundwater flow modeling (the use of several The present paper is concerned with a analytic computational elements such as sources, vertical two-dimensional steady-flow model sinks, and doublets in groundwater flow models) provided by die classical vertically aligned analyzed by conformai mapping the seepage flow source-sink pair, wherein the steady recharge line problem near a horizontal well modeled as a line with constant head is represented by die symmetry sink within a Dupuit-Forchheimer model of a line between the hypothetical source and sink, confined aquifer. He derived the approximate while the outer circumference of die gallery or resistance in the three-dimensional flow around horizontal well coincides widi a circular potential the horizontal well, as given by the following line which circumscribes eccentrically the relationship (written in this present paper's hypothetical sink located inside the gallery. By notation): the direct use of analytic geometry of circles and the algebraic manipulation of me potential 27tK(Ha-Hs) q = function, and without me need to perform (10) conformai analysis, die same results are obtained as those by Kolymbas and Wagner (2006). The where flow is depicted in Figure 1. q = inflow per unit length of horizontal well; K = hydraulic conductivity of the aquifer medium; 16 3.1 A horizontal well below a horizontal and a real sink at location (x,y) = (O.-b), which recharge boundary in a semi-infinite aquifer specifies a line of symmetry, y = 0, which is a (exact solution) horizontal recharge line with an assigned potential Let Rfvar water awlaca during high (Iowa (wat Mason) or parannial low Dow» (dry —a» on). . |gfrgJa¡ >. -15 -20 -25 -30 -15 -10 Figure 1. The circular potential lines and streamlines around a horizontal well lying below a constant recharge line, modeled by a source-sink pair (only the sink is shown for y <0). Defining the complex potential, w(z): Along die recharge line, y = 0, the value, *P = -K H, applies at any x. Also, let F be defined as follows: w(z) = -KH + -5-ln(z - lb) - -3-ln(z + ib) 2ir 271 4TTKH -4—1 (13) F = exp (15) i Lz + ibJ X + y b KH + ^h ' ( - H + ,-M«rtJ^l-«t4—11 4x [^+(y + by\ 2*1 L * J L x Jj where Deriving die equation of the curve corresponding K = permeability of me aquifer medium; to me exit potential, j=0, on the gallery wall, then H = head difference between die horizontal Equation (15) is converted to me equation of a recharge line, y = 0, and die outer circular circle: wall of me gallery (possibly including a x2+(y-b)2=F[x2+(y + b)2] (16) gravel packing); q = specific discharge (discharge per unit gallery which is rearranged into the standard form: length or horizontal well); 2 b = depm of sink centerline below die horizontal F + 1 4b F x2 + y + b recharge line, y=0. F-1 (F-1)2 It follows that the potential function is equal to: which are circular potential lines witii the following properties: 2 2 2 2 9 = -KH+^ln[x + (y-b) ]-^li,[x +(y + b) ] (14) .F+1 . depth of centerline: d = = b >b = (recharge head term) + (source term) + (sink term) (being offset or eccentric relativF-1 e to b) 2bVF radius: r = F-1 ' r 2-JY so that d" F+1 17 The last precedingrelation is a quadratic equation Figure 2. provides a comparison between the exit in the unknown : circle at zero potential and the exit oval of the free seepage surface. With the value of the potential rF-2Vf+r=0 function on the free seepage surface equal to (17) j = - Ky, and mat of the exit potential circle equal to j = 0), and for the same value of the inflow # Solving for and finally for die desired inflow discharge, q, the required size of the free surface value, q: exit oval would be smaller than that of the exit potential circle, herein indicated by Figure 4. 7 2TTKH d + Vd^r VF =exp q Horizontal recharge line at y = 0: 0 (18) 2TIKH -S q = - d + Vd2-r2 -10 In -15 -20 -25 Along the recharge line, y = 0, the recharge rate is 3.2 A horizontal well near a vertical imperme given by the vertical velocity, v: able boundary below a horizontal recharge 2 2 v(x,0) = Îj/Wo = q/(4p) my{ln[ x + (y-b) ] - boundary in a quarter-infinite aquifer 2 2 ln[x + (y+b) ]}'/2y=o. (approximate solution) Hence, v(x,0) = (-qb/p) / [x2 + b2], which is Variations of tiie source-sink pair model negative or downward as expected. can be developed using image-well superposition The computed non-circular free seepage for modeling the effects of an impervious bottom surfaces surrounding the sink also offer alternative boundary and either recharging or impermeable exit geometry and hydraulic condition of zero lateral boundaries, resulting in mathematical pressure (an isobar) instead of the constant non-circular shapes or mathematical ovals for potential lines. To the potential function, j, assign theoretical gallery wall geometries (either the particular value, j = -Ky, which is applicable potential lines or free seepage surfaces). to the free seepage surface where me pressure is zero or atmospheric and the head consists only of For a horizontal well near a vertical imper elevation: meable wall and below a horizontal recharge line, j = - Ky = - KH + q/(4p) ln[ x2 + (y-b)2]/ [ x2 + let the potential function be assigned as: (y+b)2] j = sum of four potential functions K(H-y) = q/(4p) In [ x2 + (y-b)2 ]/ [ x2 + (y+b)2 ] = two sources (at (-2a,b) and (0,b)) + two sinks (at Let G(y) = exp[4pK(H-y)/q], thus reducing the (-2a,-b) and (0, -b)). above relation to 2 2 2 2 x + (y-b) =G(y)[x + (y+b) ]. which specifies a vertical impermeable Solving for x, the geometry of the non-circular boundary located at x = -a, along which u =0, and exit oval (an isobar) can men be calculated : a line of symmetry, y=0, or a horizontal recharge x2 = [(y-b)2-G(y)(y+b)2]/[G(y)-l] (20) 18 [z-i(B + c)][z-i(B-c)] w(z) = -KH + JLln [z + 2a-ib]][z-ib] 2it [[z + 2a-i(-b)][z-i(-b)] 2K~ [z-i(-B + c)][z-i(-B-c)] 2 2 2 2 2 2 23 q[ [[(x + 2a)^(y-b) p+(y-b) ]| (2D [x + (y-B-c) ][x +(y-B + c) ] 1 ( > = (recharge head term) + (image source term at = (recharge head term) + (image source term at (0,B+c)) + (image source term at (0,B-c)) (-2a,b)) + (image source term at (0,b)) (real sink term at (0,-B+c)) + (image sink term at + (image sink term at (-2a,-b)) + (real sink term (0,-B-c)) at(0,-b)) Deriving the equation for the exit potential line, j=0: Let F = exp[4pKH/q] 3 1, then Equation 21 becomes: [(x+23)2 + (y-bfltx2 + (y-bf] = F[(x+2a)2 + (y+bfllx2 + (y+b)2]. As a reasonable approximation, assume that the vertical depth, b « horizontal distance, a, and that (x,y) is nearer to the true sink exit at (0,-b): x » 0 and y » -b, so that a reduced equation is obtained: [4a2 + 4b2][x2 + (y-b)2] = F [4a2 + 0][x2 + (y+b)2] x2 + (y-b)2 = F[x2 + (y+b)2]/[l+(b/a)2] Redefining F as F = exp[4 p K H/q] / [l+(b/a)2], then the same form of Eq. 16 is obtained. Figure 3. The composite halves (and smaller exploded views) of the Ovals of Cassini, centered 2TIKH at (0, -B+c), and located above a horizontal impermeable boundary along y = -B. A d + Vd2-r2 In. 1 + (22) Deriving the equation of the curve for the exit potential line, j=0: Let F = exp[4pKH/q]3 1. Hence, Equation 22 can be used iteratively to get q by [x2 + (y-B-c)2]^ + (y-B+c)2] = Fix2 + (y+B-c)2]^ + (y+B+c)2] (24) initializing b=d, then using the value of q to get 2 F = exp[4pKH/q] / [l+(b/a) ], then updating As an approximation for small height, c « depth, B, b = d(F-l)/(F+l), then recomputing q and F, and and (x,y) is near the well centerline: x » 0 or so on until convergence is reached. |x|«B, and y » -B, resulting in: [0 + (-2B)2][0 + (-2B)2] = F[x2 + (y+B-c)2][x2 + 3.3 A horizontal well below a horizontal (y+B+c)2]. Hence, recharge boundary and above a horizontal 16B4/F = [x2 + (y+B-c)2][x2 + (y+B+c)2] impermeable boundary in an infinite strip of (localized Ovals of Cassini near y = -B) aquifer (approximate solution) Letting 16B4/F = (m c)4 where m = a shape parameter of the ovals, the result is: Let B= thickness of the aquifer (from the x4 + 2 x2 [(y+B)2 + c2] + (y+B)4 - 2(y+B)2 c2 + c4 recharge boundary to the impervious bottom). [1-m4] = 0 Then, j = sum of four potential functions (a 4th-degree polynomial or biquadratic equipo- (which are two sets of Ovals of Cassini, with tential equation which is solvable in x). one- half of a set illustrated in Figure 3): Using 16B4/F = 16B4/exp(4pKH /q) = (m c)4, the = two sources (at (0,B+c) and (0,B-c)) + two inflow value is computed: sinks (at (0,-B+c) and (0, -B-c)) JtKH which specifies an approximately horizontal q = 2B (25) impermeable boundary along y = ±B, in which In v =0, and a perfect line of symmetry, y = 0, or mc horizontal recharge line with j = - K H. The parameter, m, has four intervals: 0 < m < 1 for small egg-shaped oval pairs, m = 1 for a lemniscate (a figure of eight), 1 < m < (2)05 =1.414 for single pinched or peanut shapes, and m > (2)05for single round ovals, as indicated in Figure 3. 19 3.4 A horizontal well below a horizontal recharge boundary and above a horizontal achieved). With two sources at (0,b) and (0,2B-b) impermeable boundary in an infinite strip of and two sinks at (0,-b) and (0, -(2B-b)), located aquifer (exact solution) relative to an exact horizontal impermeable The exact solution for this case is obtained boundary along y = ±B, in which v =0, and a by applying a conformai mapping as shown in perfect line of symmetry, y = 0, or a horizontal Figure 4 (in order to map the same z-plane in both recharge line with j = - K H, the transformation, cases 3.3 and 3.4 to another complex plane, the f(z) = exp(az), is applied where a is defined as f(z)-plane, where the exact symmetry on both follows: recharge and impermeable boundaries is 71 (X = 2B z plane Source: z = +i(2B - b) 4 Impermeable 8888«!»«« Boundary: z = x + IB Source: z2 = +ib Recharge Boundary: z = x + lO Sink: z, = -ib Impermeable Boundary: z = x - i B Sink: z3=-i(2B-b) Conformai Mapping 7t a = - 2B Impermeable Boundary f(z) = eM plane Source: f(z,) = -exp(-aib) Source: f[z2) = exp(+aib) Recharge Boundary Sink: f(z ) = -exp(+aib) 3 Sink: f(z,) = exp(-iab) Complex Potential: (e<"-e'":Xe'"-e<"') w = 6 +iy = -KH +—In aa a ! 2n (e"-e"'Xe -e ' ) tanh[a(z-ib)/2] í> -KH+-3-ta 2w tanh[a(z+ib)/2] Figure 4 The conformai mapping achieved by f(z) = exp(az) and the expression for the complex potential for a horizontal well below a horizontal recharge boundary and above a horizontal impermeable boundary (exact solution). Let Apply the transformation f(z) = exp(az) where a is defined as follows: ocz, = -iab az2 = +iab 71 a = az = -ia(2B - b) = —ire + iab 2B 3 az4 = +ia(2B - b) = +in - iab 20 The complex potential, w(z), is defined by the 4. CONCLUSION transformed sources and sinks in the f(z) plane: This paper has presented the method and (ea2-e"l2Xeai-e"4) examples for the estimation of the inflows to w = <|> + hy = -KH + —In 3 (26) 2n (e^-e^X^-e™ ) horizontal wells based on analytical models of Equation 26 simplifies to: steady groundwater flow in a fully-saturated isotropic homogeneous aquifer, using potential tanh[a(z-ib)/2] (27) flow theory and source-sink pair elements. Both w = -KH + —In exact and approximate analytical solutions are 2n tanh[cc(z + ib)/2] derived for seepage inflow rates, taking into account the effects of well radius and its The recharge rate is given by the vertical velocity centerline distance from the horizontal recharge along the line, y=0: boundary and the vertical or horizontal impermeable boundaries of the aquifer. dcp cosh(ax)sin(ab) v(x,0) = qa 71 sinh (ax) + sin (ab) 5. REFERENCES dy y=0 (28) qa v(0,0) = - Haitjema, H.M. (2006) "Accounting for resistance Ttsin(ab) _ R sin — to 3D flow near horizontal wells or galleries in a Dupuit-Forchheimer model", http:// Deriving the equation of the curve for the exit www.haitiema.com/documents/. potential line, j=0: Ilyinsky, N.B. and A. R. Kacimov (1992) "Problems of seepage to empty ditch and tanh[q(z-ib)/2] drain", Water Resources Research, 28(3), pp 0 + ivu = -KH + -^-ln 2*i tanh[a(z + ib)/2] (29) 871-877. Kolymbas, D. and Wagner, P. (2006) e Let 0 < r « b, and z = -ib + re' : "Groundwater ingress to tunnels", Tunneling 2JIKH 27tKH and Underground Space Technology, 12 pp. q = —r 2tan(ab) (in press). In In -ln-^ List, E. J. (1964) "The steady flow of precipitation ar V2B M§). to an infinite series of tile drains above an impervious layer", Journal of Geophysical For r < b « B, the value of q approaches the Research, 69(16), pp 36-46. 2TIKH (30) q~ln(2b/r) 21 EFFECTIVE MANAGEMENT OF WATER RESOURCES VIA DEMAND MANAGEMENT: SOME EXAMPLES FROM SOUTHEAST ASIA N.W. Chan1 and V. Nitivattananon2 1 Universiti Sains Malaysia, School of Humanities, Malaysia;2 Asian Institute of Technology, School of Envi ronment, Resources and Development, Thailand nwchan(a),usm. my ABSTRACT This paper examines how demand management can be employed effectively to manage water re sources sustainably in some Southeast Asian countries in a climate of dwindling water resources. In the 21st century, water is valued as "Blue Gold" and has become an increasingly critical social, economic and politi cal issue. Currently, more than two billion people in die world live in regions of water stress, and unless hu man societies radically change their consumption patterns over the next few decades or so, water demand needed for a rapidly growing population will double. Based on a worst case scenario, scientists warn that the escalating water crisis could manifest itself in terms of water conflicts, which may end up as "Water Wars". Rivers are grossly polluted and drying up in many countries. Over-abstraction and pollution are decimating groundwater. And lack of water will eventually reduce food production, leading to famine and malnutrition. While the Southeast Asian region (the focus of this paper) is considered rich in water resources, it is not free from water issues and problems. The region is rapidly developing and its population growing fast. As more and more water sources are abstracted, polluted and depleted vis-à-vis population explosion and economic development, water supply increasingly lacks behind water demand. As the total quantity of available water is finite (if not decreasing due to pollution), and demand increasing at geometrical rates, many countries are facing water shortages. In many countries in the region, balancing water supply and water demand has be come a main concern, and many are facing severe problems. Currently, many countries (or parts of the coun tries) are facing water stress when the water balance becomes negative. This research indicates that in many Southeast Asian countries, the traditional government managed (top-down water management approach) based solely on Water Supply Management (WSM) is ineffective as it tackles only one side of the coin. Vis- à-vis dwindling water resources and increasing demand, this has forced consumers to play a more active role in water conservation, notably via water demand management (WDM). Here, it should be noted that WDM is not solely on the domestic front.WD M can be most effectively employed by all water consumers, viz. indus try, businesses, government agencies, universities, schools, hotels, etc. Naturally, the role of women in con trolling water wastage as well as educating children on wise use and conservation is vital. Since domestic consumers use roughly about half of the regions' total water demand, WDM is an important tool for sustain able water management. This paper examines how water consumers can manage water via WDM in an over all comprehensive strategy for sustainable water resources management. It looks at the ways in which indus try, universities, schools, and households can use WDM to help achieve sustainable management of water resources in Southeast Asian countries. Keywords: Water Demand Management, Water Supply Management Water Conservation, Sustainable Water Resources Management Introduction is) a region of rapid economic development and The total amount of water on earth has social transformation. Globally, many Southeast remained almost constant since 3.8 billion years Asian countries are being touted either as "Asian ago when earth cooled. The estimated volume of Tiger Economies" or "Newly Industrialising total water on earth is about 1.36 billion km3 Countries (NICs)". Singapore, by most accounts, (Christopherson, 2000: 172). Hence, the amount is already a developed country. Others such as of water is finite. But population, industry and Thailand and Malaysia are also rapidly other developments have increased rapidly and are developing. Against a background of rapid infinite, and do not appear to have any limits. development, both Thailand and Malaysia have Therein lays the root of the problem in regard to experienced and are still experiencing mounting water supply and water demand, the former being environmental degradation, and Urban Environ finite whilst the latter is infinite. In the past few mental Management (UEM) problems in their decades or so, Southeast Asia has been (and still main cities. 23 The greater emphasis on rapid economic and so they may not necessarily be applicable to all coun cial development has inevitably brought about a tries. Traditional water management systems, number of UEM problems, chief of which is dete based on traditional wisdom, can supplement riorating air quality, water pollution, poor sanita modern water management systems and can be a tion and inefficient and inadequate solid waste useful tool within small communities (Centre for management (http://www.sea-uema.ait.ac.th/snp/ Science and Environment, 1997; Chan, 1999). The forum 1 /policvforum 1 .htm 7 October 2005) Con important role played by women in managing wa sidering all these problems, none other is as im ter within the family as well as within the commu portant as water as it cuts across all other environ nity is also highlighted by many (Chan, 2000a; mental issues. In most Southeast Asian countries, Hajar et al., 2002; http://www.sea-uema.ait.ac.th/ despite the richness of culture and traditional GenderReports.htm 7 October 2005). Figure 1 ways, including traditional water management shows that many countries are below the Interna systems, mere is a tendency for governments to tional Stress Line of 1700 m3 of annual renewable employ the top-down and technocentric approach, water per capita. Amongst the countries in South leaning heavily towards privatization in the man east Asia, Lao PDR has the highest water avail agement of the water sector. While such an ap ability, followed by Cambodia and Malaysia. Sin proach may be successful in some countries, gapore is the poorest. # cf ^ ^ of ^ j> Country Figure 1 : Annual renewable water resources per capita in selected countries (Note: International Stress Line is equal to 1700 m3 renewable water per person per year). Background on Water Supply and Demand in wet equatorial climate regime with rainfall all year Some Southeast Asian countries round, and no month is completely dry. Based on Taking Malaysia as an example, the an average annual rainfall of about 3,000mm, country is set to become a fully developed nation Malaysia is endowed with an estimated total by the year 2020. The pace of industrialization annual water resource of some 990 billion cubic and technological change is also rapid, as is its metres (BCM) (1 BCM = 1 million Mega litres), rates of urbanization, population growth, with total available water of 630 BCM (i.e. infrastructure development, tourism, housing and groundwater of 64 BCM plus surface runoff of others. Other things being equal, these growths 566 BCM, the remaining 360 BCM being unavail and changes are only possible with adequate water able evapo-transpiration (Hj Keizrul bin Abdullah, supply. The distribution of water resources is 1998). Currently, only 2.7 % of total available uneven, both spatially (State-wise) and runoff (surface water) is utilised. Based on these temporarily (time-wise). Malaysia experiences a statistics, it is undeniable that Malaysia is blessed 24 with an abundance of rainfall and water resources. capital city of Bangkok and other regional urban Theoretically, Malaysians enjoy a per capita centers. Traditionally, Thailand is agriculture- renewable water of more than 20,000 cubic meters based, and remains so with a total agricultural area per year. The World Resources Institute estimated of about 265,200 km2 with more than 60% of its that Malaysia has 25,178 m3/capita/year. (http:// population employed in agriculture. Agriculture earthtrends.wri.org/text/FRE/variables/694.htm 22 consumes the most water, about 71 % of total April 2004 2050 hrs). However, due to pollution, water consumption. In comparison, domestic and destruction of water catchments and other reasons, tourism industry consumption is only about 5 % this amount may be less. The water use in the year and 2 % of total water consumption respectively. 2000 is about 15.5 billion cubic meters (BCM) of The remaining 22 % are for ecological balance. which 10.4 BCM (67%) is for irrigation and 4.8 However, recent developments are expected to see BCM (31%) for domestic and industry. This fig a drop in the percentage water usage by agricul ure is expected to increase many folds by the year ture but a corresponding increase in both the in 2020 and beyond and it is imperative that Malay dustrial and domestic sectors. Rapid development sia formulates a Malaysian Vision for Water (Tan in the past decade has also seen water demand Sri Razali Ismail 2001). Despite Malaysia being increase sharply resulting in the Northeast and the endowed with such heavy year-round abundant Central Plain regions experiencing more frequent rainfall and rich water resources, many parts of droughts (http://www.fao.org/documents/ me country periodically suffer water stress and show cdr.asp?url file=/DOCCREP/004/AB776E/ water crises, increasingly so in recent decades. ab776e04.htm). In other areas, flooding also oc Despite the blame being put on El Nino and cli curs more frequently due to deforestation, global mate change, Malaysia's climate has not changed warming and urbanization. As a result of escalat substantially and most of the water problems can ing water problems, the national water resources be attributed to human mismanagement of water development budget has been increasing and takes resources (Chan, 1998a). The country's approach up a large portion of the national budget for devel is still steeply focussed on WSM but the number opment. However, current environment con of rivers and their capacities are fixed. The major straints may slow down large water resources de ity of rivers used for water supply and irrigation velopment projects in the future. In terms of ac has in fact reached their maximum capacities (Hj cess, 80 % of the urban population has access to Keizrul Abdullah 2002). Increasingly, water prob treated pipe water. In comparison, only 70 % of lems have taken central stage in the country's de the rural population is served a variety of water velopment leading some to predict an impending supply systems, viz. piped water mains, rainwater water crisis (Hamirdin Ithnin, 1997). In recent jars and tube wells. Consequently, many rural years, bad weather coupled with environmental households still have to rely on alternative or degradation due to badly planned developments other water sources. Many rural communities also have led to water crises of which the 1997/98 El manage their own water supply systems (TDRI Nino induced water crisis was most severe (Chan 1990; Neef 2004). and Kung, 2001). The federal capital and many Unlike Malaysia, which depends largely on parts of Selangor, Negri Sembilan, Melaka and surface water (97 % of water supply is from sur Penang suffered prolonged water shortages. Parts face water), groundwater is the main source of of interior Sabah and Sarawak endured severe water supply in Thailand, contributing 75 % of crop losses and thousands of people were reported domestic water supply. Groundwater also supplies to be suffering malnutrition (Chan, 1998b). Until about 20 % of Thailand's 220 cities and towns and that episode, almost all water management efforts for half of the 700 sanitary districts. Heavy mon were directed at WSM. It was after this episode soon rains annually recharge the groundwater sys that WDM emerged in the government's agenda, tem in Thailand to the tune of 40,000 million m3. and despite that, the WDM efforts were mostly ad The other source of recharge is via seepage from hoc and unsustained (Chan 2006). Only recently surface waters, mostly from rivers. Currently, it is did the government launch the firstnationa l cam estimated that more than 200,000 groundwater paign to save water in mid-2006 (http:// well projects are operated by the government and www.water.gov.mv 23 Aug 2006) the private sector, giving a total capacity of about In the case of Thailand, the country 7.55 million m3 per day (2,700 million m3 per experiences a monsoon climate with a total land year) (Gupta 1982). Despite heavy rainfall and area of approximately 512,000 km2. In 2005, the adequate water resources, Thailand is not without estimated population was about 65.4 million, with water problems. The history of water resources an annual growth rate of approximately 1.0 %. management dates back to more than eighty years, The urban population was estimated at about 11 but population explosion, rapid developments in million (16.8 %), with high concentration in the agriculture, rural development, industrialization, 25 tourism and urbanization have sharply increased fresh water supply through long-term agreements. water demands. Lack of awareness, apathetic Singapore imports about half its water attitudes, mismanagement, wastage, and requirement from Malaysia, making water a inefficient water use by various sectors and critical and strategic issue. Singapore wants to pollution has given rise to many water problems. purchase more water as a security as well as for Consequently, many areas face increasingly greater development needs, but it has not been serious problems in terms of water quantity and able to come to an agreement with Malaysia on quality. On top of that, water demand for the water price. Hence, water of late has become a agriculture is very high at 71 %. Together with bone of contention that is affecting bilateral industrial, domestic and other uses countrywide, relations between the two neighbours. Yet, despite total water demand is estimated to be about such limitations, Singapore has very efficient 86,000 million m3 per year in 2006. Increasingly water management. Its state-owned PUB is also, sectoral water use have also given rise to amongst the most well managed and efficient water use conflicts which if unattended, can lead water supplier in the region. Unlike other to serious problems (Nitivattananon et al., 2004). Southeast Asian countries that overly rely on Given mis scenario, Thailand faces serious water WSM, Singapore is quick to realize the supply constraints that are expected to stunt importance and need for WDM. Over the years, development growth as water is needed in all Singapore has had a quick start and has been spheres of development. Increasingly, water users exploring and implementing policies to improve will have to pay for water as well as for the efficiency of water usage through reduced per wastewater discharge. Water rights and allocation capita consumption and wastage, recycling, and plans will have to be set up to minimize and medi measures to diversify water supply sources (Goh, ate escalating conflicts (Nitivattananon et al., above). Singapore is aware that the high 2004). Essentially, the challenges of water dependence on Malaysia for its water is management in Thailand in the 21st century is strategically frightening and unhealthy to its long- likely to be related to allocation of water, term security interests and economic utilizing partnership and integrated approach, and development. Under tremendous pressure and the sustainable use and development (Nitivattananon lack of choice, Singapore is determined to manage 2005). Like Malaysia, Thailand's approach is also its water deficiency problem and overcome its focussed on WSM although in terms of IWRM, it vulnerability and, in so doing, this small island has been touted as die leader in the region (http:// state has managed to show its ASEAN neighbours www.adb.org/Water/Actions/THA/learning-from- and die world that it can survive and prosper histroy.asp 23 Aug 2006). In examining under a climate of water scarcity. With the support Thailand's water Vision, Sethaputra et al (2000) of its industry, citizen and otiier water users (via remarked tiiat the present stage of development tariff and strict water use regulations), the country (2000 onwards) necessitates demand-side water has managed to evolve a highly efficient water management. During this stage, the country's management system that is deemed second to water management will be characterised by none in die region, if not the world (Goh, Posted transport of water from distant sources to at http://www.greenleaf-publishing.com/gmi/ populated/urban and industrial areas, by control abstracts42/goh.html 23 Aug 2006). and regulation of wastewater and by efforts to conserve water. Water Use and Water Conservation In the case of Singapore, already The above discussion clearly defines a considered a developed country, mere are more salient point - countries with ample water serious challenges and choices to be made in resources such as Malaysia and Thailand are terms of water supply (Goh, National Institute of besieged with water problems, but a small country Education, Singapore Posted at http:// state like Singapore which does not have enough www.greenleaf-publishing.com/gmi/abstracts42/ water and relies on Malaysia for half its water is g o h . h t m 1 23Aug2006). managing its water well. It all boils down to one The country is also climatically wet with more point - Thailand and Malaysia botii rely heavily than 2,500 mm of rainfall annually, but the island on WSM but have ignored WDM, probably due to city-state has a population of 4 million and a me age-old misconception that botii countries robust economy, both needing water. Singapore is have ample water resources. Both countries are therefore considered a water-stressed country, also seasonally flooded severely, giving rise having to buy water from neighbour Malaysia to anoüier misconception diat in fact botíi since the 19th century. According to Goh (above), countries have "too much" water (Chan et al., before its separation from Malaysia in 1965 it was 2002; Nitivattananon 2005). The irony is that Malaysia can afford to sell water to Singapore, but many parts of the country such as the smaller In terms of water use, experts have outlined states of Penang, Malacca and Perlis suffer from a basic daily water requirement (DWR) of about water stress (Chan 2004). This example of two 50 litres per capita per day for the purposes of countries with richwate r resources managing their drinking, sanitation, bathing, cooking and kitchen water resources poorly, and one country with poor needs—and urged its recognition as the standard water resources managing it well, proves that the (STD) against which to measure the right to safe root of poor water management is not in terms of water (Gleick 1996 and 1999). WHO sets a level water scarcity (the natural or physical dimension) of 100 litres. For academic discussion, the average but rather mismanagement (the social or human consumption ranges from 5.4 litres (barely enough dimension). Both Thailand and Malaysia overly to live on) in parts of low rainfall countries during focus on WSM but neglect WDM, and suffers as a the dry season, such as the Sahel, to more than result because very few of its citizens even bother 700 litres per person per day in parts of the devel to save or conserve water (due to low tariffs, oped world such as the USA. Malaysians use on apathy, lack of awareness etc). Even fewer the average about 300 litres per capita per day, industries and businesses bother to install water though statistics for urban Malaysians may be recycling plants or implement water saving much higher at about 500 litres per capita per day. strategies, again because it is not mandatory to do Much of this extra water used is actually wasted so and also because water is too cheap to make it since most of it goes to flushing toilets, washing economical to install expensive recycling or rain cars, gardening and showering. In the public do fall harvesting plants. In the case of Singapore, the main, there is also a lot of water lost through non- country employs a comprehensive water revenue water (Chan, 1998a; Chan, 1998b). management strategy complementing WSM with Hence, the projected estimates of the growth in WDM. Singapore targets all water users and water demand are attributable to: 20% from popu lation growth, 40% from increased consumption "force" them to conserve and reduce water use, and 40% from leakages. In Thailand, the average recycle water and install recycling water consumption per capita is about 240 litres plants/mechanisms. Water tariffs in Singapore are per day. In Singapore, the corresponding figure is high and the penalties for wasting water is severe only about 140 litres. Figure 2 gives a good (http://www.pub.gov.sg/home/index.aspx 23 Aug illustration on the amount of water usage by 2006). Singapore mobilizes everybody to be country. involved in water conservation. Other methods employed by Singapore are: Use of membrane technology to re-treat and re-use wastewater Water Demand Management in Industry and producing Newater which is bottled for sale; Business Establishments Getting people to "Feel, touch and experience" NEWater at the NEWater Visitor Centre; Water Demand Management (WDM) is all Education and Awareness programme for schools; about managing the water consumer's water Water competition for the public and schools; usage, be it in a factory, a hotel, a school, in the Giving water awards such as "Friends of Water", office or simply at home. In the modern era, the an annual award by PUB, the national water modern way of life (for example using of long agency, that recognizes organizations and baths, toilet flushing, use of dishwashers and individuals who have contributed significantly washing machines, washing vehicles with a high towards raising awareness about water and what it speed hose, etc) is extremely wasteful in terms of takes to sustain Singapore's water supply; Offer water use (WWF-Malaysia 2004). For example, a ing Singaporeans to be a volunteer to take care of rural person in Southeast Asia probably uses only MacRitchie Reservoir Park, Singapore's oldest 100 litres of water a day. In contrast, his/her and most beloved reservoir which is used for a counterpart in the city uses more than 500 litres. range of recreational activities such as cross This is five times more than die rural person and country events, large-scale and family gatherings is certainly not acceptable, as most of the water is and excursions; Launching of "Water Efficient mainly used for flushing and washing. Hence, it is Homes" at Mountbatten Constituency and "Water imperative that people start practicing water for All exhibitions" aimed at the Clean and Green conservation at home, in the office, in the factory, Living; Use of media to advance awareness and in hotels, and in every place they go. With a education campaigns; and Archives of events and concerted and determined effort, WDM can be news relating to water issues effective in reducing water demand and solving (http://www.pub.gov.sg/home/index.aspx 23 Aug water problems. Without WDM, there will come a 2006). time when there is not enough water. In many countries, water supply will be overtaken by water 27 Country Figure 2: Per capita water consumption per day in selected countries. demand very soon. Water stress is no longer a recycling plant (e.g. for water used in cooling, air- local problem. It is now a regional issue and conditioning, washing, etc); Install rooftop rain would soon exacerbate into a national problem. fall harvesting system; Install wastewater Hence, consumers need to start learning fast on treatment plant (treated wastewater can then be how to conserve water and make water re-used); Run water saving campaigns in the conservation (and WDM) a way of life. The plant; Run water saving competitions between following discussion looks at the many ways in different departments; and Run water saving which industry and die public can save water, and competitions amongst employees all year round. in the process avert facing a national water crisis. Factories and businesses can also make adjust More importantly, WDM will help postpone me ments to their plumbing to conserve water. Many need to build more dams but save them for future such mechanisms and water saving equipment are generations. now available in the market, and they are very In the case of factories in industry, or competitive in terms of pricing, making installing business establishments, all of whom are large them profitable in the long run. These are shown water consumers, die following general measures in Table 1. can be applied to save water: Install a water 28 Table 1 : Water Demand Management Measures in Plumbing for Large Water Users Install low-volume flow control devices on shower heads and taps. Install automatic switch-off shower heads and taps. Install "half-flush" mechanisms on all cisterns. Replace existing one-flush system with half flush systems. Replace high volume cisterns (e.g. 9 litres) with low volume cisterns (e.g. 4.5 litres). Install "Push-Flush" system for urinals (men only) Install steel or copper pipes as these do not leak easily. Insulate hot water pipes to reduce the amount of water that must be run to get hot water to the faucet. Install the hot water heater as close as possible to places where hot water is needed - e.g. the bath room, kitchen and laundry areas. The closer the heater is to the faucet, the less water has to be run. For this reason, it's sometimes better to have two small water heaters located in strategic places rather than one big heater. Have the cleaners check all taps, pipes, shower heads, tub and lavatory faucets for leakages. Ensure leakages are reported immediately (e.g. the first to report a leakage can be rewarded). Leakages must be repaired immediately. Stress the importance of water conservation to employees to turn water faucets off tightly after use. Re-adjusts the float level of all toilet cisterns to reduce the amount of water necessary to flush the toilet. Install a piping system that channels treated wastewater, harvested rainwater, air-conditioning effluent water and other effluent water (called "greywater") into toilet cisterns for flushing. Table 2: WDM Measures in Canteens and Cafetarias Install a piping system from wash basin (for washing vegetables) to toilet cistern or a small pond (in factory compound) to collect the greywater for reuse. Use only biodegradable dishwashing liquid. Collect organic waste for composting. The compost can then be used for gardening. Emphasis the consumption of fresh food like salad and steamed food. These use less water to pre pare and are less oily. Hence, utensils need less water for washing. Use utensils for cooking and boiling that have tight lids to reduce loss of water through evaporation. Do not overcook as this wastes water and energy. Time the foods that are being cooked. Pressure cookers can save time and water. Do not serve bottled water as many do not finish me entire bottle. Serve water in a jug so that em ployees will pour out only what they need to drink. Unfinished jugs of water can be used for other purposes. Plan the cooking thoroughly by cutting down on the number of utensils used in preparing food. Al ways boil first and then fry. A pot used for boiling water will not need to be washed before reused for frying, but not the other way round. The dishwashing liquid should be mixed with the correct amount of water. In the case of dishwashers, wash only when there is a full load. Avoid using a dishwasher as a dishwasher uses about 50-75 litres of water per load. In the canteens of factories and business friendly and biodegradable; Carpets and rugs need establishments, Table 2 indicates the various to be vacuumed regularly so as to reduce the need measures that are applicable and effective in re- for shampooing (as this requires water); Put deco- ducing water demand. rative items (such as crafts) instead of potted In government, private sector or NGO plants in the office to reduce the need for water- offices, universities and schools, the following ing; Recycle paper, tin cans, etc as this will indi- measures would help reduce water use: Ensure the rectly reduce water demand (by other sectors); and cleaners always use a mop and never a hose to For potted plants in the office, use "greywater" wash cement, concrete, marble or other tiled from washing hands or harvested rainwater. Water floors; Cleaning liquids must be environmentally plants only when needed as over watering can kill 29 plants (WWF-Malaysia 2004). Hotels are some of the largest consumers on die 333 rooms X 198 litres saved per room, the of water in the cities and tourist destinations of total amount of water saved per day per room in Southeast Asia. Hence, it is not surprising to find the hotel is 65,934 litres. This is equivalent to that many international class hotels use more 1,978,020 litres per month. Translate mis into water than a small town. Hence, it is imperative monetary terms based on the average water tariff that hotels be made mandatory to install recycling in Penang for hotels which is 78 sen per 1,000 plants, rainfall harvesting systems, install waste litres (0.078 sen per litre), the amount saved is water treatment plants, install water-saving 1,978,020 litres X 0.078 sen = RM154,285.56 equipment, and find alternative sources of (RM1 = US$0.38). This is if the hotel is at full untreated water for gardening and washing. Water occupancy. Even if based on a 72 % occupancy used in air-conditioning and cooling systems in rate the amount saved would be RM111,085.60 hotels can also be recycled. There are certainly per month. The annual savings in this hotel based many measures hotels can do to reduce water use on 100 % and 72 % occupancy rates are and help towards WDM. Some of the measures RM1,851,426.70 and RM1,333,027.20 are shown in Table 3. respectively. Hence, it makes business sense to be Table 4 is a detailed water conservation socially and environmentally responsible. Not calculation made before and after a hotel in surprisingly, many hotels are going for ISO14,001 Malaysia implements WDM measures (Chuah certification as it requires water conservation as 2006). The hotel selected had 333 rooms and man well as other forms of environmental auditing. aged to reduce its water usage from 804 litres per ISO14,001 certification has been shown to room per day to 606 litres per room per day. This generate greater savings and profits (http:// is equivalent to a total savings of 198 litres per www.bsi-global.com/News/Releases/2002/ room per day or 99 litres per guest per day. Based August/n3f03078e4c77a.xalter?print only=l 24 Aug 2006). Table 3: Water Saving Measures for hotels (1) General Measures: Go for ISO 14001 Certification (this certification requires water auditing, amongst others). Install a water recycling plant (e.g. for water used in cooling air-conditioning, washing, etc) Install rooftop rainfall harvesting system and stormwater harvesting system. The water harvested can be used for gardening and general washing of outside compounds. Install wastewater treatment plant (treated wastewater can then be re-used) Run water saving campaigns in hotel amongst guests and employees. Run water saving competitions between different departments Run water saving competitions amongst employees and guests. Install water meters in rooms and charge guests for over-usage. Change only bed linen and towels when requested by guests, Use biodegradable soap, shampoo, washing powder, cleaning liquids, etc. Replace air-conditioners in public areas such as the lobby, garden terrace, beach front, open-air cafeteria, etc with ceiling fans. Monitor and control water quality in swimming pool. (2) Measures Relating to Plumbing: Install low-volume flow control devices on shower heads and taps. Install automatic switch-off shower heads and taps. Install "half-flush" mechanisms on all cisterns. Replace existing one-flush system with half flush systems. Replace high volume cisterns (e.g. 9 litres) with low volume cisterns (e.g. 4.5 litres). Install "Push-Flush" system for urinals (men only) Install steel or copper pipes as these do not leak easily. Insulate hot water pipes to reduce the amount of water mat must be run to get hot water to the faucet. Install the hot water heater as close as possible to places where hot water is needed - e.g. the bath room, kitchen and laundry areas. The closer the heater is to the faucet, the less water has to be run. For this reason, it's sometimes better to have two small water heaters located in strategic places rather than one big heater. • Re-pipe air-conditioning water used for cooling (after cooling, water is hot) to bathrooms. 30 Table 3: Water Saving Measures for hotels (continued) • Have the cleaners check all taps, pipes, shower heads, tub and lavatory faucets for leakages. Ensure leakages are reported immediately (e.g. the first to report a leakage can be rewarded). Leakages must be repaired immediately. • Stress the importance of water conservation to employees and guests to turn water faucets off tightly after use. • Re-adjusts the float level of all toilet cisterns to reduce the amount of water necessary to flush the toilet. • Install a piping system that channels treated wastewater, harvested rainwater, air-conditioning effluent water and other effluent water (called "greywater") into toilet cisterns for flushing. • Guests using the least amount of water (per capita daily basis) win a prize. (3) Measures in the restaurants and cafetarias: • Do not change plates, cups and cutlery unnecessarily during any 1 meal. Chinese restaurants have a bad hobby of changing plates after every course. • Install a piping system from wash basin (for washing vegetables) to toilet cistern or a small pond (in factory compound) to collect the greywater for reuse. • Use only biodegradable dishwashing liquid. • Collect organic waste for composting. The compost can then be used for gardening. • Emphasis the consumption of fresh food like salad and steamed food. These use less water to prepare and are less oily. Hence, utensils need less water for washing. • Use utensils for cooking and boiling that have tight lids to reduce loss of water through evaporation. • Do not overcook as this wastes water and energy. Time the foods that are being cooked. • Pressure cookers can save time and water. • Do not serve bottled water as many do not finish the entire bottle. Serve water in a jug so that employees will pour out only what they need to drink. Unfinished jugs of water can be used for other purposes. • Plan the cooking thoroughly by cutting down on the number of utensils used in preparing food. Always boil first and then fry. A pot used for boiling water will not need to be washed before reused for frying, but not the other way round. • The dishwashing liquid should be mixed with the correct amount of water. • In me case of dishwashers, wash only when there is a full load. • If possible, avoid using a dishwasher as a dishwasher uses about 50-75 litres of water per load. (4) Measures in the Front Office and other offices: • Ensure the cleaners always use a mop and never a hose to wash cement, concrete, marble or other tiled floors. • Cleaning liquids must be environmentally friendly and biodegradable. • Carpets and rugs need to be vacuumed regularly so as to reduce me need for shampooing (as this requires water). • Put decorative items (such as crafts) instead of potted plants in me office to reduce the need for watering. • Recycle paper, tin cans, etc as these efforts will indirectly reduce water demand (by other sectors). • For potted plants in the office, use "greywater" from washing hands or harvested rainwater. Water plants only when needed as over watering can kill plants. 31 Table 4: Water Savings via WDM Measures in a Hotel Rftfnrp. WDM After WDM Hotel average occupancy ~ 72% ~ 72% Hotel total usage (per month) 13,053 m3 9,850 m3 Hotel room usage only (%) 61.5% 61.5% Actual hotel room usage only 8,028 m3 6,058 m3 Monthly water usage per guest room 8.028 m3 == 24.11 m3 6058 = 18.19 m3 333 rooms 333 rooms Daily water usage per guest room 24.11 m3 = 804 litres 18.19 m3 =606 lit per day 30 days 30 days Daily water usage per hotel guest 402 litres per day 303 litres (Source: Chuah 2006) Another hotel with 246 rooms managed to Domestic Water Conservation reduce its water usage from 782 litres per room per day to 573 litres per room per day. This is In the home, parents, especially mothers, equivalent to a total savings of 209 litres per are important in managing the household water room per day or 104 litres per guest per day. consumption. Parents act as the primary means of Based on the 246 rooms X 209 litres saved per making water available in the family, identifying room, the total amount of water saved per day in water problems in the home such as leaks, faulty the hotel is 51,414 litres. This is equivalent to faucets, low water pressure, excessive water 1,542,420 litres per month. Translate this into charges, poor quality water, etc. They are me first monetary terms based on the water tariff in ones to know and it is therefore extremely impor Penang for hotels which is 78 sen per 1,000 litres tant that they are sensitized as far as water conser (0.078 sen per litre), the amount saved is vation is concerned. If parents are not sensitized 1,542,420 litres X 0.078 sen = RM120,308.76 per towards water conservation, then it is likely that month (RM1 = US$0.38). This is if the hotel is at fliese problems would remain unsolved and proba full occupancy. Even if based on a 72 % occu bly magnify significantly. In the home, women are pancy rate die amount saved would be the ones who wash the domes of the whole fam RM86.622.31 per month. In annual terms, the ily. Here, they can conserve water by using envi hotel would stand to save RM1,039,467.60 (72 % ronmentally-friendly detergents (which uses less occupancy) or RM1,443,705.10 (100 % occu water and are less harmful to ground water), start pancy) respectively. Hence, implementing WDM a wash only when mere is a full load, and ensure not only makes the hotel socially and environmen mat badly soiled domes get a rinse before being tally responsible but also saves the hotel a tremen washed. All these help to save water. Choosing dous amount of money. If all hotels in Penang the right kind of clothing and linen materials for (total rooms = 12,126 [http://www.penang.gov.my the household is also important as wool and many 24 Aug 2006]) were to implement WDM meas synthetic materials may need more water to wash ures like the two hotels above, the total savings in compared to cotton. Naturally, women should terms of water is 203.5 litres per room (average of always restrain themselves when it comes to a 198 litres and 209 litres) X 12,126 rooms = dishwasher. This is one mean machine when it 2,467,641 litres per day, or 74.029 million litres comes to water use. In a large extended family, per month, or 888.348 million litres per year. perhaps the need and justification for a dishwasher This amount is roughly equivalent to 44 mid-sized is there, but there isn't that many extended or even dams of 20 million litres capacity. In monetary large families nowadays. In the kitchen where 18 terms, the annual savings would be % of household water is used (Figure 3), women RM192,475.99 per day, RM5.774 million per do most, if not all the cooking. This is one area month and RM69.291 million per year. where the family can save substantial amount of 32 water. The chef can choose to cook less oily dishes (such as steaming or uncooked salad) that and hence the need to keep rivers clean (Rasagam in the long run are healthier than oily dishes. The and Chan, 1999). The children can then be treated latter uses less water both for the cooking as well to some basic water monitoring exercises whereby as for the washing up of pots and pans and they would go into the shallow river to conduct crockery. themselves. The kids as well as their mothers would enjoy this exercise tremendously as it gives them the chance to get into the river and do some thing. Many women are school teachers and this is IN HOME WATER USE an area where water education becomes important. Women teachers can teach students to conserve water the way they teach their own children. In most Southeast Asian countries, women are teach ers and this positions them well to educate children on water conservation. Women control the water budget in the house as they are engaged in watering of plants/vegetables (and even wash cars, though admittedly the men are the ones usually doing this Bathroom and they tend to waste a lot of water by using the 75% hose), wash floors and toilets, and other chores needing water. Hence, women can either save or waste water. But by virtue of their frugalityan d carefulness, women tend to save water rather than Figure 3: Percentage of Water Use in the Home. waste it. Table 5 is a check list on how to conserve the water in the home. There may be many other areas in which In the home, the bathroom is where the women can play an important role in water saving. battle for water conservation is won or lost since Since many governments in the region are now almost three-quarters of water is used here. Other preaching the use of Water Demand Management than bathing themselves, women also bathe the (WDM) to complement Water Supply Management children and should instill the need to save water (WSM), water conservation in the household will amongst the young. They should never let children determine whether WDM succeeds or fails. This is play with water during the baths, as this would send because more than half the treated water demand is the wrong message to kids, viz. that water could be from domestic households. Hence, the family unit wasted. They should try to minimize water used holds the key to WDM. Each litre of water saved when showering by ensuring that environmentally- may seem insignificant but it counts towards friendly showers are used, cut down the number of sustainable management of this precious resource. showers for themselves and their children, switch Picture the following scenarios in Table 6 where off shower when soaping and shampooing, do not Singapore, Malaysia and Thailand go on a national over-use soap or shampoo, use "organic-based" campaign to save water and assuming all citizens soap and shampoo that are less "soapy" and use heed the call. From me figures, it can be seen that less water to rinse, switch the shower on to the water saved can be huge. These three scenarios moderate shower speed and cut short their shower are taken for slight water stress (Scenario 1), time. In the extreme, women can recycle moderate water stress (Scenario 2) and severe water "greywaters" from bathrooms by collecting it in stress (Scenario 3). It is not impossible for a person tanks downstairs for use in flushing toilets to reduce his/her water consumption by 10 to 20 %, downstairs and gardening. All that is needed is a although arguably reducing it by 50 % would be too drastic and may lead to negative health and few simple plumbing connections and a tank. other ill effects. One cannot reduce the amount of Mothers can mould their children into drinking water a person needs (about 7 glasses per responsible water saving adults by starting them day) but one can certainly reduce the amount of young. They can take their children for outings to water for non-consumptive use such as washing, rivers instead of to supermarkets or shopping the number of showers taken, the number of flushes complexes. They can lead the children in "River of the toilet, watering of the plants, number of car Walk" along the banks of rivers to instill a sense washes, washing of floors or frequency one of attachment and love for rivers. Mothers can changes the water in the aquarium (http:// request help from WWP experts who will brief the www.waterwatchpenang.org 24 Aug 2006). children on the importance of water conservation Table 5: Check list on how to conserve the water in the home. (A) GENERAL: • Inspect plumbing and faucets often for leaks, and repair them immediately. • fit water conserving devices such as "half-flush" cisterns, spray taps, faucet aerators, low-pressure shower heads, automatic faucets, etc. and modify appliances to conserve water. • Washing clothes: Wait until there is a full load in the washing machine before washing. In die market, shop for environmentally-friendly machines with half load capability and with reduced water consumption. • During the day, tap water will be hot for the first few minutes when switched on. So, refrigerate water for drinking instead of running the tap while waiting for cold water. • Install quick heating heaters to prevent running the tap for long periods in the morning waiting for hot water. • Report all leakages in your housing area or anywhere when you see them to the water authority. (B) IN THE BATHROOM & TOILET: • Put a brick in the toilet cistern to reduce me amount of water used per flush. • Collect rain water with buckets or modify plumbing to collect rain water in a tank. • Recycle water used for washing hands, vegetables and fruits for watering plants. • Greywater from batíirooms upstairs can be collected via modified plumbing for flushing toilets downstairs. • Use showers instead of baths. Never use the long baths. Have showers instead, and if you are very keen on water conservation you can share your shower with your children. • Install a compost toilet. With the right materials and appropriate ventilation, it will not smell. In the old days everybody used die bucket toilet and the night soil was collected regularly by farmers to be used as fertilizers. Compost toilets do not need water as mere is no flushing and depend upon bacterial action to break down harmful bugs in die waste. • Make your organic wastes (fruit peel, rotten parts of vegetables, fruitseeds , etc) into compost radier tiian flushing them away witii water. • Use organic, biodegradable and "no rinse" soap and shampoo. (C) IN THE KITCHEN: • Wash dishes by hand, never use a dishwasher. • When washing dishes, use one sink for washing and anomer for rinsing. Do not use a running tap. AnoÜier way is to substitute sinks with bowls which are filled with less water than sinks. • Water used for cooking can be re-used as a stock or a base for soups. Any leftover stock can be kept for many days in me refrigerator. • Re-use water for washing vegetables and fruitsfo r water plants or flushingth e squatting toilets. (D) IN THE GARDEN AND GARAGE: • Car washing: Use a bucket instead of a hose. Limit die number of times of washing to once a week. It is not necessary to wash a car often. If the car is reasonably clean, just wipe it witii a piece of wet cloüi. • Garden watering: Use a bucket instead of a hose. It is best to water out of direct sunlight in the early morning or evening to reduce water loss through evaporation. • Never use sprinklers to water your plants or garden lawn. If you have a sprinkler, never switch to automatic mode as it will still go off when it is raining. • If you have a garden, grow me plants in die soil rather than in pots/containers. Plants do better in me soil and get more water from it (Hence, less watering is needed). • Grow plants mat require less water (e.g. cactus) rather man plants (water hyacinth) that need a lot of water. 34 Table 6: Water saving Scenarios in Singapore, Malaysia and Thailand Based on 10 %, 20 % and 50 % Water saving Scenarios. Country Per Capita Per Capita Scenario 1 Scenario 2 Scenario 3 Amount Water Use Water Use (10 % Wa (20 % Water (50 % Wa Saved Per Per Day Per Year ter Savings Savings Per ter Savings Year in Per Year) Year) Per Year) Country Based on 50 % Savings Singapore 140 litres 51,100 litres 5,110 litres 10,220 litres 25,550 litres 102.2 billion (4 million) litres Malaysia 310 litres 113,150 litres 11,315 litres 22,630 litres 56,575 litres 1471.0 bil (26 mil lion litres lion) Thailand 240 litres 87,600 litres 8760 litres 17,520 litres 43,800 litres 2,847.0 bil (65 mil lion litres lion) Conclusion more than half of each country's total water This paper has demonstrated that WDM is demand. Because of this huge volume, any the key towards sustainable water resources reduction in consumption can save a country a lot management in this 21st century as water of water, as shown by several water saving scenar availability becomes scarce. WDM is no longer an ios. Parents can play a vital role as they are the option. It has become a necessity. Such a situation "water managers" at home. In particular, mothers is happening all over Southeast Asia when can manage die family's water budget and be countries are developing rapidly and populations cause they use water for most of the domestic exploding. More and more people are also moving chores in the home, and they educate their chil to live in cities resulting in rapid urbanization that dren about water saving, they are vitally important puts added pressures on dwindling water sources. in water conservation. The amount of water is finite but water demand is infinite. Since there is no way we can create Acknowledgements: The authors would like to water, though desalination and re-treatment of express their thanks to the CIDA-AIT Partnership wastewater are expensive and technically prohibi 2003-2008 Southeast Asia Urban Environmental tive solutions, the lack of water will increasingly Management Applications Project, and the Urban become a critical issue in Southeast Asian Environmental Management, School of Environ countries if WSM remains the primary ment, Resources and Development, Asian Institute management approach. As long as water of Technology, for providing funding and admin consumers do not heed die call to save water, istrative support in me project from which this WSM will not be able to supply all that is paper is produced. demanded. Hence, mere is an urgent need for the water consumers(industry, businesses, universities Bibliography and schools, and households) to play a more active role in helping to conserve water resources, Chan, N.W. (1998a) "Priceless water not valued". viz. to reduce their water demand. Examples from The Star 29 July 1998, North, p. 6, 7. this paper has demonstrated how all these water consumers can apply WDM measures to save Chan, N.W. (1998b) "Water: Too cheap to prompt water. In addition, me role of the public, notably people to save". The Star 29 July 1998, North, NGOs and women are vitally important. p. 2, 3. Currently, domestic water users consume roughly 35 Chan, N.W. (1999) Water Conservation, ReUse Christopherson, R.W. (2000) Geosystems. 4th and Reduction of Water Use. Paper presented in the Edition. Upper Saddle River, New Jersey: workshop on "Sustainable Management of Water Prentice-Hall. Resources in Malaysia", 20 July, Kuala Lumpur. Organised by Global Environment Centre (GEC), Chuah L. (2006) Water Conservation in Hotels. Department of Irrigation and Drainage (DID), Paper presented at the National Seminar by Minis National Hydraulic Research Institute of Malaysia try of Energy, Water and Communications (NAHRIM), The Malaysian Water Association (KTAK) and Federation of Malaysian Consumers (MWA), Environmental Management and Research Association (FOMCA) "National Water Aware Association of Malaysia (ENSEARCH), and ness and Saving Campaign ", 31 July 2006, Kuala Selangor Waterworks Department (JBAS). Lumpur. Chan, N. W. (2000) Saving Water for the Future: Gleick, P. (1996) "Basic Water Requirements for The Responsibility of All. Paper presented at the Human Activities: Meeting Basic Needs." Water National Conference - "World Day For Water International 21: 83-92. Seminar 2000", March 20 - 21, Cititel, Penang. Jointly organised by MANCID-YPF, IHP, MMS Gleick, P. (1999) "The Human Right to Water." and DID. Water Policy 1(5): 487-503. Chan, N. W. (2004) Managing Water Resources in Goh, K.C. (Undated) National Institute of Educa the 21s' Century: Involving All Stakeholders tion, Singapore Posted at http ://www. greenleaf- Towards Sustainable Water Resources Management publishing.com/gmi/abstracts42/goh.html 23 Aug in Malaysia. Bangi: Environmental Management 2006). Programme, Centre for Graduate Studies, University Kebangsaan Malaysia. Gupta, A.D. (1982) "Groundwater Resources in me Bangkok Area: Development and Manage Chan N. W. (2006) "Water Demand Management ment Study - Comprehensive Report (1978- As the Key Towards Sustainable Water Resources 1982)", AIT Research Report No. 137. Management in Malaysia". Paper presented at the National Seminar by Ministry of Energy, Water Hajar Abdul Rahim, Radiah Yusoff, Noraini and Communications (KTAK) and Federation of Ismail and Chan, N W 2002 "The Role of Women Malaysian Consumers Association (FOMCA) in Water Conservation in Malaysia". In Ram Ian "National Water Awareness and Saving Omar, Zulfahmi Ali Rahman, Mohd Talib Latif, Campaign ", 31 July 2006, Kuala Lumpur. Tukimat Lihan and Jumaat Hj. Adam (Eds) Proceedings of the International Conference Chan, N. W., Ab Latif Ibrahim, Kung, H. T. and "Regional Symposium on Environment & Natural Liu, P. S. (2002) "Employing Nature To Combat Resources: Utilization, Management & Floods: Some Experiences From Malaysia". In Sustainable Development of the Environment and Juhary Ali, Badruddin A Rahman, Zaharí Natural Resources", Kuala Lumpur, 10-11 April Mohamed and Nizamuddin Zainuddin (Ed) 2002. Bangi: School of Environmental Science Disaster Management: Strengthening The Na and Natural Resourcces UKM, 161-170. tional Security Capabilities. Alor Setar: Penerbit Ustara, 143-158. Hamirdin Ithnin (1997) "Are we heading for a water crisis?" In Consumers Association of Penang Chan, N. W. and Kung, H. T. (2001) State of the Malaysian Environment. Consumers "Management of River Pollution as a Tool in Water Association of Penang, Penang, 142-145. Resources Management: Some Examples from Malaysia." In K L Risher (Ed) 2001 Proceedings of Hj. Keizrul Abdullah (1998; Hydrology for the Thiry-First Mississippi Water Resources Sustainable Water Resources Planning, Conference, April 10-11, 2001, Raymond, Missis Development and Management in Malaysia. Kertas sippi. Raymond: Water Resources Research Insti kerja yang dibentangkan di Persidangan tute and Mississippi State University, Mississippi Antarabangsa Terhadap "Hydrology and Water State, 108-120. Resources of Humid Tropics", 24-26 November 1998, Ipoh, Malaysia. Hj Keizrul Abdullah (2002) Integrated River Basin Nitivattananon, V., Tasingsa, S. and Ludwig, H.F. Management. Dalam Chan, N. W. (Editor) Rivers: (2004) "Planning for the Management of Towards Sustainable Development. Penerbit Conflicting Water uses in Thailand Watersheds: A University Sains Malaysia, Pulau Pinang, 3-14. Case Study," Proceedings of 30th WEDC International Conference, Vientianne, Lao PDR, http://www.adb.org/Water/Actions/TFLA/learning- Oct. 25-29, 2004. from-histroy.asp 23 Aug 2006. Rasagam, G. and Chan, N. W. (1999) The Role of http://www.bsi-global.com/News/Releases/2002/ Non-Governmental Organisations in Public August/n3f03078e4c77a.xalter?print only=l 24 Awareness and Environmental Education Aug 2006 Towards Protection, Conservation and Rehabilitation of Rivers: Some Examples from the http://earthtrends.wri.org/textyFRE/ Work of Water Watch Penang. In N W Chan variables/694.htm 22 April 2004. (Editor) Rivers: Towards Sustainable Development. Penerbit Universiti Sains Malaysia, http://www.fao.org/documents/show cdr.asp? 470-480. url file=/DOCCREP/004/AB776E/ab776e04.htm. Sethaputra, S., Thanopanuwat, S., Kumpa, L. and http://www.pub.gov.sg/home/index.aspx 23 Aug Patanee, S. (2000) Thailnd's Water Vision: A Case 2006. Study (in http://www.fao.org/documents/ 11 Aug http://www.sea-uema.ait.ac.th/snp/foruml/ 2006). policyforuml.htm 7 October 2005. Tan Sri Razali Ismail (2001) Keynote Statement. In http://www.sea-uema.ait.ac.th/GenderReports.htm 7 Sustainable Management of Water Resources in October 2005. Malaysia. Global Environment Centre, Petaling Jaya, 7 -10. http://www.water.gov.my 23 Aug 2006. TDRI (1990) 'Water shortages: managing demand http://www.waterwatchpenang.org 24 Aug 2006. to expand supply" in The 1990 TDRI Year-End Conference, Industrialization in Thailand and its Neef, A. (2004) People's participation in natural impact on the environment, 8-9 December 1990, resource management in northern Thailand - Bangkok. Paradigm shift or old wine in new bottles? Paper presented at the 4th Euroseas Conference in Paris, WWF-Malaysia (2004) Assessment on the Sus- September 1-4, 2004. tainability of Malaysia's Water Resources Utilisa tion by Using Sustainable Development Indicators Nitivattananon, V. (2005) Thailand water (SDI). Consultancy Report prepared by Water management - Issues & Challenges. Paper Watch Penang for WWF-Malaysia, 49 Jalan presented at the international forum on World SS223/15, 47400 Taman SEA, Petaling Jaya, Se- Citizens Assembly on Water, Kuala Lumpur 27- langor D.E., Malaysia. 30 October 2005 (Posted at http://wcaw.org/ upload files/13/Nitivattananon VILAS.pdf 1 December 2005) Managing Increased Water Demand in China: A Great Challenge Z. X. XU1 and J. Y. LI2 'Key Laboratory of Water and Sediment Sciences, Ministry of Education; College of Water Sciences, Beijing Normal University, Beijing 100875, China 2 China Institute of Land Surveying and Planning, Key Laboratory of Land Use, Ministry of Land and Re sources, Beijing, 100035, China ABSTRACT With the rapid growth of population, progressive urbanization, and the development of agriculture and industry, the increased demand on water resources has been and will continue to be a critical limitation on the development of economics in China. One of the critical problems that hinder the sustainable development will be the lack of renewable water resources. Over-exploitation of groundwater resources in North China, mostly to meet the irrigation demands and domestic uses, has already affected the aquifer's productivity both quantitatively and qualitatively. This paper gives an overview on the water uses and projected future water demands in China. Adaptation of an efficient and integrated policy for water utilization and relevant conservation techniques in various water consumption sectors are suggested. Recycling of water and artificial recharge of groundwater by surface water and treated wastewater should be adapted at a larger scale. Scarcity of freshwater will lead to China introducing demand-oriented water management strategy to complement or replace the existing supply-oriented water management practice. Key words: Water resources, sustainability, water demand, China 1. Introduction Plain, over-extraction of groundwater has even With the rapid development of economics resulted in the inundation by saltwater, leading to and the speedy growth of population, water supply contamination of remaining freshwater supplies. for the increasing industrial, agricultural, and It is estimated that the groundwater is being domestic purposes is becoming a critical issue in extracted several times faster than its China (Liu and He, 2000). From the 1960s replenishment rate. The use of water resources is onward, the demand for water has increased approaching the unsustainable level because of the tremendously. Of the major 600 cities in China, increased uses as a result of population growth, nearly 400 suffer from water shortage, and more industrial development, expansion of irrigated than 100 cities were in the serious state of water land and the escalating uncontrolled exploitation shortage (Jiang, 2000). With the unbridled for groundwater. In recent years, the problem development and over-exploitation on both associated with the increased demand for water surface and groundwater resources, a lot of has been further compounded by a drastic springs and wells dried up, many previous decrease in water quality, resulting from the perennial rivers and streams have become contamination by untreated industrial wastes, intermittent or dried up, and hundreds of lakes domestic sewage water, and agricultural fertilizers disappeared over the past several decades. The and pesticides. This has already and will continue Yellow, Huai, and Hai/Luan rivers, as well as the to place serious limitations on the development of Baiyang marshes, which helped to sustain Chinese economics in China. civilization over 5,000 years, are under the threat The continuing shortage of water resources from flow cessation. The streamflow in Yellow has become and will continue to be an important River, the second-longest river in China, has being factor restricting the sustainable development of declining due to the excessive water use. The economics and society in China. One of the major downstream dried up more than 700 km from the challenges for water authorities is to manage sea over 226 days in 1997. Besides me impact water supply under the spatial and temporal from anthropogenic activities, the global warming variations in water resource availability and the has also put great pressure on the water supply in large uncertainties involved. The water resources China. The situation has particularly exacerbated issue has been and will be still one of the top by prolonged drought in the arid North China, concerns for the development of economics. An adding to high population pressures and a rapidly overall investigation on the increased water de growing competition for water in agriculture and mand in China is given in this paper. On the basis industrial sectors. In part of the Northern China of investigation, some responses and measures on water supply management are proposed. 39 2. Study Area Description temperate zones. It therefore shows a diverse China has an area of 9.6 million km2, of climate from humid tropics in die south to which 33.3% is mountains, 26% plateau, 9.9% continental temperate climate with extreme cold hilly areas, 18.8% basin and desert, and 12% winter in the north and vast desert in the west. plains. Only about 11% of the land area is suitable According to the study made by me Ministry of for agricultural production. China is the most Water Resources (MWR, 1999), China is divided populous country in the world. The population in into 9 major basins/regions: Songhua/Liao River, 2000 already attained 1.265 billion, about 22% of Hai/Luan River, Huai River, Yellow River, the total in the world. The territory of China Yangtze River, Zhujiang (Pearl) River, Southeast stretches over 62 longitudes from east to west, and Region, Southwest Region, and Inland River extends over 50 latitudes from south to north basin, as shown in Fig. 1. across the equatorial belt, the tropics, the subtropics, the moderate temperate and die cold **M, I: Songliao Riv. VI: Zhujiang y. «.' VJ„ II: Hai River VII: Southeast ~%4 III: Huai River VIII: Southwest *:-.« IV: Yellow River IX: Inner River V: Yangtze River Fig. 1 Map of China showing nine basins/regions 2.1. Songhua/Liao River Basin 2.2. Hai/Luan River Basin The Songhua/Liao River basin is composed The Hai/Luan River basin, simplified as of with several major rivers including the Heilong Hai River in this paper, presents the major River, Songhua River, Wusuli River, and Liao challenge that the water authorities in China face River. For succinctness it is simplified as Songliao presently and in die future. The region accounts River basin in this study. The region covers for only 3.3% of die total area of die state, where approximately 1.24 million km2, accounting for 10% of the total population reside in. The water 12.9% of the total area in China. 9.7% of the resources availability per capita in mis region is population lives in this region with only 6.9% of only 358 m3 in 1993, which is die lowest one in total water resources. The water resource per the nine basins/regions. In 1997, it was further capita in this region was 1,704 m3 in 1993, and decreased to 310 m3, and projected to be only 270 further decreased to 1644 m3 in 1997, which is m3 in 2010 (Chen, 2000). Both the rapid growth of below the average of 2,342 m3 in China in 1993 or population and high rainfall/runoff variability also 2,200 m3 in 1997. exacerbate me water scarcity in this basin. 40 2.3. Huai River Basin sub-tropical/tropical climate, this region is The Huai River basin is quite similar with the particularly suitable to grow rice. The Hai/Luan River basin, which covers 3.5% of the precipitation/runoff variability in this basin is territory area and provides fresh water for 16.2% greater than that in the Yangtze River basin. The of the total population in China. This basin has the available water resources generally can meet highest density of population at 574 person/km2. the water demand even in dry years. But the Although 505 m3 of runoff per capita is available, precipitation/runoff can vary greatly within a year 41% higher than that in the Hai/Luan River basin, with only small share produced in winter, rational the Huai River basin is still a water-scarce region. storage will be necessary to improve the security of the water supply in this basin. 2.4. Yellow River Basin The Yellow River is the second longest river 2.7. Southeastern Region in China, being exceeded in length only by the The Southeastern Region is the smallest of Yangtze River. It covers about 8.3% of the the nine basins/regions covering only 2.1% of the country, and provides freshwater for 8.5% of the territory area with a higher population density population. Although it carries 58 km3 of surface than almost any other basins/regions, except the runoff annually, this is only 6% of the annual Hai/Huai River basin. The Southeastern Region runoff carried by the Yangtze River. Due to the has more than enough water, and the available high variability of precipitation and runoff, the water resources per capita is equal to 2,962 m3, a annual surface runoff produced by the Yellow little smaller than that in the Zhujiang River basin. River in a dry year may be only 50% or less of that produced in an average year. The seasonal 2.8. Southwestern Region flow also highly varies, with 60-80% occurred The Southwestern Region, covering major between June and September, and 40% or more rivers originating from the Tibetan Plateau, is occurred in July and August (Chen et al., 1997). composed almost entirely of high altitude prairie, Because nearly all the streamflow in dry season forest, and bare land. The runoff depth is not the has been diverted and used in recent years, the greatest in the southern China, but the available downstream of the Yellow River dried up continu water resources per capita is the greatest at 31,914 ously for 20 years during the past three decades. m3 due to its very low population density. The available water resource per capita in 1993 This region also has the smallest inter-annual was 749 m3, and further decreased to 707 m3 in variability of precipitation/runoff. The high 1997, which is also within the water scarce range. precipitation, low variability, and low demand in this region make that the Southeastern Region has 2.5. Yangtze River Basin the highest security of water supply. The Yangtze River is the largest river in China, covering 18.7% of the territory area with 2.9. Inland River Basin 34.2% of the total water resources. The available The Inland River basin is the largest basin in water resource per capita in mis basin is 2,388 m3 China, covering 36.4% of the territory area. It is in 1993 and decreased to 2,288 m3 in 1997, a little extremely arid with an annual average of only 35 higher than the average over the entire country. mm of the surface runoff depth over the entire It has the greatest share (23.8%) of the irrigated region. The basin irrigates 5.7% of the arable land area in China. However, due to the plentiful and provides freshwater for 2.1% of the total precipitation irrigated area accounts for only small population with its 130.4 km3 of available water percentage of the landuse in mis basin. The resources accounting for 4.6% of total amount in coefficient of variation in annual precipitation/ China. Because the population in this region is runoff is smaller compared to the basins in very small, the available water resource per capita northern China. This makes that the water demand reaches 5,270 m3, much higher than that in the in this basin can easily be met even in dry years. heavily populated basins in North China. Because This also makes the basin a good candidate for the water demand is relatively low, the local water transfer to basins in North China, where the available water resources in this region can water demand is difficult to be satisfied. generally satisfy the demand presently. 2.6. Zhujiang River Basin 3. Water Uses in China The Zhujiang River basin is strongly affected With the development of economics during by monsoon climate, producing a runoff depth of the past fifty years, the amount of water uses in 807 mm annually, and the available water China has increased drastically. Fig. 2 shows the resources per capita of 3,327 m3. Due to the changes of water uses in China during the past 41 several decades. It can be seen that the amount of 3.2. Industrial Water Uses water uses experienced a significant increase from From 1949 to 1993 the amount of industrial the 1950s to the 1990s. The tendency of increase water uses in China had increased by 45 times, was low during the past two decades. For from 2.0 km3 in 1949 to 91 km3 in 1993. This example, the water uses in 1949 was 103 km3, it amount further reached 114 km3 in 2001, an drastically increased to 519.8 km3 in 1993, increase of more than 25%. A drastic increase in showing an annual increase of 9%. In 2001, the industrial water uses was experienced from the amount of water uses was 556.7 km3, showing a 1950s to the 1990s. After the 1990s, the share of smaller annual increase of 7.1%. In China, most of industrial water keeps a relative stable value. For the water withdrawal - approximately 381.7 km3 example, the amount of industrial water use in or 73.4% was used by the agricultural sector, 1997 was 112 km3, it then increased to 114 km3 in mainly for irrigation (66.2%) in 1993. With 90.6 2001. But the share of industrial water use was km3 (or 17.4%), industry was the second largest nearly same for two years (20.2% vs. 20.5%). water user. Domestic water supply used only 47.5 The average consumption rate of industrial water km3, approximately 9.1% of the total water uses. in China was 22.7% in 1993. The greatest In 2001, the total water uses increased to 556.7 consumption rate was found in the Inland River km3, in which the shares by agricultural, industrial basin at 63.7%, and the lowest one was found in and domestic sectors changed to 68.7%, 20.5%, the Zhujiang River basin at 14.7%. and 10.8%. It is noted that die share of agriculture The Yangtze River basin and Southeast water decreased, but both shares of industrial and Region have experienced the rapidest increase of domestic uses increased. industrial water uses. The share of industrial water uses in the Yangtze River basin increased to 28.3% in 1997 from 24.8% in 1993, and this further increased to 29.6% in 2001. Similarly, this rate in the Southeast Region increased to 22.8% in 1997 from 16% in 1993, and reached 27.1% in 2001. In other words, nearly one third of the water withdrawal in both Yangtze River basin and the Southeast Region has been used in industrial 1949 1965 1980 1993 1997 1998 1999 2000 2001 sector. On the contrary, the shares of industrial water uses in both Yellow River basin and Hai Year River basin showed a slight decrease during the Fig. 2 The changes of water uses in China past several years. This may result from the during the past several decades serious shortage of water experienced in these two regions over the past decade, during that the domestic water had to be provided firstly. 3.1. Agricultural Water Uses Irrigation has been practiced for millennia. 3.3. Domestic Water Uses However, most irrigated lands were developed in The amount of domestic water uses depends on the 20th century, and about 70% of the total population, the level of services, and the avail withdrawals of water in the world are presently ability of water supply and drainage systems. It used for agriculture, and this even reaches 80% in also has strong association with climate con some developing countries (Cai, et al., 2003). dition. In some modern cities equipped with cen While most of the water is still used for irrigation, tralized water supply and efficient drainage sys the increase in irrigation water has been tems in the developed countries, domestic water remarkably low and remains nearly stable in withdrawal reaches 300-600 liters per day per China during the past decade. According to the person (led), and consumption does not usually estimation made by MWR (1999), the amount of exceed 5-10% of total water intake. But some un 3 irrigation water used in 1980 was 358 km ; and developed cities without centralized supply sys 3 this was decreased to 344 km in 1993, although tem may only have a specific water withdrawal of the irrigation area increased to 43.2 million ha in 100 to 150 led and consumption rates may reach 1993 from 40.9 million ha in 1980. The irrigation 40-60% of water intake (Shiklomanov, 2000). 3 water in 2001 was 348.5 km without considerable Similar with the industrial water uses, the domes change. It shows that the amount of water used for tic water uses also experienced a drastic in irrigation is nearly stable with a slight declination crease from the 1950s to the 1990s, especially over the past two decades. between the 1960s and the 1990s. In 1949, the domestic water uses was only 1 km3, with a 1% of share to the total water uses. 42 Even in 1965, this amount only increased to 2 the total demand to 663.3 km3 including 7.0 km3 km3, and the share to the total water withdrawal water for ecological uses. Fig. 3 further shows the had no any change (Jin and Young, 2001). projected water demand for different basins/ After 20 years it drastically increased to 28 km3 in regions, and the shares of different sectors. It is 1980. The greatest change occurred around the interesting to note that the share of irrigation wa beginning of the 1990s, the amount of domestic ter further decreased to 54.1% in 2010 from the water uses reached 47.5 km3 in 1993 with an 62.6% in 2001. annual increase of 4.1% (MWR, 1999). The aver age consumption rate of domestic water in China 250.0 | 1 was 32% in 1993. It is interesting to note that the greatest increase in domestic water uses occurred in both Yangtze River and Zhujiang River basins, with an annual increase of more than 0.3%. The Southeast Region also showed a greater increase. Most of the regions in northern China showed a decrease from 1993 to 1997, then a significant increase Region from 1997 to 2001. This may result from the drought occurred around 1997. Between 1980 and Fig. 3(a) Projected water use for different regions 2001, the amount of domestic water uses grew by in 2010 114.6%, from 28 to 60.1 km3 in China. 4. Projected Future Water Demand ECObgiCal Urban ,. l™ domestic Connttyside China has experienced a remarkable Livestocks V ,- domestic increase in both population and water demand 74%^X_i^^^/^ 5.8% during the past several decades. This increase ^^^^^^^^^^^^^^Ä Industry in water uses is mainly attributed to the wide ^^H|H¡[^^^ 23.3% Irrigation ^^^^^^^^^^^ expansion in agriculture practice. On the other 54.1% hand, the living standards have been raised considerably during this period. Most of the basins/regions have experienced a 2-10% annual Fig. 3(b) Shares of water demand for different increase in the domestic and industrial water sectors in 2010 demands over the last 20 years. From 1980 to 1993, substantial increase in industrial water uses Both the surface and groundwater runoff varies was encountered in the Southwest Region at the considerably among different regions and also annual increase of 12.6%. The substantial increase between different years. The driest Inland River in domestic water uses was found in the Southeast basin produces only 35 mm of runoff annually, Region at the annual increase of 8%. Most of the whereas the Southeast Region produces more man basins/regions showed a decrease tendency in 30 times as much (MWR, 1999). In 2001, the irrigation water uses in this period. The greatest water uses in Hai/Luan River basin was 39.2 km3, decrease was found in the Inland River basin at significantly larger than the 9 km3 of surface the annual rate of 1.4%. It is interesting to note runoff produced in the same year. The water that all basins/regions experienced an increase demand in the region is projected to reach 50.1 of water uses from 1980 to 1993, this tendency km3 in 2010. Presently, over-exploitation of the sustained for most basins/regions from 1993 to groundwater temporally bridges the gap between 2001, but the Hai and Yellow River basins the supply and demand of water resources in the showed a decrease tendency with the annual region. But this measure is a quite unsustainable decrease of 0.4-0.5% from 1993 to 2001. The total one, which has resulted in the continuous subsi water uses in China was 556.7 km3 in 2001. dence of both groundwater table level and land Of which, 382.5 km3 were devoted for the surface somewhere in the region. Due to the high agricultural practices including irrigation and inter- and intra-annual precipitation/runoff vari livestock uses and 174.3 km3 were utilized by ability, even the low water resources availability domestic and industrial sectors. The domestic and of 343 m3 per capita cannot be supplied with cer industrial demands in China are expected to reach 3 tainty. Most of the potential storage site has been 251.3 km by the year 2010, while the agricultural demand is expected to reach 412 km3, and developed and there will be quite few sites being suitable to construct new dams. The mere way will be through the inter-basin water transfer. This Considering the local hydrological features, multi is a region of extremely high water stress that purpose dams are still necessary. Multipurpose must import water to meet its needs. It shows that reservoir will be beneficial for mitigating losses the development in the North China Plain has resulted from both floods and droughts in already surpassed the carrying capacity of the the future. The conventional water planning ecosystem and very resiliency to water shortage. methodology that seeks to alleviate water Water demand in the Huai River basin has shortages by increasing water supplies through the greatly outstripped the available surface water development of water supply infrastructure supply. In 2001, the water demand reached 60.8 projects should be complemented or replaced witii km3, while the annual surface water resources was the demand-oriented strategy through water only about 37.5 km3. In 2001, the water demand in resources allocation and the improvements on the Yellow River basin was 39.5 km3 and less than water efficiency, and alternative water uses such 40.7 km3 of the surface runoff in die same year. as wastewater reclamation and desalination. The However, the projected water demand for 2010 establishment of a clear and well-defined national will outstrip the average surface runoff and the water policy for alleviating potential impact from basin shows a very low water security. Same with water shortage will be imperative. Water price the Hai/Luan River basin, the Yellow River basin reform is necessary, in which water should be also suffers from great variability in precipitation viewed as an economic commodity instead of a with very low available water resources per capita. public good. Public awareness relating to water The Inland River basin is the largest basin in scarcity issues is also very important. While China, covering 36.4% of the territory area. there are still many hurdles towards rational Because there are quite few population in this development and management of water resources, region, the Inland River basin holds a great these can be overcome by suitable planning and available water resources per capita at 5,270 m3, policy implementation, and this will be a great much higher than that in the heavily populated challenge to the hydrologists and water resources Hai River, Huai River, and Yellow River basins. managers in China. However, it is located in the extremely arid area with an annual average precipitation of only 35 References mm, and the coefficient of variation for the annual Cai, X. M., Rosegrant, M. W., and Ringler, C. precipitation is as great as 0.6 or 0.7. Water (2003). Physical and economic efficiency of shortage will be unavoidable in the future. water use in the river basin: Implications for efficient water management. Water Resource. 4. Discussion and Conclusion Res., 39(1), WES1:1-12. The amount of available water resources is Chen, X. D., Cao, S. L., Wang, G. A., and Zhang, not enough in China and some basins/regions have M. Q. (1997). The Hydrological Features in already shown water shortage. Precipitation/runoff Yellow River Basin (in Chinese). Yellow exhibits quite variable in both time and space. The River Water Conservancy Press, Zhengzhou, Inland River basin produces only an average China, pp.518. runoff depth of less than 35 mm annually, whereas Chen, Z. K. (2000). The sustainable use of water the Southeast Region produces over 1,000 mm. resources in China. China Water Resources (in The inter-annual variation in precipitation is also Chinese), 16(8), 42-46. significant, with the coefficient of variation as Jiang, W. L. (2000). Security strategy of the water high as 0.7. Within the year, 60% or more of the resources in China for the 21st century. China precipitation in almost all basins arrives in only Water Resources (in Chinese), 16(8), 46-53. three to four months. This variability is one of the Jin, L., and Young, W. (2001). Water use in major challenges faced by water authorities in agriculture in China: Importance, challenges, China. The water supply will likely worsen in the and implications for policy. Water Policy, 3, future with the rapid growth of the population and 215-228. the development of economics. Land-use and Liu, C. M., and He, X. W. (2000). Prediction of climate change also may exacerbate the water water resources for the first half of the 21st scare issue in some basins/regions in China. century in China, China Water Resource, An integrated and comprehensive policy (in Chinese), 16(1), 25-29. for both surface and groundwater exploitation is Ministry of Water Resources (MWR), China needed for the sustainability of the water resource (1999). Supply and Need for Water Resources development. Although there has been in China over 21st Century (in Chinese). China considerable controversy on the development of WaterPower Press, Beijing, China, pp.192. dams for decades, it seems not the case in China. Shiklomanov, I. A. (2000). Appraisal and assessment of world water resources. Water International, 25(1), 11-32. 44 Water Distribution Network Analysis for DMA Design of Ladprao Branch, Bangkok, Thailand A. Pornprommin1, S. Lipiwattanakarn1, S. Chittaladakorn1 'Kasetsart University, Thailand fenpacpdp.ku. ac. th ABSTRACT Due to high leakage in water distribution network of Bangkok area, Metropolitan Waterworks Authority (MWA) of Thailand has launched a project on leakage management by changing the existing pipe network into the new "District Meter Area" (DMA) system. The new system divides the large service area into small sectors of area called DMA by enforcing boundary valves. Thus, leakage can be monitored in each DMA more efficiently. However, changing the system may cause the problem on both pressure and flow in the network. In this study, the water distribution network analysis is performed for the DMA design of Ladprao branch. Ladprao branch is one of 15 branches under MWA and covers an area of 97 km2. The pressure and flow characteristics of the present and new systems are analyzed in the model. It is found that the model can provide valuable information of flow and pressure distribution to assist the engineer for designing the new system. Keywords: Network Analysis, DMA, Bangkok 1. INTRODUCTION In this study, the water distribution network analysis for the DMA design of Ladprao Metropolitan Waterworks Authority branch under MWA is performed. Ladprao (MWA) of Thailand has faced the problem of branch covers an area of 97 km2 and serves water leakage in water distribution network. In around 130,000 customers. The net inflow on some branches, the leakage level is up to almost February 2006 is approximately 305,000 m3/day. 40 percent of the total inflow. High leakage not Figure 1 shows me pipe network of Ladprao only induces an excess production cost to treat branch. There are two main feeders for Ladprao and supply more water, but also causes several branch, Ladprao Pump Station (P.S.) at the problems including reducing water pressure to southwest and Navamin Valve Chamber (V.C.) at customer, damaging infrastructure by creating the center. Ladprao branch is connected with void and allowing contaminant to infíltrate into otijer branches, Bangkhen branch, Sukhumvit the network. As a result, MWA launched the branch, Minburi branch and Phayathai branch at project to improve its water distribution network its north, soutii, east and west, respectively. The by establishing the new strategy to monitor and water exchange between Ladprao branch and control leakage. The new system will break down other adjacent branches is shown in the figure. It die large water distribution network into small is found that Ladprao branch supplies water to sectors of area called "District Meter Area or other branches except Minburi branch. DMA". In each sector (DMA), one or a few pipe inlets will be allowed to supply water into die Flow to area. Other pipes in DMA mat connect with out side will be closed by boundary valves. Then, flow meters will be installed at inlets of DMAs and used for monitoring the amount of water flowing into DMA. The measured inflow will be compared with water billing that shows the amount of water using by the customer in DMA. Thus, the difference between the inflow and die water billing in each DMA can be identified as water loss. If it is found that water loss is high in any DMAs, the comprehensive operation (step testing) to detect and repair leakage will be con ducted in that area. Figure 1 Water distribution network anc balance of Ladprao branch. 45 1 The concept of leakage monitoring is Dala Collection and shown in Figure 2. Flow will be measured at Water Balance Checking various points such as sources or treatment works, supply zones, flow inlets into DMAs. This j— •f measurement will allow the engineer to understand and operate the system in smaller GIS Preparation Model Calibration areas, and allows more precise demand prediction, leakage management and control to take place. (Farley and Trow, 2003) Closing boundary valves V in order to divide the area into small sectors will Data Transfer Network Analysis for change the characteristics of flow and pressure in to Network Model DMA Design the network and may cause negative effects to die system such as low pressure in some area. Thus, determining DMA size, number of inlets and Figure 3 Methodology of network analysis for DMA boundary valves and their positions must be design. carefully designed to prevent any undesired situations. For the primary DMA design, the calculating pipe friction and the maximum number of properties in DMA (e.g. 1,000 - 3,000 acceptable iteration error is set to be 0.005. The properties), the existing natural boundaries and the steps to develop the network model are shown in characteristics of pipe network are taken into Figure 3. As the first step, GIS data including consideration. However, it is necessary to per pipes and customers is collected and checked form die water distribution network analysis in (Figure 4). Then, me GIS data is prepared for order to investigate the primary DMA design transferring to network model. The information in more detail. The network analysis can provide of pipe size and type is necessary, for example, a me information of flow and pressure of die exist 400-mm cast-iron pipe. Customer information is ing system and the effect on the new system when obtained from monthly water billing comprising DMA boundaries are closed. (Farley, 2001) of user demand and user type, for example, a business-type customer using water of 1,500 SubrJstrier meter measure« 3 flow into a «mailer area m /montii. After that, the prepared GIS data is Intajce and e.g. 1000 properties tlB l nl transferred into the network model by using " ™ Otetrtet meter measure« *wte (low into dtetrictii AQUIS sub-model. Figure 5 shows the transfer Bu* meter «.g. 1000-3000 propemw hito eupply zone of points of customers into the nearest nodes on distribution pipes. Also, leakage is distributed in each node. From the record from MWA, Ladprao branch has die leakage level of 24.1 percent of the total inflow. In order to calculate me hourly water balance of Ladprao branch, me measurement of flow at Ladprao P.S., Navamin V.C., and main trunks connected with omer adjacent branches is performed. Figure 6 shows me hourly net inflow Figure 2 Metering hierarchy and DMA design options. of Ladprao branch from Monday to Friday on (Farley and Trow, 2003) March 20-24, 2006. It is found that the flow of each weekday has a similar pattern and the peak 2. Methodology inflow is around 19,000 m3/hr at 8:00. The second The water distribution network model in peak inflow occurs in the evening at 19:00 with this study considers pipe size larger or equal to the approximated flow of 15,000 m3/hr. The next 100-mm diameter due to the limitation of GIS step is to calibrate the model. Data loggers had data and computational time. The AQUIS been installed on about 60 fire hydrants in order to software is used in tiiis study. The software can measure pressure time series for 7 days (Figure analyze water supply networks in both steady- 7a). The measured pressure was compared witii state conditions and dynamic conditions. More the computed pressure from the network model details on me AQUIS model can be found in for calibration. An important parameter used for AQUIS Version 2.0 user's manual (Seven the calibration is the pipe roughness coefficient. Technologies, 2002). In this study, me model is An example of the pressure time series of fire based on the steady-state hydraulic simulation. hydrant FH010 is shown in Figure 7b. The The Colebrook-White equation is employed for full line is the computed pressure after calibration 46 and the dash line is the measured one. The agree ment between them is satisfactory. At this stage, the calibrated model can be used to investigate the present system and it can be developed further for analyzing the new system when the DMA system is established. Figure 5 GIS data transfer to network model. • »• • J« Figure 4 Pipes and customers in GIS 25000.00 20000.00 15000.00 10000.00 —x— Mon —as— Tue 5000.00 —•—Wed —i—Thu —-— Fri —•— Average (m3/hr)] 0.00 oooooooooooooooooooooooo oooooooooooooooooooooooo oo os o —• (N ci 9999999999-r-7VT-7-7-7'7 o— o— o— o— ó— o— ooo oooooooooooooo« -" N N N No o o o o o o ooo ooooooooooooooo Figure 6 Hourly net inflow characteristics of Ladprao branch. 47 LadprowFinalCaliR.mdl Ltgwd — Nodc-FHDIO Ladpruw - Prassun • F HO tO \ ^^v ^ N 0.D0D - £3* 24JXJD nSBO MOM 30 DDO 32JOD MJU) MÜOO WAD «JOD 4ZJXB 44JJ0D 40.000 «.ODD a) b) Figure 7 Concept of model calibration: a) Data logger installed on firehydrant . b) Pressure time series of fire hydrant FH010. (A full line is from the model, and a dash line is from the measurement.) 3. Results and Discussion countries usually having pressure higher than 20 m, it is found that the pressure of the water distri 3.1 Present System bution system in Bangkok is relatively low. In The pressure distribution in Ladprao Ladprao branch, most of the area has the pressure branch at 8:00 (the maximum hourly inflow) is between 6 - 12 m. In some area where pipes are shown in Figure 8. The pressure around the two not well-connected or remote from main trunks, main feeders (Ladprao P.S. and Navamin V.C.) pressure drops and becomes less than 6 m. The is highest. The water is delivered via main result from the network model shows that in the trunks, where pressure is still high, and then present situation there is some low pressure area flows through distribution pipes to the nodes that needs to do some pipe improvement before with customer's demand inside. Comparing me new system will be established. with developed Pressure ¡oiwe} Time: 81-OfcJ VI Figure 8 Pressure distribution in Ladprao branch of the present system at 8:00 where • n • • • ™ red <4 m, LJ yellow4-6m, ™ green 6-8 m, ™ turquoise 8- 12 m, ™ blue > 12 m. 48 3.2 New System water is allowed to flow into the DMA via only After analyzing the present system, the pipe im the designated inlets. Figure 9 shows die results provement in some area is recommended and the of die flow direction and the pressure distribution primary DMA design is revised. The model will of DMA no.12-06-03 at 8:00. This DMA has 2 be used to investigate the effect of the new sys inlets from a trunk main at its west (DM-12-06- tem. Since the model in this study is in the scale 03-01 and DM-12-06-03-02) with the pressure of of one branch, die information of flow and pres 8.22 and 8.19 m, respectively. The flow has to sure change at tiie end points of the main trunks run from tiie west to the south of DMA, and then which are connected Ladprao branch with other it will go inside me DMA. Thus, the critical pres branches is lack. As a result, it is difficult to sure point (CPP) of 4.00 m is found at the north judge how the flow and pressure will be changed east of tiie DMA. With die use of die network at Ladprao branch boundaries. Thus, our crucial model, die effect of die new system in both flow assumption is that the pressure at die boundaries is and pressure can be analyzed. This information assumed to be the same as die previous analysis. will help an engineer to find die appropriate solu Then, boundary valves used for dividing each tion to improve the network. DMA are added and closed in the model, and Figure 9 Flow direction and pressure distribution in DMA no.12-06-03 where the arrow shows the flow direction and the symbols ^^™ and ^^* are an opened inlet point and a closed boundary valve, respectively. Figure 10 shows the pressure distribution of the DMAs are set up is to increase pressure in main new system after pipe improvement. There is no trunks. In contrast, it is found that pressure rises area with pressure under 4 m, but pressure drops in the DMA near the center of die branch where in many areas where it is far from a main trunk. the Navamin V.C. feeder is located. This is However, the pressure drop is found to be in die because die boundary valves are closed and that acceptable range and die policy of the Ladprao DMA does not have to provide water to otiier branch after DMAs. 49 Pressure [mwc| Time: 01-08:1 Vic Figure 10 Pressure distribution in Ladprao branch of the new system at 8:00 where * red < 4 m, *-• yellow 4-6 m, ™ green 6-8 m, * turquoise 8-12 m, ™ blue > 12 m. 4. Conclusion Reference The water distribution network analysis of [1] Farley, M and Trow, S., "Losses in water Ladprao branch under Metropolitan Waterworks distribution networks: a practitioner's guide Authority of Thailand is performed for designing to assessment, monitoring and control", the new "District Meter Area" (DMA) system. IWA Publishing, 2003. The analysis shows that it provides the informa [2] Farley, M., "Leakage management and tion of pressure and flow of the existing system control: A best practice training manual", helping to improve the pipe network before estab WHO, 2001. lishing the new system. Then, the network analy [3] Seven Technologies, "AQUIS version 2.0 user sis is used to investigate the new system. The manual", 2002. assumption of using the same pressure at the main trunks connected with other branches is employed since there is no information at the boundaries. The results of pressure and flow in each DMA are determined and investigated. This information helps an engineer to find the appropriate solution for correcting problems. In conclusion, the net work model shows its valuable capability for the DMA design. 50 Water Rights and water allocation in Andean basins J. Molina1, E. Villarroel2, J. Alurralde2, A. Apaza1, F. Soria1 1 Hydraulics and Hydrology Institute, University Mayor of San Andres, Bolivia 2Sustainable Water, Bolivia imolina ihhdp.acelerate.com Introduction Studied area After a period of water-related conflicts, the The project was carried out on the Taquina Interinstitutional Water Council (CONIAG) was and Khora-Tiquipaya catchments, located in the created in 2002, with representatives from the Tunari mountain range, to the north of the Central Bolivian government and the civil society. The main Valley of Cochabamba, Center Bolivia. The goal of the CONIAG is the construction of the catchments selected belong to the Caine River basin, National Water Policy and Legislation. As a tributary of Grande River (Fig. 1), which feds some contribution to CONIAG's action frame, the of the most ancient and biggest irrigation areas of International Development Research Centre of the country at altitudes between 2500masl and Canada (IDRC) financed the research project 2700 masl, therefore its importance. Figure 2 shows Regulation of Water Rights in Bolivia, which was the irrigation areas and the sources of water of the carried out by the Commission on Integrated Water system Tiqupaya-Colcapirhua. Management in Bolivia (CGIAB), an association of academic, public and non-governmental institutions, Climatic conditions in the interandean region under the administration of the Sustainable Develop are predominantly semiarid, with annual potential ment Consortium in the Andean Ecoregion évapotranspiration rate higher than the precipitation (CONDESAN). rate, mild weather and average temperature of 12°C to 19°C. Such conditions allow agricultural The scope of the project was to investigate activities almost during all year if irrigation is the impact of a variety of allocation models and available. Agricultural land use is predominant, with water rights regulations at a basin level, in order to interferences of urban and industrial use. In the identify and propose water right schemes that could Tunari the weather is subhumid, with altitude in die be incorporated in the legal frame. The research was range of 2800masl and 4300masl, average tempera developed from January 2003 to March 2005, as a ture of 7°C to 15°C with frequent frosts. Within the case study on previously selected catchments using studied system, artificial transboundary channels a river basin management model. Such tool allows transport die water from Sayto Khocha and Chankas dealing with a large quantity of information in order Lagoons subbasins, which belong to the neighbor to assess water management practices and scenarios Misicuni River basin (tributary of Beni River), simulation, considering temporal and spatial directly to die Tiquipaya irrigation system. variability in supply and water demand and competition among users. Moreover, such model Figure 1: Geographic location of Taquina and would allow the identification of conflictive points Tiquipaya basins and critical situations, and therefore the formulation B50000 of "transparent" water allocation schemes. Finally, the model would be beneficial to future river basin authorities to develop water resources management plans, as well as a supporting tool for institutions and social organizations working on water management. The MIKE BASIN model (DHI, 2002) was chosen. 51 Figure 2: Irrigation area and catchments of the system Tiquipaya-Colcapirhua Water users CODE Montecillo Moa ' Tolavi Toi Villa VE Esperanza Tumvm Tíq Santiaguilla San 1 Rumi RM Maya | Putucu P 1 Capa CK Khachi [ Misicalle Mis | Linde Lia Collpapampa CoN [ Chilimarca CM 1 Canarancho Ca [ Sirpita Sir 1 Chiquicollo Ch | Molinos Mol I Cuatro Esquinas CE 1 Bruno BM Mofeo Cofia CC [ Cofia Favorable weather and living conditions Water allocations and its representation increase settlements in the region, hence growing water usage competition and conflicts among Water management at a basin level can be irrigation and urban and industrial users limit the conceived as an intention to identify the access to water sources, increasing system's best possible utilization of the available water operative problems. As a consequence, dynamic resources given certain social, legal, technical and systems such as the Tiquipaya-Colcapirhua have environmental constraints. The basin is commonly developed, with strong organizations (Tiquipaya and used as the management unit, due to its properties to Colcapirhua Irrigation Systems Asociations receive and concentrate meteoric water, and also due ASIRITIC) and experienced water users. to the strong relationship and interdependence Furthermore, the system becomes one of the most among water utilization and users and between users complex systems in the country in relation to its and their environment. modest size (1625 hectares), using water from three regulated water sources (reservoirs), two The research project uses a water rights non-regulated (natural streams) and one regulated approach for the analysis, for both current and future source external to the system. Additional sources conditions that would be me result of the application such as groundwater (although its importance for of proposed water legislation. Due to the necessity irrigation has decreased) are used. Due to the to represent current and hypothetic situations and reasons established above, the Tiquipaya- due to the large quantity of information expected, Colcapirhua system could be seen as an interesting mathematical models were applied. Water allocation research spot for water management in Andean schemes within a water system are studied through basins. the application of the model considering a variety of conceptual frames, incorporating socio-human and physic-biologic components in different levels. 52 Figure 3 shows a simplified representation of Figure 4. Representation of a water system in a water management model. Input data comprises MIKE BASIN. time series data of catchment runoff and demands on water, allocation and system operation rules. The .Dlgitlied river network main model output data is the amount of water allocated to each user as a function of time. Major variables include lower order variables or subthemes, e.g. water supply might include contributions as precipitation (climate), surface and groundwater flow, transferred water, water quality and existing or projected infrastructure that would modify the temporal and/or spatial distribution of the supply of water. Water demand is expressed for different uses (domestic, industrial, irrigation, MIKE BASIN is chosen after upon economic and energy, etc), including ecosystems. Allocation rules technical criteria analysis. The model is a GIS are based on water rights originated from a legal based application that represents the water system frame or traditional uses. Water allocation is also as a river network, where branches represent river influenced by its economical value, environmental streams and nodes represent confluences, control costs associated with its utilization, investment points and locations where water activities occur recovery and operation rules within the water (e.g. extractions, derivations, return flows, etc) or system (i.e. time and duration for reservoir's points where simulation results are required. releases). Even though the main output is the amount of water that every user receives (with its Water supply respective temporal variation), additional results can be obtained through the application of a model, such Surface water supply estimation providing as user's level of satisfaction, system's water deficit the systems of the sub-basins Taquina and and losses, reservoir's operation rules, etc. Tiquipaya is made through a hydrological study based on existing information by the application Figure 3. Inputs and outputs of a water management of rainfall-runoff models (Molina and Soria, model 2004). Water availability is determined in the system as continuous series of monthly discharge for the period 1972 to 2002 (Table 1). Although , V contributed volume from La Angostura reservoir Allocation rules (water rights) is not included since it does not belong to the system, estimated values are obtained through field data and the river basin management model. 1 1 Water ' \> I Management |<-^ ' Water supply I 1^ I model |^J | demand r, Water allocation 53 Table 1: Average monthly runoff of the system sources, in thousands m3 Tiquipaya Water Taquina Taquina Saytu Lagun Khora- Source Chankas Chutakahua Lagoon River Khocha Mayu Machumita Jan 252 160 429 962 217 251 858 Feb 345 222 554 1331 223 367 1158 Mar 406 269 587 1363 238 360 1081 Apr 238 158 314 669 75 192 512 May 192 127 238 469 27 136 312 Jun 121 73 192 368 13 112 228 Jul 78 42 163 308 8.0 93 182 Aug 53 27 137 238 5.4 76 145 Sep 37 20 106 176 5.2 63 117 Oct 34 19 107 147 8.2 60 107 Nov 45 27 109 140 21 64 132 Dec 105 66 201 284 83 116 316 Total 1922 1219 3154 6560 946 1855 5050 Table 2: Average potential irrigation demand in the system Tiquipaya-Colcapirhua, in thousands m3 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Mon 45.8 34 38 34.6 13.9 9.2 16.7 27.4 71.5 133 178 103 Toi 13.1 9.7 17.2 23.7 19.9 7.7 11.3 13.7 24.8 31.9 41.6 24.4 VE 17.2 8.8 11.3 13.3 17.4 12.8 19.2 26 46.1 61 73.4 42.7 Tiq 5.3 1.5 0.3 0.3 0.3 0.2 1.6 0.4 3.4 10.5 17.2 12.2 San 31.3 15.3 18.7 23.1 25.7 17 27.2 33.6 63.8 92.2 119 72.3 RM 12.4 5.9 6.7 8.2 13.2 8.8 13.8 18.9 34.3 48.3 57.7 34.1 P 23.4 13.9 17.2 18.1 13 9.8 15.5 20 41.9 63.9 84.7 50.3 CK 199 110 155 186 166 122 181 225 406 525 680 401 Mis 19.6 12.1 19 25.1 26.3 14.1 20.8 27.2 48.4 62.9 76.8 44.3 Lin 8.2 5.4 7.1 8.5 12.6 8.5 12.2 19.7 32.6 40.8 44 24.5 CoN 97.8 61.6 84.6 104 80.3 44.4 69.6 93.4 175 250 328 199 Chil 6.9 4.6 5.5 5.3 2.6 1.1 2.4 2.3 8.9 18.8 26.8 15.6 Ca 16.8 7.3 6.9 8.3 9.6 6.6 11.8 14.2 28.9 47 62.9 39.9 Sir 50.8 28.6 38.5 45.4 44.9 33.1 49.4 67 118 153 190 111 Ch 64.4 40.4 64.6 79.3 74.1 50.1 70.4 92.4 159 189 231 129 Mol 34.8 34.1 36.8 34.7 2.9 2.1 4.7 31.9 56.4 91.9 112 70.1 CE 37 18.7 22.9 27.4 23.3 16.8 27.2 32.5 62.7 91.1 124 76.6 BM 19.5 12 17.6 22.4 23.4 13.7 20.3 28.1 49.2 64.1 77.4 45 ce 35.3 18.9 25.1 29.6 25.5 19.3 29.3 36.4 66.9 90 119 71.3 Total 739 443 593 697 595 397 604 810 1499 2064 2642 1567 54 Table 3: Demand of main drinking systems Water rights Figure 5 shows a scheme of the types, Average subjects and ways to express water rights for the Equivalent daily water Total Drinking discharge subsystems Machumita and Lagun Mayu. Users usage demand system 3 rate (liter/ per user (m / (mVmonth) Similarities are observed as in the way to express sec) month) water rights as a function of time units (hours) for Tiquipaya 1500 68 102000 39.4 individual and combined rights. Even tíiough collective water rights would be applicable to the Montecillo 280 30 8400 3.2 source, existing differences associated to the type of source appear: collective rights are expressed in terms of total volume for reservoirs and in terms of Irrigation water demand was calculated proportions of available water. based on crop surveying carried out by the research team for the agricultural campaign 2003-2004. The Figure 5b. Type, subject and irrigation water results are presented as monthly series. Table 2 rights in Tiquipaya. Lagun Mayu irrigation system shows mean values per community. Drinking water demand in die urban area of Tiquipaya and Monte cillo was estimated based on survey data from the WATER RIGHTS SCHEME: Municipallity of Tiquipaya and campaigns carried LAGUN MAYU IRRIGATION SYSTEM on by the project's work team with die help of personnel of the Centro Aguas institute. Right type Right's subject Right's expression Figure 5a. Type, subject and irrigation water Total Volumen rights in Tiquipaya. Machu Mita irrigation system COLECTIVE of Lagun Mayu WATER RIGHTS SCHEME: expression Fraction of COMBINED: the time -Individual or Family duration of 5/6 of the total -User's group the reservoir discharge of release Qhora River A fraction of INDIVIDUAL the irrigation OR FAMILY A fraction of time allocated the total time to the group COMBINED: (hours, of users -Individual or minutes) of a family turn (cycle) -User's association Scenarios and results A fraction of System schematization should represent the INDIVIDUAL the irrigation activities at the level of detail mat is required to OR FAMILY time allocated to fulfill project goals. The local community was the group of defined as the irrigation user. Irrigation communities users and drinking water systems are listed respectively on table 2 and 3. Results are presented for two 55 scenarios. The first is a representation of the current of water amount provided by the source have strong situation of the Tiquipaya-Colcapirhua system, with influence, but at the seasonal level variations in supply, demand and water rights described above. demand and allocation rules can be determinant. For Simulation requires additional data such as: many users, the seasonal variation is the most a) topographic characteristics of reservoirs; important. b) system losses, implying deficiencies in reservoir takes due to filtrations through dams; c) operation Water deficit is a useful parameter to rules, including duration, dates and volume of evaluate if water distribution in a system is equitable reservoir releases. and satisfies requirements of all users or, on the contrary, a high concentration of water rights on few The second scenario considers the same data users exists. It also allows the identification of users for water supply and demand as for die first or areas tiiat would be beneficiated in future projects scenario, as well as the same operation rules and mat would exploit new sources. loss types. The main difference is the way water rights are expressed and water is allocated. It was For the second scenario and with the intended to represent the content of the preliminary exception of Montecillo community, locates projects of the Bolivian Water Law elaborated in the upstream, average deficit increases in all period 1997-1999, including the last version, which communities with respect to the first scenario. It is in the most relevant aspects says: "The access to evident a strong decrease in die number of different uses and water resources exploitation communities with moderate deficit (12 to 25%). On will be carried on through given concessions to the other hand, a strong increase in die number of individuals or public collective or private persons by communities witii high deficit (35 to 45%) is means of administrative resolution declared by the observed. For the Tiquipaya system, the application Water Regulation Body (Superintendence), which of the proposed water law would increase the for effects of the current law, will be denominated differences among communities respect to water Water Title, conferring in a clear and stable way the deficit, and therefore, in the level of satisfaction of right of use, joy and aptitude for specific and/or water users. As a consequence, the scenario tends to multiple use of die resource". "Concessions will be increase inequality, competition and conflicts for established in units of volume per unit of time in water usage. agreement to regulations". The model gives results on other variables Allocation per volume requires the definition that allow assessment of the system's behavior. For of a priority order on every source, which is example, tables 3 and 4 show that only 23% (0.28 essential during dry or under scarcity conditions. A million m3) of total available water in Chankas "head-waters" priority criterion was used: users subsystem (1.22 million m3) reaches the irrigation located upstream receive water first, according to zone. A significative part (0.126 million m3) are lost their rights. Downstream users receive water after due to infiltration in the reservoir and losses along upstream users have fulfilled their demands or the diversion channel, which includes water taken rights. Any remaining water continues to flow by communities situated along the channel mat do downstream to be used by the next water user. not have proper water rights. Another amount (0.157 million m3) escapes continuously through the Figure 6 shows the relative average annual reservoir intake and is transported by the diversion deficit per community as percentage for botii channel to the reservoir Sinergia and SEMAPA scenarios, according to a colour gradation. Relative companies, that use tiiat water for hydropower water deficit is defined as potential demand minus generation and urban water supply. Sinergia/ water actually consumed divided by potential SEMAPA also benefit from the water that spills demand. MIKE BASIN calculates relative deficit for from the dam (0.631 million m3) during the rainy every time step and for die whole time series. season. Therefore both companies receive 0.78 Deficit values change witii variations in time of million m3 from Chankas, which turn them into the supply and demand, and also with allocation and main users of the system, although they have no operation rules. On an inter-annual level, variations rights at all on Chankas subsystem. Figure 6. Relative average deficit in percent per community CURRENT SCENARIO SCENARIO FOR THE PROPOSED WATER LAW ngrwnormaul aaücitdeficit. pepecr year 25-38 36- There are other scenarios born from the Table 3: Gross supply, used water, and loss per analysis of the current system operation. For source (thousands m3) example, losses and technical deficiencies may give origin to scenarios where those problems are solved SOURCE Used Grou lupply Loss (*) or at least reduced. Of the various scenarios which water Machu Mita have practical meaning we mention: a) Increase of 6,559.5 4,170.9 0.0 water availability through the new Batea Laguna Chutakahua 946.1 123.9 0.0 project; b) Control of reservoirs losses; c) Demand Lagun Mayu management measures; d) Water system 3,153.6 800.0 252.3 optimization. They also show the potential use of Chankas 1,220.4 280.3 126.1 the model to assess and eventually improve water Saytu Khocha management in die Tiquipaya system. 1,920.5 795.5 378.4 Angostura 1,222.2 1,222.2 0.0 Total without Angostura 13,800.2 6,170.5 756.9 Total with Angostura 15,022.4 7,392.8 756.9 (*) Infiltration losses and channel seepage until the head of the subsystems. Do not include losses through intakes or reservoir spills 57 Table 4: Balance of inputs and outputs of Chankas From die technical point of view, it has been subsystem showed that the methodologies and tools used to develop the current study contributed to make INPUTS transparent the operation of the water system, to Gross supply (basin contribution) 1220.4 identify water management problems and therefore the solutions. The model allowed a qualitative and OUTPUTS quantitative identification of water users with Water used by Montecillo 50.3 insufficient water rights to satisfy their water Water used by Misicalle 17.7 requirements, over whom new projects that increase Water used by Cuatro Esquinas 18.7 water availability have to be focused. The consequences of system operation problems, such as Water used by Sirpita 106.1 losses in intakes and channel and dam seepage, Water used by Capa Khachi 87.5 were also assessed, showing at the same time TOTAL WATER USAGE 280.3 that in several cases they translate into effective Infiltration and conduction loss 126.1 transference of water rights from one subsystem to another. When die analysis of scenarios is carried Evaporation and others 25.6 out wimin die model, it is possible to asses various Escape through the dam intake 157.7 alternatives to solve water related problems, which Spillway 630.7 in combination with socio-economical analysis, Water available for Sinergia/SEMAPA 788.4 would allow their comparison and ranking. As a whole, the study represents a significant Conclusions contribution to water planning and integral water management in me Tiquipaya system and basin. From the first steps of the construction of the water and irrigation legal frames, the grant of References water rights became one of the most complex tiiemes. A majority of the irrigation organizations DHI Water & Environment, 2002. MIKE BASIN, in the country are aware of the difficulties that Guide and user's tutorial. Denmark, 27 p. could arise from the preliminary project of the Molina, J. and Soria, F., 2004. Hydrological study: Bolivian Water Law. Results from the present water availability, regulation project. La Paz, study show some of the consequences that would January 2004, 74 p come from the application of the mentioned law project and also a visualization of the diversity and complexity in the allocation of water rights. As a consequence, it was decided upon maintaining such diversity and habits in the irrigation normative. 58 Membrane Filtration Removing Salts and Arsenic in Drinking Water for Rural Areas Pikul Wanichapichart", Wiriya Duangsuwanb, Darunee Bhongsuwan", Pusadee Mohamad", and Porntip Sridang' aFacuity of Science, bFacuity of Agroindustry, and "Faculty of Engineering Prince of Songkla University, Hat Yai, Songkhla, Thailand 90112. ABSTRACT A mobile membrane filtration unit was constructed to filter brackish water (3,300 ppm) and arsenic contamination pipeline water in Satingpra and Ronpiboon District, respectively. For arsenic filtration, it was installed at the community center of Moo 2, Ronpiboon Sub-district for 3 months. It appeared that both salinity and arsenic level varied due to seasonal changes. In hot season, both levels were much higher than the acceptable level for drinking water. Arsenic content in water varied from 0.10 ppm in December to 0.16 ppm at the end of March and permeate from the membrane unit was arsenic free (<0.01 ppm). The rate of water production at the beginning was 98 L/hr for brackish water and 208 L/hr for arsenic contaminated water, and was decreased to some extent due to fouling and the quality of feed water. However, the rate was resumed after washing the membrane for about 2hrs. The cost of drinking water estimated on a 3 month basis was 0.10 Baht/L. An ultra-filtration unit was also constructed to filter surface water in Kuraburi District, Pang-Nga Province. The unit produced 2,000 L/hr drinking water at a much lower costs. Keywords: Membrane technology, Drinking water, Brackish water, Arsenic, Surface water, Underground water, Satingpra District, Ronpiboon District Introduction During the pass decades, me cost of filtration As being realized, industrial growth is an membranes has been reduced. This offers an indicator for economic trend of the country and it opportunity to introduce membrane technology to serves needs resulting from an increasing world rural areas of Thailand, particularly where the population. On the other hand, waste treatment pipelines are lacking and chemical contamination in increasingly plays important roll to maintain healthy natural water is severe. Normally, clean water for societies with rich natural resources, as it used to be daily consumption of villagers is from preserved in the past. Problems arise whenever these two are rain. In the case of drought, underground and well un-balanced, hence new technologies emerge, and water is a major source of water supply. After the one of these is membrane technology. It has been living style has been altered and diverged more to commercialized since 1960, using a pressure driven serve industrial needs, man-made contamination of process to separate particles, 0.01-0.001 nm, from a natural water resources is in progress at a much feed solution by filtering through membrane pores. faster rate compared to methods improving water It should be noted that me pressure used ranges from quality and, in return, resulting in environmental 100 kPa to 2-3 MPa, for micro-filtration (MF), harm to health. Moreover, it appears that geological ultra-filtration (UF), nano-filtration (NF) and reverse built also leads to contamination of natural water. osmosis (RO), which is selected according to Examples are brackish water found in the particle size. Ranges of membrane pore size and underground or shallow wells nearby coastal line the related pressure required for filtration is shown and arsenic found in die surface and the in Figure 1. Besides, the driving force can be underground water reserves in Ronpiboon District at temperature, concentration and electrochemical high levels after tin mining was deserted. These gradient, depending on processes and separation have been evidenced in Satingpra District, purposes. Among mese, the feed solution can be oil, Songkhla Province and in Ronpiboon District, alcohol, water, or gas and the membrane used must Nakornsrithammarat Province, respectively [3,4]. be appropriate not only to particle size but also to These affect not only agricultural products, but also the type of feed solutions. Details for several me need for adequate clean water for daily processes in membrane technology can be studied consumption. However, arsenic (As) poisoning in from [1,2]. drinking water seems to be a more severe problem 59 as far as health is concerned. In Ronpiboon, soon as die arsenicosis was revealed and also arsenicosis was firstly reported in 1987, then As provided die communities with underground water contamination in soil, vegetation, meat, stream supplies. Later, it appeared that As was also found at sediments, surface water and underground water high level [4]. In addition, As removal for portable was followed during 1990-1999. More details are drinking water was demonstrated by adding some provided by a summarized work by Kantachote et chemicals to enhance As precipitation and then al.[5]. Government projects have been launched as draining with activated carbon [6]. Pressura Bar pore sizG %i m Reverse Osmosis (RO) 30-60 10-4 - 10 ; Nano Filtration (NF) 20-40 lO'J - 10» Ultra Filtration (UF) 1-10 10-? - 10-! Micro Filtration (MF) <1 IO ' - 10 ' Reténtate • Bacteria, tat (concentrate) • Proteins „„^JP'W 1 t\ «Kfcfer • Lactose F*«<*J • Minerals (salts) I ^ • Water Permeate (fíltrate) Figure 1 Pressure levels (left arrow) in bar and corresponding membrane pore size (right arrow) for MF, UF, NF and RO membranes . Note that 1 bar = 100 kPa. This work presents another trial for water using membrane technology. producing drinking water using membrane technology without adding extra chemicals into raw Materials and Methods water. Two filtration units were constructed; one Figure 2 shows a NF unit was constructed with NF and the other with UF membranes. The which consists of one micro-filtration (MF) of 5.0 process is totally pressure driven with a small-scale mm (USFilter) average pore size and one 4x40 inch mobile unit. In fact, sparse distribution of houses in spiral wound NF membranes, of polyamide thin film villages makes it inevitably expensive to install the composite (Dow Chemical) of 5.0 nm pore size. pipelines. Therefore, it seems to leave some extra The dimension of this unit was 0.8 m xl.O m xl.3 space for the mobile unit to perform well. It is m. and it was supported with wheels so as to move small enough to move around and meet village around easily. Two pressure gauges are required so requirement, and it is simple enough to handle with that trans-membrane pressure drop can be observed, some training. Work is extended to Tsunami victim indicating of membrane fouling. The pump is of 2.2 in Pang-Nga Province with careful chosen kW pressuring feed water to pass through a NF membrane type. Primary concern of this project is to membrane and also helps sucking feed water to the educate the villagers on producing clean drinking MF membrane. 60 Figure 2 Diagram represents pro duction process for drink ing water, using MF and NF membranes. For ultra-filtration shown in Figure 3, the natural brackish water collected from Satingpra pump was only 750W and the NF membrane was Hospital, Songkhla Province. Rejection of salt is replaced with a UF membrane (China) comprising of 12,000 hollow fibres made of cellulose acetate ( C ^ and being encased in a housing unit. An additional %R = xlOO CB MF membrane of 1.0 mm was inserted after the 5.0 determined by V , where CP and mm one to prolong working period of the UF CF is salt concentration in permeate and feed, membrane before being fouled. Particles of 0.01- respectively. Salt concentration was measured in 0.05 mm were retained on the UF membrane terms of solution conductivity. In case of As surface, including bacteria. The diameter of these contamination, water samples were collected weekly capillaries ranged from 0.5-1.4 mm. from Moo 2 Ronpiboon District, Nakornsritham- To reduce salinity, a comparison study was marat Province and the amount was determined made between 4,000 ppm NaCl feed solution and using ICP-OES and chemical analysis. OUT ~L-_ rtni f^ IN *Er 5ji l|i UF Figure 3 An Ultra-filtration unit with MF membranes of 5 urn and 1 urn for filtering larger particles. A UV light unit ensures bacterial free. 61 Results and Discussion Ministry of Public Health. It should be noted that On salinity problem most shallow reservoirs are iron free. Therefore, Table 1 shows the data collection at Klongri the filter unit can be directly applied to water in Sub-District, where brackish water is the major these areas. Since salinity is not the major problem, problem. Samples were taken from the outlets at it is possible to minimize the operation cost by using houses in the areas. It appears that the level of salt in ultra-filtration membrane instead, and the pressure Nov.-Dec. (rainy season) is lower that in Mar.- April driving the system can be reduced. (hot season). Table 2 shows that rate of production An example of ultra-filtration unit was the set (permeate volume) is increased with the increased up near a pond at Kuraburi District in Pang-Nga pressure, resulting in greater % salt rejection. High Province (Figure 3). The unit was simple and much rejection appearing in brackish water could be less expensive. In fact, the feed was rain water explained that the sample contained larger oozing out from the under ground and the reservoir suspension particles; such as divalent and trivalent was large enough to serve a small community. ionic salts, compared to the made-up NaCl solution. The feed tank filled with pond water of drinking This was found to be the case, since chloride, sulfate water quality and the production rate was about and manganese level in feed water was high (see 2,000 L/hr. This unit can be connected directly to Table 3) and this explained why slightly lower tap water, of which its quality meets the requirement production rate from brackish water was obtained. to be filtered by the unit. This system is mainly for Besides the salinity, the quality of pipeline water water free from toxic substances and it only removes produced from shallow ponds was also studied and bacteria and viruses from feed water. It should be reported in Table 4. It shows mat some samples noted that the additional UV light may not be do not meet the standardization required from the necessary. Table 1 Salinity (ppm) found in pipeline water at Klongri Sub-district, Satingpra District, Songkhla Province. Place Year 1999 Year 2000 Oct. Nov. Dec. Jan. Feb. Mar. April May Jun. Moo 5 3,000 2,965 1,868 1,920 2,550 3,050 3,350 - 1,540 Moo 8 1,520 1,435 1,446 1,520 1,650 1,650 1,680 - 1,360 *WHO salinity standard < 500 ppm Table 2 Comparing permeate volume between NaCl solution and underground brackish water at Klongri Moo 8, Satingpra District. Feed Feed Pressure Permeate volume % Salt rejection (MPa) (L/h) NaCl solution 0.46 10.Ü0.2 63.6±0.1 (4,000 ppm) 0.77 38.1±0.4 60.5±0.8 2.25 92.7±0.4 77.5Ü.8 Brackish underground water 0.46 7.6±0.3 75.0±7.6 0.77 37.2±0.4 69.7±5.1 (3,300 ppm) 2.25 87.6±0.8 93.0±6.7 62 Table 3 Water quality before and after being filtered under different pressures. Elements Feed Permeate Standard* 0.29 MPa 0.48 Mpa 1.26 Mpa Copper (ppm) 0.01 0.01 0.00 0.00 1.0 Manganese (ppm) 0.93 0.02 0.01 0.01 0.3 Chloride (ppm) 2,060 410 215 152 250 Sulfate (ppm) 870 26 26 17 250 TDS (ppm) 4,300 1,400 1,200 500 500 pH 8.31 7.01 7.23 6.93 6.5-8.5 Ministry of Public Health, 1991 Table 4 The quality of pipelines water providing for communities in Satingpra District, Songkhla Province. Data Color Turbidity Hardness Iron Area» (Ft-Co) fflTW (mm (msm Standard levels* 5-15 5-20 100 0.5-1.0 1. Choompon Moo 4 414.38 36.64 62.90 2.53 2. Klongri Moo 8 17.22 3.37 243.67 0.04 3. Kookude Moo 4 322.78 24.84 200.92 0.05 4. Kookude Moo 6 66.15 7.64 30.59 0.02 5. Wat Kookude 0 0.94 371.27 0.01 6. Municipality 80.72 11.77 66.50 O.ll 7. Tahhin Moo 4 151.78 13.09 32.75 0.06 8. Tahhin Moo 5 103.54 6.53 303.35 0.05 9. Tahhin Moo 6 34.43 2J9 142.43 0.04 •For bottled water regulated by Ministry of Public Health, #61, 1991 On arsenic problem basis. Figure 4 shows a decline in production rate, indicating that it will half of its original rate after 72 Since pipeline water was drawn directly from weeks, if no cleaning is made during the operating a mountain reservoir, its quality was examined and period. At the end of a 10-week period, the rate shown in Table 5 . There are hardly any toxic reduced to 198 L/h, and after washing the membrane elements except arsenic, particularly in dry season for 2hrs in sodium hydroxide solution, the original when it is concentrated. The level of As in feed rate was resumed. water varied from less than 0.01 ppm to 0.16 ppm, Cost Estimation during these three months. After being filtered, This study estimated the cost of drinking permeate was found to be arsenic free (<0.01 ppm) water in Ronpiboon District to be about 0.10 Baht/L, in all cases. Due to the clearness of rain water, the regardless of bottling and services. However, it is unit produced drinking water at 208 L/h, at a much reduced about 10 times if a UF membrane is used higher rate compared to the case produced from and raw water fed to the system is from surface brackish water. For operating benefit, the amount of water of smaller trace of impurities. water produced per hour was recorded per week 63 Conclusion It should be noted that the majority of local people Ideally, a mobile unit is suitable for small keep their way of using reserved rain water. The unit villages since it can be moved around to serve com supports self-reliance; i.e. adequate drinking water munity need. The cost of drinking water produced in their own village before moving it around for by NF unit is 0.10 Baht/L in Ronpiboon area and it serving nearby communities. For the UF unit, a can be reduced if the quality of raw water is community center is an ideal place so that everyone improved. One unit can serve several villages can reach the supply. Besides, the communities when As becomes high level in hot season and learned to use simple test kits examining the quality this coincides with the lack of reserved rain water. of drinking water. Table 5 The quality of water quality before and after being filtered through NF membrane. Data were collected from Dec. 2004 to Mar. 2005. Suspending ele Pipeline water Drinking water Standard ments Dec. Jan. Feb. Mar Dec. Jan. Feb. Mar pH - 4.6 5.2 4.8 - 4.7 5.8 5.1 6.5-8.5 ±0.8 ±0.2 ±1.2 TDS (ppm) - 15 - - - 3.0 - - 500 Turbidity (NTU) - 0 1.5 0.9 - 0 1.6 0 20FAU ±0.4 ±0.1 Color - 0 2.2 0.8 - 0 0 0 20 Hz ±1.2 Hardness 5.34 5.34 8.5 6.0 0.87 <1.0 0 0 100 (ppm, CaC03) ±4.3 ±3.5 Iron (ppm) 0.15 0.10 0.03 0.08 0.02 <0.01 0 - 0.3 ±0.01 ±0.06 Chloride (ppm) 15 <5.0 - - <10 <5.0 - - 250 Sulphate (ppm) <25 <25 - - <25 <25 - - 250 Nitrate <0.01 1.0 - - 0.2 0.6 - - 4.0 (ppm, Nitrogen) Total arsenic (ppm) <0.01 <0.01 0.11 0.12 <0.01 <0.01 <0.01 <0.01 0.01 ±0.01 ±0.03 220 - 210 - 200 - -^ 190 - S 180 y = -1.4868x +211.38 170 160 150 -i 1 r- -1 4 6 8 10 12 No. of week Figure 4 A linear decline of production rate with time of 5 day work/week. 64 Acknowledgement [4] Suwanmanee A., Arrykul S., and Kooptanon K. This project was funded by three 1990. Arsenic in soils and stream sedi sources; Prince of Songkla University on mak ments: Ronphibul Distriact, Nakorn Si th ing the mobile filtration unit, the Health Thammarat Province. In the 16 Confer System Institute of Southern Region Thailand ence on Science and Technology, Thailand, on the arsenic removal in pipeline water in D05: 147-148. Ronpiboon District, Nakomsritham- [5] Kantachote D., Areekul S., Chongsuvivatwong, marat Province, and World Vision Organization V., Bunnual P., and Naidu R., 2006. on UF system in Kuraburi District, Pang-Nga "Extent and severity of arsenic poisoning Province. in Thailand" In Managing arsenic in the environment: From soil to human health. References Ed. Naidu R. et al. CSIRO Publishing, [1] Baker RW. Membrane Technology and Australia 2006: 656 pages. Applications. McGraw-Hill, 2000: 514 [6] Areekul S., Maharachapong N., Kooptanond K. pages. (1997) Arsenic removal from portable [2] Sorensen TS. Surface Chemistry and Electro water. In Proceedings of Toxic Metal of chemistry of Membranes. Marcel Dekker, Pak Panang and Pattani River Basins, Inc., 1999: 1015 pages. Research and Development Office, Prince [3] Wanichapichart P., and Duangsuwan W. (2003) of Songkla University, Oct. 17: 35-45. Development of prototype for brackish water filtration using membranes, Research and Development Office, Prince of Songkla University. 65 Basin water use accounting method with application to the Mekong Basin Mac Kirby, Mohammed Mainuddin, Geoff Podger and Lu Zhang CSIRO Land and Water, Canberra, Australia ABSTRACT The Challenge Program on Water and Food undertakes research to maximise water productivity in several of the world's major river basins including the Mekong. The research must be underpinned by information on how much water there is in a basin, where it goes and how it is used. There should, furthermore, be an understanding of future constraints (such as the impact of climate change), opportunities (such as increased diversions for irrigation) and trade-offs (such as changed land use improving dryland productivity but leaving less water for downstream use). We describe a water use account for the Mekong that provides monthly estimates of major water uses. We have used it for historical estimates, but in principle it can also be used for prediction. Starting with rainfall, the account tracks the partitioning of water into runoff, and évapotranspiration by dryland vegetation. The runoff is tracked as it becomes flow down the rivers, with losses (such as evaporation and seepage) and gains (such as tributary inflows), storages in lakes and reservoirs, diversion for irrigation or other purposes, floods in lowland floodplains and finally discharges to the sea. The account estimates die water use by the major irrigation industries and other uses. The account helps develop understanding of die water uses in the Mekong Basin, and the likely consequences of large changes, such as climate change, land use change, increased diversions and irri gation water use, and changed storages. The water use account is developed as an Excel spreadsheet. It is a tool for integrated water resources management, and provides a sound basis for integrating hydrology, environment, social and economic issues and policy and institutional issues in a river basin. Introduction growth;. the contribution of economic sectors to environmental problems; the implications of The international Challenge Program on environmental policy measures (such as regulation, Water and Food aims to explore threats, charges and incentives); to identify the status of opportunities and trade-offs in water access and water resources and the consequences of impact on agricultural productivity and hence management actions; and, identifying the scope for poverty / livelihoods and die environment for savings and improvements in productivity. several major river basins around the world. The However, those accounts are static, providing a Mekong Basin is one of its focal river basins. The snapshot for a single year or an average year. program must be founded on sound information Furthermore, they do not link water movement to about how much water there is, when and where it use. is available, and how it is used. It requires the means In this paper, we describe a water use to explore trade-offs amongst uses, opportunities accounting metiiod. In contrast to me static national such as increased irrigation, and threats to the and basin water use accounts referred to above, our water resource such as land use change and climate accounts are dynamic, with a monthly timestep, and change. Furthermore, it requires the means to assess thus account for seasonal and annual variability. die interactions between water and food, poverty They can also examine dynamic effects such as and the environment. These needs are addressed by climate change, land use change, changes to dam water use accounting (Molden, 1997; Molden et al., operation, etc. The accounts are assembled in Excel, 2001a). and are quick and easy to develop, modify and run. Water use accounting is used at national We have applied this accounting method to several (ABS, 2004; Lenzen, 2004) and basin (Molden, major river basins including the Murray-Darling and 1997; Molden et al., 2001a) scales to allow the Limpopo (Kirby et al., 2006). Here we describe assessment of the consequences of economic the application to the Mekong. 67 We emphasise that the account is a high level, Outline of water use in the Mekong Basin whole-of-basin account: necessarily, it averages or The water use and hydrology of the Mekong glosses over much detail. It is not a hydrology Basin are described in MRC (2003, 2005). The model for river planning management, nor is it a Mekong Basin (Figure 1) covers 795,000 km2, and detailed account such as might be used, for example, is drained by the 4200 km long River Mekong. The to determine small zones of high seepage loss from basin is mostly long and thin, particularly in the a river channel (e.g. Gippel, 2006). upper, Chinese part, and the Mekong is fed mostly In contrast to water use accounting, by many short tributaries draining small catchments. hydrologie modelling is generally developed with a The largest catchments are the Mun-Chi (about narrower focus of river or catchment planning and 100,000 km2), the Se San (80,000 km2) and the management. The hydrologie models of the Mekong Tonle Sap (85,000 km2). Basin (such as those of Kite, 2001, and Podger et al., 2004) generally do not deal with all water uses, are often too complex (in spatial, temporal and/or process resolution) to use for analysis of broad scale trade-offs, and several deal only with part of the basin (such as those of Herath and Yang, 2000; Takeuchi et al., 2000; Fujii et al., 2003; Kummu et al., 2005). Nevertheless, the hydrology of the basin is an important element of the water use account. The Decision Support System (MRC, 2004) (see accom panying Box) is the most comprehensive and care fully calibrated hydrology model of the Mekong Basin, and was accompanied by much gap filling and error checking of discharge records and other information. We have used it to develop and test the hydrology part of the water use account. The Mekong Decision Support System The Decision Support System is based on SWAT (Neitsch et. al., 2002), IQQM (Simons, et. al. 1996) and ISIS. SWAT simulates catchment runoff based on estimates of daily rainfall, potential évapotranspiration (PET), the topography, soils and land cover. IQQM then routes these flows through the river system, making allowance for diversions for irrigation and other consumptive demands, and for control Figure 1: The Mekong River Basin, showing the structures such as dams. The ISIS hydrodynamic major drainage network, the riparian countries, and model represents the complex interactions caused three locations (Luang Prabang, Tonle Sap and by tidal influences, flow reversal in the Tonal Sap Phnom Penh) at which observed and calculated River and over-bank flow in me flood season flows are compared in the paper. with the varying inflows from the IQQM model at Kratie. The Decision Support Framework has been successfully used as the planning and The source of the Mekong is fed by trans-boundary analytical tool to assess various snowmelt. The Lower Mekong is fed by runoff, scenarios by the MRC (Podger et. al., 2004; characterised by a pronounced wet and dry season. Jirayoot and Trung, 2005). It is also the only The peak flow from the Upper Mekong more or less modelling package that has been accepted by all coincides with the peak inflows from runoff into the MRC member states (MRC, 2005). Lower Mekong. Furthermore, the wet season affects the whole of Lower Mekong more or less simultane Flow is stored in dams and other storages and, ously. The consequence is that the Mekong has a during high flows, in die channels and floodplains. very pronounced annual flow cycle, with the high Water is lost from rivers (especially downstream season flow being 15-30 times the low season sections in rivers in arid or semi-arid zones) by flow. Furthermore, the high season flow occurs evaporation and seepage, or by the consumption as along the whole length of the Mekong at more or évapotranspiration of a proportion of flood less the same time, with only a short lag between discharge onto the floodplain. Water is diverted upstream and downstream. from the rivers mainly for use in irrigation, and The floodplain of the lower basin is unused water flows to the sea. extensively flooded during the high flows / wet The account is based on a monthly timestep, season. The floods take water from the main channel which we consider adequate for our purpose. above Phnom Penh and divert it into the Tonle Sap, The account links known quantities in the across the floodplain back to the river below the water balance, such as rain and streamflow meas Phnom Penh, or to the delta. Some of the flood ured at gauging stations, with simple, physically water is consumed as évapotranspiration, and does plausible models, guided by the data. not return to the river. When the Mekong is at peak flow, its water Rainfall / évapotranspiration / runoff level is above that of the Tonle Sap River which The partitioning is derived from the drains the Tonle Sap (Great Lake). Hence water is reasoning of Budyko (1974) (which applies to pushed up the Tonle Sap River and is stored in the average annual runoff), with the addition of a lake. This reverse flow reverts to normal flow when storage of which varies from month to month. me Mekong flow recedes, and the Tonle Sap River Rainfall (P) plus irrigation (Ir) is first partitioned at then drains the stored water plus additional water the surface into the runoff (Ro) and infiltration (7), from runoff within the Tonle Sap catchment. The where conservation must be observed: storage of water wimin the lake is of great importance to local fisheries and livelihoods. P + Ir-I-Ro = 0 (1) Water use in the upper Mekong basin is dominated by évapotranspiration from forests Rainfall plus irrigation is the supply limit, and grasslands, with more than a third of the whereas the unfilled portion of a generalised surface precipitation becoming runoff. In the mountainous eastern part of the basin (mainly Laos), water use is storage, AS^,,,,*, is the capacity limit governing the also dominated by évapotranspiration from forests partition and includes soil storage and small surface and runoff which account for more than three stores. A Budyko-like equation is used to smooth quarters of the rainfall, with the remainder being the transition from the supply limit to the capacity mainly cropping. The western part of the basin, in limit: northeast Thailand, is mainly under cultivation, and évapotranspiration from rainfed croplands accounts (2) for about three quarters of the water use. Rainfed / _ ÍM^Lf cropping and irrigation are important water uses in ^ma* Ul + ((P + /r)/ASîmaxr)J die lower part of me basin, with irrigation being especially important in the delta. Figure 2 shows mat with larger values of the parameter a\ this equation makes a sharper Method transition from the supply limit to see capacity limit. The water use account is a top-down Thus, given precipitation, irrigation and the model (Sivapalan et al., 2003), based on parameter al, the infiltration into the generalised simple lumped partitioning of rainfall into surface store is found from equation (2) and the évapotranspiration and runoff at the catchment level. runoff from equation (1). The évapotranspiration is further partitioned into the proportion accounted for by each vegetation type or land use, including évapotranspiration from wetlands and evaporation from open water. Runoff flows into the rivers, with downstream flow calculated by a simple water balance. 69 River flow and storage upstream of Kratie / River flow is modelled as a series of reaches, / —- with mass balance observed between reaches. The / ' -""-'' _ " // ^'" large difference between the high and low flows implies considerable storage within the river 1f / ''' channels. Furthermore, floods also imply / / // considerable storage, particularly further // 1/ ^______downstream. / _-——~~~——~~~ ' Thus, the reach outflow, Q„, is given by the inflow, Qi, plus any tributary flows, Q,, plus the runoff from the adjacent catchment, R (as x 1 1 i 1 1 0 0.0 05 2.0 25 calculated above), plus a baseflow component, Bfi ^/r)/ASmax less any diversion (for industrial or agricultural use), D, less any losses (evaporation, seepage), L, plus Figure 2. Behaviour of the runoff infiltration the change in reach storage DSr: partition equation with different values of the parameter a\. Qo = Q+Q+R(*Bf-D-L+târ (5) The évapotranspiration depends on the The reach storage is taken to be a function of the potential évapotranspiration (ET^, capacity limit) inflow: and the surface storage (Ss, supply limit). Although soil and other surface stores are not differentiated, c2 S = C\Q, (6) the implication is that evaporation occurs from small r ponds, puddles, and the soil surface, whereas where c\ and cl are parameters. The change in transpiration comes from deeper soil storage. A reach storage is the difference between reach storage similar equation to the above, with a second at two timesteps: adjustable parameter a2, is used to smooth the l-Al transition from the supply limit to the capacity limit: AS, = Sí - S; (7) The reach storage is recovered as river flow ET ((s: (3) during recession. Outlfow from one reach becomes ET„, inflow to the next reach. Where tributaries join a \ f4c reach, the inflow is the sum of the outflows from the joining reach and the tributaries. This equation also behaves as shown by the The monthly baseflow, Bß was considered to figure with me obvious changes to the parameters. be equal to the monthly average drainage-to- In all the examples described in this paper, we shall baseflow component, DB- This implies that the use dl = al, so the rainfall-runoff model has two groundwater levels are sustainable (inflows balance adjustable parameters. outflows over a long period). The implications of other assumptions can be calculated. The surface storage is increased by the infiltration and decreased by the évapotranspiration Flow and flood spill downstream of Kratie and a drainage-to-baseflow component, DB: Downstream of Kratie, floods spill from the river on to a wide floodplain and, in general, do not return to the river (Fujii et al, 2003; Morishita et al, Si S'-&' + I-ET - £>„ (4) 2004). Some water flows to the Tonle Sap, some re-enters the Mekong downstream of Phnom Penh (that is, in a downstream reach), some flow is where t is time and At is the timestep (one month). directed to the delta region, and some is presumably Baseflow is a small component of the total flows in the Mekong, so the baseflow component was assumed to be a small value, constant with time, varying from catchment to catchment. 70 consumed on the floodplain as évapotranspiration. Note mat the height of the Mekong, HM, is The reach outflow is given by: calculated from the flow at Kratie, QMK. c8 and c9 8 are parameters. When (HM + cl) > HTS, QTS is Ö, = Q, + Q, + Ro - D - L - F0 + F, ( ) negative, indicating flow reversal. where F¡ is overland flood inflow from an upstream Lake storage, S¡, is given by the storage at the reach (which equals zero for the first reach down previous timestep, plus runoff from the Tonle Sap stream of Kratie, since this is the first reach to catchment, minus losses (evaporation, etc), minus contribute overland flood flow to downstream flow in the Tonle Sap River. reaches), F0 is overland flood outflow, given by the river height, H, above a threshold value, H : t S¡ = S}-" +Ro-L-Q TS (13) F0 = C3(H-H,) H>H, (9) F„ = 0, H (KCij ETpotj - Pe,j) (14) Irr,Dtmmid i j Tonle Sap and reverse flow IE. Flow in the Tonle Sap River, QTS, and consequently storage in the lake, depends on the where KCiJ is the crop factor for the rth crop, IE¡ is difference in height between the Tonle Sap River the irrigation efficiency and Pe,,i ei] is the effective and me Mekong. It is also assumed that the flow rainfall. Note that if Peieiji > K^CijCi , ETLpotpBOI„ Irr,irrDemandtj = 0. capacity of the Tonle Sap River increases with The irrigation demand per unit area for that crop, increasing IrrDemandA„ is summed for me following 12 months, Ora = C6(HT. HM - CV HTS height. Thus: in order that a full year's demand may be compared with a full year's supply: (11) Irrn. = 5>iDemandij where c6 and cl are parameters, and .7=1.12 (15) HTS and HM are the heights *i of the Tonle Sap River and the Mekong. The terms in the brack The total irrigation demand per unit area for n crops ets account for the flow dependence on height dif for the subsequent 12 months, IrrDemamfr, is: ference, whereas the term accounts for the increas (16) ing flow capacity of the Tonle Sap River with in Irrn. - s ^ creasing height. The cl parameter accounts for the " A Sift fact that the absolute heights in die two rivers are not calculated. Rather, relative heights are calculated where Aamm is the maximum area available for crop , , from the volume of water /' and A is the maximum area available for all Hm = cS{S, + c9fNn 5 TIns3X stored in die Tonle Sap lake, S¡ irrigated crops. The actual area mat may be supplied H M = c9QZ and the flow in the Mekong as: for an irrigated crop, A,¡, depends on the available supply of water, and is given by: (12) Wt. A,, = MNlA,, 4* (17) Irrn Allm 71 where WAvai¡ is the volume of water that may be period of the outflows and changes to storage less available from flow and/or storage, and the MIN the sum of the inflows. We did not calibrate monthly function limits the area irrigated. The volume, D¡, or seasonal behaviour. diverted to supply crop i is: 2. We made the calculated annual average diversions equal to independently measured values n — A Trr (18) where we had them, by adjusting Kc„ Ailmax or ATImax 1 ^i ~ ^il "DemandAi in equations (14), (16) and (18). Again, we did not and the individual diversions to each crop are calibrate monthly or seasonal behaviour. summed to give the total diversion, D: Balance checks (19) The spreadsheet has two checks of the D = Dl+D1+. • D. overall water balance for each sub-basin. The first Partitioning of dryland évapotranspiration by check is that the sum of the monthly rain over the land use / vegetation type full period equals the sum of the monthly Equation (3) gives an estimate of the monthly évapotranspiration plus the sum of the monthly évapotranspiration for each catchment which is runoff plus the difference in the surface storage, Ss, constrained by and consistent with the measured between the beginning and end of the period. outflows. This can be partitioned into the The second check is that for each sub basin évapotranspiration from each land use / vegetation the sum of the monthly inflows equals the sum of type in several ways, using vegetation water use the monthly losses to discharge, evaporation from modelling principles. The FAO crop factor approach storages and diversions plus the difference in is a suitable candidate, since it is a simple model storages between the beginning and end of the closely based on observed crop water use, and has period. been applied all over the world. As well as providing a better estimate of the partition, it would Results: Mekong water use account also provide an independent check of the rainfall- The Mekong was divided into 12 sub-basins. evapotranspiration-runoff partitioning of the simple Climate and flow data for each sub-basin were hydrological model. At this stage, we have used a partly supplied by the Mekong River Commission, simple pro-rata partitioning based on land use areas and partly gathered from other sources as part of a derived from remotely sensed data. project within the Challenge Program on Water and Food. Land use data were obtained from the IWMI website (www.iwmi.cgiar.org). We show here die ET. tLlrr (20) n A flow modelling for an upstream location and a downstream location on me Mekong, and for the Tonle Sap catchment where AiD is the rth dryland land use, ATD is the total dryland area, and ETTD is the total dryland évapotranspiration. Upstream location: Luang Prabang We emphasise that this simple partitioning is The reach from Chiang Saen to Luang not a restriction in the method. It is merely an Prabang drains an area of about 53,000 km2 (not expedient used here for this demonstration. Using including the Muong Ngoy catchment, which was something like the FAO CROPWAT approach is modelled separately). The behaviour of mis reach quite easy to implement. was modelled using equations (5) to (7). The calculated flows were matched to die Calibration observed flows in two stages. First, the rainfall - We used two main calibration steps. runoff partition was used to derive a runoff record in 1. The runoff into any reach must equal the sum of which, over the full length of the flow record, the the outflow, losses, diversions and changes to stor runoff equalled the difference in the observed in age minus the sum of the inflows. This is true for flows and outflows less the inflows from the Muong any period, from a single month to the full length of Ngoy catchment. Parameter a lin equation (2) (with the record being considered. We set the sum of the a2 = a\ in equation (3)) was adjusted to achieve this runoff (by varying Smax, and a\ - which we made match. Then, the calculated monthly outflows were equal to a2 - in equations (2) and (3)) over the full matched to the observed outflows using Solver to period to be modelled to equal the sum over the full 72 minimise the sums of squares of differences flow contribution (ie the difference between reach between observed and calculated, while varying inflow and outflow in Figure 3, right) shows parameters c\ and c2 in the equations above. The occasional apparently negative flows, which indicate results are shown in Figure 3. The model was fitted reach inflow greater than outflow in some months, to the total flows (Figure 3, left). The observed local due to water held up in river storage. These are modelled reasonably well. 40000 15000 — obs o cale •| 30000 o E a," 20000 I lililí D) m g 10000 UlliI I I 0 i J vAJu v vUvvJ v \J 1985 UvU199V0 1995 2000 1985 1990 1995 2000 Year Year Figure 3. Measured and modelled flows at Luang Prabang, 1985 to 1999. Left: hydrograph of observed and mod elled outflows. Right: comparison of difference between observed inflows and outflows, and difference between modelled inflows and outflows. Tonle Sap Downstream location: Phnom Penh The reach from Kratie to Phnom Penh floods The Tonle Sap catchment is about 85,000 during the wet season. Evapotranspiration is km2. Several small rivers drain into the Tonle Sap assumed to consume part of the flood and part (Great Lake), which in turn is drained by the Tonle returns to the river system. Sap River which discharges into the Mekong at This reach was modelled using equations (8) Phnom Penh. Applying the Tonle Sap flow model in to (10). In the same manner as the upstream reach, equations (11) to (13), with parameters c6 to c9 the rainfall - runoff partition was used to derive a optimised using Solver in Excel, gives the flow in runoff record in which, over the full length of the the river. Figure 4 shows the comparison between flow record, the runoff equalled the difference in the observed and modelled flows. The flow reversal is observed inflows and outflows. Then, the calculated monthly outflows were matched to the observed outflows using Solver to minimise the sums of o squares of differences between observed and E calculated, while varying parameters cl, cA and H, o in the equations above. The results are shown in E Figure 5. The model was fitted to the total flows 2> (Figure 5, left). The observed local flow (0 contribution was calculated from me difference .c o between reach inflow and outflow (Figure 5, right). 73 positive peak followed by an apparent negative negative peak, indicating a loss from the channel). peak, flowed by a second small positive peak. This On recession, the flood stops and there is a second results from an increase in river height with the small positive flow contribution. The model onset of the wet season (the first small positive reproduces this feature. peak), followed by a flood spill (the apparent 120000 20000 o E E o E 1985 1990 1995 2000 1985 1990 1995 2000 Year Year Figure 5. Measured and modelled flows at Phnom Penh (Mekong main channel below the confluence with the Tonle Sap River), 1985 to 1999. Left: hydrograph of observed and modelled outflows. Right: difference between observed inflows and outflows, and difference between modelled inflows and outflows. Water use by major land uses The water use by major rainfed and irrigation land uses was calculated by equations (14) to (19). The monthly calculations aggregated to an annual average water use are shown in Figure 6. The figure shows the dominance of forestry and runoff in the eastern parts of the basin and of cropping and irrigation in die southern and south-western parts of the basin. Discussion The water use account spreadsheet is an example of top-down modelling in mat it describes the overall behaviour of a basin based on observed responses, which Sivapalan et al. (2003) regard as the defining feature of a top-down approach. It is at a level of detail appropriate to an overview of the Mekong Basin. Individual model components are inferred from the data, ramer than pre-conceived. The elements of the water account (flow, storage, water use) are linked to time-series graphs (such as the hydrographs in Figures 3 to 5, compar ing observed and calculated flows). Systematic learning about catchment and basin behaviour is facilitated, gaps and deficiencies in data are readily Figure 6: Major water use for main catchments within the apparent, and hypothesis testing is quick and easy. Mekong Basin. The area of each pie chart is proportional to the volume of mean annual rainfall in each catchment. 74 Parameter estimation can be automated through the Temperature is expected to increase by 3CC and the Solver function in Excel, though care is required in rainfall is expected to increase by 7 %. Figures for its use: the user must be satisfied that the underlying potential évapotranspiration were not given, but we sub-model is reasonable and that the optimum is assume that effect of increased temperature (which sensible. We use this method in some of our will increase potential évapotranspiration) is offset calibrations to fit parameters. by the increased rainfall and presumably cloudiness A significant advantage of a water use (which will decrease potential évapotranspiration). account for a whole basin is that there are often We thus assume that the potential évapotranspiration many sources of data with which to constrain or will not change. We assume that the rainfall is calibrate a model. We have used many flow gauges, increased everywhere by 7 %. Figure 7 shows that known annual discharge, and the Mekong Decision the modelled flood peaks are greater under this Support System (IQQM) estimates of diversion scenario. These results are broadly similar to those volumes for irrigation districts. Furthermore, the ofHoanhetal. (2003). requirement to balance all gains, losses and changes to storage, both across the basin and for each and every component, places severe constraints on permissible use, flows and storages. When several current tributaries contribute to a main channel, the climate change calculated flow in each is constrained so that, even if o one or more is ungauged, tight limits can be placed E "g 100000 on the flow from each. This is even more the case if o there are independent estimates of vegetation water E Q> use. In another context, Raupach et al. (2001) noted S5 co 50000 the usefulness of mass balances in providing .c o physical constraints to material flows, particularly 10 when several flow calculations are linked (their con text was several entities - carbon, energy, water, nutrients - in one place, whereas here we deal with 1985 1990 1995 2000 one entity - water - in several places). Year Sivapalan et al. (2003) note problems and caveats with the top-down approach. Finer scale Figure 7: Comparison of modelled historical flows processes are glossed over, and the user must be and flows assuming increased rain due to climate confident that key features are not ignored, and that change, at the border between Cambodia and large scale models are physically reasonable Vietnam, 1985 to 1999. interpretations of the processes. There are dangers in generalisations and extrapolations to new situations. However, our main motivation for Thus, the water use accounts should be used to developing water use accounts is to study the investigate scenarios that are but modest impacts on agricultural productivity, economics and perturbations of the conditions for which they are livelihoods. Many of the purely physical changes tested and calibrated. mentioned above, particularly flow regimes, could The water use account spreadsheet provides a be examined as well or better with the Mekong basin overview of major natural, dryland and Decision Support System SWAT-IQQM-ISIS irrigated water uses, flows, storage, major losses and catchment - hydrology model. We do not propose discharge. It provides a basis for examining the that water use account spreadsheets should be used impact of physical changes to the system and in place of such models. for interactions with agricultural productivity, Molden et al. (2001b) and Sakthivadivel and economics and livelihoods. Molden (2001) show that basin water use As an example of use of the spreadsheets to accounting is central to linking institutions to water examine physical impacts, Figure 7 shows the resources development and conservation. They impact on downstream flows of a 7 % increase in develop static water account that aggregate water rainfall. The IPCC Fourth Assessment described in uses across whole basins. Such accounts do not the estimated impact of climate change in the readily indicate which parts of a basin (if any) Southeast Asian region, and noted that in 2080 the might be most vulnerable to change or in need of 75 institutional attention, nor do they indicate issues References such as seasonal shortages, floods, or agricultural or Allen, R.G., Pereira, L.S., Raes, D., and Smith, M., ecosystem productivity. Biltonnen et al. (2003) 1998. Crop évapotranspiration - Guidelines for show that water use accounting is central to water computing crop water requirements - FAO policy development of the Mae Klong Basin in Irrigation and Drainage Paper 56. FAO - Food Thailand. In contrast to Sakthivadivel and Molden and Agriculture Organization of the United (2001) and Molden et al. (2001), they develop Nations, Rome. accounts for different parts of the basin, though the ABS, 2004. Water Account Australia 2000-01. accounts are nevertheless static. Australian Bureau of Statistics, Canberra. Our accounts are dynamic and thus suited to Biltonnen, E., Kwanyuen, B., Kositsakulchai, E., investigation of a wider range of issues. We have and Pattani, S., 2003. Development of water- also developed water use accounts for the Murray- management institutions in the Mae Klong River Darling, Karkheh and Limpopo river basins (Kirby Basin, Thailand. In: Governance for Integrated et al., 2006). We have used water use accounts, Water-Resources Management in a River-Basin particularly in the Murray-Darling Basin, as the Context, Eds., Bruns, B., and Bandaragoda, D.J. basis for assessments of water allocation under International Water Management Institute, various policy scenarios (eg Qureshi et al, 2006a, Colombo. 2006b), and water use in optimal decision making Budyko, M.I., 1974. Climate and Life. Academic, about water trading and environmental water use San Diego, 508pp (Kirby et al., 2006). In our future work we aim to Fujii, H., Garsdal, H., Ward, P., Ishii, M., Morishita, use the water use account outlined in this paper to K., and Boivin, T. (2003). Hydrological Roles of investigate impacts of changes to water availability the Cambodian Floodplain of die Mekong River. and use on poverty and livelihoods in the Mekong. International Journal of River Basin Conclusions Management, 1(3): 1-14. Water use accounts are a powerful way of Gippel, C.J., 2006. Scoping study of operational describing the overall water use and flow behaviour changes to reduce river and storage losses along of a river basin. They capture the main aspects of the Eiver Murray system. Fluvial Systems Ply die behaviour, both spatially and temporally Ltd., Stockton, NSW. The Living Murray, (seasonally, annually), and the balance between Murray-Darling Basin Commission, Canberra. different types of water use (dryland, irrigated, Herath, S. and Yang, D. (2000). Distributed forest, wetland and other water uses). Hydrological Modelling in Mekong Basin. The water use accounts spreadsheets we have Hydrological and Environmental Modelling in developed are useful for systematic learning and the Mekong Basin, Al-Soufi, R. W. (ed), hypothesis testing, and also help the user rapidly Mekong River Commission, Phnom Penh. identify gaps and limitations in the data. They can Hoanh, CT., Guttman, H., Droogers, P., and Aerts, be applied in cases where data are limited, and it is J., 2003. Climate, food, and environment. possible to construct a reasonable account based on Mekong basin in southeast Asia. Contribution to data available on the Internet. the Project ADAPT - Adaptation strategies to The water use accounts spreadsheets provide changing environments. International Water basin overviews of water uses, and provide a basis Management Institute, Colombo. for examining the impact of physical changes to the Jirayoot, K. and Trung, L. D. (2005). Decision system and for interactions with agricultural Support Framework - The Transboundary productivity, economics and livelihoods. We Analysis Tool Developed by Mekong River emphasis, however, that they are not detailed Commission. Proceedings of the International catchment hydrology models, and are not suited to Symposium on Role of Water Sciences in Trans- river planning and management, nor to boundary River Basin Management, 10-12 investigations of small-scale, detailed effects. March 2005, Ubon Ratchathani, Thailand, Herath, S., Dutta, D., Weesakul, U., and Das Acknowledgements Gupta, A. (eds), United Nations University. Albert van Dijk and Cuan Petheram helped Kirby, M., Qureshi M.E., Mainuddin, M., and with data and ideas on water accounting. Funding Dyack, B., 2006. Catchment behaviour and was provided by the Challenge Program on Water countercyclical water trade: an integrated model. and Food. Natural Resources Modelling (in press). 76 Kirby, M., Mainuddin, M., Ahmad, M-u-D., Podger, G.M., Beecham, R.E., Blackmore, D., Marchand, P., and Zhang, L., 2006. Water use Stein, R. and Perry, C. 2004. Modelled account spreadsheets with examples of some Observations on Development Scenarios in major river basins. 9th International River the Lower Mekong Basin, Mekong Water Symposium, September 3-6, 2006, Brisbane. Resouces Assistance Strategy, World Bank, Kite, G. (2001). Modelling the Mekong: Nov. 2004, Vientiane. Hydrological Simulation for Environmental Qureshi M, E, Connor, J, and Kirby, M, 2006a. Impact Studies. Journal of Hydrology, 253: 1-13. Costing environmental flow acquisition Kummu, M., Koponen, J. and Sarkkula, J. (2005). strategies in the Murray Darling Basin, Assessing Impacts of the Mekong Development Australia. Integrated Water Resources in the Tonle Sap Lake. Proceedings of the Management and the challenges of sustain International Symposium on Role of Water able development. Marrakesh, Morocco, 23- Sciences in Transboundary River Basin 25 May 2006. Management, 10-12 March 2005, Ubon Qureshi, M.E. Proctor, W. and Kirby, M., Ratchathani, Thailand, Herath, S., Dutta, D., 2006b. Economic assessment of water trade Weesakul, U., and Das Gupta, A. (eds), United restrictions in the Murray Darling Basin. In Nations University. ternational conference on regional and Lenzen, M., 2004. Nature, preparation and use of urban modelling, Brussels, June 1-3, 2006. water accounts in Australia. Technical Report Raupach, M.R, Kirby, J.M., Barrett D.J. and 04/2, Cooperative Research Centre for Briggs, P.R. Balances of Water, Carbon, Catchment Hydrology, Melbourne. Nitrogen and Phosphorus in Australian Molden, D., 1997. Accounting for water use and Landscapes: (1) Project Description and productivity. SWIM Paper no 1. International Results. CSIRO Land and Water Technical Water Management Institute, Colombo. Report 40/01 (2001). Molden, D., Sakthivadivel, R., and Habib, Z., Sakthivadivel, R., and Molden, D., 2001. Link 2001a. Basin-level use and productivity of water: ing Water Accounting Analysis to Institu examples from south Asia. IWMI Research tions: Synthesis of Five Country Studies. Report 49, International Water Management Chapter 2 in: Integrated Water Resources Institute, Colombo. Management in a River-Bain Context. Eds. Molden, D., Sakthivadivel, R., and Samad, M., Bruns, B., Bandaragoda, D.J., and 2001b. Accounting for changes in water use Samad, M. International Water Management and the need for institutional adaptation. In: Institute, Colombo. Intersectoral Management of River Basins, Simons, M., G. Podger and R. Cooke, IQQM - ed. C. L. Abernethy. International Water A hydrologie modelling tool for water Management Institute, Colombo. resource and salinity management, Morishita, K.; Garsdal, H.; Masumoto, T. Environmental Software, Vol 11, Nos. 1-3, Hydrological monitoring system for the pp 185-192, 1996. nd Cambodian Floodplains. 2004. Proc. of the 2 Sivapalan, M., Bloschl, G., Zhang, L., and Asia Pacific Association of Hydrology and Vertessy, R., 2003. Downward approach to Water Resources Conference, 1: 191-199. hydrological prediction. Hydrological Proc MRC, 2003. State of the basin report: 2003. Mekong esses, 17, 2101-2111. River Commission, Phnom Penh. Takeuchi, K., Tianqi, A. and Ishidaira, H. MRC, 2004. Decision Support Framework. Water (2000). Hydrological Simulation of the Me Utilisation Project Component A: Final Report. kong Basin by BTOPMC. Hydrological and Volume 12: Technical Reference Report. DSF Environmental Modelling in the Mekong Ba 630 iSIS Models. (Phnom Penh). sin, Al-Soufi, R. W. (ed), Mekong River MRC, 2005. Overview of the hydrology of the Commission, Phnom Penh. Mekong basin. Mekong River Commission, Vientiane. Neitsch, S.L., Arnold, J.G., Kiniry, J.R., Srinivasan, R., and Wwilliams, J.R. 2002. Soil Water Assessment tool User Manual. Grassland, Soil and Water Research Laboratory, Agricultural Research Service, Temple, Texas. A Framework to Assess Model Structural Stability through a Single-Objective Global Optimization Method Giha LEE. Yasuto TACHIKAWA, Kaoru TAKARA Disaster Prevention Research Institute, Kyoto University, Uji, Kyoto 611-0011, Japan leegihaiöflood. dpri. kvolo-u. ac.ip Introduction runoff simulation. However, their research is limited The problem of model structural uncertainty to improve a classical calibration strategy, i.e., with advanced automatic calibration methods is an single-objective optimization algorithm coupled issue of increasing interest in recent researches with their own conceptual models (e.g., SAC-SMA (Yapo et al, 1996; Gupta et al, 1998; Boyle et al, model). Hence, it is questionable that a physically 2000, Vrugt et al, 2003). Gupta et al (1998) based distributed model, which has a different pointed out that a subjective selection of objective model structure to reflect real rainfall-runoff functions (e.g., SLS, RMSE, HMLE) for calibration processes from a conceptual model, results in the of conceptual hydrologie models lead to an overemphasis on particular portion of predicted overemphasis on a certain aspect of the response hydrographs or whether optimal parameter sets (e.g., peak flows), while neglecting the model change or not according to objective functions. As performance with regard to another aspect (e.g., low reported in previous studies (Yapo et al, 1996, flows). In other words, different parameter 1998; Gupta et al, 1998; Boyle et al, 2000), the combinations can exist according to various result of variation of optimal parameter combination objective functions due to the presence of structural calibrated by a single-objective optimization method uncertainty. Structure error is unavoidable problem can be employed as one of the well-founded in hydrological modeling since hydrologie models indicators to account for model structural stability. are conversion and simplification of reality, thus no This study is conducted to investigate matter how sophisticated and accurate they may be, answers to the following questions: 1) What kinds of those models only represent aspects of conceptuali models are stable in terms of model structure zation or empiricism of modelers. In consequence, for description of rainfall-runoff process? (i.e., output time series of hydrologie models are as Definition of model stability). 2) How can modelers reliable as hypothesis, structure of models, and or engineers identify model stability and suitability? quantity and quality of available forcing data, and (Methodology for identifying model stability parameter estimates (Gupta et al, 1999). and suitability). A framework is outlined as an Hydrologists have concentrated their effort attempt to assess the model structural stability using on development of more powerful model calibration single-objective global optimization method. The scheme to assess the suitability of model structure Shuffled Complex Evolution (SCE-UA) algorithm is for representing the natural system and for used to calibrate a conceptual lumped model, identifying model structural inadequacy (Gupta et Storage Function Model (SFM) and a physically al, 1998; Boyle et al, 2000). Their new scheme based distributed model, CDRMV3 using five his well explains the inherent multi-objective nature of torical flood events (see Table 1) from Kamishiiba the problem and the role of model errors in rainfall- catchment located in Kyushu area. Parameters Sept. 1997 June 1999 Aug. 1999 Sept. 1999 Sept. 2005 Peak discharge (mVs) 1203 210 489 644 1840 3 Initial discharge (m / 33 59 82 34 11 s) Total amount of rain 496 463 473 339 831 fall (mm) Total amount of dis 415 238 237 256 780 charge (mm) Runoff ratio 0.84 0.51 0.50 0.77 0.94 Table 1 The characteristics of each year heavy rainfall and flood discharge at Kamishiiba catchment. 79 Methodology for Model Structure Analysis Decision of parameters to be optimized Sslocuonof ——* StnsftMty ObjocoV« Function modal structura analysis solocöon (OFs) Analysis of paranwtsf $T"'*$!tsrÀ m****1&mnr\~ Indexa,-.Index^ transferability from /¡turnt ö**"* MuuiUiBMvf}-* Inde&*\-<,îndeg£* ovsfitto ovont QualttaoVo and Quantitativ« Assossrmr* of modol «o f >*• mm\M^^W\*Jndei^,--,lnde^^\ staietural stability M**MoM Structure: 6%?* ÎS*ÊMN °*f> von r on, 9> Fig. 1 Schematic illustration of a framework to assess model structural stability optimal parameter set; Index' guideline index for assessment of model structural stability; / = objective Function; 7 = storm event. Our purpose of this study is to establish a 2) An ideally-structured model can be regarded as a framework for how to assess the model structural stable model which has the identical best stability. This work is summarized by two parameter set regardless of objective functions main procedures. The first step is a qualitative and have high parameter transferability from identification of model stability according to selection of objective functions. The second nEventX ^ ^ riEveniN °OF, ~ ' ' ' ~ "oF„ procedure is a quantitative assessment of model event to event, i.e., structural stability through the analysis of parameter 3) Therefore, variation of optimal parameters transferability with a development of benchmarks or calibrated by using an appropriate single- guideline indexes. Figure 1 illustrates the schematic objective optimization scheme can be one of the process of me framework for assessment of model indicators for assessment of model structural structural stability. stability. This assessment procedures are based on the After considering synthetically all concepts following ideas: above mentioned, we conclude that a more reliable 1) If hydrologie forcing input data (e.g., rainfall, model structure leads to the constant optimal stream flow) for model calibration using efficient parameter set without regard of any objective func- and robust optimization algoridim are not erroneous, calibrated parameter set can reflect S}Event\ nEvent\ ÜOK •ÜQF. explicitly the structural stability of hydrologie tions selected subjectively, i.e.. model. Moreover, such model structure maintains high degree of accuracy for predicted hydrographs when 80 applying parameter set for various type and \f*r, if Y/^KSA magnitude of floods in the same study site, i.e., 1 = [ r, if Yr>RsA (2) ûEvenll srEventN where, S = water storage; re = effective rainfall ' ~t>OF i . As a result, model structural intensity; r = rainfall intensity; q = runoff; t = time stability is evaluated by the ability enable to reduce step; k = storage coefficient; p = coefficient of die influence of objective functions and flood events. Two different types of model structures are T, applied to compare model structural stability for nonlinearity; / = primary runoff ratio; = lag verification of our framework and each model is calibrated by SCE-UA optimization strategy with SA time; and = cumulative observed rainfall from three objective functions. Applied models and the beginning of storm. optimization method, objective functions are introduced in following sub-sections. Physically based distributed model, CDRMV3 Hydrologie Models CDRMV3 is a physically based distributed hydrologie model developed by Kojima et al. (2003) Conceptual lumped model, Storage Function including discharge-stage relationship with Model (SFM) saturated-unsaturated flow (Tachikawa et al, 2004). This model is a lumped model with die The model solves the one-dimensional kinematic reflection of nonlinear characteristics of hydrologie wave equation with the discharge-stage equation response behavior. SFM is used for the rainfall- using the Lax-Wendroff finite difference scheme runoff simulation in a small watershed usually less according to orderly nodes and edges, edge man five hundred square kilometers in Japan. The connection based on flow direction map (see Figure form of SFM is expressed as: 2). All géomorphologie information are extracted from 250m based DEM data. Channel routing is also dS_ carried out by the kinematic routing scheme = r.(t-T,)-q, S = kqf (1) dt as well as calculation of slope elements reflecting contributing areas. <*) Radar Rainfall or Gag« Rainfall - Gao-data Proeosslng I Rainfall-Runoff Simulation surfJco flow i , * .SUPSUf Hllldpp« H—Jfl Channol (Option) routing mmmmwmmmtmmm untmmtm3mmmmmmDischarg e at Nodot and Edgo« Fig. 2 Schematic representation of CDRMV3 (a) Modular structure of CDRMV3 (b) Distributed grid rainfall data (c) Close-up of edges and nodes extracted DEM (d) Model structure for me hillslope soil layer and discharge- stage relationship. 81 The model assumes that a permeable soil layer models, apart from the Genetic Algorithm. This covers the hillslope as illustrated in Figure 2(d). evolutionary approach method has been performed The soil layer consists of a capillary layer which by a number of researchers on a variety of models unsaturated flow occurs in and a non-capillary layer with outstanding positive results (Gupta et al., 1999) in which saturated flow occurs. According to this and has proved to be an efficient, powerful method mechanism, if the depth of water is higher than the for the automatic optimization (Gan and Bifu, 1996; soil depth, then overland flow occurs. The Yu et al, 2001). Basically, this scheme is discharge-stage relationship is expressed by synthesized by following three concepts: equation (3) corresponding to water levels (see Fig (1) combination of simplex procedure with the ure 2(d)) defined as: concepts of controlled random search approaches; (2) competitive evolution; and (3) complex vcdc{hidcy, o dc ds 2 is the depth of the capillary soil layer and SLS^iqf-q.iß)) (5) is soil depth. Detailed explanations of model structure appear in Tachikawa et al. (2004). q, Shuffled Complex Evolution (SCE-UA) where is observed stream flow value at time t; Algorithm The reasons for difficulties in automatic is model simulated stream flow value at calibration with respect to the response surface of 9 parametric structures, i.e., existence of numerous time t using parameter set ; N is the number of local optima, non-smooth response surface, flow values available. non-convex shape around global optimum, motivate The main reason for the popularity of SLS a development of a global search algorithm in has been its direct applicability to any model. The rainfall-runoff modeling (Sorroshian and Gupta, selection of SLS as an objective function implies 1995). Evolutionary algorithms are probably the assumptions concerning the probability distribution most commonly applied global optimization of the errors: 1) the residuals are independent and methods in rainfall-runoff simulation. The Shuffled identically distributed; 2) the residual distribution Complex Evolution Algorithm (SCE, Duan et al, has homogeneous variance; and 3) the residuals are 1992; 1993; 1994), one of the computer-based normally distributed (Yapo et al, 1996). automatic optimization algorithms, is a single- Moreover, the largest disadvantage of objective optimization method designed to handle SLS is the fact that the differences between the high-parameter dimensionality encountered observed and predicted values are calculated as in calibration of a nonlinear hydrologie simulation squared values. This shortcoming results that larger values in a time series are overestimated 82 whereas lower values are neglected (Legates and *,=0, -ft(0) McCabe, 1999). Krause et al. (2005) proposed the where is the model residual at modified index of agreement (MIA, Willmott, 1984) to overcome the inefficiency of these measures with time t; is the weight assigned to time t, squared errors. This objective function is calculated w,=/,2(M) /, =qT' as: computed as . Where is the expected true flow at time t, is the unknown transformation parameter which stabilizes the (6) „obs til iT q, MIA = 1 • variance. Fulton (1982) recommended using + obs „mean q,m-<¡r q, -q, mstead of to approximate q, where In this study, above mentioned three objec tive functions are used in the calibration trials and is a mean value of observed time series. the analysis of hydrologie model stability. Sorooshian and Dracup (1980) remarked that Result and Discussions the problem often encountered in rainfall-runoff modeling is the fact that the residuals variance The impact of objective functions to model per increases with increasing flow values, i.e., the formance assumption of homoscedascity cannot be justified. When we plot the residuals versus predicted runoff, The plots of comparison between the simulated and we can easily find out whether the variance of the the observed hydrographs according to the three residuals increases with increasing flow values, i.e., objective functions (SLS, MIA and HMLE) are il the problem heteroscedascity. Generally, the lustrated in Figure 3. commonly practiced methods of handling this heteroscedastic error cases in natural system have From Figure 3, we notice that: been tiirough the application of Weighted Least 1) In SFM cases, the simulated hydrographs based Squares (WLS) function and transformation on the parameters calibrated by SLS are close to functions (Sorooshian and Dracup, 1980). They the observed ones while other parameters suggested the Heteroscedastic Maximum Likelihood optimized by HMLE, MIA lead to less magnitude Error (HMLE), which enables the estimation of the than the real measured stream flows under big most likely weights through the use of the maximum flood events (e.g., Event 1, 4, 5). estimation theory. This new procedure can eliminate 2) The calibrated parameters based on MIA, HMLE some of the subjectivity involved in the selection of result in underestimated outcomes in big flood transformation and/or a weighting scheme and but simulated results obtained from small floods yields a more balanced performance over the entire reproduce closely to the observed (see Figure 3 hydrograph (Yapo et al, 1996; Gupta et al., 1998). It (a))- is calculated as: 3) In CDRMV3 cases, all simulation results shown in Figure 3(b) are close to observed discharge for (7) any objective functions. This result implies that the problem of subjectivity related with selection of objective functions for model calibration based <=\ min HMLE •• N on single-objective optimization algorithm can be e.x ignored for distributed hydrological modeling used here. 83 (b) CDRMV3 24 (a) SFM « gf-^r-^/ 3 EvenCIMM«t 41 MOO «FM SLS - am HULE - SFM MIA - nnuax Thntpv) n K 120 Ui 1U lní"Vir *" "VIT" ~^Y~ IrVVM" "I—^VTT " T" 1" SFtttJSLS Event 2 CDRMV3 SLS SFtHJÎÊILE CDRUV3 HÊTLB Event 2 SFMJtttA — - . CORUV3 MIA MMMobtwrad t&& 200 X x 1» 1 V JK A**V£*>''*^ 1 ^~^Jf *w ^9^%W Vs, Í00 ^v **«. 50 1» f«W f« 72 « 120 144 Tin* (ht) 130 143 ^T vwu VyA/" r T WT" "WAT' SFM SIS - CO«»« SLS - SFM HULE - com/mjimj- - SfWJUM - CDRUV3 MIA »nmnobnrml - Event 3 £«00 Event 3 I«» _Jw^ \ too '-JF**_~~«*^ 120 14> nu ut F r""""^^ Cùmm sis - CDRMV3 HJKLE - CDBKV3_KW Event 4 4» TlmOv-J "^ 2000 CDRMV3 SCS - im CDRMVttJMKE - 1600 CORMV3J4IA - Event 5 OSMM»ofcMfV«d - ÏMOO É12M |ÏOOO | «M S WO 400 »0 ^^S^v, 120 144 Tkmfhi) Fig. 3 Comparison between the simulated and die observed hydrographs according to three objective functions; (a) SFM cases, (b) CDRMV3 cases. 84 Peak,,, The Assessment of Parameter Transferability where is the simulated peak discharge, from Event to Event Despite of successfully calibration as Peak, shown in Figure 3, it is still questionable that best is the observe peak discharge and is parameter set obtained from the event selected the mean observed discharge. PDR measures the subjectively can be applicable to other arbitrary tendency of the simulated peak discharge to be events in study site. If we expect good simulation larger or smaller than the observed peak discharge; results from transferred best parameter set, we can the optimal value is 1.0, larger values than 1.0 regard such model as a stable model structure, indicate an overestimation of the simulated peak which has high parameter transferability. The each discharge and smaller values than 1.0 indicate an performance from transferred parameter sets are underestimation of the simulated peak discharge. NS evaluated by peak discharge ratio (PDR) and Nash- measures a relative magnitude of the residual Sutcliffe(NS) statistics of the residuals as guideline variance to the variance of the observed stream indexes for measurement of parameter transferabil flows; the optimal value is 1.0. ity, defined as: The quantified results of parameter transferability PDR=PeaK„IPeak, (8) are plotted in Figure 4. As shown in Figure 4(a) and Table 2, the conceptual lumped model has low 2 parameter transferability while the physically based NS = l-£(tf- -qt(ß)f l^ift -?—) (9) model has high parameter transferability from event to event. Nevertheless, a successful transfer of parameters in CDRMV3 model, the results simu lated by optimal parameters of Event 2 over entire cases are inaccurate; NS values (a) SFM (b) CDRMV3 AppHM Event event! Bmaa Event} 0V0flf2 ffvem? 1 IS a 6 Ô ê -âà 0 a + a 0 9 9 ? §** • »°8 a g i t a ' 9 io.e , D * * tO.6 0 * •i. £ 'H t \o.4 o SLS ', sus ., u mut + UILM ' MM a MM C 11345 11346 1Î145 1134$ 11345 13341 1334t 11145 11345 11345 AppUwtEvM Appv*aenat ËHM2 evoti ftmMf ÊV9IH5 Bveatl «veno AWMJ BmM even« SLS h ' StS -> mu + 4MLS + ¡S < MM ü MM D + so ?» • tr u !" 8 £ ÔQ 9 3 s, P 9 ^ Z '- %> "» ? L V o Ç f S s œ ¿r n S S? •' -t i ? 0.4 1234t 12345 11345 11145 1234t 1734t 1134S 1134t 1334t 1134t PêrtmurS« Fig. 4 Plots for assessment of parameter transferability from event to event; each point indicates the evaluated NS, PDR due to parameter transfer. 85 are usually less than 0.7 and PDR values are Conclusions underestimated/overestimated irregularly. This In this paper, we have demonstrated a result indicates that Event 2 input data during framework for assessment of model structural calibration has poor information. Moreover, this stability through a single global optimization finding is allowable hydrologist or modelers to method (SCE-UA) and comparison of two distinguish a high or low quality data. In hydrologie models (SFM, CDRMV3). The results CDRMV3, all of the measured points described in under our framework lead to following conclusions Figure 4 due to three objective functions converge that either conceptual lumped model or physically into one position while points obtained from SFM based model is suitable for rainfall-runoff simulation are scattered very irregularly. if based on available informative data during model Unfortunately, the constant single optimal calibration, and that the simulated results of parameter set over all storm events is not observed CDRMV3 are not affected by objective functions ¿yEvenA . . ûEventN while the computed results of SFM are fluctuated "OR ^ • • • ^ C0E in the study site, i.e., . In the according to objective functions. Then, we test pa physically based distributed model, the different rameter transferability from event to event in the parameter combination also can lead to acceptable study site. The structural stability of CDRMV3 is model performances with proper values of NS or superior to SFM in terms of parameter transfer. jEvenA ^ fv rEveniN However, even though a physically based distrib PDR, i.e., " . This finding is uted model leads to high parameter transferability, strongly connected with "equifinality" (Beven, problems of uncertainty still remain to be unsolved. 2001). Therefore, the perfect framework of model The principal reason is that the identification of an structural stability still require analysis of appropriate model structure and the identification of uncertainty sources in hydrological processing and appropriate parameter set within this structure are its effect on the predicted output variable. difficult due to a range of uncertainties involved in the modeling process, which are also unavoidably propagated into die model output. Therefore, the analysis of uncertainty in the modeling process must go side by side with identification of stable model structure. Table 2 Evaluated NS and PDR for testing parameter transferability ApoM SHI Sat I;-»>»*I.<;.»-;MI:;«; : MI¡¡':;1!,:„MI;;-*"'.II¡¡':;1" '.n;;*;*ii¿<:*:«inji • M NS MM FOR MA Eventl 0.84 JtM 0,87 0*0 _SÄi 042 0*8 102 059 IX» 0*8 IOS Evert2 0.58 0.4? 0.56 0.48 083 043 0*8 056 085 0.72 088 0.75 tittti 0« 0.77 —AIS 089 OJO 0.70 _&B IOS 0*7 092 _aä 0*4 0-W SM _fiSI Jal JH4 0.7S ML 1.11 0*7 106 . J*S 1*7 EvartS 091 o*o 0.ÏÏ 0.7« ~~o~M 0.70 ö*? IM 0*6 uo 0*4 1*7 Evert! 0.37 1.39 0.35 1.34 0.25 1.41 0.69 155 087 1.25 0.68 1.27 Event2 ose 1.00 OBS 1.02 0*2 1.11 0.96 092 083 1.13 083 1 14 Ever« 0.34 1.39 0 48 1.32 0.46 1.33 0.63 138 0.70 1.22 087 1.28 Event4 ose 1.26 0.20 0.80 0.35 1.32 0*8 1.35 058 1.40 0 55 1.41 Eventt Events -a.ee 0*8 -0.57 0*2 DJB 1*2 0.49 1.33 0.57 1.33 0.53 1.34 Evertl 0.72 1.06 0.84 asa OBS 0.93 oss 0.86 0.95 DBS 0*5 088 Event2 0.77 0.67 0.77 0.67 0.64 0.71 0.79 0.84 0.86 0.66 0.87 0*9 Events 0.95 0 88 0.94 0.62 0.95 084 0.96 0.92 0.96 0.90 0.98 0*1 Event4 0.77 UM 0.54 059 085 088 0.94 0.67 0*3 0.84 0.93 0.BS E*ent3 Events 0 35 1JJ1 0.39 087 059 087 0.92 1.05 0*4 1*3 0.93 1*6 Evertl 0.92 1.11 0.92 0 92 080 0.97 0.95 0.74 0.96 0.75 0 96 0.7B Event2 0.76 0.57 076 0.57 D.79 054 0.60 0.42 0.72 0.47 0.75 0 49 Event3 0.64 053 0.84 083 0 84 0.84 ösr D.B2 0.92 0.73 0*4 0.77 Event4 0.93 0.97 0.68 0.51 0 91 0.86 0.99 0.87 0.96 083 0.98 088 Event4 Events 0.67 088 0.72 084 0.82 070 0.92 0.97 0*4 0.92 0.92 0*6 EvenM 0.93 0.94 0.91 0.78 0.91 0.77 0.B2 0.96 083 0.98 0.65 1.00 Event2 0 65 0.48 0.65 0.46 0.71 0.53 0.49 0.59 0.62 0.78 0.65 0.79 Eventa DB7 0.72 0*3 0.BS 084 0.67 0.87 1.00 0.67 D.B7 0*6 089 Event« • 94 050 0.75 0.57 089 0.72 0.B6 1.06 083 1.01 085 1.04 Events Events 0.96 052 0.91 0.74 0.90 0.72 0.96 1.02 0*5 1.02 0.95 1.07 «VE . 0.63 0.90 0.6 D77 0.70 OBS o*o 0.9S 0 83 0*4 0.83 0.97 86 References Krause, P. and Flügel, W. A. 2005. Integrated research on the hydrological process dynamics Beven, K.J. 2001. Rainfall-Runoff Modeling : the from the Wilde Gera catchment in Germany, primer, Wiley, Chichester. IAHS Conference, Bergen. Boyle, D. P., Gupta, H. V. and Sorooshian, S. 2000. Legates, D. R. and McCabe Jr., G. J. 1999. Toward improved calibration of hydrologie Evaluating the use of "goodness-of-feet" models: Combining the strengths of manual measures in hydrologie and hydroclimatic and automatic methods, Water Resours. Res., model validation, Water Resour. Res., 35(1), 36(12), pp. 3663-3674 pp. 233-241 Duan, Q., Sorooshian, S. and Gupta, V.K. 1992. Sorooshian, S. and Dracup, J.A. 1980. Stochastic Effective and efficient global optimization for parameter estimation procedures for conceptual rainfall-runoff models, Water hydrologie rainfall-runoff models: Correlated Resours. Res., 28(4), pp. 1015-1031. and heteroscedastic error cases, Water Duan, Q., Gupta, V.K. and Sorooshian, S. 1993. Resours. Res., 16(2), pp. 430-442. Shuffled complex evolution approach for Sorooshian, S. and Gupta, V.K. 1995. Model effective and efficient global minimization, J. calibration, Computer Models of Watershed Optimization Theory and Application, 76(3), Hydrology (ed. Singh, V.P.), Water Resources pp. 501-521. Publications, Colorado, pp. 23-68. Duan, Q., Sorooshian, S. and Gupta, V.K. 1994. Tachikawa, Y., Nagatani, G., and Takara, K. 2004. Optimal use of the SCE-UA global Development of stage-discharge relationship optimization method for calibrating watershed equation incorporating saturated - unsaturated models, J. Hydrol., 158, pp. 265-284. flow mechanism, Annual Journal of Hydraulic Gan, T. Y. and Biftu, G. F. 1996. Automatic Engineering, JSCE, 48, pp. 7-12. calibration of conceptual rainfall-runoff Vrugt, J. A., Gupta, H. V., Bouten, W. and models: Optimization algorithms, catchment Sorroshian, S. 2003. A shuffled complex conditions, and model structure, Water evolution Metropolis algorithm for Resour. Res., 32(12), pp. 3513-3524. optimization and uncertainty assessment of Gupta, H.V., Sorooshian, S. and Yapo, P.O. 1998. hydrologie model parameters, Water Resours. Toward improved calibration of hydrologie Res., 39(8), 1201 models: Multiple and noncommensurable Willmott, C. J. 1984. On the evaluation of model measures of information, Water Resours. Res., performance in the physical geography, 34(4), pp. 751-763. in Spatial Statistics and Models (ed. Gaile and Gupta, H.V., Sorroshian, S and Yapo, P. O. 1999. Willmott), D. Reidel, Norwell, Mass. Status of automatic calibration for hydrologie Yapo, P. O., Gupta, H. V. and Sorrooshian, S. 1996. models: comparison with multilevel Automatic calibration of conceptual rainfall- expert calibration, Journal of Hydrologie runoff models: Sensitivity to calibration data, Engineering, ASCE, Vol. 4, No. 2, pp. 135- J.Hydol., 181, pp. 23-48 143. Yapo, P. O., Gupta, H. V. and Sorrooshian, S. 1998. Kojima, T. and Takara, K. 2003. A grid-cell based Multi-objective global optimization for distributed flood runoff model and its hydrologie models, J. Hydol., 204, pp. 83-97 performance, Weather radar information and Yu, P.S., Yang, T.C. and Chen, S.J. 2001. distributed hydrological modeling, IAHS Comparison of uncertainty analysis methods Publ. No. 282, pp 234-240 for a distributed rainfall-runoff model, J. Hy drol 244, pp. 43-59. 87 Derivation of Rainfall Intensity-Duration-Frequency Relationships for Short-Duration Rainfall from Daily Data Le Minh NHAT. Y. Tachikawa, T. Sayama, K. Takara Disaster Prevention Research Institute, Kyoto University, Uji, Kyoto 611-0011, Japan nhat(5),flood.dpri.kvoto-u,ac.ip ABSTRACT In this paper, the properties of time scale invariance of rainfall are investigated and applied to Intensity- Duration-Frequency (IDF) relationships. The hypothesis of simple scaling implies in direct and empirically verifiable relations among the moments of several orders of rainfall intensities in different durations. Using these relations, it is possible to analytically derive IDF relationships for short-duration rainfall from the statistical characteristics of daily data only. The simple scaling model has been applied to precipitation data observed at the Introduction 3. To derive a IDF relationship for short-duration The intensity-duration-frequency (IDF) rainfall from daily data, and relationship of heavy storms is one of the most 4. To compare with traditional method and discuss important hydrologie tools utilized by engineers for the results. designing flood alleviation and drainage structures in urban and urban areas. Local IDF Equations are The paper is organized as follows: The next often estimated on the basis of records of intensities two sections present the theoretical background abstracted from rainfall depths of different as related to the time scale invariance properties durations, observed at a given recording rainfall of short-duration rainfall. In the sequence, these gauging station. In some regions, mere may exist a properties are verified for data observed at rainfall number of recording rainfall gauging stations gauging stations located in the Yodo catchments, in operating for a time period sufficiently long to yield Japan. The IDF relationship, derived from daily a reliable estimation of IDF relationships; in many rainfall data, is then compared to the traditional other regions, however, these stations are either estimated curve. Conclusions are given in the last non-existent or their sample sizes are too small in section. developing countries. Because daily precipitation data is the most accessible and abundant source of The generalized IDF relationship rainfall information, it seems natural, at least for the In recent years, the study of phenomena with regions where data at higher time resolution are scale variance has grown from applications in scarce, to develop and apply methods to derive me physics phenomena such as statistical theories of IDF characteristics of short-duration events from turbulence field theory (Gupta and Waymire, 1990) daily rainfall statistics. In this regard and in contrast to hydrologie phenomena such as stochastic rainfall to earlier empirical disaggregation techniques, the modeling and intensity-duration-frequency (IDF) works of Burlando & Rosso (1996), Menabde et al. curve formulation. While empirical equations have (1999) and Pao-Shan Yu et al.(2004) are examples been used for nearly one hundred years to explain of methodologies in which the theories of the form of IDF curves, scale invariance has helped scaling properties and employed to infer the IDF to understand these relationships. Sherman (1905) characteristics of short-duration rainfall from daily first developed a generalized IDF relationship, and data. many other versions of this relationship have been This paper aims developed in the years since. All forms of the 1. To present the properties of time scale invariance generalized IDF relationships assume that rainfall of rainfall, deptii or intensity is inversely related to the duration 2. To apply them to a location of station in Yodo of a storm raised to a power, or scale factor. cathment of Japan, 89 The IDF formulas are the empirical equations w representing a relationship among maximum rainfall 1 = intensity (as dependant variable) and other parameters of interest such as rainfall duration and frequency (as independent variables). There are several commonly used functions found in the where /' denotes the rainfall intensity for duration d literature of hydrology applications (Chow et al., and w, v, 6, and rj represent non-negative 1988). Four basic forms of equations used to coefficients. A numerical exercise proposed by describe the rainfall intensity duration relationship Koutsoyannis et al. (1998) shows that the errors are summarized: resulting from imposing v=l in Equation (5) are much smaller than the typical parameter and Talbot Equation: quantile estimation errors from limited size samples of rainfall data. Considering the specification of v£l • _ ü results in over-parametrization of Equation 1 = m (5), Koutsoyannis et al. (1998) suggested the JTY following equation as a general expression of IDF Bernard Equation: relationships for a given return period: i = JL- (2) 1= W (6) (d+0? Kimijima Equation: Rigorously, the coefficients w, 6 and rj in / = a (3) Equation (6) are not independent from the return de+b period. However, because the IDF curves for different return periods cannot intercept each other, Sherman Equation: such a dependence cannot be arbitrary; this restriction imposes limits to the variation range of i = —^ (4) parameters w, 6 and tj. For instance, if fwl, 61, r\l) and {w2, 62, rj2} denote two id + bf different parameter sets for return periods Tl and where i is the rainfall intensity (mm/hour); d is the T2 < Tl respectively, then Koutsoyannis et al. duration (minutes); a, b and e are the constant (1998) suggest the following possible restrictions to parameters related to the metrological conditions. the parameter space: Although many previous studies depend on curve-fitting techniques, studies generalizing IDF 6i=62=6>0;0 90 estimated for many locations: For example The scaling exponent, Hq, can estimated L.M.Nhat et al. (2006) established the IDF curves from the slope of linear regression relationship for precipitation in the monsoon area of Vietnam. between the log-transformed values of moment (j0„ £ fj i T) and scale The scaling of rainfall intensity theory In this section, a general theoretical parameters (log X) for various order of moment (q). framework for the proposed model is introduced. This is definition of "strict sense" simple scaling Let the random variable 1(d) the maximum annual (Gupta & Waymine 1990). A less restrictive value of local rainfall intensity over a duration d. It definition is "wide sense" simple scaling with d = 1 is defined as: hour and Xd= X, given by , l+d/2 £ [/,'] = AK« E[l¡] (14) ¡(a)- •• max (9) 0ir<,lyear l-d/2 The scaling exponent H is not constant with the order of moment (q), as Equation (14), but it varies where X (£) is a time continuous stochastic process as in: representing rainfall intensity and d is point in time. Suppose that 1(d) represents the Annual Maximum E[I<]= A'ME[I¡] (15) Rainfall Intensity (AMRI) of duration d, defined by the maximum value of moving average of width d The K(q) is function of the moment order. The of the continuous rainfall process. The random procedure is adopted to test the suitability of scale variable 1(d) has a cumulative probability invariant model to describe rainfall process, in here distribution, which is given by briefly described in Figure 1. The same moment Ap 1 are plotted on the logarithmic chart versus 1 (10) die scale X for different moments' order q¡. The Pr{ld (16) ÍJX» (11) Id = \DJ (17) D H Defining ^ = _ as the scaling ratio E[l«]=K(q)d < (18) d For many parametric forms, Equation (18) may be s H l(dY ' = Ä l{Ad) (12) expressed in terms of standard variant, as in where H is a non-integer scaling exponent factor. t~Md (19) Fdii) = F The equality " <**' " refers to identical probability distributions in = both side of the equations. Where F(.) is a function independent of d. Under The Equation (12) may be rewritten in terms this form, it can be deduced from Equation (18) that of the moments of order q about the origin, denoted q by E[/d ]; in these terms, the resulting expression is fid -A flxd (20) E[I<]= A*E[II,] (13) (21) 91 Log [moment] Simple scaling Multi-scaling Figure. 1 Simple and multiscaling in term of statistical moments. Fist step, moments of different orders q are plotted as function of scale in a log-log plot. From the slope, values of the function K(q) are obtained. If K(q) is linear, the process is simple scaling. If K(q) is non-linear, die proc ess is multiscaling. Substituting Equation (19), (20) and (21) into The simple scaling property, as formalized Equation (8) and investing with respect to /, one by Equation (17), can be empirically verified by obtains: replacing the population moments by the corresponding sample moments. On the other hand, MÁMY+ 92 1 10 100 Duration (hour) Figure 2 Relationship between sample moments of order q and duration The regression between moment E[Ixd ] and relationship confirms the hypothesis of " simple duration d, for 1 h to 24 h, may also be used to check scaling in wide sense", as defined by Equation (17). the validity of Equation (17). In fact, by represent ing the exponent by K (q) =-r\q and plotting it The slope of the regression line between K(q) and q against q, Figure 3 shows there is an almost perfect linearity between the two variables. Such a linear is TI=0.6058 for Hikone and T|=0.678 for Hirakata station, as an estimated for the scale factor. 0 I Hikone Station -0.5 ..__rSv^ -1 -1.5 •9 2" -2 -Z5 y = -0.6058X- 3.0103 -3 R2=0.99É » -"Ni > -3.5 Figure 3 Relationship between K(q) and the sample moment order q The IDF relationship for short duration (28) rainfall can be deduced form daily data by applying ld,T - H Equation (22) with T|=0.6058 for the Hikone station d and with the estimates of HD and a^ with D=24h. From 24-hour data collected at the Hilkone and: . M + (TF-1{\-1/T) (29) l recording gauging station, the sample of 21 years of d,T drj 24hours annual maximum rainfall intensity yields, the estimates /j,D.24 = 4.615 and (XD=24 ~ 2.604 . 31.56-17.8/»(-/»(l-l/r)) i = (30) Back with these estimates and the Gumbel inverse ,0.605 function, the deduced IDF relationship for the location of Hikone may be written as Equation (30) with/i=/lV= 31.56 and «r^V „=17.81 93 Another traditional way of constructed rainfall IDF The cumulative density function (CDF) curves by Equation (2) (Bernard Equation). Frequency analysis techniques are used to develop r (32) F (JC ) = 1 - exp a the relationship between the rainfall intensity, storm V duration, and return periods from rainfall data. P Analysis of distribution for rainfall frequency Where aand ßare the parameters. The EVI is based on the EVI (Gumbel) distribution. distribution used to calculate the rainfall intensity at The probability distribution function for EVI is different rainfall durations and return periods and the maximum rainfall intensity for considered a 'x-a^ (31) durations and 2, 5, 10, 20, 50,100 and 200 years /(*) = -exp ß return periods, have been determined. The set of IDF curves can be estimated by Bernard Equation shown in Figure 4. 10 15 Duration(hour) Figure 4 The Rainfall Intensity-Duration-Frequency (IDF) curves for Hikone station by Bernard Equation. The Rainfall Intensity Duration Frequency as calculated by Equation (30) and plots as the curves for Hikone station can be reconstructed be return period varies from 5 to 100 year returns. The scaling methods by Equation (30), it is show small scale factor T|, along with parameters a and ju in differences, with higher values for increasing return Equation (29), may be interpreted as periods. Figure 5 shows how the IDF relationships, regional climatic characterize. tc « x 13 |f x Cir*tor*eM: 94 Figure 5 The Rainfall Intensity Frequency Curves at Hikone station, Yodo Catchments, Japan by scaling method. Conclusions Chow, V.T., Maidment, D.R. & Mays, L.W. (1988). In the paper, a simple analytical formulation Applied Hydrology, McGraw-Hill. for rainfall IDF relationship, which utilizes the David M. Hershfied (1961). Estimating the Probable scaling behavior is presented. The proposed Maximum Precipitation, Journal of the approach is based on scaling properties of rainfall Hydraulic Division, Proceeding of the ASCE, time series. The hourly IDF curves were derived in HY5, pp. 99-116 a normalized form apart from EVI (Gumbel) Le Minn NHAT, Yasuto TACHIKAWA and Kaoru distribution fitted to the maximum rainfall intensity TAKARA: Establishment of Intensity- for several durations between 1 hour and 24 hours Duration-Frequency curves for precipitation in collected in 2 stations at Yodo catchments of Japan. the monsoon area of Vietnam, Annuals of The IDF curves for short duration (hourly) Disas. Prev. Res. Inst., Kyoto Univ., No. 49B, were derived from 24-hour data. The simple 2006 submitted. scaling property verified by local data; then IDF Menabde M., A. Seed & G. Pegram. A simple relationships are deduced from daily rainfall which scaling model for extreme rainfall. Water show good results as compared to IDF curves Resources Research, Vol. 35, No. 1, pp. 335- obtained from at-site short-duration rainfall data. 339, 1999. Kothyari, U.C. and Grade, R.J. (1992). Rainfall in References tensity duration frequency formula for India, J. Bell, F.C. (1969). Generalized rainfall duration Hydr. Engrg., ASCE, 118(2), pp. 323-336. frequency relationships. Journal of Hydraulic Koutsoyiannis, D., Manetas, A. (1998). A Div., ASCE, 95(l),pp. 311-327. mathematical framework for studying rainfall Burlando P. & R. Rosso. Scaling and multiscaling intensity-duration-frequency relationships, models of depth-duration-frequency curves for Journal of Hydrology, 206, pp. 118-135. storm precipitation. Journal of Hydrology, 187, P.S. Yu, T.C. Yang and C. S. Lin, "Regional rainfall p. 45-65, 1996. intensity formulas based on scaling property of Chen, C.L. (1983). Rainfall intensity-duration - rainfall", Journal of Hydrology, 295 (1-4), frequency formulas, Journal of Hydraulic pp.108-123, 2004. Engineering, ASCE, 109(12), pp. 1603-1621. Chow, V. T. (1964). Handbook of Applied Hydrology, McGraw-Hill, New York, pp. 1-1450. 95 Stochastic Modeling of Rainfall Maxima Using Neyman-Scott Rainfall Model Carlo Mondoftedo, Yasuto Tachikawa, Kaoru Takara Disaster Prevention Research Institute, Kyoto University, Uji, Kyoto 611-0011, Japan mondonedo(wflood. dpri. kvoto-u. ac. ip ABSTRACT When discharge records are inadequate in length, synthetic rainfall can be generated from stochastic tech niques to supply crucial flood control decision variables (following the rules of rainfall-runoff modeling) in de sign and operation of a flood control structure. The Neyman-Scott Poisson Rectangular Pulse Rainfall Model was applied here to generate synthetic rainfall maxima. Applying this model required historical moments of the rain fall in me study areas. Several test sets, consisting of different numbers of historical moments were prepared to estimate NSM parameters for the regional rainfall of Kamishiiba (Kyushu), Naha (Okinawa), and Sapporo (Hokkaido). Based on these NSM parameters, random numbers of different types were generated from the inverse CDF method to generate me synthetic rainfall of each area. The resulting maxima were adequate for the majority of cases except when a mix of rainfall sources was prominent in several months of die year. In general, an ideal set of parameters was determined but the methodical testing of combinations of moments employed in the proposed test sets was recommended for future applications in regions elsewhere. Introduction Independent Poisson Marks Model (IPMM) (Eagleson, 1972), Poisson Rectangular Pulse Flood control decision variables connected to Models (PRPM) (Rodriguez-Iturbe et al, 1987), and design and operation depend on an appropriate Clustered Poisson Rectangular Pulse Models availability of rainfall records. Ideally, such (CPRPM) (Rodriguez-Iturbe et al, 1987 and decision variables should be determined based on a Burlando and Rosso, 1993). The Neyman-Scott quantile of flood discharge. Unfortunately, Poisson Rectangular Pulse Rainfall Model (NSM discharge records may be short (less than 30 years) here, for brevity) (Rodriguez-Iturbe, 1987), a model and/or nonstationary due to land use change and/or from me latter type, was the key technique used in river intrusive construction. In mis case, synthetic this study. rainfall time series generation coupled with rainfall- All previously mentioned models were based runoff modeling proves to be a valuable tool. on me meory of point processes. Essentially, in For the general purpose of the former, stochastic point processes, probabilities can be mapped to techniques can be employed to generate syntiietic random occurrences of point events, rainfall in this rainfall and thus, synthetic quantiles. The case. For instance, bodi me IPMM and PRPM generation of such synthetic rainfall records is thus initiate rainfall arrivals as Poisson occurrences (see an effective decision-making aid to water resources Burlando and Rosso, 1993, for an example). engineers. Random variables such as duration and intensity of Several stochastic methods under me point rainfall are considered exponential in distribution. process (Cox and Isham, 1980) approach are The two models differ by the overlapping possible available for syntiietic rainfall generation. Such to the rectangular pulses of PRPM omerwise absent synthetic data can tiien be used in design storm in IPMM. Applications of mese models have evaluation for small retaining and impoundment appeared in the literature (Rodriguez-Iturbe et al, structures as well as sewer systems (Cowpertwait, 1987 and Burlando and Rosso, 1993) with the short 1996). For such purposes, several families coming that the models developed could not be of stochastic models are available such as me consistent witii more man one aggregation period. 97 Cluster arrival ^ cluster waiting time time cluster size (cells) + 4^ 3 cells T^. 4 cells time cell arri vals 3rd cell 'cell rd 4in cell 3 cell time 2nd cell time Fig 1. Schematic diagram of the Neyman-Scott Model. The models under the category of CPRPRM necessary for risk analysis in flood control decision are an application of point processes whereby the making. In so doing, longer NSM rainfall records, synthetic data generated can be consistent with more although synthetic, can be available as reliable bases than one aggregation level (i.e.: synthetic hourly and of quantile events. daily rainfall is consistent with historical The historical data used in this study were counterparts). In essence, rainfall arrives in clusters obtained from three locations to incorporate the of random arrival time and number, each cluster effect on the quantiles of rainfall generated by fronts consisting of rain cells of random birth, intensity, and/or typhoons. Sixteen yearly records (1988 to and duration. In addition, the Neyman-Scott model 2003) of hourly rainfall were taken from Kamishiiba (NSM) is a CPRPRM in which rain cells arrive (Kyushu Island), while 26 years (1976-2002) each subsequent to the arrival of a cluster's arrival. of hourly rainfall were taken from Naha (Okinawa) Unlike previous studies, the emphasis here is and Sapporo (Hokkaido). to investigate the ability of the NSM to preserve the historical quantile rainfall depths of 1-hour and 24-hour duration, which may be basic information 98 p[N=ri] = probability that the number of clusters Neyman-Scott Poisson Rectangular Pulse Rain N is equal to n fall Model ts =storm arrival time J(ts) = probability that the arrival of a storm Figure 1 show the random processes involved origin is ts in the concept of the Neyman-Scott model. This p[C=c] = probability that the number of cells basic version consists of essentially five probability of a storm C is equal to c distributions. In this NSM, clusters of cells He = mean number of cells in a storm are linked integrally to a storm origin with fltj) = probability that the arrival of a cell from mean occurrence rate X, regarded as a Poisson the storm origin is f¿ process, where waiting times between clusters are \lß = mean displacement of a cell from the exponential in X. The arrivals of these clusters are storm origin shown in the first time line of Fig. 1. Each storm X'c) = probability that the intensity of a cell can have a random number of cells described by a is equal to ic geometric distribution (with all clusters containing fa = mean intensity of a cell at least one cell), as shown in the second time line. X'c) = probability that the duration of a Relative to the cluster origin, the random arrival of cell's life is equal to tc each cell is based on an exponential distribution, MS = mean cell life span as shown in the third time line. Each cell has a corresponding independent identically distributed Inherent in the model is the assumption of (iid) random intensity and duration, also based on stationarity in the mean and variance. In applying the exponential distribution, shown in the fourth the model therefore, it would be beneficial to have and fifth time line, respectively. The total rainfall as long a historical rainfall record as possible. intensity is then the superposition of the effects of Based on the method of moments, the historical these random cell intensities, as shown in the sixth rainfall record can be expressed in terms of the time line. A succinct representation of the model parameters as (Rodriguez-Iturbe, 1987): previously mentioned distributions can be written as: E(YlM) = A(Mc)h/(öMx) (7) p(N = n). (1) 3 ./YM\ _ 4»/ -llß A(h)-o%(h)} UuA(h) m ; 1 1 W /(0 = l/Aexp(-AU (2) \ ' fr^ip-s ) M/S „fyw vM\ *4M.(M) AU2-ifc4(*,*)-*fe(»,*)]f9) iC = c] = (i-i/^r (3) in which: Sh Al(h)=Sh-l + e f(td)=ßexp{-ßd) (4) B,(h)=ßh-\ + e--Sh /('c)=1/>",eXP(-1/>",/c) (5) 1 ¿2(A,*) = 0.5(l-g-*)V*<*- > 2 (i 1) 52(Ä^)=0.5(l-e"^) ^ - f(tc) = Sexp(-Stc) (6) where: where: / = time interval counter v= mean number of occurrences h = integer specifying time step interval of data = XT; T is the time period in consideration (i for 1 hour, 24 for 1 day, etc.) À = mean arrival rate of a storm = rainfall depth in the i-th time of interval h 99 Five parameters are required in the NSM: X, 5, u« u„ and ß. As shown previously, these parameters = mean rainfall depth record at h-hours are directly connected to sample moments of the var(^(A)) historical rainfall records in the form of equations = variance of rainfall record at h-hours (7)-(9). Several nontrivial combinations of these equations are used to form systems to be solved nu cov = covariance or rainfall record at merically. Normally, these systems are solved for h-hours at lag k the required parameters by minimization of an ob jective function. In this study, the following objec Parameter Estimation tive function was used. Although the model parameters can be M f estimated by maximum likelihood, this study (10) followed previously adopted techniques from the w. method of moments. Indeed, a major limitation of 7=1 V the former method is the lack of the features of where: aggregated rainfall over a certain interval of time from the sample data sets, which was the case in this = j-th NS moment equation of rainfall study. Hence, the method of moments was adopted depth Y, (from equations (7) -(9)). here instead. WJ = historical moment value from rainfall record. M= number of equations to be adopted in the estimation. Table 1. Proposed test sets for NSM parameter estimation problem. Hours of Aggregation to be Used Test Set Moments 1 6 12 24 48 Mean O I Variance O O Covariance* O o Mean o II Variance o o o Covariance* o o o Mean o in Variance o O o Covariance* o o o Mean o IV Variance o o O Covariance* o o o Mean o V Variance o o o o Covariance* o o o o Mean o VI Variance o o o o Covariance* o o o o *Covariance at lag 1 This choice of (10) was made here to ensure and Napolitano (1999), and Favre et al. (2004), to that large numerical values do not dominate the cite a few. These combinations range from the most fitting procedure (Favre et al., 2004). To apply (10), basic (5 equations in the objective function to solve one must adopt a combination of equation (7) - (9), for the 5 parameters are used) to the more thorough depending on the target use of the resulting NSM. (more than 5 equations in the objective function to Such combinations include those used in the studies solve for the 5 parameters are used). For instance, of Rodriguez-Iturbe et al. (1987), Burlando and the determination the five parameters of the NS Rosso (1993), Cowpertwait et al. (1996), Calenda model can include the following equations, desig nated here as Test Set I: 100 1 .hourly mean of rainfall depth ((7) cast in h = 1 hr) where: 2.variance of hourly rainfall depth ((8) cast in h = 1 hr) E, = multiplier = 75 = 16,807 3.variance of daily rainfall depth ((8) cast in m = modulus = 231 - 1 = 2,147,483,647 h = 24 hrs) mod = modulus operator 4.1ag-l covariance of hourly rainfall depth Ij = previous random integer between 0 ((9) cast in h = 1 hr, and lag k = 1) and m-\ 5.1ag-l covariance of daily rainfall depth Ij+i = succeeding random integer between 0 and ((9) cast in h = 24 hr, and lag k = 1) m-\. For this study, the hourly and daily maxima The inverse CDF method (ICDFM) was used of the synthetic data should be sufficiently close to for generating continuous random variables while a those of the historical sample. Since the above- look-up table implementation of this method was mentioned set does not directly employ these used for generating discrete random variables (see maxima in the estimation of the parameters, it would Gentle, 1998). be appropriate to assume that Test Set I may not cover this requirement for all possible rainfall Application of the Neyman-Scott Rectangular conditions (i.e.: the temporal storm structure of Pulse Rainfall Model rainfall may vary by location and season). To cover possible dependencies of the target maxima on the Parameters were estimated on a monthly short-term and long-term moments of the historical basis in what was considered the season of heavy data (if such dependencies exist), six combination rainfall of June to October. This was done to test sets were adopted, as shown in Table 1. maintain the stationarity assumption of NSM. Three The minimization technique adopted here locations were selected to consider the effects of included a means to employ the constraints shown in rainfall in a typhoon-frequented area (Naha, Table 2. Napolitano and Calenda (1999), in their Okinawa), a front-frequented area (Sapporo, study on unbiased parameter estimates for the NSM, Hokkaido), and both (Kamishiiba, Kyushu). adopted these ranges. The application of mese Parameters obtained from each set for each month of ranges in this current study include the use of the each area were used to generate synthetic rainfall Nelder-Mead Simplex (Press, et al, 1992) with the exception of Sapporo, where Test Sets IV and Levenberg-Marquardt (Press, et al, 1992) and VI did not seem to yield reasonable parameter minimization techniques in tandem for initial and estimates (see discussion on Sapporo below). refined estimation, respectively. One hundred synthetic records were generated for each parameter set of each month. Table 2 Range of NSM parameters used for optimization. Parameter Min Max Synthetic moments were calculated from each set )i(l/h) ÔÔÔÎ ÖÖ50 and compared with the historical counterparts. For u, 2.0 100.0 b (1/h) 001 0 50 each test of each month, a search for monthly hourly Mmm/h) °3° 15° and daily maxima was conducted, enabling a n,(i/h) 010 yo Kolmogorov-Smirnoff (KS) test between historical maximum and synthetic maximum. The KS Random Number Generation tests used here was based on the set of equations (12) - (13). The KS probability here is a variant of To use the NSM parameters, a uniform the original KS test mat uses the Kuiper statistic, V, deviate generator (random numbers within (0,1)) defined in this application as the sum of the was developed from the method described in Press maximum distance of a cumulative frequency et al (1992) which made use of the Park-Miller distribution function of the synthetic maxima above "minimal standard" generator based on the simple and below the cumulative frequency distribution of multiplicative congruential algorithm: die historical maxima, yielding a more sensitive test at the extreme ends of the CDF. /,„=#, (mod m) <»> QM=2±(4jY-iy^ (i2) 7=1 101 ß«(o)=i. aggregated data. This weaker correlation structure (13) of the historical data at the 48-hour level seemed to affect the results of Test IV and Test VI more than ßßW=0 the Test V case. Appendix 1 and 2 contain partial historical moments and NSM parameter estimates The value of the KS probability PKs is then obtained, respectively. given in equation (14). High values PKs disproves The moments of the synthetic record were the null hypothesis; the two samples compared may compared to the historical records by calculating the originate from the same population. residual in equation (16). Ideally, for each synthetic record, one must obtain a value as close to zero as possible for the hourly and daily data. Generally, P (y>obsen>ed)=Q \jÑ~ +Q.l5S + 024lV^Jo) (14) KS KS l the residual of the synthetic daily records were larger than the residual of the synthetic hourly N-J^ (15) records, as shown in Table 3. Nh+Ns V where: 1- (16) D = maximum deviation between cumulative 4-1 frequency distributions of historical and V synthetic rainfall maxima. where: Nh = length of historical record. R¡ = residual of the 1-th hourly aggregated data Ns = length of synthetic record. SM;, y = Mean of synthetic rainfall record at the 1-th aggregation level = This version of the KS Test was taken from Press, et SMij Variance of synthetic rainfall record at al (1992). the 1-th aggregation level Parameter Estimates SMif3 = correlation at lag-1 of synthetic rainfall In general, the value of the objective function record at me 1-th aggregation level (10) was quite higher when it was adopted in HMij = Mean of historical rainfall record at estimation of NSM parameters in Test IV, V, and the 1-th aggregation level HM/2 = Variance of historical rainfall record at VI. It was observed that the correlation coefficient the 1-th aggregation level of the historical data dropped from about 0.6 at the HMi3 = correlation at lag-1 of historical original hourly level to around ±0.1 at the 48-hourly rainfall record at the 1-th aggregation level Table 3. Residuals of synthetic moments to historical moments KAMISiniBA JUNE JULY AUGUST SEPTEMBER OCTOBER TEST HOURLY DAILY HOURLY DAILY HOURLY DAILY HOURLY DAILY HOURLY DAILY I 0.000 0.006 0.000 0.001 0.001 0.002 0.001 0.007 0.017 0.020 n 0.001 0.004 0.000 0.001 0.001 0.001 0.001 0.007 0.023 0.001 m 0.005 0.007 0.002 0.005 0.004 0.004 0004 0.005 0.013 0.026 rv 0.013 0.427 0.001 0.042 0.003 0.005 0.000 0.016 0.001 0.001 V 0.000 0.025 0.008 0020 0.003 0.021 0.001 0.005 0.039 0.168 VI 0016 0 080 0.003 0.037 0.001 0 051 0.001 0.002 0.026 0.057 NAHA JUNE JULY AUGUST SEPTEMBER OCTOBER TEST HOURLY DAILY HOURLY DAILY HOURLY DAILY HOURLY DAILY HOURLY DAILY 0.012 0.031 0.006 0.006 0.001 0.002 0.012 0.013 0.004 0.011 n 0.012 0.010 0.006 0.050 0005 0.012 0.003 0.004 0.004 0.008 m 0.003 0.004 0.005 0.064 0.012 0.015 0.001 0.004 0.007 0.029 0.012 0.032 0.006 0.017 0.011 0.046 0001 0.001 0012 0.034 Vrv 0.005 0.018 0.003 0.153 0.004 0.027 0.002 0.006 0.002 0.005 VI 0.006 0.257 0.003 0.052 0.034 0.159 0.001 0.025 0.002 0011 SAPPORO JUNE JULY AUGUST SEPTEMBER OCTOBER TEST HOURLY DAILY HOURLY DAILY HOURLY DAILY HOURLY DAILY HOURLY DAILY I 0.243 0.234 0.001 0.002 0.001 0.095 0.005 0.083 0.002 0.002 n 0.230 0.173 0.011 0.160 0.001 0.012 0.001 0.031 0.002 0.018 m 0.241 0.354 0.006 0.002 0.001 0.004 0.000 0006 0.001 0.017 rv -not at 3plied- -not applied- -not applied- -not applied- -not applied- V 0.005 0.026 O.OOl] 0016 0.003| 0.054 0.0051 0.052 O.OOll 0.043 VI -not applied- -not applicd- -not applied- -not applied- -not applied- Kamishiiba, Kyushu (1988-2003) 102 The Kyushu Region exhibits mixes of rainfall It was predetermined that a 95% value of the sources, although typhoons are more erratic in KS probability was a practical value for disproving occurrence in the summer. Several rainfall moments the null hypothesis (although other tests can be are shown in Appendix 1, taken from the employed). Thus, at 95% KS probability, the Kamishiiba Observatory in Kyushu. As expected, maximums of synthetic data appear to come from the residuals between synthetic and historical the same population as those of the historic data. moments for this area are quite small, shown in The results of hourly and daily synthetic maxima of Table 3. This indicates the ability of NSM to the study areas were grouped into three categories preserve the historical moments in the generated (see Figure 2): synthetic rainfall. This applies to all the months of (1) Both hourly and daily maxima display a KS the rainy season in the area. This was the original probability PKS ~ 95%. objective in the conceptualization of the model and (2) Both hourly and daily maxima display a KS was considered (in this study) secondary to the probability PKS < 95%. effectiveness of the model to yield synthetic maxima (3) Only the hourly or daily maxima display a KS similar to the historical maxima. probability PKS ~ 95%. Hourly Maximum Rainfall Daily Maximum Rainfall j 0 40 ¡ 0 30 E 0 20 O 0 10 0.00 Rainfall (mm) Rainfall (mm) (—Synthetic Record —Historical Record] 1—Synthetic Record —Historical Record | Jul Test I (PKS = 0.99536) Jul Test I (PKS = 0.999) Case (1) Hourly Maximum Rainfall Dally Maximum Rainfall ¡ 0.10 0.00 100 150 Rainfall (mm) Rainfall (mm) |— Synthetic Record —Historical Record] j—-Synthetic Record —Historical Record! Jul Test IV (PKS = 0.88421) Jul Test IV (PKS 0.50603) Case (2) Hourly Maximum Rainfall Dally Maximum Rainfall 1 00 £ 0 90 f, 0.B0 y o 70 £ 0.60 a 0.50 I 0.40 I 0 30 I 0 20 g 0 10 0.00 Rainfall (mm) [—Synthetic Record —Historical Record] |— Synthetic Record —Historical Record | Jul Test IV (PKS = 0.99946) Jul Test IV (PKS = 0.86387) Case (3) Fig. 2. Sample KS plot of hourly and daily rainfall maxima from Kamishiiba. Results are grouped into three major categories such that Case (1) results show KS probabilities > 95% for both hourly and daily sets of maxima, Case (2) results show KS probabilities < 95% for both hourly and daily sets of maxima, and Case (3) results show KS probabilities > 95% for at least the hourly or the daily set of maxima. 103 Case (1) KS results were the ideal for this result is in the mixed rainfall sources of this period. study, when both synthetic moments and maxima Essentially, in these months, rainfall events tend to agree with the historical counterparts. Case (2) be convective pockets of short bursts with possible results were considered total failure of the NSM to medium to high intensity. It is possible within model the historical maxima. Case (3) results were these episodes to concur with a passing of a considered partial failures as one time scale's typhoon, complicating the temporal rain cell struc synthetic moments and maxima agree with its his ture of the actual cluster. It is this mixed source of torical counterpart, while another time scale fails to rainfall that the current NSM may not be able to model historical maxima. Both cases (2) and (3) detect, causing it to fail in this respect. Thus, were considered unacceptable for the purpose of this although the moments of the overall historical study. record is not affected (Table 3), specially in Tests I, While most tests yielded maxima of passing II, and III, the varying effect of this mix of rainfall KS probabilities, the majority of cases (2) and (3) origins appeared to affect the results of the KS appear in the main summer months of August and probabilities of these months (Table 4). September. A possible explanation for this odd Table 4. Kolmogorov-Smimoff Tests for synthetic maxima. KAMISHHBA JUNE JULY AUGUST SEPTEMBER OCTOBER TEST HOURLY DAILY HOURLY DAILY HOURLY DAILY HOURLY DAILY HOURLY DAILY I 0 998 0.998 0.995 0 999 0.941 0.927 0.953 0.793 0.997 0.982 II 0.994 0.996 0.999 0 999 0874 0.806 0.894 0.793 0.884 0.506 III 0.884 0.996 0.999 0 997 0.830 0.853 0.972 0.842 0 995 0.999 rv 0.449 0.991 0.999 0.864 0.382 0.853 0972 0.565 0.726 0.998 V 0.994 0.998 0.998 0 988 0 609 0.853 0.985 0.595 0 999 0999 VI 0.968 0999 0.999 0.994 0.726 0.853 0993 0.842 0.999 0.948 NAHA JUNE JULY AUGUST SEPTEMBER OCTOBER TEST HOURLY DAILY HOURLY DAILY HOURLY DAILY HOURLY DAILY HOURLY DAILY 0.844 0.126 0.743 0.601 0 999 0 976 0999 0.999 0.536 0.965 n 0.993 0.241 0.957 0980 0 999 0 999 0999 0.928 0.630 0.993 m 0.949 0.481 0.994 0.979 0.999 0 998 0.999 0.972 0.354 0.966 rv 0.982 0.196 0.957 0 999 0.991 0 856 0.996 0.959 0.920 0.579 V 0.936 0697 0.992 0.816 0.999 0 994 0999 0.959 0.927 0 723 VI 0982 0.077 0.999 0 996 0.772 0.798 0.995 0.971 0.920 0848 SAPPORO JUNE JULY AUGUST SEPTEMBER OCTOBER TEST HOURLY DAILY HOURLY DAILY HOURLY DATLY HOURLY DAILY HOURLY DAILY 0.997 0.805 0.995 0.966 0 998 0.953 0.993 0.909 0.603 1.000 II 0.891 0.971 0 998 0.999 0.998 0.406 0 964 0.957 0.999 0 999 m 0.947 0998 0.973 0.999 0 999 0.953 0.999 0.971 0.999 0.947 rv -not applied- -not applied- -not applied- -not applied- -not applied- V 0.999| 0.944 0.999| 0 999 0.999| 0.544 0999| 0997 0.999| 0.999 VI -not applied- -not applied- -not applied- -not applied- -not applied- A reformulation of the NSM, following previously, as the region experiences more the concept proposed by Cowpertwait (1994), may rainfallin August than in September (refer to rectify the poor results for the August and Appendix 1), although typhoons tend to be more September synthetic rainfall of Kamishiiba. This frequent in the latter (based on the Japan Meteoro formulation includes "light rain cells" of long logical Agency). The effect of the mixed sources expected duration and "heavy rain cells" of short can thus be considered more prominent in August expected duration. Known to be consistent with rain than in September, yielding the failing case (2) field observations, the inclusion of these cells in results. Cowpertwait's (1994) version of NSM may be more In the months when the KS tests are appropriate for these mixed conditions. successful, the information supplied to the The August KS probabilities within this parameter estimation is minimal (Test I and II), or region, the area with the highest rainfall averages maximal (Test V), the model generates case (1) among this study's study areas, are all of case (2). results. This is an indication that the rainfall, when The KS probabilities of September on the other hand it is more consistent in origin such that the current exhibit both the case (2) and case (3) results. cluster structure is valid, tends to be correlated well This supports the source-mixing concept described within a possible two-day period. 104 Naha, Okinawa (1976-2002) tends to be the month when rainfall maxima origin The Okinawa Region is quite known to be tends to be mixed. In this region however, a case more typhoon-frequented than the remaining parts (1) type KS probability can be observed on every of Japan. In the periods when the seasons change, month, witii the exception of June. In this period say from Spring to Summer (May to June), as well however, rainfall is least frequent (with an average as Summer to Autumn (October to November), it magnitude of 2 mm per day) and was not considered may exhibit mixes of frontal and typhoon rainfall. a crucial month. It was thus excluded in the Similar to the Kamishiiba results, certain analysis of the results. months of the synthetic record generated for Naha It was not possible for most cases to apply display poor KS probabilities, spread out evenly in the NSM at the Test IV and VI condition. At this the case (2) and case (3) categories. However, test's level of aggregation, 48 hours, the rainfall was these two months in question, July and October, are considered practically weakly correlated. This is separated by three months where the model yields because the historical data in this region tends to case (1) results quite often. Why this occurs is a show a change in sign of correlations (with a very crucial change from the Kamishiiba case. small magnitude) at this aggregation level. For this Throughout the July-September period, reason, the Test IV and VI results were determined rainfall was considered to be of a consistent unreliable for the Hokkaido Region. source (possibly convective) and thus, the current In mis region one test condition always formulation of the NSM used here was adequate for yielded reliable synthetic maxima, Test III in this modeling maximums. The October synthetic case. It was possible to have more case (1) results in maxima can be explained similarly to the some months than others though. However, the Kamishiiba case, in which summer convective assumption of stationarity in mis region, due to less rainfall episodes are mixed with the passages of erratic typhoon episodes, is quite valid, and thus the typhoons. However, as the second quarter of the target result in which a NSM can correctly model year is considered the transition from the cold to the maxima was obtained. For this reason, it was the warm season of Japan (based on the Japan also considered more practical in the future to merge Meteorological Agency), the Naha Region possibly data in mis region when rainfall occurs, rather than experiences a mix of the convective pockets and the to model rainfall in months. In this case, it would be passage of cold fronts in June. The effect of this crucial to be aware of the precipitation conditions as passage is similar to the October results in the snow sometimes begin in October in this region summer of Naha, and the August-September results (although infrequently), for which the NSM was not of in the summer of Kamishiiba. conceptualized. The mixed sources of this region are fairly Upon supplying the medium-term moments less pronounced as the case (2) and (3) results are of Test III (12-hour moments), the NSM becomes spread out evenly throughout the two months in most effective. Although the ideal test set was question. However, the same reformulation similar in this region and in the previous region, proposed in the Kamishiiba case is considered these resulting tests may be considered isolated and appropriate here as well. With the subclassification coincidental. Further testing would prove to be of rain cells in the mentioned reformulation, it appropriate at this point. would be possible to accommodate the mixing effects of both June and October. Ideal Test Case The NSM tends to be consistent in the Test II In essence, the exercise showed that a range and III case for mis region. This indicates that the of parameters, varying by the degree of information model becomes more effective when short-term supplied in the form of moments, should be moments (6-hour and 12-hour moments) were sup prepared in estimating NSM parameters for the plied. It is possible that the general tendency of purpose at hand. It is not clear at this point that a aggregated rainfall in this region, without mixed- single test set is applicable to all regions. In sources of rain cells, tends to be correlated well addition, even more sophisticated test sets can be within a possible daily period. supplied by using higher order autocorrelations or the so-called "zero-probabilities" or periods with no Sapporo, Hokkaido (1976-2002) rainfall (see Cowpertwait, 1996). The Hokkaido Region of Japan is known to be less affected by typhoons. Indeed, only June 105 Concluding remarks References 1. The Neyman-Scott Rectangular Pulse Rainfall Model, or NSM, was applied in this study to model Burlando, P. and Rosso, R., 1993. Stochastic mod rainfall maxima of certain regional rainfall in Japan: els of temporal rainfall: Reproducibility, Esti Kamishiiba (Kyushu), Nana (Okinawa), and mation and Prediction of Extreme Events, Sto Sapporo (Hokkaido). chastic hydrology and its use in water 2. Since the information used to estimate the resources systems simulation and optimiza parameters of NSM did not include the historical tion. (Marco, J. et al, ed.). NATO ASI Series. maxima, several sets of parameters, varying by the Kluwer Academic Pub. degree of information supplied (in terms of the Calenda, G. and Napolitano, F., 1999. Parameter number of historical moments) was prepared. estimation of Neyman-Scott processes for tem 3. Based on these NSM parameters, random poral point rainfall simulation. Journal of Hy numbers different types were generated from the drology Vol. 225 pp. 45-66. inverse CDF method to generate the synthetic Cowpertwait, P. 1994. A generalized point proc rainfall. ess for rainfall. Proc. R. Soc. London A 447, 4. Kyushu and Okinawa exhibited periods when pp. 23-27. rainfall is a mix of origins of frontal, convective, Cowpertwait, P. et al, 1996. Neyman-Scott model and typhoon type. The presence of this mix could ing of rainfall time series: 1. Fitting procedures not be accommodated in the current formulation of for hourly and daily data. 2. Regionalization NSM. and disaggregation procedures. Journal 5. Regardless of period or region, the moments were of Hydrology Vol. 175 (1-4), pp. 17-46, 47-65. modeled effectively by NSM. This was the original Cox, D. and Isham, V., 1980. Point processes. purpose of die model conception and was treated Chapman and Hall, London. here as a secondary objective. Cox, D. and Isham, V., 1998. Stochastic spatial 6. In all cases, Kolmogorov-Smirnoff Tests were temporal models for rain. Stochastic meth used to check whether die null hypothesis should be ods in hydrology: rain, landforms and floods. ignored. The KS probability of 95% was taken to (Barndorff-Nielsen, O. et al, ed). World Scien disprove the null hypotíiesis, it proves Üiat two tific. samples belong to the same population. Eagleson, P., 1972. Dynamics of flood fre 7. The KS Test results were grouped into three quency, Water Resources Research, vol 8, no. 4, cases: pp. 878-898. (l)Both hourly and daily maxima display a KS Favre, A. C, et al, 2004. Unbiased Parameter Esti probability PKS " 95%. mation of die Neyman-Scott Model for rainfall (2) Both hourly and daily maxima display a KS simulation with related confidence interval. probability PKs < 95%. Journal of Hydrology Vol. 286, pp. 168-178. (3) Only die hourly or daily maxima display a KS Gentle, J., 1998. Random number generation and probability PKS " 95%. Monte Carlo methods, Springer-Verlag, New York. 8. As a result of the periods of "source-mixing" in 3, Press, W. H. et al, 1992. Numerical Recipes in For die KS Tests of several months appeared to have tran, 2nd Ed. Cambridge University Press, Lon case (2) and (3) results. For Kamishiiba, these don. months were determined to be August to September. Rodriguez-Iturbe, I. et al, 1987. Some models for For Naha, diese montiis were determined to be June rainfall based on stochastic point processes. and October. Proc. R. Soc. London A 410, pp. 269-288. 9. It may be possible to capture die effect of this "mixed sources" by incorporating richer autocorrelations in the estimation procedure, or by using expressions for dry probabilities (Cowpertwait, 1996). 106 Appendix 1. Partial list of historical rainfall record moments. Area/Moment Month Kamishiiba June July August September October Hourly mean 0.7765 0.5826 0.5654 0.4694 0.1699 Hourly variance 2.5060 2.5315 2.5881 2.3487 1.1388 Hourly correlation, lag 1 0.6714 0.6716 0.6949 0.7207 0.7170 Dairy mean 18.6354 13.9819 13.5685 11.2667 4.0786 Daily variance 32.9315 37.5997 37.8424 34.0762 14.2534 Dairy correlation, lag 1 0.1516 0.3480 0.3805 0.3165 0.3324 Naha June July August September October Hourly mean 0.2528 0.2019 0.3326 0.3529 0.2100 Hourly variance 1.7434 1.7456 2.1507 2.4051 1.7380 Hourly correlation, lag 1 0.3812 0.4893 0.5124 0.5917 0.5706 Daily mean 6.0679 4.8459 7.9821 8.4704 5.0406 Daily variance 15.8101 15.9282 24.2758 29.2576 19.5216 Daily correlation, lag 1 0.1995 0.3765 0.2177 0.3303 0.2242 Sapporo June July August September October Hourly mean 0.0975 0.0975 0.1714 0.1807 0.1575 Hourly variance 0.6983 0.6983 1.0955 1.0267 0.8369 Hourly correlation, lag 1 0.5092 0.5177 0.6363 0.6844 0.6362 Daily mean 2.2593 2.3393 4.1135 4.3358 3.7802 Daily variance 6.9383 7.1159 14.2919 12.7698 10.0246 Daily correlation, lag 1 0.1010 0.0955 0.1375 0.1004 0.1206 107 S 1 -9 2 J¡ 4 JS o I 3 ja 3 O i « 08 g es Ë s a Sapporo, ido là ! S S S ce H > S C t- «•4 P d H S > H > a H S S H B fe S H e Vi d S ^ H S v\ d fe Ed v\ > f- ? i s | TEST | TEST EST EST | TEST EST EST r» CO CO d o o ON CO O d o o px CS VI d o o o CO es O O Tt T d o o r» CO _H O ON px r- d o NO d o o o CO es d o d o d o es es o o d ON d o (S ri o o es Tt CO »n o © d o O ON NO ON o CO o ri »n o o px < d o (S o 0 Tt ri vi ON © o cs Tt CO d ON CO ^ r» o -^ px ON o ,—1 d »— o Tt d — r- NO es O Tt »n o NO Tt d o (S px NO O o Tt d o co rs »n Tt CS ON VO NO o CS TI co co »n d o m o o o VO o CO O Px m d O o •—' NO d ri Tt o r- d •—' NO NO px m d »-* NO VO o r- d in px CO CS O »-H H M ri ^ NO px in ri px CS ON CO ON CS o ri ON n Tj- o Tt o ri r- cs ri ri px O ON P- m »n ri ON o ri CO ri o O CO ON o CS VO CO NO ri cs ~ px NO vi vi p o o es' Ti CO es ri ON »n o o o o o ri CO px o m ri VO ro o o vi o o o o o m px px CO ri »n F- o '—' co o f- px' CO ^H VO Tf cs »n »n Tt vb px O »ri VO NO TI VO o Tt O Ol ON VO Tt- ON r- o TJ- »ri V) co NO co o vi co ri co px ,_i vi co px o ON m o px CO O O px es od "-' Ol Tt o ON vi »-* o px CO NO CS ^H in in »n CS Tt- ON" NO T NO CO d »—' o VI CO Oi CO es NO »n o px ri OÑ Tt zL - r^ SS 7 ~ "co" ~ ~ — öT ~ ~ _Ë= •s? ~ CO ON NO NO Tt o CO •—' px o p- *°. CN i- o -* VI o vo O and Sediment Transport from Watersheds Guillermo Q. Tabios III, Ph.D. Department of Civil Engineering and National Hydraulic Research Center, College of Engineering, University of the Philippines, Diliman, Quezon City 1101, Philippines ABSTRACT This paper presents results of a simulation study to study the effect of storm rainfall movement on soil erosion and sediment transport from watersheds. The moving storm rainfall fields are stochastically generated using a Neyman-Scott process, space-time rainfall model. This model is used to generate spatially- and tempo rally-varying storm rainfall fields at different speeds and directions. For the different rainfall fields generated, the amount of soil erosion from overland flow planes and amount of sediment transported in the channels are calcu lated using the KINEROS2 model of the Agricultural Research Service of the U.S. Department of Agriculture. The KINEROS2 model is an event oriented, physically-based model to simulate die various watershed processes especially interception, infiltration, surface runoff and erosion from overland flow planes and channels for a given storm rainfall event. Two different watershed shapes are considered: a del-shaped watershed and an elongated watershed. This study shows that there are differences in flow and sediment hydrographs at different speeds and directions of the storm rainfall movement for the two watershed shapes considered. 1. INTRODUCTION spatially and temporally-varying storm rainfall- Aside from catchment characteristics such as fields at different speeds and directions. For the size, geometry, topography and land use, it is well different rainfall fields generated, the amount of soil recognized that the shape and distribution of erosion from overland flow planes and amount of streamflow hydrograph depend on the spatial and sediment transported in the channels are calculated. temporal dynamic characteristics of rainfall. Two different watershed shapes are considered: Subsequently, the erosion and sediment transport a del-shaped watershed and an elongated watershed. process is influenced by the rainfall dynamics. In particular, the effect of rainfall dynamics on the 2.WATERSHED STORM HYDROGRAPH, hydrograph as well as erosion and sediment EROSION AND SEDIMENT TRANSPORT transport is basically in the form of mass, MODELING momentum and energy inputs that are function of The KINEROS2 model is used in this study rainfall intensity, duration and spatial extent. to calculate the watershed storm hydrograph and This paper investigates the effect of storm associated erosion and sediment transport rates from rainfall movement on soil erosion and sediment the watershed. This model was developed by the transport from watersheds. In this study, the Agricultural Research Services of the U.S. KINEROS2 model of the Agricultural Research Department of Agriculture. This model is a Service of the U.S. Department of Agriculture is single-event, physically-based model to simulate the used which is an event oriented, physically-based following various watershed processes: 1) intercep model to simulate the various watershed processes tion; 2) infiltration; 3) surface runoff; and, especially interception, infiltration, surface runoff 4) erosion from overland flow planes and channels. and erosion from overland flow planes and channels In this model, the storm hydrograph routing is by for a given storm rainfall event. In particular, soil kinematic wave equation through rectangular erosion from overland flow planes is due to splash overland flow planes and trapezoidal channels. The or inter-rill erosion caused by raindrop energy and soil erosion from overland flow planes is due to hydraulic or rill erosion caused by flowing water. splash erosion caused by raindrop energy and The moving storm rainfall fields are stochastically hydraulic or rill erosion caused by flowing water. generated using a Neyman-Scott process, space-time The finite difference method is used to solve the rainfall model. This model is used to generate overland flow, channel flow, erosion and sediment transport equations. For details of this model, the 109 reader referred to the website of the Southwest Wa in terms of bifurcation ratio and channel length tershed Research Center of the Agricultural Re ratio, respectively common to small watersheds. search Service of the U.S. Department of Agricul Also considered in the network construction is that ture. the drainage density of about 1.5 per km and that the In KTNEROS2, the overland flow and channel channel frequency (ratio of channels and watershed flow equations are based on kinematic wave routing. area) is 70 percent of the drainage density. In this In particular, the continuity equation is written as: case, a total of 17 streams (or 34 overland flow planes) were created for the del-shaped watershed as shown in Figure 1 and a total of 15 stream (or 30 ^ + M =iL dt dx overland flow planes) were created for the elongated The momentum equation simplifies to: watershed as shown in Figure 2. Both watersheds 2 So = Sf have a drainage area of 259 km . Also indicated in these figures are the in which Q is discharge, A is the flow area, qL is The channel slopes also follow Horton's law lateral inflow, S0 is bed slope and SL is friction slope. of stream slopes which requires mat stream slope The continuity equation can be written homogene within a channel order are approximately constant ous in Q or A using the equation: and they increase with decreasing channel order and that the overland flow plane slopes are related to the Q = aAß adjacent channel slopes according to Strahler For instance, based on the Manning's equation: (1950). and for a rectangular channel. The Manning's roughness coefficient for the The erosion and sediment transport in overland channel is assumed to be 0.04 while that roughness flow plane and channel are described as follows. in the catchment is 0.30. The overland flow plane The sediment continuity equation in the overland channels are assumed to be rectangular. In both flow plane is given by: overland flow plane and channel, three sediment sizes are considered with sizes 0.005, 0.05 and 0.25 dAC , 8QC es(x,t)-eh(x,t) = q(x,t) mm. dt dx — • Del-ShanpH M ""! u **'* L \\-l ?st \\J A. __._!_ -J v< £ > ^-! - s^ where Q is discharge, C is sediment concentration, A » _/._5 \1 [\.H >'V _ 7 rL U >j/¡ i S'1 /l S •*~ ->3r / ,' j 4 ^3/ «** _, is flow area, es is splash erosion rate, eh is hydraulic L ?----5"II. ! V/r '' erosion and q is lateral sediment inflow. ii V r X \T\.\¿%*,- ; j The || tfûb- \ fl * "Ï •>l*f25SÍ M t chh 2 splash erosion term 7 14 S.-' , \ es~cf-cf e is written as: 11 \-~<\ iz ^-"i: y.*t-t in S""3"• !* j "* *"• f ad! ia Z • *SV "" i -7 in which r is rainfall, h is water depth, c/ is a con ^1 stant related to soil and surface properties and ch is a •* Storm I o^h—_^. coefficient representing the damping effectiveness or^n díimrlinn wT of surface water. 0 2 4 t t 10 2S 1* 14 11 11 20 22 24 The hydraulic erosion (deposition if C > Cm) term is the _ rate of exchange Figure 1 Del-shaped watershed configuration of sediment eh c#( m ' between flowing water and soil over which it flows. where Cm is concentration at equilibrium transport capacity (i.e., modified Engelund-Hansen equation) and cg as coefficient. 3. CATCHMENT DESIGN OF WATERSHEDS For this study, two particular basin shapes are created referred to here as del-shaped watershed and 110 in 1» u ai c y^\ ~s Elongated is the time of birth of a raincell, is the " 71 • S 10 S-^ ï-,^^^ f Xciy ) n ri ^ y- c Zt VÏ tfr ^ 259 sq. kms location of raincells at birth, is the cen AÍ ih !4 '-, A tt?^r^ troid of the CPR, *' y is the rainband velocity : !, '$ TO 7 30 overland flow planes 20 - - fi '¡Ê V " i,22- B r <• 'F™ \ ^ i 1 2^ j 3 sediment sizes: relative to the fixed spatial origin on the ground, and \l/'r (0.005,0.05,0.25 mm) Ac- 17 r - (i I[SS * is a specified mean raincell duration. In the *5 v. SZ¿ S p 12 P-. 3 Jl ' 2 tl •''¡I/ f'f } f { } 10 i ÏC »f y 7 above equation, and can take the • ll*v following form: 1^|3 / 1/ Í3 ' T J 4 AM -,£ ; ¿ Storm 2 3 E jfTTL. 0 2 4 1 S 10 12 14 1t 11 Figure 2 Elongated watershed configuration. in which is the cellular birth rate and w is the 4. STOCHASTIC, SPACE-TIME RAINFALL cluster spread factor. MODELING OF STORM RAINFALL Associated to each raincell is its rainfall In this study, the storm rainfall is generated 8 r intensity from a cell of age and at a radius using the stochastic, space-time rainfall model based on Neyman-Scott process. This model was i(S,r)=io ea8e?/2d for S>0andr=0 developed by Gupta and Waymire (1979) and subsequently improved by Waymire et al (1984). In this model, the storm systems referred to as from the cell center assumed to be of the form: rainbands arrive as a Poisson process with mean 10 parameter . The location of the rainband is where is the rainfall intensity at the cell center relative to some fixed spatial origin at the ground at the time of its birth, a is an attenuation XM YM with specified centroid [ and m rectangular coordinates]. A rainband contains a coefficient in time, and D .is an attenuation distinct number of cluster potential regions (CPR) coefficient in space. The rainfall intensity is and that the centroids of these CPR constitute a assumed to be spatially symmetric around the cell center and its intensity exponentially decays in time. spatial Poisson process with parameter PL relative to the fixed spatial origin of the rainband. To obtain the rainfall depth ' ° at Associated to each CPR are raincells that are (i-l)AT /AT randomly bom in time and located relative to the some fixed tune interval, to , and centroid of the CPR. The number of raincells e -,•<•>/ ID' y R ,( x o. y o ) ds associated to a CPR is random variable which ( i - n 4 T is assigned a Poisson distribution with mean rate E[v] . Raincells within a moving rainband are ! ! ! ! r (-) = (x-x,y + (y-y0) + [2u,(x-xo) + 2Uy(y-yJ]S + (¿+u y)s born in time and space according to the probability density function: [x ,y ] at some point in the ground located at 0 0 Eq.(8) can be integrated such that where is the time of arrival of the rainband, 111 The rainfall data used to estimate the 5. SIMULATION RESULTS parameters of the space-time model are daily rainfall Figure 3 shows the hydrograph and sedigraph from storms during the typhoon months of July, at overland flow plane 12L of the del-shaped basin August and September in the Philippines. Storm (see Figure 1) for different storm directions (storm rainfall data from 1984 to 1992 are available at five speed = 10 km/hr). The storm in the upward rainfall stations in the vicinity of Pasac Delta area in direction result in the highest peak flow and the Pampanga Province of the Philippines. The sediment yield. However, as shown in Figure 4 parameters of the space-time rainfall model were depicting the hydrograph and sedigraph at the outlet estimated by method of moments except for the of the del-shaped watershed, the storm in the A direction across the watershed results in the higher mean parameter of the rainband arrivals and peak flow and sediment yield. Figure 5 shows the sediment-discharge rating curve at the outlet of the Ac the mean cell duration which were estimated del-shaped watershed which likewise shows the directly from the data. The Newton-Raphson storm across result in the highest flows while the nonlinear least-squares method was used to solve storm upward result in the lowest flows. the moment equations based on the station means, Figure 6 shows the hydrograph and sedigraph variances and covariances between stations. Table 1 at overland flow plane 9R of the elongated below shows the parameters of the space-time rain watershed (see Figure 2) for different storm speeds fall model. (storm upward direction). Figure 7 shows the corresponding hydrograph and sedigraph as well as Table 1 Space-time rainfall model parameters Figure 8 shows the sediment-discharge rating curve at the outlet of the elongated watershed. In this Parameter Value case, the storm with the lowest speed of 5 km/hr Xu (mean rate of storm arrivals) 0 225 hr"1 result in the highest peak flow and subsequently 2 PL (mean number of CPR) 0 4654 CPR/km high erosion and sediment transport rates. EI v ] (mean number of cells per CPR) 1.7844 cells/CPR Table 2 summarizes the peak flows and cr,,ay (cluster spread factor) 0.7278 km sediment yield for the del-shaped and elongated ß (mean raincellu birth rate) 1.9569 hr' watersheds for storm of different speeds and a (attenuation of rain lifespan) 1 5862 hr"' directions. It is shown here that the storm in the D (attenuation of rain spatial extent) 4.1902 km upward direction result in the higher peak flows and ¡0 (rain intensity) 31.196 mm/hr sediment yields. In investigating the effect of storm Ac (mean raincell duration) 0.7 hr speeds, the storm with lowest speed likewise result in the highest peak flow and sediments yield. For purposes of this study, three storm directions are investigated which are: 1) upward direction; 2) Dd-shaped Watershed Hydrograph and Sedigraph at Overland Flow Plane 12L across direction; and 3) downward direction.. These for different storm directions (storm speed = 10 km/hr) are parameterized in the rainfall model through the spatial origin of the rainband ' as well as the storm speeds . The storm origins for the three storm directions are indicated in Figures 1 and 2, for the del-shaped and elongated watersheds, respectively. The effect of varying storm speeds is also investigated in this study for storm speeds of 5 kph, 10 kph, and 25 kph. 0 4M »00 1200 1100 In all cases investigated here, the vertical and Tim» (ménutM) horizontal extent of rainbands are taken equal to Figure 3 Hydrograph and sedigraph at overland 100km thus representing a synoptic storm with flow plane 12L of del-shaped basin (see Figure 1) spatial scale of about 10,000 km2. for different storm directions (storm speed =10 km/ hr). 112 Del-shaped Watershed Elongated Watershed Hydrograph and Sedigraph at Outlet Hydrograph and Sedigraph at Outlet for different storm directions (storm speed = 10 km/tar) at different storm speeds (5,10 and 25 km/hr), upward direction. Storni Spaad f kmffir - - 8torm SpMd 10 hmJhr Storni 8|M«d 2i kmjhr Tims (minutos) Figure 4 Hydrograph and sedigraph at outlet of del- Figure 7 Hydrograph and sedigraph at outlet of shaped watershed (see Figure 1) for different storm elongated watershed (see Figure 2) for different directions (storm speed =10 km/hr). speeds (storm upward direction). Del-shaped Watershed Elongated Watershed Sediment-Discharge Rating Curve at Outlet Sedimen-Discharge Rating Curve at Outlet for different storm directions (storm speed = 10 km/hr) at different storm speeds (S, 10 and 25 km/hr), upward direction. Storni SpMd S Imvnr Storni spaad 10 kmtir Storm Spoad 20 kmAir / ^~* // if % I 3 Dtahugo [m'JjMO) Figure 8 Sediment-discharge rating curve at outlet of elongated watershed (see Figure 2) for different speeds (storm upward direction). Figure 5 Sediment-discharge rating curve at outlet of del-shaped watershed (see Figure 1) for different Table 2 Summary of peak flows and sediments for storm directions (storm speed =10 km/hr). the del-shaped shaped and elongated watersheds for Elongated Watershed Hydrograph and Sedigraph at Overland Flow Plane 9R PeakFlow Peak Scdhncnt Scdjn»* Vkld at different storm speeds (5,10 and 25 km/hr), upward direction. (mA3Aec) D» charge (tona/hs) (tom/icc) Storni Spaad i kmtir Del-Shaped Watenhed Stem SpMd 10 kmffir Storm Upward* 718.72 125.56 79.48 Storm SpMd 21 hmJhr Storm Downward* 812.84 137 J» 84.67 Storm Acroft* 859.93 145.12 90.65 Elongated Watershed Storm Upward* 664.73 75.32 51.86 Storm Downward* 574.22 62.78 49.54 Storm Acrott* 729.46 83.07 55.76 Storm Upward (5 kmAir) 686.34 78.85 55.11 Storm Upward (25 km/hr) 599.72 67.42 47.06 * storm «peed •tlOkmAir Tim« ftnlnutot) Figure 6 Hydrograph and sedigraph at overland flow plane 9R of elongated watershed (see Figure 2) for different speeds (storm upward direction). 113 5. CONCLUSIONS 6. REFERENCES It is shown in this study that there are differ ences in the flood hydrographs, soil erosion rates Gupta, V.K. and Waymire, E. (1979), A stochastic and sediment yields in both del-shaped and elon kinematic study of subsynoptic space-time gated watersheds for different directions and speeds rainfall, Water Resources Research, Vol. 15, of storm rainfall. These results signify that the dy No. 3, pp. 637-644. namic characteristics of the storm such as its speed Horton, R. E., 1945, Erosional development of and direction are far too important to be neglected. streams and their drainage basin: hydrophysi- It is still a common practice in flood, erosion and cal approach to quantitative morphology, Bul sediment transport studies to use storm hydrographs letin of the Geological Society of America, obtained from point rainfall values projected over an Vol. 56, pp. 275, 370, March. area, thus ignoring the effect of storm movement. Strahler, A. N., 1950, Equilibrium theory of ero sional slopes approached by frequency distri bution analysis American Journal of Science, Vol. 248, pp. 673-696, October. Waymire, E., Gupta, V.K. and Rodriguez-Iturbe, I. (1984). A spectral theory of rainfall intensity at the meso-$ scale, Water Resources Re search, Vol. 20, No. 10, pp. 1453-1465. 114 Redistributed Hydrological Modeling in Changjiang River Yan Huang*, Yaowu Min, Wenfa Yang Bureau of Hydrology, Changjiang Water Resources Commission, Water Resources Ministry, China ABSTRACT To improve current Flood Forecasting (FF) technology, recently, in addition to the traditional lumped models such as API and XAJ, there are distributed hydrological models applied at Changjiang River, namely the URBS (Australian origin) and PREVAH (Swiss origin). The meteorological model MM5 is also used to produce Quantitative Precipitation Forecasting (QPF) as the input for the hydrological models. The application of such FF systems is meant to improve the flood forecasting accuracy as well as to improve hydrological modeling at the un-gauged areas, e.g. Three Gorges Area. However, it is found that the Digital Elevation Model (DEM) based distributed hydrological models are insufficient to represent hydraulic constructions (e.g. a dike or a retention basin), which create uncertainties in especially runoff generation. On the contrary, the traditional lumped models can solve such problem by taking the hydraulic construction into special consideration, but its linear reservoir based routing is inadequate if the calibrated parameters are transferred to other places. To cope with the drawback in each type of models, a new method is developed in BOH CWRC, named as Redistributed Hydrological Model. The essential idea is to supplement the limitations of each model by combining runoff generation process from API and XAJ with the flow routing process of the distributed model URBS. The paper describes briefly the idea of Redistributed Hydrological Modeling and some applications of both distributed and Redistributed Models. The results show a great potential to adapt the Redistributed Model in future hydrological modeling for real time flood forecasting in both gauged and un-gauged areas. Introduction gauged area with similar geographic and With the rapid social-economic development, environmental conditions, where the parameters can there are increasing needs to improve FF (flood be calibrated using historical data. forecasting) technology to support flood To obtain better hydrological forecasting, management in Changjiang (Yangtze River). The meteorological forecasting is also included in the needs also come from the difficulties in FF for areas current FFS to provide precipitation forecast for the that are not of sufficient gauging, i.e. without pre hydrological models. Currently a numerical cipitation gauges or/and flow gauges. Improvement meteorological model MM5 (Anthes and Warner, of FF technology is in general valued in terms of 1978) produces Quantitative Precipitation accuracy and efficiency (time and resources Forecasting (QPF) are used. It provides rainfall allocation). prediction with spatial distribution associated with The current FFS (flood forecasting system) regions or resolutions defined in the lumped or for Changjiang includes various traditional distributed hydrological models. lumped hydrological models such as Antecedent However, there are appealing disadvantages Precipitation Index (API) method (Köhler, 1951) for these two aforementioned modeling systems. It using Unit Hydrograph (UH) method (Sherman, is found that the applied distributed hydrological 1942) and XinAnJiang (XAJ) model (Zhao, 1984). models, URBS and PREVAH, are more suitable for Recently, aiming for improving FF technology, areas which have little human activity impact. For attempts have been made to include several areas that have considerable human activity impact, distributed hydrological models such URBS the current Digital Elevation Model (DEM) based (Australian origin) and PREVAH (Swiss origin). distributed hydrological model are insufficient to Both types of models are also applied at areas of few represent hydraulic constructions (e.g. a dike or a gauges, for example in Three Gorges area, where retention basin), which create uncertainties in rainfall is observed but not many gauge stations are especially runoff generation. On the contrary, the set up for the small branches. Parameters for these lumped models can solve such problem by taking un-gauged (in particularly for those without flow the hydraulic construction into special consideration, gauge) small catchments were transferred from but dieir liner reservoir based routing procedures are 115 insufficient. That is, with similar land covers, the As shown in Figure 1, the lumped model can better runoff generation coefficients (or net rainfall) can be represent runoff generation, and the distributed reliable with lumped approach, whereas the routing model can better represent the flood routing part, procedures is relatively uncertain as it depends on and therefore is more reliable when transferred to an the catchment topography/slope, or DEM un-gauged area. which cannot be correctly represented by simply transferring me flood routing coefficients from anotíier area. Such uncertainty distribution (or contribution) can be illustrated qualitatively in Figure 1. Land cover Soil DEM Hydraulic constructions \y ir ir Runoff generation Flood routing i ' ^ ' i model Flow hydrograph Distributed model Figure 1 Illustrative uncertainty contributions to flow simulation in hydrological model To cope with the drawback in each type of Han River, which is one of the largest tributaries of model, a new method is developed in BOH CWRC, Changjiang River. The results show that the which is named as Redistributed Hydrological distributed models are in general acceptable but the Model. The essential idea of the Redistributed newly developed Redistributed Hydrological Model Model is to supplement the limitations of each performs relatively more flexible for un-gauged model wim the strong point from the other model, area. i.e. to combine the runoff generation process from API and XAJ with the flow routing process of the The hydrological modeling system distributed model, URBS. The details of the In current FF systems, in parallel to the coupling are given in Section 2.2. interactive expert-based hydrological models which Another reason of combining these two are mainly developed based on conventional models models are that most areas in Changjiang river have such as API and XAJ model, the distributed models been well calibrated in the past using API and XAJ. URBS and PREVAH are also used to simulate This can reduce work load and maintain the same runoff and route it through die catchment. Next to it, reliability when the distributed model is introduced in the newly developed Redistributed Model, die to simulate the hydrological process in the API model is used to generate runoff (or net rainfall) catchments. The ultimate goal is to improve the and the flood routing procedure of URBS model is flood forecasting accuracy for effective and efficient incorporated to route the flow to die edge of the flood management at Changjiang river basin. catchments, fn both methods, meteorological In mis paper, the distributed hydrological simulation and prediction are included as the model model and Redistributed Hydrological Model are input, which produce precipitation information briefly introduced. Using these two systems, some directly from satellite image witiiout rainfall gauges, real time hydrological forecasts have been made at and provide input for the hydrological models. Daning River in Three Gorges area, and at Details of me models are presented as follows. 116 1 Meteorological model (MM5) Australian origin model URBS, and the Swiss origin In order to obtain real time precipitation model PREVAH. Each model was introduced into forecast/simulation, a meteorological model, MM5, CWRC via international cooperation projects. They for mid-term forecast, is adapted. MM5 is developed are both in the test period in parallel to the tradi jointly by USA National Weather Service and tional hydrological models which are API and University of California, which was first developed XAJ based. from a mesoscale model used by Anthes at Penn State in the early '70's (Anthes and Warner, 1978). 2.1 URBS The QPF is carried out separately and not linked URBS is an Australian-origin networked (i.e. with the hydrological model automatically in terms sub-catchment based) runoff-routing model that of computer programming structure. estimates flood hydrographs by routing rainfall Start from 2004, BOH CWRC introduced excess through a module representing the catchment MM5 to carry out regular rainfall prediction. storage. In URBS, the storages are arranged to There are three layers included in the model witii represent the drainage network of the catchment. resolution of 63km, 21km and 7km. The lead time is The distributed nature of storage within the from 24, 48 and 72 hours. The model provides catchment is represented by a separate series of ASCII file which can be accessed directly by concentrated storages for the main stream and for distributed models such as URBS and PREVAH. major tributaries to provide a degree of physical realism. The storages in the model are generally 2 Distributed hydrological modeling non-linear. The model is outlined in Figure 2 as Currently there are two distributed below. hydrological models implemented in the FFS, the 1 r Rainfall ~\ Runofl A Surface Excels M Rmitfar \ „ „ i Runoff (or auront*«] Hydrograph runoff) y^r —> '•JÊÊËfêz \ Baseflow \ Baseflow J Generator J * Hydrograph Figure 2 Schematic representation of URBS model In the rainfall ~ runoff process, rainfall initial loss of '//' mm before any rainfall becomes excess is first estimated from rainfall data using the runoff. After this, a continuing loss rate of 'c/' mm so-called loss models before it is applied to the per hour is applied to the rainfall, subject to die limit runoff-routing component of the model to compute of the soil infiltration capacity. The loss rates can be the surface runoff hydrograph. Among several specified 'globally' to the entire catchment or options, the initial and continuing loss model was 'individually' to each sub-catchment. Baseflow, if adopted in the FFS for the Changjiang tributary significant, is estimated separately and added to the catchment. This loss model assumes that there is an surface runoff hydrograph to provide the total catchment hydrograph. 117 In the runoff routing in URBS, the so-called 2.2 PREVAH 'Split' model is adopted in the FFS for the The hydrological model PREVAH Changjiang tributary catchments. In the Split model (Precipitation-Runoff-Evapotranspiration-Hydrotope the rainfall excess for each sub-catchment is first model) has been developed for sound estimation of determined by subtracting losses from the rainfall the hydrological dynamics within mountainous hyetograph. The rainfall excess is then routed catchments (Gurtz et al. 1999). The spatial through a conceptual catchment storage to determine discretization of PREVAH relies on the aggregation the local runoff hydrograph for the sub-catchment. of GIS information (e.g. DEM, land cover and soil The storage ~ discharge relationship for catchment maps, Figure 3) into hydrologie response units routing is: HRUs. Its parameters are based on physical principals. It consists of several subsystems for 0) runoff generation including snow model, glacier 1 (1 + t/)2 model, interception model, a model of soil water storage and depletion by évapotranspiration. The module for soil water storage and 3 Where, Scalch is the catchment storage (m h/s); depletion by évapotranspiration relies on the is the catchment lag parameter; A is the area of HBV-model and the Penman-Monteith equation sub-catchment (km2); U is the fraction urbanization (Monteith, 1965, Gurtz et al., 1999). Snow and of sub-catchment; F is the fraction of sub-catchment glacier melt are calculated using a modified forested; and m is the catchment non-linearity temperature-index approach, including potential parameter. In the equation (1), ß is determined direct clear sky solar radiation (Hock, 1999; Zappa during model calibration and is a global parameter. et al., 2003). The runoff generation module uses The local runoff hydrograph is then concepts from the well established HBV-model combined with runoff from die upstream (Bergström, 1976; Lindström et al., 1997), adapted sub-catchment and routed through a channel storage to a spatially distributed application (Gurtz et al., to obtain die outflow hydrograph for die 2003). sub-catchment. Channel routing is based on the non-linear Muskingum Model (Cunge, J.A., 1969). The model accumulates base flow tiirough the network and does not route it through as for direct runoff. Interpolated meteorology Spatial information precipitation, air temperature, Elevation, Aspect, Landuse, PREVAH global radiation, water vapour Slope, Soil Properties, ... HRU-Table pressure,sunshine duration 4- Precipitation correction Site adjustment of radiation and temperature y..y->}»s!»m.,m -Jam,::- Evapotranspiration Snowmett / Icemett Interception Soli moisture Runoff generation Calibration and Verification Figure 3 Model structure of PREVAH 118 Each runoff component is transformed by its places where geometrical might be similar but land specific storage coefficient valid for the whole cover and soil condition might be significantly sub-catchment. The percolation into groundwater different. Thus, the flood forecasting department of storages is calculated dependent on soil conductivity BOH further developed the Redistributed Model by and moisture content of the upper storages. The combining it with the well-calibrated lumped calculation of flood-routing is based on the models such as API and XAJ. combination of linear storages and translation This is realized as URBS allows the user to components. Figure 3 outlines the model structure of select one of its several standard loss models. URBS PREVAH including the input of spatially model also allows the user to undertake continuous interpolated meteorological data. soil moisture accounting using an interface to any external soil moisture accounting model such as 3 Redistributed Hydrological Model XAJ and API, where the well calibrated XAJ and As aforementioned in section 1, the DEM API produce runoff using precipitation information based distributed hydrological model (e.g. URBS) at its predefined sub-catchments which are different has a drawback at the runoff generation processes from the sub-catchments that are defined in URBS. for its inconvenience to be transferred to other The model structure is shown in Figure 4. Sub-catchments 1 Previous parameters API or XAJ Net rainfall Sub-catchments 2 URBS Flow forecasting Figure 4 Structure of the Redistributed Hydrological Model URBS API and URBS XAJ The theory of API and XAJ can be found in Application of PREVAH at Three Gorges Area many literatures (e.g. Köhler, 1951; Zhao, 1984). 1 Modeling area Both methods have similar principle in net rainfall As shown in Figure 5, the Three Gorges area, calculation, i.e. depending on the moisture condition between Yichang station - the gauge station at main and soil capacity and conductivity, net rainfall is channel of Changjiang River and upstream stations generated once the storage(s) is filled up. The idea is of Chongqing at main channel and Wulong at Wu to combine URBS with API or XAJ, to provide River branch, has total area of about 56000 km2. URBS with net rainfall, then the rainfall is proc Among the various small catchments, Daning River essed into different flow component in URBS, and is one of the few catchments that has better and is routed through sub-catchments with flow routing representative data condition, and therefore it is procedure provided in URBS. chosen as the calibration area. By separating the two processes, rainfall gen To cope with such problem, PREVAH is eration and flow routing, it is also easy to transfer in calibrated with Daning River data and then trans case the model is needed to be transferred to another ferred to the neighboring catchments. The Yichang catchment. station is used as the major control to calibrate the 119 whole Three Gorges and to provide Three Gorges and karst limestone rock outcrops. Numerous reservoir with more accurate inflow. groundwater flow conduits exist within these limestone formations and mus produce a steady base The Daning River is a minor tributary of the flow in the river. Data were available for 10 rainfalls Changjiang River located in the Three Gorges reach. and 4 stream gauging stations spread across the It has a total catchment area of about 4180 km2 and a catchment. Modeling was undertaken for the upper stream length of about 162 km. The catchment rises half of the catchment (2001 km2) down to Wuxi, from elevations of less than 100 m at its confluence where the most downstream gauging station is with the Changjiang River to elevations of over located. 2000 m. The catchment is quite steep, has a thin Yichang Wulong Figure 5 Daning River and Three Gorges area 2 Model set up observed ground level, different pressure With spatial resolution of 630 meters and information and the weather forecasting results from time step of 1 day, in PREVAH, 34 sub-basins were T213 model which is currently running in the defined for the Three Gorges area. Time series of national meteorological center. T213 model results period during May 1st to September 30th 2005 have 240 hours lead time, and the resolution 62 km (several minor floods) are used for model horizontal grid size, and vertically 31 layers. Figure calibration, and verified for the corresponding 6 shows a 24hr lead time rainfall prediction. periods in 2004 (major flood) and 2006 (hardly no It has been found that MM5 is more accurate nd flood up to August 22 ). Time step of precipitation for rainfall process, location and its quantitative input is 6 hour, internal time step of PREVAH is range. However it has been found that rainfall inten one hour. sity and precipitation area is relatively larger than the measurement. According to the analysis, 65% 3 Results forecast are acceptable, 20% rainfall processes are 3.1 Quantitative precipitation forecast using missing and 15% are wrong forecast. This indicates MM5 The meteorological model MM5 is applied to a large room for model improvement for QPF. provide precipitation forecast for the distributed models. The inputs of MM5 models are the regular 120 too'e 110E i»e 301K 30« UWE 110E 12TE 0.1 in u lojo iu KO sut 10O.O 200.0 aooLO Figure 6 24hr QPF at Changjiang River basin using MM5 3.2 Discharge simulation Figure 7 plots the discharge hydrograph between As the data at Daning River is more model simulation and measurement at Yichang representative, the parameters are firstly calibrated station. The model shows a correlation coefficient at Daning River using discharge at Wuxi station. of 0.98 between the model simulation and Then the parameters are transferred to the rest of the measurement. Three Gorges area. Discharge at Yichang station is used to optimize the parameters at the Three Gorges area. 50000 45000 — PREVAH • Measurement 20000 5-Aug-05 10-Aug-05 15-Aug-05 20-Aug-05 25-Aug-05 30-Aug-05 4-Sep-05 9-Sep-05 14-Sep-05 Time Figure 7 Comparison between model simulation and measurement at Yichang station 121 URBS and Redistributed Model at Han River numerous small storages but these are not generally 1. Modeling area represented in the model. The area of some As one of the branches in Han river and one sub-catchments was reduced to represent the effects of the largest tributaries of Changjiang River, the of storages. Tang River joins the mainstream of the Han River about 140 km downstream of Danjiangkou Reservoir (after combining with the Bai River). Forecast Unit 3 (FU3) has an area of 7,000 km2 and includes all of the catchment upstream of the Guotan (67624) gauging station. The catchment is primarily flat agricultural land, with a highly artificial drainage system of channels and storages. 2. Model set up The URBS model uses input from 10 rainfall stations and diere are 5 flow gauging stations with available data. A list of tiie input stations is shown below. The spatial distribution of rainfall stations is adequate. Good quality flow data is generally available at Guotan (67624) and Tanghe (67614), which are used as calibration station for URBS model. The Tang River URBS model is used in die FFS to forecast tributary inflows into die Han River hydraulic model. In original URBS, die Tang River basin is divided into 31 sub-catchments, as shown Figure 8 Tan River Catchment in Han River in Figure 8. The Tang River catchment includes • Measurement URBS URBS_API -- URBS_XAJ 07-08-06 07-16-06 07-24-06 08-01-06 08-09-06 08-17-06 Time Figure 9 Comparison between measurement and model 122 Rating curves have been adopted at a number show good fitness to the measurement at Three of gauging stations in the catchment. Water level is Gorges area and Tang River catchment in Han generally recorded more frequently than discharge. River. In particular in the Three Gorges area, the The model includes some small dams to simulate model PREVAH has not been entirely calibrated for catchment storages. Transmission losses were all of the catchments. Parameters were transferred applied at a number of locations. from Daning River which has better representative In the Redistributed Model with URBS, i.e. data compared with other catchments. The better the URBS flow routing procedure combined with performance of URBSAPI and URBSXAJ also net-rainfall generation procedures of API and XAJ, show the feasibility of combining PREVAH with has slightly different sub-catchment boundaries. the previously calibrated API or XAJ model when it Instead of 31 sub-catchments in URBS, there are is to be applied at other areas. only 5 sub-catchments defined in API and XAJ. The Meteorological modeling is important and latter can first of all, reduce largely the computation can be essential for the application of both load and time, and secondly, analyses show it will distributed model and Redistributed Model. not reduce the accuracy by using fewer For un-gauged area where rainfall is missing, sub-catchments. meteorological modeling can provide both real time and forecasted precipitation for the defined 3. Results sub-catchments. For gauged area where point The URBS is calibrated using data of year rainfall measurement is available, meteorological 1998-2000. For the URBS_API and URBS_XAJ , modeling can provide QPF with better accuracy die net-rainfall generation parameters used in Tang with data assimilation technology which can update River in BOH's routine FF work are adapted, while the model set up with real time rainfall monitoring. the calibrated flood routing parameters in URBS are In this research, such applications have been carried kept. out but not presented here due to the limit to the The calibrated model was applied during paper length. flood season of 2006. Figure 2 plots the comparison The research shows a strong support to between measurements at the gauge station Guotan. separate runoff generation process from flood No big difference is found between these three routing procedures with better means, for example, models. The general performance shows a by combining lumped models such as API and XAJ correlation coefficient of 0.86, 0.905 and 0.912 for which work with predefined sub-catchments, with URBS, URBSAPI and URBS XAJ, respectively. the flood routing procedure that is DEM based and This indicates a slightly better performance of with another set of sub-catchments. This is very URBS_API and URBS XAJ, however modeling necessary for two reasons. First of all, the lumped results using URBS merely also provides acceptable models API and XAJ work with predefined accuracy for Tang River. The findings further prove sub-catchments with identified human affected area, the usefulness of employing distributed hydrological which can not be sufficiently represented by DEM models for real time flood forecasting in Han River. based distributed models. Moreover, API and XAJ During the flood season in 2006, similar set have been long used in many Chinese water up and performance are obtained for the other institutions for their design and planning work. catchments of Han River. Application of the Thus, such improvement would allow the use Redistributed Method, i.e. using the previously of their existing available parameters. Improvement calibrated API and XAJ parameters, has largely on distributed models can be realized by including reduced the work load for model calibration of another GIS layer on which runoff generation are URBS, and maintains a reasonable forecasting separated geographically from flood routing performance, especially at the net-rainfall generation procedure, which is however rarely adopted in most process. of the current distributed models. Acknowledgement Conclusions The paper is based on the results from a joint Distributed Hydrological Model shows a project between Chinese and Australian participants great potential to be applied as an operational including China's Ministry of Finance and hydrological model for both gauged and un-gauged Commerce and Ministry of Water Resources, area flood forecasting. Both PREVAH and URBS Australian Agency for International Development (AusAID), Coffey Croup of companies as the 123 Australian Managing Agent (AMC). The paper has Hock R., 1999. A distributed temperature-index also received assistance from Mr. Massimiliano ice- and snowmelt model including potential Zappa and Mr. Thomas Bosshard, from the Swiss direct solar radiation. Journal ofGlaciology 45, Federal Research Institute. They are the project 101-111. colleagues of "the Changjiang Flood Forecasting Köhler, M.A., and Linsley, R.K., 1951. Predicting Assistant Project" which is assisted by the Swiss runoff from storm rainfall. Res. Paper 34, U.S. government through the Swiss Agency for Weather Bureau, Washington, D. C. Development and Cooperation and by the CWRC. Lindström, G., Johansson, B., Persson, M., Gardelin, M., and Bergström, S., 1997. Reference Development and test of the distributed HBV- Anthes, R. A., and Warner, T. T., 1978. 96 hydrological model. Journal of Hydrology, Development of hydrodynamic models suitable 201,272-288. for air pollution and other mesometeorological Monteith, J. L., 1965. Evaporation and environment. studies. Mon. Wea. Rev., 106, 1045-1078. Symp. Soc. Exp. Biol., 19, 205-234. Bergström, S, 1976. Development and Application Sherman, L.K., 1942. 'The unit hydrograph of a Conceptual Runoff Model for Scandinavian method', Chapter XIE of Hydrology, ed. O.E. Catchments. Bulletin Series A, No. 52, Meinzer, pp. 514-525. University of Lund, 1976. Zappa, M., Pos, F., Strasser, U., Warmerdam, P., Cunge, J.A., 1969. On the subject of a flood Gurtz. J., 2003. Seasonal water balance of an propagation method. Journal of Hydraulics Alpine catchment as evaluated by different Research. IAHR, 7, pp.205-230. methods for spatially distributed snowmelt Gurtz, J., Baltensweiler, A. and Lang, H., 1999. modelling. Nordic Hydrology 34, 179-202. Spatially distributed hydrotope-based modelling Zhao, R.J., 1984. Hydrological modelling. Water of évapotranspiration and runoff in Conservancy and hydropower publication, 1984 mountainous basins. Hydrol. Process., 13, [M] Beijing. (MAfè, 1984, íJrLͧlt7jc£^ 2751-2768. [M]« im-. 7KfÜ*;bÜJ/féíto ) Gurtz, J., Zappa, M., Jasper, K., Lang, H., Verbunt, M., Badoux, A., Vitvar, T., 2003. A comparative study in modelling runoff and its components in two mountainous catchments. Hydrological Processes 17, 297-311. 124 Status of Disaster Database and Limitations for Planning T. Merabtene1, J. Yoshitani1, A. Pathirana1 1 International Center for Water Hazard and Risk Management (ICHARM), Public Works Research Institute (PWRI), Japan m-tarek(q),pwri. go. ¡p ABSTRACT Over the last century, there has been a significant rise in water-related disasters, affecting an increasing number of people, particularly those living in developing countries. Resulting damages to property and losses of life and livelihoods compromise the gains of development. Agencies in charge of disaster management have recognized the importance of a holistic disaster planning and management for achieving disaster resilience for vulnerable communities and to substantially reduce poverty and disaster losses by 2015 as endorsed by the Hyogo Framework for Action. To achieve this end, a reliable disaster database is one of the most crucial and invaluable tool for risk management professionals working to develop innovative approaches and applications to reduce the risk and impact of disaster. This research reviews the state of the art of disaster database and discusses their limitations for effective disaster planning at each stage the disaster management cycle. The research also looks on the means to overcome the existing lack of international census regarding the best practices for collecting effective and reliable data on disasters with flood disasters in focus. The findings also highlights the emerging needs for more structural program to collect accurate reliable data, local needs and assessment on water-related disasters in various regions of the world for the database to be qualified as policy effective information for disaster risk management. Introduction The "Hyogo Framework for Action 2005- Long experience has shown that setting 2015: Building the Resilience of Nations and targets is vitally important for providing incentives Communities to Disasters", offers guiding to mobilize actions on key issues of development. principals, priorities for action, and practical means Recognizing the need to eradicate extreme poverty for achieving disaster resilience for vulnerable and speed up socio-economic development, the communities. The five priorities for action include: 2000 United Nations General Assembly Millennium 1) Ensure that disaster risk reduction is a national Meeting established eight Millennium Development and a local priority with a strong institutional basis Goals (MDGs), with targets to be achieved by 2015. for implementation; 2) Identify, assess and monitor Although disaster reduction was not defined as one disaster risks and enhance early warning; 3) Use of the MDGs, the 2nd World Water Development knowledge, innovation and education to build a Report (WWDR) clearly highlighted that goals such culture of safety and resilience at all levels; as poverty eradication, promotion of environmental 4)Reduce the underlying risk factors; and 5) sustainability or development of global partnership Strengthen disaster preparedness for effective for development can never be achieved without response at all levels. The second priority action having disaster management as integrated part of above of the HFA identifies that the starting point Integrated Water Resources Management and water for reducing disaster risk lies in the knowledge of governance strategies. This consensus was also the hazards and the physical, social, economic and confirmed by the Hyogo Framework for Action environmental vulnerabilities to disasters that most (HFA) that have set the goal to substantially reduce societies face. To achieve this end, three key disaster losses by 2015 - in lives, and in the social, components were recognized, namely to (a) develop, economic and environmental assets of communities update periodically and widely disseminate risk and countries. The HFA represents a new land mark maps and related information to decision-makers, to boost disaster management higher in the political the general public and communities at risk in agenda of many countries since it was adopted by an appropriate format; (b) develop systems of 168 governments at the World Conference on indicators of disaster risk and vulnerability at Disaster Reduction, in Kobe, Japan 2005. national and sub-national scales that will enable 125 decision-makers to assess the impact of disasters on disasters. As of 26 July 2007, the database contained social, economic and environmental conditions and 16,035 entries covering the period from 1900 to the disseminate the results to decision makers, the present. The database is publicly accessible and public and populations at risk; and (c) record, recognized as one of the most accepted international analyze, summarize and disseminate statistical disaster databases. For an event to be added to the information on disaster occurrence, impacts and database one their four criteria must apply. Amongst losses, on a regular bases through international, disaster databases, EM-DAT provides one of the regional, national and local mechanisms (ISDR, most comprehensive and transparent explanations of 2006). The international community also recognizes the methodology employed (Tschoegl, 2006). the need to develop indicators to track progress on disaster risk reduction activities. 2. Disaster Database Project, University of In order to deploy a set of policy oriented Richmond strategies in line with the above priority actions it is The Disaster Database Project, at the a primary need to confidently assess, define and University of Richmond, includes disaster address the challenges of risk identification, information of natural and man-made disasters management and communication. At the top of these including conflict-based disasters. The database also challenges is the systematic collection of reliable attempts to disaggregate the disaster to give detailed disaster information (impact and cause) as description of the factors involved at the different invaluable tool for sustainable disaster planning and stage of a disaster cycle. It is publicly accessible and risk management. includes over 1500 entries dating back to 2000BCE where entries are also based on defined criteria. The International disaster databases main sources of information are government reports, The efforts to account the effects and losses newspapers, and scholarly texts (Richmond, 2006). associated to disasters have been a major focus of disaster prevention agency at national and regional 3. Munich Reinsurance Company: NatCat levels. On the other hand many initiatives have NatCat is a private international level disaster taken place to develop a comprehensive database maintained by Munich Re insurance international disaster databases with the ultimate company. There are over 20,000 entries dating back goal to improve international response and actions to 79AD, with major updates since 1980. Recorded to disaster mitigation. Private initiatives with less information includes human and economic impacts humanitarian perspective and goal are also present and disaster losses, as well as scientific data such as and widely recognized. As can be seen from many wind speed and geocoding. The database is mainly comparative studies on disaster databases (Tschoegl exclusive to Munich Re clients; however the 2006, Arakida & Murata 2003, Guha-Sapir 2002), database is also partially accessible to the public, there is wide variability in the description of disaster mainly in graphic form via the company web site, events and in the specifics of their impacts. Even and through the Munich Re annual report on disaster reports of modern events may vary widely in such (Münchner, 2003). details as affected people, fatality counts and eco nomic impacts. It is not unusual for more than one 2.4. Swiss Reinsurance Company: Sigma technical investigation to be conducted of a major Swiss Reinsurance Company maintains the disaster with each investigation report providing Sigma database on natural and man-made disasters differing accounts of both the event and causation. dating back to 1970. There are approximately 7000 Hereafter, only major international disaster entries in the database. Entries are subject to set of databases are reported, other international, regional criteria and it requires that at least one is satisfied and national disaster databases are also reviewed such as at least 20 deaths or 50 injured or 2000 elsewhere (Tschoegl 2006, Merabtene 2005). homeless or a fixed amount of insured losses by sectors. The primary source of data is unclear and 1. OFDA/CRED Emergency Disaster Database the lack of public accessibility to die Sigma database (EM-DAT) makes it difficult to report on die ability to search The Emergency Disasters Database the database [http://www.swissre.com]. (EM-DAT) managed by the Centre for Research on the Epidemiology of Disasters (CRED) at the Catholic University of Louvain, Belgium, is an international database on natural and technological 3. water-related disaster planning and Limitation by flood and windstorm alone during the period of international databases 1960 to 2004 was as high as 300 million people, that As shown in Figure 1, of all natural disasters, is about 7 million people every year in average water-related disasters (i.e., floods, droughts and (Merabtene & Yoshitani 2005). Many countries epidemic water related at the top of the list) are have developed at varying levels their respective prevailing in number and their overwhelming capacities for disasters risk reduction and have put consequences for the society and the environment forward numerous preventive and mitigations cannot be overemphasized. The disaster database actions. Nevertheless, the complex dynamics of EM-DAT reported that the number of people killed water hazards determinants, both natural (such as by flood and windstorms during the period from climate variability) and human-made (such as 1960 to 2004 was as high as one million people and population explosion in hazardous area), have in average about 20,000 people were affected every increased the vulnerability of people and property to year. Figure 2 depicts the transition of the number of the wide diversity of water related disasters. The flood disaster, the number of killed people and the observed perennial toll on human lives and property number of affected people from 1980 to July 2006. challenges the effectiveness and sustainability of From January 2006 to date, nearly 1000 people died, adopted policies for water-related disaster reduction 15 thousands people affected and over 1.6 billion and calls to search for innovative and holistic worth of economic damages were reported in strategies for water-related disasters mitigation and EM-DAT database. The number of affected people risk management. 800 Flood Wind Storm 700 Drought Water Epidemics 600 Landslde Famine e 500 Wave & Surge Other Disasters ^ 400 4- O "I 300 200 100 0 1960-1964 1965-1969 1970-1974 1975-1979 1980-1984 1985-1989 1990-1994 199S-1999 2000-2004 Souw. Data Iront the Center for Epidemiology of Disasters ¡0FDA-CRED) in louvatn {Belgiuml. Analysis by the Public Works Research Institute IWVRI) in Tsukuba (Japan). 2Ô05.4 Figure 1. Trend of water disasters by type of hazards, 1960-2004, as compared to other natural disaster (Contribution of ICHARM to the World Water Development Report 2) 127 1,000.000,000 {] .-. Il II I II II Il II UiMtttttlilii 111 II 111111 11111 I 11 I I ¡ ! i Figure 2. Annual number of flood disasters and concurrent number of death and total affected people from 1980 to July 2006. Sources: EM-DAT, analysis by ICHARM [Data for 2006 are the disaster statistics for the period January 01, 2006 to July 24, 2006] To this end building a common platform to As proven throughout the results of our share knowledge and expertise among nations and technical analysis of the EM-DAT water-related experts has become more crucial than ever before. database (Merabtene & Yoshitani, 2005), there are Our experience in water-related disasters mitigation numerous overwhelming difficulties underlying the in Japan and other Asian countries confirms assessment of disaster impacts. There is definitely diat there is a prevailing need for the water urgent need to undertake a more in-depth survey and community to develop a more in-depth and common analysis in order to draw conclusive concepts for understanding of the mechanisms underlying the vulnerability analysis and water-related risk man risks associated with each type of water hazards. agement in general. For instance, there is no clear The importance of coherent and comprehensive evidence to fully support the largely acknowledge water-related disaster databases developed on the perspective tíiat the global trend of death toll from basis of deep analysis of selected disasters as a flood disasters and water-related disasters is fundamental tool for information in the learning, decreasing (see also Figure 2). Our assessment also planning and decision processes has become called the attention to the pragmatic aspect of data increasingly evident to confidently engage in processing based on different time scales and disaster management at both proactive and reactive analysis periods, which might often lead to stages. In other words, to analyze the risk and serves controversial conclusions if die results are not the needs of managers and practitioners at each level handled and interpreted with extreme care. of the disaster management cycle namely Undoubtedly, for us to have a complete pre-disaster mitigation and preparedness and after picture on the usefulness of current databases for disaster response and recovery plans, the database disaster planning a more in-depth event based must be reliable and accurate with regards to the analysis is required. Nevertheless, in order to put causes of the disaster, the effects of the disasters and further clear emphasis on the complexity of such more important the causes of die effects (i.e., cause quality assessment of water-related disasters of death for instance). The database shall also allow databases in particular, reference is made to Figures assessment of trends and future risks as well as to 2 and Figure 3. When we produced Figures 2 and serve the development of risk management and Figure 3 we have excluded all extreme values that policy effectiveness indicators (see also Nair 2006). were repeated over several years (especially those 128 with exactly the same numbers such as the droughts affected people are reported. However, the meaning in India or the famines in Korea Democratic Popular behind these numbers is questionable, as some death Republic). Such values are taken into consideration associated to others famine disasters are not part of only once or twice at most depending on the decade the famine database. Though separating the real of occurrence. impact of drought from other factors is still a big - Figure 3 looks at the database in terms of the top challenge, attention is directed to the famine data in one hundred most disastrous events classified with Figure 3 showing zero people killed by the largest regards to the largest numbers of affected people by famine disasters, in contrast to the famine data water disasters form 1960 to 2004; while, plotted in terms of largest killed people (Figure 4) - Figure 4 looks at the database in terms of the top showing nearly zero affected people (save the North one hundred most disastrous events with regards to Korea famine in 2002). Thus, even after isolating the number of dead people for the period 1960 to large extreme events there is still tremendous work 2004. to be carried out on event-basis to accurately A classification as of Figure 3 shows that identify the statistical trends of water-related most devastating water disasters are drought, flood, disasters as to respond to the demand of disaster windstorm and famine. For drought, the planners and managers at each stage of the disaster figure includes only 6 events where both killed and cycle or even just to develop indicators to compare between regions or countries. Number of affected Top 100 disasterous events classified in term of largest numbers of affected people and killed people 1,000,000,000 100,000,000 10,000,000 1,000 100 Drought Flood Windstorm Famine Figure 3. Extreme one hundred water-related disasters classified with regards to the largest number of affected people from 1960 to 2004. Misinforming are the flood and windstorm data as flood disasters. This inaccuracy applies to both old classified in Figure 3 which shows a more logical disasters as well as to more recent disasters. For reporting based on the ratio between the numbers of instance, to cite only few, when looking at the flood killed and affected people Nevertheless it is by no events of 1970 in Bangladesh (10 million people mean a valuable conclusion, especially when we affected), 1979 in India (26 million people affected), refer to the data of flood and windstorm as classified and 2002 in the China Republic (20 million people in Figure 4. Figure 4 clearly shows that there is affected) the EM-DAT database reported zero killed also inaccuracy and undependability in reporting people for the three disasters. This complex large 129 Number of killed and affected people Top 100 disasterous events classified in term of largest numbers of killed people 1,000,000,000 I no killed • total affected 100,000,000 10,000,000 1,000,000 100,000 10,000 1,000 100 10 •SK:s§stJs;siSsg?™s;gt SHJESssSäSSSKsHSHsSlS? &*&* ¡5SSK Famine Drought Flood Windstorm Wave/surge water Epidemic Slide Figure 4. Extreme one hundred water-related disasters classified with regard to the largest number of killed people form 1960 to 2004. diversity of interrelation between the numbers of Conclusion killed people and affected people, especially under There are numbers of pragmatic issues the classification shown in Figure 4, can be further for proper implementation of integrated investigated for wave/surge, famine, epidemic multidisciplinary approach in water resources water-related, slide and windstorm and extended to management. These issues are being stressed and other international and regional databases. Needless discussed at many scientific and political gatherings to re-emphasis that die consistency of reported data and are mainly reflected as an integral part of the can only be verified through a comprehensive action plan of the 4th phase of the International comparative study with national disasters data Hydrological Program (IHP-VI Plan: 2002-2007). bases. However, we all recognize that for achiev In water-related disasters further problematic issues ing this basic assessment goal mere is a need for a are found within die risk management process and strong political and scientific will to share data and cycle. The major issue is die reliability of the open the national databases. risk quantification and assessment which we The analysis above has brought to us a far acknowledge to be among me most important steps clearer understanding of the complexity and in risk management and also die most difficult and limitation of current international database to be prone to error regarding the diversity of die type of used as a platform for water-related disaster water-related disasters and concurrent impacts. The planning or for even international emergency reliability of risk assessment does not only relay on response. The results show that the conclusions on die quality of reporting, but also on the methodology the trends of death tolls and affected people by any to monitor and assess disasters' impacts at tiie first type of water-related disasters should be questioned stage. A direct problem resulting from such with even more vigor in the light of more rigorous pragmatic assessment is at least of two folds, first, the uncertainties surrounding tiie accuracy of die scientific approach in data selection and data numbers quantifying tiie impacts, second, tiie processing. meaning behind the numbers resulting from me risk 130 assessment upon which most mitigation strategies References: are built on. It is therefore extremely valuable Arakida M. and Murata M. (2003): "Disaster infor to create a strong international platform for mational common means with high promptness cooperation with primary focus to share data and and reliability: The GLobal unique disaster expertise on water-related disasters and related risk IDEntifier number (GLIDE)" in Japanese, management issues. Asia Disaster Reduction Center publication. The publicly accessible disaster databases are CRED 2006, Proceedings of the Workshop to im general with regards to the natural and man made prove the compilation of reliable data on disas disasters reported. Despite, the invaluable ter occurrence and impact: 2-4 April, 2006 - information presented, it is clear that there is a lack Bangkok, Thailand. of international consensus regarding best practices Guha-Sapir, D., Below, R., 2002, "The Quality for collecting reliable data on natural disaster. and Accuracy of Disaster Data: A Comparative In addition to the difficulty of collecting disaster Analysis of Three Global Data Sets. The information (causes and impacts) data, there ProVention Consortium, The World Bank; Oc also remains huge variability in methodologies, tober, [http ://www.proventionconsortuim.org/ definitions, tools and sourcing (CRED 2006). files/sapir.pdf.l In this direction, the International Centre for Water ISDR 2006, "Hyogo Framework for Action 2005- Hazard and Risk Management (ICHARM) is 2015: Building the Resilience of Nations and dedicating great efforts to complement the gaps for Communities to Disasters". World Conference accumulating accurate reliable for water-related on Disaster Reduction; January 18-22; disaster. At the first stage of this initiative focus will Kobe, Japan. be given to flood disaster as it is the largest natural Merabtene T., and Yoshitani J., 2005, "Technical disaster with regards to threat to life and property. Report on Global Trends of Water-related Dis The goal is to build homogenous and comparative asters", Technical Memorandum of Public databases where disaster information is acquired in Works Research Institute (PWRI), ISSN 0386 standard and rigorous way for planning at local -5878, No. 3985. levels. The approach is to involve a network of Münchner Rückversicherungs-Gesellschaft, 2003, affiliated experts that are committed to creating a "NatCat Service A guide to the Munich Re da water-related effective information database that is tabase for natural catastrophes." [http:// policy relevant to serve disaster reduction at each www.munichre.com/publications/3 02- stage of the disaster cycle. 03901 en.pdf?rdm= 17651.1 Setting measurable goals and targets the Richmond, 2006, The Disaster Database Project: International Centre for Water Hazard and Risk database on me internet, University of Management (ICHARM) will work with its partners Richmond, c2002-2006. [http:// to adopt a standardized approach for damage learning.richmond.edu/disaster/index.cfin] evaluation and reporting mechanism. ICHARM will Nair S. S. (UNDP India), 2006, "Systematic Help develop an appropriate framework of database for disaster risk reduction", Workshop collaboration to widen the scope of the collection to improve the compilation of reliable data on of water-related disaster information, as well as to disaster occurrence and impact: 2-4 April, 2006 provide linkage and in depth study of available - Bangkok, Thailand. databases and will ensure improved access to its Tschoegl, L., 2006, "An Analytical Review of findings. Since its Establishment in March 2006, Selected Data Sets on Natural Disasters and ICHARM is dedicated to diligently work for filling Impacts". the gap in this area and promote the development and adoption of standardized methodology for disasters impacts quantification 131 ICHARM's Research Strategy toward Effective Implementations of Flood Forecasting and Warning System in Asia K. Fukami1, H. Inomata1, P. Hapuarachchi1 and R. Oki2 1 International Centre for Water Hazard and Risk Management (ICHARM) under the auspices of UNESCO, Public Works Research Institute (PWRI), Japan; 2 Earth Observation Research and Application Center, Japan Aerospace Exploration Agency (JAXA) k-fukam i(a)pwri. go. ip ABSTRACT We are now required to "identify, assess and monitor disaster risks and enhance early warning" for flood disaster reduction. Quick implementation of flood forecasting and warning system is a key component to mitigate flood disasters effectively in the flood-prone areas of the world. The same is true for Asian countries. Several key steps are to work properly so mat flood forecasting really results in flood disaster mitigation: 1) monitoring of meteorological & hydrological conditions, 2) flood forecasting simulations, 3) analysis of forecasts and judgment of hazardous risk, 4) dissemination of warning, and 5) crisis management at the site such as flood fighting, evacuation, etc. If any of the steps does not work properly, then the effort for flood forecasting would not lead to real flood disaster mitigation. There are, however, quite a few obstacles in each step for me flood-forecasting flow. In order to cope with the problem, ICHARM has initiated the two research initiatives: 1) the development of satellite-based global quasi-real-time rainfall product for flood forecasting and warning on a riverbasi n scale as a joint research with the Japan Aerospace Exploration Agency (JAXA) and 2) the development of a common basis for quick & efficient implementations of flood forecasting and warning systems even in poorly-gauged basins as a joint research with Infrastructure Development Institute (IDI), and nine major civil-engineering consulting companies in Japan. The details of the researches are described. Introduction - floods in Asia As stated in die Hyogo Framework for Action 2005-2015, "Building the Resilience of Nations and Communities to Disasters", adopted at the World Conference on Disaster Reduction in January 2005 at Kobe, Hyogo, Japan1', to "identify, assess and monitor disaster risks and enhance early warning" is one of the priorities for action in the next ten years. Early warning is indispensable not only for tsunami disasters but also for flood ones because floods are very repetitive phenomena in many regions and therefore have been causing the Fig. 1 Trend of water-related disasters by type ( 1960- major portion of casualties and economic losses 2004) among natural disasters in the world. Fig.l shows third of the world's total water-related disasters. the recent trend of water-related disasters by type for Besides, the number is still increasing. This trend the period of 1960-2004. The definition of should have some relationship with the facts in Asia "disaster" here is the ones with the death toll of mat many regions in Asia are suffering from "too more than 10 or victims of more than 100. The much" water brought by monsoon or tropical number of disasters by flood or storm is increasing, cyclones/typhoons, plenty of sediment yields, and especially in the recent decade. As a result, flood is population increase, especially cities located in the most major cause of natural disasters in the flood-prone areas. Although flood had not only world. Fig.2 shows the recent trend of water-related negative impacts but also beneficial aspects in terms disasters by continent for the period of 1960-2004. of water & aquatic-life resources, especially in Water-related disasters in Asia now accounts for one the case of annual seasonal floods, flood disaster 133 Those problems would make it difficult that flood (!««•< *l wtltffU\t4 4H*Mt *1 CHIIaot) UM0»1M4) forecasting and/or warning system fulfill the mission to reduce flood disasters. In this context, ICHARM extracted the following major research themes to cope with every problem at each step above: a)Development of satellite-based global real-time rainfall map for flood forecasting and warning on a river basin scale b)Development of rainfall forecasting system for ungauged basins c)Development of a common basis for quick & Fig.2 Trend of water-related disasters by continent efficient implementations of flood forecasting and (1960-2004) warning systems even in poorly-gauged basins reduction is crucial for Asian countries to achieve a d)Development of a guideline for flood warning sustainable development. dissemination to meet local flood-plain needs in different natural/social/monitoring Typical Problems for flood forecasting & warn e)Development of a guideline for integrated flood ing system management combined with other structural & Flood forecasting and warning system is very non-structural measures important for reducing flood disasters, especially in At the first stage of the initiating ICHARM developing countries since structural measures are research activities in terms of flood forecasting and very limited there. It is, however, necessary to make warning, ICHARM has already started the the following steps work properly so as to lead to researches a) & c). The latter part of this paper real flood disaster reduction: introduces the current plan and situation of the two l)Monitoring of meteorological & hydrological research themes: conditions 2)Flood forecasting ca Development of satellite-based global real-time 3)Analysis of forecasts and the judgment of rainfall map for flood forecasting hazardous risk A reliable and useful flood forecasting 4)Dissemination of warning requires on-line real-time rainfall data. Although 5)Crisis management (flood fighting, evacuation, the actual definition of "real time" should be etc.) dependent upon the its requirements and the availability of on-line data at the target area, the However, we often see the following typical prob availability of ground-based real-time rainfall data lems for each step: does not seem enough for flood forecasting in many rivers in Asia, despite of the high risk of flood l)Low density of gauging stations, low sustainabil- hazards. Even if the real-time telemetry system for ity of maintenance of observatories, lack of his raingauges and water level gauges was once torical hydrologie database, etc. implemented in the past, it is often the case that the 2)Lack of real-time hydrologie data, and therefore telemetry system is no more working well because difficult situation to construct and run forecasting of poor maintenances for the ground-based scattered & warning system and complicated system. Besides, in the case of 3)Lack of historical hydrologie & statistical data of basins of transboundary rivers, it is difficult to flood events and damages, therefore difficult establish a general telemetry system to cover the situation to judge hazardous risk compared with entire basin. Then real-time hydrologie data in the real-time information and/or simulations, etc. upstream country does not reach the downstream 4)Lack of disaster-management community and country properly. Based on this understanding, communication network, incompatibility of flood ICHARM has taken notice of the possibility to use information with local society and needs, etc. satellite-based global rainfall products in the world 5)Improper governance, insufficient institutional for flood forecasting/warning purpose. Table 1 cooperation, etc. shows two examples of such available databases. Their products are made from the combination of the microwave sensors' products and geo-stationary 134 Table 1. Examples of satellite-based rainfall products 80.000 70,000 -- Qsim(3B42V5) Product name 3B42RT CMORPH 60.000 — Qsim(3B42V6) Builder NASA/GSFC NOAA/CPC k Coverage 50N-50S 60N~60S 50.000 — Qsim(Ground Rainfall) < , ' Spatial resol. 0.25" 8km 0.25" 0.25" 0.5* co 40.000 • Qobs (monthly) ••/. Temporal resolution 3 hours 30 minutes 3 hours 1 day 3hours £ 30.000 Delay of data delivery 10 hours 15 hours 20,000 Timing of data Every 30 Every 3 Every one Every 3 Every 3 hours min. hours day hours 10,000 updating (UTC) (UTC) (UTC) (UTC) (UTC) ß& 0 -^^^A H 1 h^= Coordinate system WGS Recent 4 Dec. 2002 Jan. 2004 Dec. 2002 Data archive Dec. 1997 ~ days 1998 TRMM-TMI, DMSP- Fig.3. Comparison of runoff simulations with ground- SSM/I. Data source Aqua- DMSP-SSM/1, TRMM-TMI. Aqua-AMSU-B, and IR based and satellite-based rainfall data at Pakse in the AMSR-E. 2 AMSU-B Mekong River (545,000km ) using the BTOPMC andIR developed by Yamanashi Univ. infrared data, so the relatively high accuracy, rainfall (mm) which micro-wave radiometer originally has, is not guaranteed. Besides, we should pay attention to sampling errors generated by sparse snapshots from satellites. In this context, ICHARM has started a joint research project with Japan Aerospace Exploration Agency (JAXA) to develop our original quasi-real-time satellite-based global rainfall product using currently existing satellite data, with finer resolution in time and space (3 hours, around 0.1 deg.) and more detailed information about its Fig.4. Comparison of runoff simulations with ground- accuracy compared with ground-based raingauge based and satellite-based rainfall data at Oomagari in data, under die cooperation of a JST-CREST the Kitakami River (Area = 6,400km2) using the research group titled "Production of High Precision storage function method (MLIT & IDI, Japan) and High Resolution Global Precipitation Map by Using Satellite Data" led by Prof. Okamoto of the 2 Osaka Prefecture Univ. ' The update information of calculations with satellite data, but this result is very the study will be presented at the Session. encouraging for us on the future of "satellite-based At present, the temporal and spatial hydrology". This is one motivation to start this kind resolution of satellite-based rainfall products is of research at ICHARM. never enough fine for flood forecasting in relatively small-scale rivers, especially for flash flood, they Development of Integrated Flood Analysis Sys seems significant for river basins more than around tem (IFAS)3) 2 5,000 km , especially for continental rivers. Fig.3 In order to implement flood forecasting and shows an example of runoff analyses with 3B42V6 warning system, flood runoff modeling and analysis of NASA. The BTOPMC developed by the Yama are fundamental basis. It is, however, still difficult nashi University was used as a hydrologie model to apply advanced research results on hydrologie calibrated by ground-based rainfall data. If the old modeling and analytical methods directly to flood version 3B42V5 was used, then me result was poor. runoff analyses in "poorly-gauged" river basins, However, die coincidence between simulations and especially in developing countries, because the observations was much better when the 3B42V6 limitation of the availability of hydrologie and was used. This shows the accuracy of the new geophysical data. Regarding hydrologie data, not version of satellite-based rainfall data was much only historical but also real-time data is important improved. Fig.4 shows another example of runoff for tiiat purpose. Therefore, one problem for calculations with 3B42RT at the Oomagari station implementing flood forecasting and warning system (6,840km2) of the Kitakami River. The error of in developing countries is how to develop a flood peak discharge was around 17% and the error of forecasting model for a river basin without enough total runoff was just 5%. It is not guaranteed to historical and real-time database. Another problem get such a relatively good result on flood runoff is how to implement state-of-the-art GIS and 135 hydrologie analytical methods and tools with user- distance- squared method, or the Krigging method. friendly interfaces at minimal cost. In order to cope The satellite-based rainfall data assimilation is a with such issues, the Incorporated Administrative future subject for IFAS. Agency Public Works Research Institute (PWRI) of The scarcity of hydrologie observational Japan (as of 2005) started a cooperative research network means the lack of historical hydrologie project (FY2005-06) to develop an integrated flood database. The parameters of a rainfall-runoff analysis system (IFAS) for poorly-gauged basins hydrologie model are usually determined by with the Infrastructure Development Institute (IDI) historical hydrologie (rainfall & river discharge) and major private consultancy companies in Japan data. Therefore, the lack of hydrologie data is a in the field of civil engineering. question of vital importance for flood forecasting and warning. Of course, if we do not have any The following questions are to be coped with for die data at all, it is almost impossible to make flood development of IFAS: forecasting. Here, the authors assume the situation that there are some data but not enough. In such a 1) How can we prepare the required (historical and situation, the usage of distributed-parameter real-time) hydrologie and geophysical data for hydrologie model, the parameters of which are rainfall-runoff modeling as the base of flood related with geophysical conditions, is to be forecasting and warning system in "poorly- considered. gauged" river basins, especially in developing The authors tested several distributed- countries? parameter hydrologie models for IFAS. The criteria 2) What kind of hydrologie (rainfall-runoff) model to select a default hydrologie model for IFAS were is suitable for flood forecasting system under the as follows: condition of such limited data availability? • Wide availability in the world 3) How should the system be to contribute the • Data requirement, performance implementation of flood forecasting and warning • Availability of default parameter set system in developing countries? • Linkage with GIS • Stability of parameters with temporal and Our cooperative research group took the following spatial scale approaches for the issues above: • Easy to apply to real river watersheds and flood forecasting there A) To prepare the capacity to utilize satellite-based • Light model to run rainfall information, which is available all over • Free right for the developers (the joint research the world, as a quasi-real-time hydrologie data group) to use, modify and distribute input to a rainfall-runoff model? B) To prepare the capacity to analyze parameters of As a result, the PWRI-distributed model was rainfall-runoff model with globally available GIS selected tentatively as a default hydrologie model database. (Fig.5)4). The PWRI-distributed model was excellent C)To prepare a modeling framework and user- in terms of the data requirement & performance, friendly interfaces which can be used for any stability, easiness & lightness to use, linkage with grid-based parameter-distributed hydrologie GIS and free rightt o use for the authors. The model model as the engine of the flood forecasting and does not have enough applications in the world, but warning system? this can be compensated with ongoing researches D)To make the system small & light as much as considering a variety of experiences in Japan. possible by focusing on flood forecasting and IFAS prepared not only satellite-based & runoff analyses ground-based rainfall data input modules but also GIS tools to analyze geophysical data and estimate Regarding A), the authors has already parameters of the hydrologie model. Topographical mentioned above. A rainfall-runoff model for and river channel network parameters are estimated IFAS is supposed to be a mesh-(grid-) based one as from USGS-GTOPO30. Parameters for land- described later, and therefore, any kind of atmosphere interactions are estimated from the rainfall source data is input to the model as land-use data of USGS-GLCC and, if required, mesh-based data. Ground-based rainfall data can satellite imagery such as MODIS. Parameters for also be converted to mesh-based data with soil moisture accounting are estimated from the Thiessen - polygon method, the inverse soil texture data such as UNEP / DEWA / GRID. engine module, and 5) calculation results output i Ï&H>Ï module. Main module is for initiating and managing the whole functions. One whole 5./ Q.r management process is set as a project file and then the module can be copied. The multiple main modules can be run. Rainfall data input module can Si Qi handle both satellite-based and ground-based rainfall data. Data filling and temporal & spatial So interpolation can be done. The data are converted -fr y y y into mesh-based rainfall data as mentioned above. GIS data input and analysis module is to import external GIS data, to make flow direction & river \ <3o channel network automatically, to arrange GIS data -m*>v for hydrologie model and to evaluate the parameters of the hydrologie model. Runoff calculation engine 5, -*> Q. module is the rainfall-runoff model. As for the -K IFAS Ver.l, the PWRI-distributed hydrologie model hL was selected tentatively as a default engine as C,3 mentioned above. The hydrologie model can be replaced by any other conceptual or mesh-based Upper layer: distributed-parameter hydrologie model. Q : surface runoff (Manning eq.) sf Calculation results output module is to graphically Q¡: subsurface runoff (Darcy) display the outputs of the dataseis and the outputs of Q0: percolation (Darcy) the runoff calculation engine. The results are shown Lower layer: Qgi: Unconfined groundwater runoff by plan view, three-dimensional view, tables & Qg2: Confined groundwater runoff graphs. It is also possible to animate the outputs River routing: dynamically with a certain time step. IFAS is Kinematic-wave method developed by Microsoft-Visual C/C++, Visual Basic and ACESS on the Windows. The SAGA-GIS, a open-source free GIS5), is used as a GIS analysis Fig.5. PWRI-distributed model engine. Regarding the engine for database Parameters for groundwater are estimated referring management, Microsoft-Access is used. to geologic data such as CGWM. IFAS secures its potential applicability to any river basin in the world Concluding remarks through the preparation of modules to estimate ICHARM's activities are sub-divided into parameters from such globally-available GIS three categories: research, training (capacity databases and the combination with a GIS-based building) and information networking. A set of hydrologie model. studies on flood forecasting and warning described Although the PWRI-distributed model was here is one of the major three research themes at chosen as the first default hydrologie model to ICHARM. Through such research activities, develop, check and test the IFAS and their internal ICHARM likes to contribute to the flood disaster interface modules, as described above, IFAS is to be mitigation in Asia and the world. open also for any other mesh-based distributed- parameter hydrologie model since there is certainly Appendix 1 a hydrologie model most fit to each region of the If we can always use rainfall data retrieved world. Such flexibility and easy maintenance for from microwave sensor data, then the accuracy of upgrading a hydrologie model is to be secured by an flood runoff analyses would be much more object-oriented modeling with design-patterns improved. At present, most of satellite-based aiming for enhancement of reusability and rainfall products using present microwave self-intelligibility. radiometers and geostationary-infrared data. There Considering the concept of IFAS described is a joint plan of JAXA, NASA, and ESA for Global above, IFAS is composed of four main modules: 1) Precipitation Measurement (GPM) project in the main module, 2) rainfall data input module, 3) GIS future6),7). In this plan, there are constellations of data input and analysis module, 4) runoff calculation satellites in polar orbit, each carrying microwave 137 Then GPM will implement high-frequency (every by microwave sensors and more reliable three hours) global precipitation measurement only satellite-based rainfall product (Fig.6). Core Satellite Constellation Dual-frequency Precipitation Radar Satellites Microwave Radiometer 8 Satellites with Microwave ^Observation of rainfall with more Radiometers accurate and higher resolution »More frequent Observation »Adjustment of data from constellation satellites JAXA (Japan) Cooperation :NOAA(US), Dual-frequency Precipitation Radar NASA(US), ESA(EU), NASA(US) CNES/ISRO(France/lndia) and Satellite Bus, Micro-wave others Radiometer Global Observation every 3 hours Fig.6 Concept of Global Precipitation Measurement (GPM) Project Appendix 2 3) K.Fukami, N.Fujiwara, M.Ishikawa, M.Kitano, International Flood Network (IFNet) has T.Kitamura, T.Shimizu, S.Hironaka, S.Nakamura, recently started a trial service of the Global Flood T.Goto, M.Nagai and S.Tomita: Development of an Alert System (GFAS) on their WWW page8' using Integrated Flood Runoff Analysis System for the NASA-3B42RT data with some features for Poorly-Gauged Basins, Proceedings of the 7th end-users such as the data provision in text for end International Conference on Hydroinfomatics (HIC users, the indication of heavy rainfall with 2006), vol.4, pp.2845-2852, 4-8 September 2006, corresponding return period, e-mail notification of Nice, France. the heavy rainfall, etc. Therefore, the IFNet-GFAS 4) T. Suzuki, and A.Terakawa: Development of is substantially a heavy rainfall alarm as trigger physics-based distributed model for operational hy- information for flood forecasting and warning in drological forecasting, Civil Engineering Journal, each country. vol.38, No.10, pp.121-126, 1996. (in Japanese). 5)http://www.saga-gis.uni-goettingen.de/html/ References index.php 6) http://gpm.gsfc.nasa.gov/index.html 1) "Hyogo Framework for Action 2005-2015: 7) http://www.eorc.iaxa.ip/GPM/index_e.htm Building the resilience of nations and communities 8)http://gfas. internationalfloodnetwork.org/ to disasters", World Conference on Disaster gfas-web/ Reduction, Kobe, Hyogo, Japan (2005) 9)ICHARM homepage:http:// 2) http://www.radar.aero.osakafu-u.ac.jp/~gsmap/ www.icharm.pwri.go.jp/ index english.html 138 Modified Remote Sensing Information Model of Water Erosion on Hillslopes D.Y. Shen, K. Takara Kyoto University, Japan shendayons(a),rap. apri. kyoto-u. ac.jp: takaratcpflood. dpri. kvoto-u. ac. ip Introduction Section 2 will introduce the modeling idea of In terms of main types of erosion models Remote Sensing Information Models. (Lane et al., 1988), empirically based models (e.g. Universal Soil Loss Equation (USLE)), partially Modeling idea conceptually based and partially empirically based Remote Sensing Information Model is models (e.g. Unit Sediment Graph (USG)), and created using the method of imaging remote sensing partially physically based and partially empirically information and geographic information. After based models (e.g. Water Erosion Prediction Project image registration, the model will be used for (WEPP) model) are Mathematically Formal Logic calculation pixel by pixel via inputting variables, (MFL) based erosion models while Remote Sensing which are interpreted from both remote sensing Information Models such as Remote Sensing information and imaged geographic information Information Model of Water Erosion on Hillslopes (Ma, 1997). (RSIMWEH) are Mathematically Dialectical Logic (MDL) based erosion models (Shen & Takara, 2006) as shown in Fig. 1. Empirically based models Partially conceptually based and partially MFL based erosion models empirically based models Partially physically based and partially empirically based models MDL based erosion models Remote Sensing Information Models Fig. 1. Development of water erosion models (Shen et al., 2002). RSIMWEH is a MDL based erosion model, Coefficient and exponent are variables; which coincides with the three main laws of dialec Theoretically represents all affecting factors; tics as presented by Engels (1940) in The Dialectics Setting model parameters as variables can repre of Nature. These laws are the transformation of sent complicated spatio-temporal changes of a quantity into quality, the unity of opposites, and the geoscientific phenomenon or process, mus can negation of the negation. Main characteristics of further explain the different calibrated values of MDL-based erosion modeling are listed as follows model parameters while applying the model to (Shen & Takara, 2006): research areas with different spatio-temporal scales. 139 RSIMWEH was created based on dimensional The methodology of RSIMWEH is shown in Fig. 2. analysis and written as follows: Firstly, mapping and grading to provide /-/„ a basis for scientific sampling. After that, initializing E,i — xo hslfism2afexp{-x,Sj (i) and calibrating model parameters based on meas ured data collected from sample points, derived data from models, and remote sensed imagery data. Fi where s (mm) is depth of soil erosion, (mm nally, soil erosion is computed pixel by pixel. 1 min" ) is rainfall intensity, ° is critical value of /. f Details of initializing and are as erosive rainfall intensity, (mm) is depth of follows: overland flow, •^1 ? ^2 ' 3 ' 4 coefficient and are geoscientific (i-i Y ... *0 ' Xl ' X2 ' XZ ' X4 j , . And mdices. m the model are set as variables, which explicitly show MDL modeling E = c(i30y (3) idea. Methodology where ' are pending coefficients and ' are *i- ••nm mapping pending exponents. Then let *\ ••nm grading n I-In (4) V •7Im sampling = c(hoY sampling Finally we get I-I W / (5) solving = ( 3o)" i.e. calibrating (6) computing I0=J{IMIJ Fig.2. Methodology of RSIMWEH (Shen & Takara, 2006). where is a pending coefficient and is a pending exponent. 140 f Model evaluation (ü) initialization Evaluation of RSIMWEH and WEPP (Foster, 1982) was carried out using the data from the 1 erosion plot experiments carried out in Wufendigou f =K K watershed, Inner Mongolia, China. Each erosion Set , where is soil erodibility defined plot is of 11.05m length and 4.52m widdi. Table 1 in USLE (Wischmeier & Smith, 1978), which shows the initial plant in each plot (Jin et al., 1993). depends on the organic matter, texture, permeability Then we used data measured from 8 rainfall events and profile structure of the soil. occurred in 1987 in the research area to do (iii) initialization modeling. Figures 3 and 4 show the simulation results. When comparing RSIMWEH with WEPP Based on MDL based modeling idea, model, we find their simulation results are quite integrates both indeterminate factors and unknown similar as shown in Fig. 3 and Fig. 4. Note that we get the simulation results after the calibration of both RSIMWEH and WEPP. factors of soil erosion. In this study, is initialized as 1.0. Table 1. Initial plants in the erosion plots. Plot ID Initial plant 1 Bromegrass 2 Astragalus adsurgens Pall 3 Bromegrass 4 Natural Pasture 5 Natural Pasture 12 3 4 5 6 7 8 9 10111213141516171819202122232425262728293031323334353637383940 measured erosion sample points RSIMWEH simulated erosion Fig. 3. Erosion simulated by RSIMWEH. 141 o> 4 1 2 3 4 5 6 7 8 9 101112 1314151617181920 212223 24 25 26 272829303132 3334 353637 383940 measured erosion sample points WEPP simulated erosion Fig. 4. Erosion simulated by WEPP. Conclusion References: This paper firstly summarizes the development of water erosion models, and then Foster, G. R., 1982. Modelling the erosion process, focuses on the modeling idea and methodology of in Haan, C. T., Johnson, H. P., and Brakensiek, D. RSIMWEH, a MDL based erosion model, which L. (Eds.), Hydrologie Modelling of Small Water has the advantages of both physical models and sheds. ASAE Monograph No. 5 American Soc. Of empirical models. The model was created by Agrie. Engr., St Joseph, Michigan, 295-380. integrating dimensional analysis and statistical Jin, Z. P., Wang, Z. W., Gao, Z. M. & Qi, N., 1993. analysis. Each model factor has clear geographical Study on laws of water and soil loss in Wufendi- and physical meaning. Moreover, model parameters gou experimental plot and on prediction of soil are set as variables, which well represent erosion by water in Huangfuchuan watershed. In: complicated spatio-temporal changes of a Collection of experimental investigation of com geoscientific phenomenon or process, thus can plete development of agriculture, forestry and hus further explain the different calibrated values of bandry by comprehensive treatment on water and model parameters while applying the model to soil loss in Huangfuchuan watershed, Loess Pla research areas with different spatio-temporal scales. teau, China Agricultural Scientech Press, Beijing, (in Chinese) Acknowledgements: This work is supported by Lane, L. J., Shirley, E. D. & Singh, V. P. 1988. Mod JSPS (Japan Society for the Promotion of Science) elling erosion on hillslopes. In: Anderson, M.G. Fellowship and the MEXT Grant-in-Aid for (Ed.), Modelling Geomorphological Systems: Scientific Research (No. 16004360). Special thanks 287-308. New York: John Wiley & Sons. to Prof. Ma, A.N. and Prof. Lin, H. for their help and Li, S. L., Cai, Q. G & Wu, S. A. 1999. Calculation advice. We are also grateful to Senior Engineer Mr. on soil losses by rainfall. Journal of Sediment Re Jin, Z.P. from the Inner Mongolia Institute of Water search (1): 49-54. (in Chinese) Resources Research in China for kindly providing Ma, A. N., 1997. Remote Sensing Information Mod related data. els. Peking University Press, Beijing, 165pp. (in Chinese) 142 Shen, D. Y, Ma, A. N., Mao, S. J. & Yang, P., 2002. Development of modeling erosion by water on hill-slopes. Journal of Soil and Water Conserva tion, 16(6):43-45. (in Chinese) Shen, D. Y & Takara K., 2006. A modelling and 3-D simulation system for water erosion on hill- slopes—M3DSSWEH. Journal of Hydraulic Re search (in press). Wischmeier, W. H. & Smith, D. D., 1978. Predicting rainfall erosion losses - a guide to conservation planning. Agrie. Handbook No. 537, US Dept. of Agrie, Washington, D.C. 143 Capacity Building-Application of Geoinformatics for Disaster Lai Samarakoon1, Takashi Moriyama2, Chu Ishida2 'Geoinformatics Centre-Asian Institute of Technology, Thailand, 2Disaster Management Support System Office- Japan Aerospace Exploration Agency, Japan ABSTRACT Natural disasters are extreme, sudden events triggered by environmental factors causing unexpected and shocking impacts on societies. Earthquakes, windstorms, floods, and disease all strike anywhere on earth, often without warning. Asia is not exceptional in the receiving end of most of these devastating disasters in recent years. Among them Tsunami disaster in 2004 still haunts the region and until today societies are struggling to recover from the huge loss. Apart from this rare event, one of the most devastating disasters in Asia is flood. This causes death and huge economical loss in most of Asian countries, and this has become a regular annual calamity. It is well known fact that natural disasters are taking place everyday, anywhere and it is the risk reduction that has to address to mitigate natural disaster calamities. The challenge of natural disaster mitigation is to reduce the current natural disaster damage and limit the increase in future damage while at the same time continuing to manage and address issues associated with sustainable occupation and use of natural disaster prone areas. General mitigation program may includes pre disaster preparedness programs to post disaster relief and development programs, but in either case empowerment of societies with information and technical know-how plays a major role in mitigating a given natural hazard. Japan Aerospace Exploration Agency with the collaboration of Geoinformatics Center of Asian Institute of Technology is engaged on technology transfer activities of the use of Geoinformatics in disaster mitigation activities, specifically in Asian region. Several project based training programs referred to as Mini-Projects are being conducted at the Geoinformatics Center with the participation of several countries including Bangladesh, China, Cambodia, Nepal, Sri Lanka, Philippines and Vietnam. The focus areas of ongoing capacity building activities are in the area of flood mitigation, landslide hazard zonation and identification of drought prone areas. This paper present the vision, target of capacity building programs and brief summary of ongoing capacity building activities, implementation plan, achievements and finally a case study conducted in Bangladesh to identify safer location for temporary housing of people during a flooding season. KEY WORDS: Capacity Building, J AXA, Disaster, Remote Sensing, GIS INTRODUCTION Japan Aerospace Exploration Agency training programs were carried out at GIC inviting (JAXA) previously known as National Space participants from the region who are working in Development Agency (NASDA) of Japan has been national agencies. Structured courses were contributing to capacity building in remote sensing conducted for two weeks at GIC with full and related space technologies in Asian region with sponsorship of JAXA. The sponsorship included the cooperation of Geoinformatics Center (GIC) airfare to AIT in Bangkok, accommodation and previously named as GIS Application Centre (GAC) living expenses in Bangkok, and tuition fee at of Asian Institute of Technology since 1995. With GIC. This activity was continued until year 2003 the success of the first course conducted with the satisfactorily training more than 400 people. The collaboration of GIC, JAXA continued to support other type of training program referred to as the region with more training courses covering Caravan training programs are being conducted remote sensing, GIS, GPS and application of these locally with the collaboration of local agencies. It is technologies. JAXA supported capacity building expected that this program could offer opportunities and information sharing in the region was carried to a larger audience to increase awareness in remote out in number of initiatives since 1995 such as sensing, GIS and GPS as they conduct locally. structured training programs, Caravan training Generally, the duration of Caravan training is programs, Mini-Projects and Workshops. Structured for five days targeting a topic that is relevant to the 145 country concerned. Having helped to create a GIC to use remote sensing, GIS and GPS technolo favourable environment for use of remote sensing gies and other relevant information in brining in best information and GIS tools, a new type of program solution for the objective selected by them. was launched to enhance technical competency in This develops self-confidence of participants as the adopting these technologies in operational basis. program structure allows them to develop individual With mis aim, a new capacity building program skill by working on a project that is relevant to their called 'Mini-Project' was launched. Mini-Project individual organizations. Table 1 shows topics are selected by two agencies referred to as Mini-Projects carried out in the year 2005, which is "user agency" and "service providing agency" and the second year for the new initiative. Most of them it is expected both of these agencies nominate are successful but it is not possible to say that all participants to work on the selected topic together project yielded good results. Success depends on the with GIC staff at least for a year with short participants' basic knowledge, level of education term visits. Specific training is provided at GIC, and enthusiasm. Further, available satellite data, fieldwork will be carried out together and field data and time that could spend to integrate if necessary local support is provided. Ample time the phenomenon with Geoinformatics plays an is provided to participants to work independently at important role in the success ratio. Table 1. Themes of Mini-Projects conducted in year 2005 Theme Country Water Induced Disaster Management - A case study on Application of Remote Sensing and Bangladesh GIS Techniques for Flood Mitigation Cambodia Land Use/Land Cover Changes and Flood Risk assessment in Cambodia Using RS & GIS Nepal Application of Remote Sensing and GIS for Earthquake Disaster Mitigation in Kathmandu Integration of RS & GIS with Flood Simulation Models for Flood Hazard Mapping and Nepal Mitigation - A Case Study of Bagmati River Rice Area Mapping and Backscatter Analysis Using Multi-temporal Radarsat Images in the Philippines Rainfed Areas of Pangasinan and Nueva Ecija Modeling the Spatial Occurrences of Rain-Induced Landslides and Identifying Potential Philippines Landslide Hazard Zones Using RS/GIS as a Tool Sri Lanka Application of RS & GIS Technology for Landslide Susceptibility Assessment Vietnam Application of Multi-Temporal Satellite Data for Land-Use/Land-Cover Change and Flood Mapping in the Coastal Zone of Vietnam Vietnam Application of Conventional and Spatial Data in Detection of Underground Karsticic Formations to Store Excess Extreme Floodwater Flows in the Red River Delta in Vietnam 146 Remote Sensing and GIS for Flood Vulnerability 2200 mm in the northeast. About 85% of the rainfall Mapping in Bangladesh - A case Study occurs during monsoon i.e. from June to September. One of the Mini-Project conducted in year Flood comes from three sources: direct rainfall, over 2005 is highlighted in this paper as a case study of bank spills from the major boundary riversan d over application of Geoinformatics in water induced bank spills from the international or regional rivers. disaster managements application. The objectives of In order to evaluate potential of satellite data in the study were to prepare flood hazard map using flood area mapping, attempted was made to use various satellite data, develop a criteria and identify various sensor data which are readily available with suitable locations for shelters, identify best route for reasonable cost. Also, as the area is frequently evacuation as a mitigation approach. overcastted, it was considered to compare the potential of SAR data with more familiar optical 1. Study area and Data data. Following are the data used for the study. Study area, Munshiganj district, is located • Landsat TM 2002 February south of Dhaka city (Figure 1). The area of the • JERS SAR (L Band) 1996 June district is about 919 sq km. The population of the • ADEOS AVNIR 1996 November study area is about 1,294,000 according to the • RADARSAT (Scan SAR Wide mode) 2004 July population census of 2001 (BBS 2004). The main • SRTM2000 physiographic units of the area include the • Population 2001 taken from the Bangladesh floodplains of the Padma, the Jamuna, the Meghna Bureau of Statistics and the Old Brahmaputra rivers. Average annual • Administrative boundaries, roads, water bodies, rainfall ranges from 1400 mm in the southwest to Agricultural land, Metal roads, School/College and Hospital from LGED Figure 1. Location map of the study area 2. Methodology 3. Analysis procedure and derivation of Conceptual flowchart of the analysis is products shown in Figure 2. This figure shows the main input Land Use map: Vector data layers of land information, analysis method and output. Data cover of the study area were generated for presented in the previous section were used in preparation of land use map. The data layers have generating various interim products such as land use been generated from aerial photographs taken during maps, flood maps, elevation map, shelter location, 1999-2000. It was difficult to generate algorithm for flood vulnerability map etc. before completing the creating automated land cover classification as the analysis identifying best routes for evacuating to land use of the study area was heterogeneous. identified shelters during a normal flood season. Therefore, visual interpretation was carried out using ADEOS AVNIR and Landsat TM for land cover mapping. 147 Satellite Data/Aerial photo Schools/Colleges/ Hospitals Population location (Pop. affected by flood) Easiest Path & Evacuation Plan Figure 2. Flowchart depicting major steps of analysis Flood area and Food depth map: This was were estimated using published population in each generated combining optical and SAR data. Attempt land cover category. was made to combine three type of sensor data as a Flood vulnerability map: Flood way to demonstrate the practical usage of satellite vulnerability map was generated comparing data during rainy season where it is impossible to population affected by flooding. In order to create acquire optical sensor data. ADEOS AVNIR, JERS this, flood map was compared with population SAR along with and RADASAT were used for density map re-created by re-classifying population interpretation of the flooded and non-flooded areas. densities to land use categories. The final product Using optical images and field observations, was a polygon map showing spatial distribution of accuracy was clarified and ambiguities were flood vulnerable areas and their population corrected. Further, flood map and SRTM, which is densities. In order to use this information in freely available as a global dataset from NASA were evacuation planning process, centeriod of each overlaid and compared in estimating flood depth. polygon was chosen to represent the number of Population by landuse: Population people in each polygon. The criteria used for information received from statistical department was selecting shelters for evacuation was based on the in Union (sub-district) basis. This data merely states accessibility and availability of schools or hospitals the number of people registered in a given Union, in non-flooded areas and distance to roads. but does not reveal real spatial distribution of Cost Surface: Cost is the "trouble" or population. It is unrealistic to say that population "difficulty" to reach a shelter from a place of living. density is uniform through out a given sub-district as If a person is required to select a shelter among the inhabited area could be a fraction of the total several nearest of them, it is necessary to consider land area of a district. The question of estimating a number of factors in selecting the optimal shelter for realistic population distribution was attempted subjects to move in a disastrous situation. Selection incorporating land use categories with population of could not only base on the distance but need to sub-districts. In order for establishing different consider flooded area, flood depth, accessibility etc population densities with respect to land use classes, during an evacuation plan in a flood mitigation population density ratio of 3 major landuse classes work. In the present study four factors mat could 148 influence the selection of a shelter was used to 40:30:20:10, for flood depth, road density, road create the cost. Those are flood deptíi, road accessibility and slope respectively. The detail of density, road accessibility and slope. The percentage weighting and reclassify score are shown in Table 2. of influence by each factor was assumed as Table2. Weighting influence's factors and scoring's factors for cost surface analysis Factors %lnfluence Classes Interpretation score Flood depth 40 Non Flood No flood 1 Flood depth <= lm. Low flood depth 2 Flood depth >l-3m. Moderate flood depth 3 Flood depth >3-5m. High flood depth 4 Flood depth >5m & water Very High flood depth 5 Road surface 30 Metal road area > 10% High density 1 density Metal road area <=10% or Soft road Moderate density 2 area> 10% Soft road area <=10% Low density 3 All road area = 0% No road density 4 Road 20 Inside Metal &Soft road buffer Very high accessibility 1 accessibility Inside Metal buffer 500 m. High accessibility 2 Inside Soft road buffer 100 m. Moderate accessibility 3 Outside Metal &Soft road buffer Low accessibility 4 Slope 10 Slope 0 - 5 % Low steep slope 1 Slope 5-10% Moderate steep slope 2 Slope > 10% High steep slope 3 Shelter's serviceable area: Here geographi 4. Results cal area that a shelter can serve the population was Figure 3 and 4 shows landuse map and considered. The simplest way to achieve this is to the annual flood hazard map, respectively. There are look into the aerial distance from point of origin to four categorize of land use: agriculture, built-up, the shelter location. In this study, the serviceable settlement and water bodies. During normal flood area of each center was mapped based on the total season 61 % of the study area is affected by flood accumulated cost by moving subjects to each and the affect is very prominent in west of the study shelter. Geographical area of each shelter was based area. on the area that provided the least accumulated cost In Figure 5 shows Cost Surface classified in moving subjects during a normal flood season. into three classes; High, Moderate and Low. The This was carried out by cost allocating tool in low cost is areas mostly in high density of road and ArcGIS. outside flood areas. Proposed 89 flood shelters and Best evacuation path: Calculation of the Cost Distance are shown in Figure 6. The Cost best way to evacuate (least-cost path) was done by Distance Surface illustrates smooth continues using Cost distance tool in ArcGIS. The analysis surface of accumulated cost value. The least cost is considers destination and origin together with cost found near the shelters and increased with the surface. The tool calculates accumulated cost of distance. travelling from any particular location to each shelter. 149 Figure 7 shows the best path for evacuating maps could develop products mat surely help the people in each serviceable area of 89 shelters. When decision maker in identifying areas that need special comparing the evacuation path with road networks attention and their spatial distribution in a case of a and flood prone areas, it was found that this analysis disaster. Application of presented here based is rather precise. The best path provides less on Bangladesh context is only a case study of difficulty to travel to a shelter. It was found that application of satellite data and various GIS data in most of them are at road networks and away from locating affected areas due to foods in the study area flooded areas. Several paths were observed as travel and possible areas that are safer during a flood. The paths even they are flooded due to low flood methods and tools presented here can be used in a levels and unavailability of roads. Figure 8 is an similar situation in any other country for similar enlargement of the result providing better view of type of analysis and coupling with better data such roads, flooded areas and location of shelters to as ALOS-PRISM could generate more accurate justify the accuracy of the model. flood hazards maps due to precision of elevation information. Conclusion JAXA is conducting resourceful, productive Acknowledgement and timely training program with the cooperation Authors would like to acknowledge the use of Geoinfornatics Center of Asian Institute of of data of "Mini-Project on Bangladesh flood" by Technology. It showed mat training programs based supported Japan Aerospace Exploration Agency on project approach refer to as Mini-Projects brings (JAXA) 2004-2005 as well as the use of Bangladesh better results as they servers target areas of national data which prepared by Bangladesh Space Research programs of participating countries. Further, it is and Remote Sensing Organization (SPARRSO). evident that integration of GIS together with flood 151 Scope of Flood Hazard Mapping in Developing Countries Shigenobu Tanaka, Rabindra Osti and Toshikazu Tokioka International Technical Exchange Team, International Centre for Water Hazard and Risk Management, Public Work Research Institute, Japan, osti55(a)pwri. go. jp ABSTRACT Since there is always a possibility that the structural measures can not resist the flood above their designed capacity, it is apparent that we should promote non-structural measures. Flood Hazard Map (FHM) especially prepared to facilitate safe and smooth evacuation of people and belongings during flood crisis has been found very effective in Japan with long history of its development. Despite the fact that FHM is developed as a supplementary tool in Japan, it should promptly be promoted as a key component of disaster management in developing countries. In order to promote FHM effectively in developing countries, the foremost task is to build the human resources that capable of executing hazard mapping and its dissemination. In other hand, the preparation technique and the features of FHM that currently practiced in Japan may not be suitable for developing countries unless the local natural, socioeconomic and environmental circumstances are deeply studied and correctly reflected on the FHM. The International Centre for Water Hazard and Risk Management (ICHARM) under the auspices of UNESCO has initiated an approach that bridges the gap between Japanese experience and the current need of developing countries. The trainings and seminars are planned to conduct in a regular basis to enhance the capacity of professionals from developing countries. Moreover, ICHARM aims to address the problems concerning development and application of FHM in developing countries through cross cutting scientific researches and advanced information networking. Key words: Flood Hazard Map, Developing Countries, Trainings, Alliance and, ICHARM Background Flood Hazard Map (FHM), which can provide Among all, flood tops the list among natural information including the past flood track records, disasters for human and economic losses worldwide, anticipated flood, potential evacuation routes, and there are evidences that they are becoming more evacuation places etc to the local residents are frequent and severe. Especially the impacts of flood indispensable for emergency response and for disasters are much serious in developing countries long-term flood disaster management. Keep an eye and the majority of floods' victims are mainly the on the fact that the floods have also some advan poor people. The deteriorating consequences of tages to the community and environment (Kazama such flood disasters in developing countries are et al. 2002), the strategy for co-existence with flood exacerbating through self-exited poverty-cycle and a move from direct use of floods to indirect and phenomenon (Osti, 2004) with several activities controlled use is emphasized (Takeuchi, 2002), closely related to the cause and effect chain at which can be characterized by FHM of the grass root level (Osti et al., 2006). This trend of concerned area. In addition, the role of FHM is destructions due to floods in developing countries immense in preparing flood policy (Faber, 1997) will continue unless suitable coping mechanisms are and in managing all potential flood management well established in advance. However, due to the strategies such as elaborated by Petry (2002) i.e. a) lack of awareness, resources and suitable approach, keep the flood away from people, b) keep people the problem couldn't be solved as in the pace of away from flood and c) accept and act with flood. developed countries. This particular vulnerability of Since making FHM is inexpensive and quick, it developing countries underlines the urgent need to would be the best way to develop social resilience to cope with extremities in concerned areas. promote relatively fast, technically tolerable, Although FHM has been introduced in developing environment friendly and socially accepted countries, there is no overwhelming response to its cost effective structural as well as non-structural implementation and effectiveness. In order to be countermeasures. effective, knowledge and experience of local people 153 should be incorporated and the methodology to Necessity and Importance adopt such tool should be promoted (Osti, 2005). Conventional flood disaster prevention meth In the context of alleviating flood disaster impact, ods, which mainly rely on the construction of flood FHM can play a vital role and research & develop protection facilities, can reduce significantly the ment should highlight its added values and propose damage caused by flood, however because of chang a way forward of its adaptation in such scenarios. ing environments the safety provided by such struc This paper outlines the development and application tural measures is uncertain. Therefore, the role of practices of FHM in Japan, describes its necessity non-structural measures along with structural coun- and importance in the context of developing termeasures in different flooding conditions as countries and elaborates the role of the International described in Tsunami and Strom Surge Hazard Map Centre for Water Hazard & Risk Management Manual (MLIT, 2003) has been emphasized (Fig. 1). (ICHARM) in bridging the current gap between Different names of flood hazard map and different Japanese techniques and need of developing countries. ih g si Probability of occurrence * él Limit of disaster prevention Disaster prévention using structures (tfrwçturigl) '//////fet*^ Dhastcr prevention usina Information tnoii-structural) -Scale of external force No Certain degree of damage Serious damage damage Figure 1. Relationship between structural and non-structural measures for preventing flood disaster approaches such as GIS based hazard map, based on Japanese national mapping project was formally flood affected frequency, based on flood depth and commenced in 1994 under the Ministry of velocity etc have been practiced worldwide. While Construction (MLIT, 2003). The Government of the main purpose of a hazard map is to show the Japan officially publicized Past Inundated Area extent of flooding under a given scenario and Maps for about 500 rivers and also Flood-Prone provide information for smooth evacuation, it can Area Maps during the period 1993 to 1994 also be used as a basis for river management, especially for the rivers under ministerial integrated flood risk assessment, joint risk jurisdiction. Later, municipalities were authorized to assessment, flood education and land-use as well as start production and implementation of FHM in resources management. moderate and small rivers in their vicinity. By the There is no clear idea of the development end of fiscal year 2005, total 440 out of 1800 history of FHM, however the flood hazard aspect municipalities have already prepared and publicized map was initially practiced in Japan in 1947 and 154 the Flood Hazard Maps. Although it is not the case of developing countries, FHM in Japan has been emphasized as a secondary or supplementary countermeasure in the following circumstances. a) Construction of structural measures is time consuming and until the completion of such measures, there is a need for an alternative b) In case if the flow in river exceeds the design water level of assigned structures, FHM can be a great help for safe evacuation c) Once the frequency and magnitude of floods are reduced by structural measures, rapid urbanization will take place in the vicinity of rivers and in case if flooding due to levee breach occurs, damage will be higher and evacuation is inevitable d) Due to the construction of levees and embank ments, people feel overconfidence and may not response well during accidental flood events therefore they need to be aware and trained to evacuate properly in flood crisis In order to achieve those objectives, flood hazard map was developed in such a way that it includes not only inundation areas and depth but Figure 2. Existing situation in HatYai, Thailand a) also information such as evacuation centers & people are enjoying on watching raising water level in routes, disaster management centers, dangerous river, b) red warning signal for evacuation is patched spots, communication channels and systems, on a notice board, Photos by: S. Tanaka 2005. evacuation criteria, tips for evacuation including emergency kits, and mechanisms and symptoms of hazards for educational purpose. The effectiveness of FHM was observed in Koriyama city, Japan after however due to several reasons the development the flood 1998 and it was found that the residents could not achieve its desired goal (Fig. 2). There who evacuated safely one hour earlier than their fore, the importance of FHM can be seen as an entry counterparts in reference to provided FHM were 1.5 to the holistic development of the regions. times in numbers (MLIT, 2003). Despite the fact that FHM is developed as a Among different, facilitating smooth and safe supplementary tool in Japan, it should be promoted evacuation is the most important aspect of FHM as a primary measure in developing countries as a) it development. Evacuation is always very painful is less time consuming, effective, inexpensive and exercise to the affected population and generating environment friendly, b) structural measures can not self motivated evacuation is very challenging task. easily be promoted due to the lack of resources, In another hand, timing of evacuation is much c) planned structures cover only a tiny part of critical as premature and late evacuation may lead to flood-prone community, d) adopted structural hardship and damages. Moreover, assigning right measures are very weak due to poor construction time for evacuation depends on available and management, e) it would be the firsthand information and experience. In the case of Tsunami working tool for poor and vulnerable communities 2003 in Kesennuma city in Japan, more than 90% of and, f) it has long term benefit on land-use and local people spontaneously thought about the resources management (Osti, 2006). Although possibility of floods after the big scale earthquake change of current land-use pattern is somehow observed in the region (MLIT, 2004). Although difficult, it is hoped that the coming generations will many people evacuated safely in reference to the adhere to an idea. Some countries have already provided hazard map, for many others, who embraced an idea of integrated flood management, however are very conscious of disaster, did not take 155 Research on site specific issues Results of Research Identifying site ^specific issues Extensive research activities Reporting activities Sharing activities with each country Enhancing cooperation with external All Trainees: Reporting the work done by the former trainee from their home country Figure 3. ICHARM's working principles an evacuation action promptly. It indicates that the suitable for developing countries unless die local realization of potential flood and having good natural, socioeconomic and environmental knowledge on FHM are not only enough but the circumstances are deeply studied and correctly methods for enhancing people's awareness level and reflected on the FHM. Keep an eye on the above for giving them information regarding evacuation mentioned problems, ICHARM has initiated an should be improved to bridge knowledge and action. approach mat will help to develop the technique and Therefore, FHM should always be integrated with knowledge relevant to developing countries, so that awareness campaign and reliable warning systems. the effective implementation of FHM would be Although it is die fact that the FHM is the possible. best alternative, mere are several anticipated problems for its development and application in Role of the ICHARM developing countries such as (a) data deficit such as ICHARM is a newly established organization hydrometeorological and topographical, (b) lack of with lots of commitments to resolve the water re human, technical, physical and financial resources, lated disaster problems worldwide. Its main objec (c) centralized approach, (d) poor literacy rate, (e) tive is to formulate strategies mat help agencies to improper communication mechanism, (f) public and workout better in advance and to assist their imple political wills, (g) over representation of agencies, mentation process Üirough moral and practical sup (h) unplanned cities/towns/villages, (i) information ports. The strategies will be prescribed reflecting on FHM that practiced in Japan, may not useful for the differences in targeted societies. The implemen different localities or regions. tation assistance of ICHARM would be in the form In order to promote FHM effectively in of capacity building of actively involved profession developing countries, the foremost task is to build als in me field of water hazard and risk manage the human resources that capable of executing ment. Figure 3 illustrates me ICHARM's three hazard mapping and its dissemination. In other fundamental working principles and some of major hand, the preparation technique and the features activities. of FHM that currently practiced in Japan may not 156 Figure 4. Sixteen trainees from 8 countries per year and established network. Especially for the promotion of FHM in and other local experts, who know well the local developing countries, the ICHARM has initiated an circumstances and have responsibility to manage approach to fulfill the current needs through action problems. The networking among ex-trainees oriented research and training programmes. provides a platform to share ideas, solve problems, A month long Flood Hazard Mapping training discuss and disseminate information, where the course has been organized every year for responsi ICHARM will continue to play a coordinating ble public servants from developing countries since and backstopping role in enhancing their active 2004 (Fig. 4). Till date, 32 professionals from East participation. As a part of follow up process, Asian countries especially from China, Thailand, ICHARM has realized the need of regional Vietnam, Laos, Cambodia, Malaysia, Philippines seminars, which are now being planned and will be and Indonesia have received such trainings and organized in a regular basis especially to a) give many others from around the world affiliated with strength to the existing networking, b) update the other related training programs have also acquired recent innovations and, c) give feed-back to the short introductory classes on the importance of progress made by the authorities in the regions. One FHM. As a part of FHM promotion, trainees are of such seminars is scheduled to be held in February requested to draft an action plan and to finalize 2007 in Malaysia. In addition to the former trainees, it later after consulting with their concerned proposed seminar expects the participation of wide organization. Current progress on flood hazard range of experts in flood hazard management from mapping in participants home countries such as different organizations e.g. governments, research FHM under consideration in Lampang Municipality, institutes, Universities etc. In addition to trainings Thailand indicate that they are performing well in and seminars, ICHARM aims to address the their stand. Alliance for localism is a motto of mentioned problems in developing countries ICHARM and an established local networking through cross cutting scientific researches and advanced information networking. Especially for among ex-trainees (Fig. 4) would make it possible to field research in flood hazard mapping, south East share knowledge for better practical application Asian countries have already been selected and of FHM. In order to produce fruitful outcomes, respective governments are yet to be consulted for ICHARM aims to cooperate closely with trainees coordination. 157 Conclusions Reference Damages due to floods are widespread worldwide. Increasing frequency and magnitudes of Faber S. (1997), Flood Policy and Management: A floods, degrading river basins and declining Post-Galloway Progress Report, River Voices, a economic growth as well as poverty concerns make Quarterly Progress Report of River Network the problem more severe and complex, which USA, Vol. 8 No2. now can not only be addressed by widely used Kazama, S., Y. Muto, K. Nakatsuji, and K. Inoue conventional practices especially by structural (2002), Study on the 2000 Flood in the Lower approach. In other hand, the majority of victims of Mekong by Field Survey and Numerical flood events are mostly the poor people from Simulation, Proceeding of 13th congress the developing countries, which are lacking on APD/IAHR, Vol.1, pp.534-539. resources to cope with flood problems. In MLIT (2003), Flood hazard map manual for tech preventing human and economic losses due to nology transfer, Infrastructure Development floods, an application of non-structural measures has Institute, Ministry of Land, Infrastructure and gained noteworthy recognitions for their simplicity, Transport, Japan. ease of use and effectiveness. Flood hazard map, MLIT (2004), Tsunami and storm surge hazard map which can provide information including the past manual, Port and Harbors Bureau, Ministry of flood track records, flood anticipation, potential Land, Infrastructure and Transport, Japan. evacuation routes, evacuation places etc to the local Osti, R. (2004), Community participation and residents has been found indispensable for agencies role for the implication of water emergency response in Japan. Its potential use in induced disaster management; protecting and developing countries highlights the urgent need of enhancing the poor, International Journal of its promotion. In order to be effective, the Disaster Prevention and Management, Emerald techniques of FHM development should be UK, Vol. 13, No 1, pp. 6-12. developed in accordance with local requirements, Osti, R. (2005), Indigenous practices on water where Japanese experience will be the reference. harvesting in semi arid environment of ICHARM has initiated an approach that bridges the Nepal, International Journal of Sustainable gap between Japanese experience and the need of Development & World Ecology, Vol. 12, No 1, developing countries through action oriented pp. 13-20. research and training programmes. ICHARM has Osti, R., Tanaka, S. and Tokioka, T. (2006), Flood been organizing training courses for engineers from Hazard Mapping in Developing Countries: developing countries since 2004 and the established Problem and Prospect, International Journal of network among trainees has been functioning well. Disaster Prevention and Management, Emerald As a part of follow up activities, ICHARM has UK, Vol. 15 Issue 5 (In Press). planned to organize regional seminars aiming to Petry, B. (2002), Coping with floods: complemen update the knowledge of trainees and to tarities of structural and non-structural encourage concerned organizations for effective measures, Proceeding of the 2nd Int. Sym. on implementation of FHM. Research institutes, Flood Defense, Bejing, China, pp. 60-70. universities and otiier organizations, which are also Takeuchi, K. (2002), Flood and society: a involving in flood management in the regions, are never-ending evolutional relation, Proceeding, cordially invited to share their findings in these of die 2nd Int. Sym. on Flood Defense, Bejing, seminars and upcoming conferences organized by China, pp. 15-22. ICHARM. Alliance for localism is a motto of WWAP (2006), UN world water development ICHARM and it aims to promote coordination report, World Water Assessment Programme- among regional and international partners to WWAP, France. facilitate the FHM implementation process. 158 The Prediction of Flooding Area in the Pasak River Basin by Using Mathematical Model : A Case Study on Land Use N. Hungspreug ' and A. Penghuaro2 1 Department of Environmental Science, Faculty of Science and Technology, Thammasat university, Rangsit campus, Thailand [email protected] 2 Department of Water Resources, Thailand ABSTRACT The study on the prediction of flooding area in Pasak river basin by mathematical modelling was carried out based on land use in 2002 and the flood risk map was prepared. The Pasak river basin was divided into 11 sub-basins with a telemetering station in each sub-basin. Two cases of analysis were carried out. In the first case, five factors:-drainage density, stream slope, soil hydraulic conductivity at saturation point, land use and maximum flow-rate were selected to study their impact on flooding. Then weighting and sub sequently classifying the level of impacts into 5 levels by using. Geographic Information Systems (GIS) Also, the statistical relationship between land use and flooding area was determined. Secondly, the mathematical model, MIKE 11, was used to study flooding conditions from Petchaboon telemetering station Muang district, Petchaboon province down to the Pasak Jolasid Dam telemetering station Pattanakhom district, Lopburi province. The calibration and verification of die mathematical model were carried out by using water level and flow rate data. The Manning coefficients were ranged from 0.028 to 0.029. The results of the study showed that the land use of Pasak river basin was divided into 7 types:-field crops (5,958.63 km.2), forest (4,669.58 km.2), paddy field (2,912.44 km.2), orchard and perennial crop (970.50 km.2), urban and built-up land (432.41 km.2), water bodies (188.59 km.2) and miscellaneous land (113.59 km.2). In the first case, if was found that almost all of the area of Vichienburi district, Petchaboon province (833.19 km.2) was the most serious flood risk area whereas Chaibadarn district, Lopburi province (883.78 km.2) was second. This was due to the fact that Vichienburi district had lower drainage density (0.25-0.30 km./ km2.) and lower soil hydraulic conductivity at saturation point (20 cm./mn.) that cause drainage problem and surface overflow. There was a relationship between the most flood risk area with paddy field with a correlation coefficient; (r) of 0.667. In the second case, the prediction of maximum flowrate in 2003 showed that the paddy field was flooded about 70.10% of the total flood risk area. In 2007, the prediction of flooding based on the maximum flowrate data from the Royal Irrigation Department (1993) showed that the flooded paddy field would be 66.70% of the total flooded area. The management of the flood risk area should pay attention to land use especially the paddy field and the drainage system should be promoted to lessen the impact from flooding. Key word : Flooding Area, Pasak River Basin, Mathematical Model and Geographic Information System :GIS Introduction Purposes Pasak river basin covers are an area 15,403 This study has the following purposes: sq. km. The basin is feather shaped with a total (1) to study land uses based on 2002 data and to length of 568 km. and a maximum width of 45 prepare flood risk map and (2) to utilize the km. The basin is flanked by mountain ranges mathematical modeling to study the flooding which cause frequent flooding. The average conditions and to utilize the calibrated model for annual rainfall in the basin is 1,213 mm. with 87% flood forecasting. occuring in the wet season. The average annual runoff is 2,897 million cu.m. The obvious factors resulting in flooding are flow rate, stream slope, land use, drainage density soil characteristics, etc. 159 Theory /Issues drainage density, stream slope, coefficient of soil The study on flooding in the Pasak river hydraulic conductivity at saturation point, land use basin was carried out in 2 cases, namely, Case 1 and maximum monthly runoff. The binary logistic by using Geographic Information System regression was used to prioritize me impacts of GIS (Arc View 3.2a) and Case 2 by using flooding due to those factors. The results from mathematical modelling (MIKE 11 and MIKE 11 these factors were overlaid and resulted in various GIS) with the governing equations of conservation levels of the flood risk map. Moreover, these of mass and momentum as follows: outputs were used to study the relationship between land uses with the flooded areas. Hydrodynamic model: In the approach 2, data collection were cross-sectional areas of river from TS 04 to TS 10, topographical contour, drainage system, and dQ ÔA hydrological data including hourly water level, -^ +— = q hourly flow rate, daily rainfall and runoff data dx dt during January until June 2003. These data were collected at 11 telemetering stations starting from 2 \ the upstream area at Muang district, Petchaboon Q ^ jdh^ gQ\Q\J L a + gA — + —r — = 0 province downwards to the downstream area of dt dx dx C2AR the basin at Pasak Jolasid dam, Patananikom district, Lopburi province. They were input in the Where A = cross-section area (m2) hydrodynamic model of MIKE 11 for model C = Chezy resistance coefficient (m1/2/s) calibration and the determination of the Manning g = gravitational acceleration (m/s2) coefficient. The adopted model was used to h = hydraulic head (m) analyze the hydraulic phenomena in the basin Q = flow rate (m3/s) which were subsequently input in the MIKE11 R = hydraulic radius (m) GIS in order to determine the flooding extent. a = momentum distribution coefficient q = lateral inflow (m3/s) t = time (second) Methodology The methodology used in the study on the prediction of flooding areas in the Pasak river basin comprises of 2 approaches, namely: (1) The use of geographic information system (Arc View 3.2a) and (2) The use of mathematical modelling MIKE11& MIKE GIS. The steps employed in approach 1 was shown in Figure 1. and approach 2 in Figure 2. In the approach 1, data collected were catchment area, drainage pattern, topographical contours (Hungspreug et al.,2003 a and b). They were input in the Arc View 3.2a. The land use data was classified into 7 types, namely, community area, upland crop area, paddy fields, orchard plantation and perennial crop, forest, water body, and miscellaneous area. Five factors were considered to have impact on flooding, i.e., 160 Approach 1 GIS 1. Data collection land use data (2002) and - land use classification 1 2. Data verification 3. Analysis sub-division into 11 sub-basin and - land use classification 4. study on factors and criteria to determine the flooding 5. Prioritization of factors causing flooding causing flooding by binary logistic regression 6. Data overlay 7. Establish the relationship between land use and flooded areas Flood Risk Map Figure 1 Steps in using GIS to predict flooding areas. 161 Approach 2 GIS 1. Data collection - cross-section of river - contour- drainage system - hydrological data as -water level, flow rate -rainfall & runoff data 2. Data input in the hydrodynamic model 3. Model calibration 4. Model verification 5. Adopted hydrodynamic model 6. Application of MIKE 11 with MIKE GIS c ., 1 ) Flood Risk Area Figure 2 Steps in using mathematical model for prediction of flooding areas Results Land use : province and at Chaibadarn district, Lopburi The land use in the Pasak river basin could province with the flooded areas of 833 and 883 be classified into 7 types with the coverage areas sq.km., respectively as shown in Figure 4. The as in Table 1 and in Figure 3. relationship between land use and flooding areas showed that the high flood risk area was the paddy Relationship between land use and flooding field with the correlation coefficient (r) of 0.667 at The use of GIS showed mat high flood the confidence level of 95% risk areas were at Vichianburi district, Petchaboon 162 Table 1 Land use types in the Pasak river basin land use type coverage areas (sq.km.) % upland crop 5,958 38.7 forest 4,670 30.3 paddy field 2,912 18.9 orchard and perennial crop 970 6.3 community area 432 2.8 water body 189 1.2 miscellaneous 114 0.7 Total 15,403 100.00 163 Moreover, the use of hydrodynamic model 16% of the total flooded area. The details of the showed that the maximum flow rate in the river flooded areas on various land uses were shown in resulted in the increase in the quantity of water to Table 2. It could be observed mat the most overflow the banks and flooded the surrounding flooded area had a drainage density of 0.25-0.30 areas. The paddy fields were most flooded with km/sq.km., a slope of 2.1-3.1%, soil hydraulic the area of 428 sq.km. or 70.1% of the total conductivity at saturation point greater than 20 flooded area of 611 sq.km. The upland crop area cm/min. and with maximum dischanrge more than was second with the flooded area of 99 sq.km. or 100-199 cms. land use types areas predicted flooded % of total (sq.km) area (sq.km.) flooded area paddy field 1,783 428.7 70.1 upland crop 4,358 99.4 16.2 water body 172 29.6 4.8 community area 194 24.5 4.0 orchard and perennial crop 458 23.8 3.9 miscellaneous 37 3.3 0.6 forest 2,049 2.3 0.4 Total 9,053 611.5 100.0 Table 3 Criteria for weighting the levels of flood risk Factors Range Weight Drainage density < 0.25880 5 (km/sq.km.) 0.2501-0.300 4 0.3001-0.3500 3 0.3501-0.4000 2 > 0.4000 1 stream slope (%) <1 5 1.1-2.0 4 2.1-3.0 3 3.1-4.0 2 >4.0 1 maximum flow rate >400 5 (cms.) 300-400 4 200-299 3 100-199 2 < 100 1 soil hydraulic conductivity <1 5 at saturation point (cm/ 1.1-5.9 4 min) 6.0-10.9 3 11.0-20.0 2 >20.0 1 land use Paddy field 5 community 4 Water body 3 Upland crop 2 forest and perennial crop 1 Note* weight 5 means very high risk, f means high risk, 3 means medium risk, 2 means low risk, 1 means very low risk 164 The flooded area and the maximum Flood risk map of the Pasak river basin discharge along the Pasak river were shown in In order to establish the flood risk map, the Figure 5. It could be observed that the flooded criteria used in the study of critical areas for area as determined by the hydraulic modeling was water management in the Pasak river basin by much less than that determined by the satellite Thammasat University, 1999 was applied in this map. This was due to the fact that the flooded area study. Five factors were used as criteria in determined by the model was primarily due to the weighting the level of flood risks as shown in overspill of the water from the natural channel. Table 3. They were applied on the satellite map at every sub-basins in order to prepare the flood risk map as shown in Figure 4. Figure 4 Flooded area from satellite (2002) interpretted by GIS Arc View 3.2a 165 Figure 5 Flooded area and maximum discharge along the Pasak river as determined by the hydrodynamic model 166 Conclusion The most highly flood risk area in the Pasak rive basin was in Vichianburi district, Petchaburi province and followed by Chaibadarn district, Lopburi province with the flooded areas of 60 and 63% of the sub-basin areas, respectively, These areas were mostly paddy field. The drainage system of these paddy fields should be promoted in order to mitigate flooding. Moreover, the land use planning in the Pasak river basin should take into consideration cropping patterns and cropping calendar as they had the influence on the amount of water in the river basin. References Danish Hydraulic Institute, A Modelling System for Rivers and Channels User Guide. Agen Alle's, k-2970 Horsholm' Denmark, 2003. Hungspreug N., Anuragsa B., Siiripittayangkool S. and Paikea Y. The Relationship between the Concentration of Chlordane and Endosulfan Residues in Water and Sediment with Land Use in Pasak River Basin. Asian Water Qual 2003:19-23 October 2003 a. Hungspreug N., Anuragsa B., Chanhom S., and Srenetr V. Temporal and Spatial Variation of the Coliforms Bacteria in Pasak River. Thailand. Asian Water Qual 2003.19-23 October 2003 b. Royal Irrigation Department, Feasibility Study and Environmental Impact Assessment on Pasak Dam Project. 1993. Thammasat University.The Project on Determination of Critical Areas for Water Management in the Pasak River Basin. Department of Environmental Science, Faculty of Science and Technology. 1999. Combating Reservoir Sedimentation: A Challenge for Sustainability T. Tingsanchali1. N.M. Khan2 12School of Engineering and Technology, Asian Institute of Technology, Thailand tawatchCäj.ait. ac. th ABSTRACT Importance of reservoirs for supplying fresh water has increased significantly at present. In order to move a step forward towards making the present day un-sustainable reservoirs into sustainable infrastructure, this article reviews the sediment erosion and reservoir sedimentation estimates at global scale and applies the Reservoir Conservation Tool (RESCON) for one of the large reservoir: Tarbela Dam in Pakistan. The article points out the wide variations of the estimates both for soil erosion and reservoir sedimentation and shows how the various reservoir level and duration of flushing will effect in achieving the sediment balance ratio in the Tarbela reservoir. It is found that the flushing can be done efficiently with Sediment Balance Ratio (SBR) greater than 1. The long term sustainable capacity of the reservoir is about 27% with respect to its original capacity. Keywords: Reservoir, sedimentation, erosion, RESCON, Tarbela Dam Background The exponential increase of the reservoirs River's banks have been known as cradles and dams has helped the world to fulfill the food of civilization as most of the large civilizations and fiber requirements of unmatched population have flourished around the banks of major rivers. bulge in last century in addition to providing the As the need of food and security increased people hydroelectric power for rapid economic growth of learned to control and manage this mighty the countries. Over 45,000 large dams are resource through dams. The signs of earliest uses providing irrigation supplies to 30 to 40% of of dams and reservoirs to harness the basic need world's 268 million hectare irrigated land (WCD, of life 'the water' are found as old as 5,000 years 2000). Discounting conjunctive use of ground and in Jordan, Egypt and surrounding civilizations surface water, dams/reservoirs are contributing to (WCD, 2000). Use of dams and reservoirs have 12-16% of world's food production. Reservoirs become more frequent in the past thousands years. and dams have helped to create green revolution The construction of the dams was at its full swing in mid of the last century and are providing 19% in the mid of twenty century and had got the of the world's energy supplies too. highest momentum in 1960-70's (Figure-1) when approximately three dams were commissioned on each day on average. 2,000 • N America S S America B N. Europe • S. Europe 1,800 • Sub Sharan Africa H N. Africa 1,600 m China m S.Asia 1,400 m C.Asia m S.E.Asia m Pacific Rim n Mid. East *È 1,200 g, 1,000 I« 800 600 400 200 F^ , p—y H 0 <1900 1900 1910 1920 1930 1940 1950 Decade Figure 1 : Storage volume constructed per decade (White, 2005) 169 *- -US Irrigated Urad (1910-1) 9 IB Reservoir Storage (1910=1) 8 t; 8 * i i 'i i i • i 1900 1910 1S20 1830 1940 1950 1960 1970 1880 1990 2000 10 • Global Irrigated Land (1910 =1) 140 s -Global Reservoir Storage (1910=1) .120 8 7 ; 100 6 ä 9 4 3 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 Figure 3: Comparison of reservoir storage increase [White, 2005] with irrigated area in the World [Goklany, 2002] Goklany (2002) mentions how U.S. has Similarly at global scale, for a population been able to fulfill the food and fiberneed s of her increase of 251% over 1910-1995, a mere increase population growth of 184% over the period from of cropped area by 95% was not sufficient if the 1910-1995. He states that although the cropped irrigated area increase during the same period was area was reduced by 7% in this period and the not of me order of 435%. The cropped area per cropped area per capita was reduced by 70%, the capita at global scale was reduced by 50% yet, the country was able to cater it through more fresh world over, food and fiber is being provided to water available, which increased the irrigated area over 6 billion population of the world, again by 353 percent by the same time period. This thanks to an increase in the irrigated land per increased irrigation facilities were of course not capita by 50%. The close correlation between the possible without die construction of thousands of increase in reservoirs storage and increase in reservoirs in US in last century (Figure 1) which irrigated area in US and in die World is shown in helped the country to become largest exporter of figure 2 and figure3 . corn and rice, despite reduction in cropped area and cropped area/capita. 170 This data of USA and the World clearly the needs of the population, certainly for the next shows that increased water availability due to 30 years or so. However there are likely to be development of storage reservoirs and related shortfalls in water storage in South America, infrastructure played important role in last century Africa and Asia and this may restrict the to fulfill the food and fiber needs of rapidly aspirations of the population in these regions for growing population. Of course technological improvements to their standard of living'. development in other fields helped in this regard In the present circumstances, the authors too (better seeds, agricultural practices etc) but it are of the view that the current reservoirs should is clear that all these technologies were not be operated with consideration of conservation sufficient without availability of sufficient of storage rather than maximizing the power supplies of fresh water provided by the reservoirs. generation and meeting immediate demands of Due to lack of suitable sites and due to water supply for irrigation, power, navigation, environmental concerns, the development of new recreation. This change in priority is required to reservoirs, on one hand, is reduced (rather stand convert the current day unsustainable and still in US and other developed countries Labadie irreplaceable infrastructure, the reservoirs, into (2004)), while on the other hand the existing sustainable infrastructure, serving even the reservoirs are under threat because of what is coming generations. called as 'insidia solida' (Silvio and Hotchkiss, In order to have a step towards the much 1995). Insidious aspect of the problem is that 1 to emphasized and needed goal of sustainability of 2% loss of capacity of the reservoirs, usually go reservoirs, this article looks into the extent of unnoticed for several years. reservoir sedimentation and sediment erosion. The synergetic effect of reduced Later this article applies the latest developed development and increased loss of existing storage reservoir conservation tool, RESCON can be easily seen by ever growing conflicts (Kawashima et al. 2003) on one of the worst among the competing sectors namely hydropower, affected large reservoir, the Tarbela dam Pakistan, irrigation, water supply, ecology, navigation and to find the effects of various parameters on recreation. As the possibilities of developing new flushing the sediments from the reservoir. Finally dam projects are becoming bleaker day by day, some new research areas are identified for further the need to use the existing dams and reservoirs exploration in the years to come. optimally and sustainability can not be overemphasized. It will result in protecting the Extent of Sedimentation Problem in the World fresh water resources intact and will enable the Mahmood (1987), Atkinson (1996), Morris water managers to reduce poverty of the world to and Fan (1997) and White (2001, 2005) have some extent through provision of freshwater . specifically addressed the engineering and design The need for protecting the reservoirs is aspects of reservoir sedimentation. As the problem being echoed at international fora too. While of reservoir sedimentation has its roots in quantity specifying the New Policy Framework for Large of sediments reaching to the reservoirs, it will be Dams, WCD report 2000 emphasizes that; necessary to review die estimates of the soil '...opportunities exist to optimize benefits from the erosion at global scale, many existing dams, (by) addressing the a. Assessment of Global Soil Erosion outstanding social issues and strengthening Sediment erosion is a complex process environmental mitigation and restoration which is not yet properly quantified rather than measures '. being understood. The erosion is dependent on Similar message was delivered in one of precipitation, geology, geographical location, soil the sessions of 3rd World Water Forum; type, land cover, soil management practices, land 'Whereas 20lh century focused on reservoir devel use, climate and seismology of the area. In the opment, the 21s' century will necessarily focus on words of Lai (2003), 'most available statistics on sediment management, the objective will be to the extent and severity of soil erosion is convert today's inventory of non-sustainable res subjective, qualitative, obsolete, crude and ervoirs into sustainable infrastructures for future unreliable'. The uncertainty can be judged from generations '. the fact that the reported sediment yield figures for Although the fresh water storage of Amazon river basin (World's largest river) are existing reservoirs is reducing annually all over 600 tons/year to 1,200 tons/year amounting to a the world, yet some continents are more variation of 100% (Lima et al. 2005). Similarly, vulnerable, as concluded by White(2005); 'In Walling and Web (1996) have listed 14 estimates North America and Europe the stored water of total suspended sediment transport to oceans availability is likely, in general terms, to satisfy varying from 8.3 billion tons/year to 51.1 billion tons/year. Potential causes of errors and wide Syvitski (1992). One of such map is reproduced in estimated ranges are varying quality of measuring figure 4. Maps of later years, like Milliman & the sediment loads round the World, lack of Mead (1983), Walling & Web (1983), Lvovich et availability of sediment data for many basins, al. (1991) are in better agreement with each other recycling of data without confirmation, in depicting the severity of sediment erosion as measurements at various locations with respect to compared to those by Fournier (1960) and geomorphology of basin, no data for small Strakhov (1967). catchments, lack of bed load measurements and Drainage basin data including the sediment lack of data of floodevent s which results in heavy yield of 62 of the basins have been collected by sediment fluxes. Milliman and Meade (1983) and later updated by Despite of these problems several estimates Milliman & Syvitski (1992) and Milliman et al. are available for sediment erosion, and are broadly 1995. It has estimated a global sediment flux of classified into two categories. One which is 13.5xl09 tons/year for the mentioned riverbasins . developed from suspended sediments data at the It is updated for the sediments trapped in river mouths resulting into quantity of sediments reservoirs to 14xl09 tons/years and later further flushed to oceans per annum by the World river. increased to 20x109 tons/year to account for the Starting with a modest data from 60 number of ungauged basins. rivers (Lopatin, 1952), present day estimates Walling and Web (1996) & Lai (2003) account for 400 large rivers (Milliman et al. have preferred the figure of 20x109 tons/year as 1995), covering 66% of land area of the world. the sediment flux to ocean, by conforming from Second type of the estimates use the gauging the information from Global Assessment of Soil stations data within the basins, to measure the Degradation (GLASOD) survey which has denudation rate (mm/year) and/or specific estimated that 1,094 million ha of the land surface sediment erosion in terms of tons/km2/year. Both of the earth is currently experiencing serious soil of the kinds of estimates are accompanied by the degradation as a result of water erosion (Oldeman Global sediment flux maps (Milliman & et al. 1991). Using average severe erosion rate of Mead, 1983) and Global soil erosion maps as 100 tons/ha/year and sediment delivery ratio of produced by Fournier (1960), Strakhov (1967), 20%, one can check that soil erosion of 20x109 Walling and Web (1983, 1996), Milliman & Mead tons is a reasonable estimate. (1983), Lvovich et al. (1991) and Milliman & Suspended sediment yield: t km•» yr-"1 5 20 200 10ÔO SOOO Figure 4: Global soil erosion map (Walling and Web, 1983) 172 b. Reservoir Sedimentation taking 13% as the delivery ratio as suggested by No definite estimate of reservoir Walling & Web (1996, 1983), the average sedimentation is available till date. The estimates sediments erosion at global scale comes out to be as presented by Mahmood (1987), White (2001) 153.8xl09 t/year. If the 30% trapping by the and Atkinson (1996) present a wide range of reservoirs is taken (as mentioned by 2nd World reservoir storage loss. Mahmood (1987) roughly water development report (WWAP, 2006)), it estimates the storage loss as 1% of the world's results into the sediment trapped into reservoirs as 3 gross reservoir storage in 1986 (i.e. 4,000 km ) 33.5 km3/year (taking sediment density as 1.385 amounting to an economic loss of approximately t/m3). This figure results into annual storage loss US $ 6 billion each year. He did not specify the of 0.49% of the total of reservoir storage of 6,815 method of coming up to this estimate of 1%. The km3 (White, 2005). This reservoir storage lost is same percentage of storage loss is quoted by Yoon same as that mentioned by the White (2001, 2005) (1992), Atkinson (1996) and by Morris & Fan and other contemporary researchers that is 0.48% (1987). per annum. White (2001) has updated this figure, The cost of replacement storage is also through review of data of 2,300 dams, in 31 much debatable. For example Mahmood (1987) countries and information from various sources. calculated it to be US$ 6 Billion per annum, White He estimated a total loss of storage of 567 km3 (2001) mentioned it to be about US$ 360 Million since start of twentieth century out of gross per annum and Palmieri et al. (2003) quoted it to reported storage volume of 5,976 km3. It amounts be US$ 13 Billion per annum. Out of these three to approximately 10% of gross storage volume references, only Mahmood was kind enough to available in year 2000. The year 2001's annual outline the basis of his estimation, stating a unit loss of storage as reported by White (2001) is cost of replacement of 0.12 US$/m\ To come up 0.48% per year, which amounts to a financial loss with some agreement, we consider the marginal of £200 million annually, nearly, half of that cost of construction of reservoir storages (taking reported by Mahmood (1987). latest figures from ongoing Mangla Raising White's effort to quantify the reservoir Project and for Miranai Dam project in Pakistan), sedimentation is quite remarkable, as he listed which is 0.28 US$/m3 (Wapda, 2006). It results each of the significant country and dam in his into a replacement cost of US$ 9.3 Billion/annum landmark contribution. After careful review of the for annual loss of 0.5% of current storage of 6,815 list of dams used by White, an observation can be km3. This cost is not in agreement with above made, which may answer the difference between three references. We can not comment on two the reservoir sedimentation figure quoted by references, as they have not mentioned the unit White (2001) and that by Mahmood (1987). costs of replacement. But as far as Mahmood's The observation is that the numbers of dams estimate of US$ 6 Billion is concerned, it seems to considered for China are just 28 (with annual be reasonable at the price level of 1987 and as sedimentation of 1.19%) while that for USA is such die estimate of US$ 9.3 Billion seems 1,105 dams. Although we understand, that it is reasonable too at current price levels. difficult to list all the dams in China and USA (as As Mahmood failed to provide any US has more than 70,000 dams and China also calculations for his assumption of 1% of reservoir have more than 40,000 dams). But failing to storage loss per annum so it can only be incorporate sediments of the reservoirs of China considered as intuition or an engineer's estimate. will ultimately give a biased (on lower side) Irony is that it has been referred quite frequently reservoir sedimentation rate, as it is quoted in the in the literature. In contrary the reservoir literature that China has severe sedimentation sedimentation estimate of White (2001) using the problem amounting to 2.3% as compared to USA data of 2,300 reservoirs is confirmed by the which has annual average sedimentation of 0.2% independent calculations by the authors (as only. presented in proceeding paragraphs). Thus the The reservoir sedimentation can be authors are of the view that current level of calculated using the information as quoted by reservoir sedimentation is 0.5% of me gross recently published 2nd world water development storage of reservoirs, and its current replacement report (WWAP, 2006), which states that the dams cost is of the order of US$ 9.3 Billion/year. and reservoir stops 30% of the sediments from 9 reaching to oceans. If we use the figure of 20x10 Methods to Resolve the Problem of Reservoir tons of sediments being flushed by the rivers to Sedimentation ocean (as suggested by Walling & Web (1996), The methods of controlling the reservoir Milliman & Syvitski (1992) and Lai (2003)), and sedimentation can be divided into two categories; 173 Preventive and Curative. The preventive methods for irrigation flows and low level flushing if include reduction of sediment yield in the opted. The 30 years operation of the reservoir watershed and preventing their entry and/or resulted in a capacity loss of 28% with a huge deposition into the reservoir. Curative measures under water delta, whose pivot point is just at 10 include the evacuation of sediments which are km from the dam toe (Figure-5). The liquefaction already deposited in the reservoir through of the delta in case of an earthquake, pose a flushing, dredging and hydrosuction. Morris & serious threat to the serviceability of the dam, as it Fan (1997) have elaborated each of the metiiod in may overwhelm the tunnels intake and may choke detail with case studies of seven reservoir projects, the inlets. mentioning the problem of sedimentation and how some of them were resolved. b. Background of RESCON Model The overall aim of the RESCON model Application of RESCON for Tarbela Dam (Palmieri et al., 2003 and Kawashima et al., 2003) is to select a sediment management strategy that is a. Tarbela Dam technically feasible and also maximizes net Tarbela Dam Pakistan is the largest dam in economic benefits of a reservoir using four Pakistan built across Indus River in 1976. Its explicit options of sediment management, namely; original capacity was 14.3 Km3 and the length of Flushing; Hydrosuction; Traditional Dredging; reservoir is 88 km. The dam has a height of 145m and Trucking. In addition, the "do-nothing" above the bed level. The dam has five tunnels, alternative, (i.e., no sediment removal) where three tunnels (No 1, 2 & 3) are equipped with eventual decommissioning will be required, is also power houses with generation capacity of 3,470 analyzed. The program may be used for existing MW. Rest two tunnels (No 4 & 5 ) are reserved dams as well as proposed dams. FSL=472.56 fi DL in 2002=4f7 5L Original DL=39& Tun. 3&4 =354 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 Distance from dam (km) Figure 5: Longitudinal section of Tarbela reservoir bed profile As depicted in Figure 6, the RESCON checked for technically feasibility, by comparing model requires project-specific technical and the amount of sediments evacuated with respect economic data in addition to environmental and to amount of sediments inflow in a year. social safeguards parameters. The important The trucking and dredging options are always parameters with respect to flushing are considered as technically feasible. For the baseline 'Representative flushing discharge', 'Duration case of no sediment removal the model assumes of flushing', 'Frequency of flushing events' in two alternatives: run-of-river and eventual addition to the parameters related to reservoir decommissioning. All options are taken through geometry, outlet capacities, sediments and water an optimization routine to find economically flow. Before detailed economic calculations, the optimum values for each. The results of economic flushing scenario and hydrosuction scenario are optimization are compared and ranked for final decision by the user. 174 The amount of sediments flushed is calculated where; using the Tsinghua University equation (1), as specified in Atkinson (1996). d = discount factor, defined as l/(l+r), where r is rate of discount (as specified by user) f C2 = initial cost of construction for proposed dam Q.=v 06 (1) w (= 0 for existing dam) (as by user) where V = salvage value (US$) Qs = Sediment load during flushing(tones/sec ) T = terminal year 3 3 Qf = Water discharge during flushing(m /sec) St = remaining reservoir capacity (volume) m S = Longitudinal Slope of channel eroded during in year t flushing M = trapped annual incoming sediment m3, and 3 W = width of channel eroded during flushing(m) , Xt = sediment removed in year t, m c 5 as calculated from 12.8 Qf y/ = is constant depending upon sediments type, The time path of the control variable Xt is and the values are provided in Atkinson selected so that the optimization is achieved for (1996). flushing, and then for hydrosuction. The optimized Net Present Value of each sediment The sediment evacuated through evacuation option is presented as output to the hydrosuction is calculated using the Hotchkiss and user for final selection of the method of Xi (1995) equation. evacuation. The economical optimization is done to optimize the net benefits, using the optimal c. Application of RESCON Model control theory (Seierstad and Sydsaeter, 1987). In this study, efficiency of the flushing is The optimization function (2) is to; explored for various scenarios of water discharges, reservoir drawdown levels, and 1 duration of flushing. The efficiency of flushing Max^NBtxi' -C2 + VJ (2) t=o is judged by Sediment Balance Ratio (SBR) alone, while Long Term Capacity Ratio (LTCR) subject to St+i = S,- M + X, (3) provides insight into sustainable storage capacity USER INPUT: (A) TECHNICAL 4ECÛH0MIC DATA {BJ ENVHWNMENTAL & SOCIAL SAFBäUARD RATWGS AM) WUCY ECONOMIC ANALYSIS RESULTS* CONCLUSIONS Figure 6: Structure of RESCON (Palmieri et al. 2003) 175 of the reservoir. SBR is the ratio of sediment two flushing discharges for each reservoir level evacuated from the reservoir to the sediment (depending whether 2 tunnels or 3 tunnels are inflow to the reservoir. For a successful flushing used) and 3 values of flushing durations (20, 30 & (technically feasible flushing operation) the SBR 40 days). must be greater than 1. LTCR is a parameter showing possibility of achieving long term d. Results of RESCON Model sustainability of the reservoir capacity. It is ratio The result of the RESCON model in table of the volume of the channel eroded in result of 1, shows that in all the cases, SBR is greater than flushing to the original capacity of the reservoir. 1 while LTCR is up to 27%. The results compare Normally, a value greater than 0.35 is considered well with LTCR= 36% as that calculated by as acceptable, but is not a final signal of success Kawashima et al. (2003), for the same reservoir. or failure of the flushing. It shuold be used in Kawashima et al. used flow and sediments data conjunction with SBR to decide about the available up to 1996 for the Tarbela reservoir, possiblity and viablity of the flushing option. while the current study employed data available Overall 12 scenarios are studied (Table 1) using till 2003. A number of parameters used by two reservoir levels (the original dead level of the Kawashima et al. (like reservoir bed width, reservoir i.e. 396.2m and current dead level of the average sediment inflow, unit cost of dam reservoir i.e. 471.6m above msl) during flushing, construction, does not match with the data used Table 1: Sediment Balance Ratio (SBR) & Long Term Capacity Ratio (LTCR) for various scenarios, Tarbela dam Water level Flushing Duration during dis of flush Sedi Long flushing charge ing ment Term Sce Balance Capacity nario m above Ratio Ratio no. msl m3/sec Days (SBR) (LTCR 1 396.2 3,013 20 3.49 0.21 2 396.2 3,013 30 5.24 0.21 3 396.2 3,013 40 6.98 0.21 4 396.2 4,864 20 6.51 0.27 5 396.2 4,864 30 9.76 0.27 6 396.2 4,864 40 13.01 0.27 7 417.6 3,787 20 3.17 0.17 8 417.6 3,787 30 4.71 0.17 9 417.6 3,787 40 6.34 0.17 10 417.6 5,985 20 5.74 0.21 11 417.6 5,985 30 8.62 0.21 12 417.6 5,985 40 11.49 0.21 in this study. The data in this study is collected achieved men more flushing discharges should be from source (the Tarbela reservoir), while allowed (through power tunnels or through Kawashima et al. depended on the printed reports, planned future tunnels) and the depth of water without visiting the reservoir site. should be further reduced to achieve the riverine The low values of the LTCR, in present study, are conditions (full drawdown conditions) during because of wider section of the reservoir used and flushing. low discharges allowed through existing outlets. Higher values of the SBR signifies that the Conclusions flushing operation may be opted to evacuate the Reservoirs are playing a vital role in the sediments out of the reservoir, but the continuous development of economies and in providing the flushing over the year should not be considered to basic need of food, fiber and energy. It has been achieve the original capacity of the reservoir seen that world has recognized the need to convert (LTCR =1) ramer it is restricted to a maximum these reservoirs into sustainable infrastructure value of 0.27. If more long term capacity is to be through innovative ideas and tools. Wide variation 176 in sediment erosion values at Global scale has Kawashima, S., Johndrow T. B., Annandale, G. been noticed and after in depth discussion, a W. and Shah, F., (2003), Reservoir value of sediment flux of 20x109 tons/annum is Conservation Vol II: Rescon Model and recommended. Similarly the difference in values User Manual, The World Bank of reservoir sedimentation is discussed and finally Washington DC. concluded that a value of 0.5% is more Labadie, J.W. (2004), Optimal operation of reasonable, with a replacement cost of US$ 9.3 multireservoir systems: State-of-the-art Billion/annum. review, Journal of Water Resources A recent planning tool RESCON has been Planning and Management, March/April applied to one of the reservoir in Pakistan 2004,pp93-111 (the Tarbela reservoir) using latest data available, Lai R. (2003), Soil erosion and the global carbon and it is found that the reservoir can achieve a budget, Environmental International 29, sediment balance ratio of more than 1 even with a 437-450. flushing for 20 days in a year with water depth at Lima, J.E.F.W., Lopes, W.T.A., Carvalho, N.D., 3 417 m above msl and discharge of 3,787 m /sec. Vieira, M.R. and da Silva, E.M., (2005), LTCR values obtained are not quite favorable, Suspeneded sediment fluxes in the large requiring further increase of flushing discharges river basins of Brazil in IAHS Pub. 291, and lowering of water level during flushing. pp355:363 As three of the current tunnels of the Lopatin, G.C. (1952), Detritus in the rivers of the reservoir are being used for power generation and USSR, Zap. Vses. Geogr. Obsch. 14. plans are there to equip the fourth one also with Geografgiz, Moscow, (cf. Walling & turbines, it is feared that the machinery of the Web, 1996) turbines will be adversely affected if the flushing Lvovich, M.I., Karasik, G. Ya., Bratsiva, N.L, discharges are passed through power tunnels. In Medvedeva, G.P. and Maleshko, A.V. this regard it is proposed that the operator of this (1991), Contemporary intensity of the reservoir (and other hydropower projects) should World land intercontinental erosion, learn to live with the sediments by designing and USSR Academy of Sciences, Moscow. running the turbine machinery (runners, cooling Mahmood, K. (1987), 'Reservoir Sedimentation: water system, valves) in such a way that higher Impact, Extent, and Mitigation', World quantities of sediments should be allowed to pass Bank Technical Paper Number 71, through the turbines. It can be done through use of Washington DC. better quality (hard) materials available for Milliman J.D. & Syvitski J.P.M. (1992) turbines & tunnel linings and through reduced Geomorphic/tectonic control of velocity of flow. The high density of sediment sediments discharge to oceans: the laden water can compensate the reduced velocity importance of small mountainous rivers. of flow (to reduce chances of abrasion and J. Geol. 100, 325-344. harmful effects to the machinery) to achieve Milliman, J.D., and Meade R.H, (1983), World the same power output with high evacuation wide delivery of river sediments to of sediments. Of course, it needs further Ocean, Journal of Geology, 91(1). experimentation to come up with concrete Milliman, J.D., Rutkowski, C. & Meybeck, M. parameters and factors. (1995), River discharge to the sea, A global river index (GLORI). LOICZ Core References Project office, Texel. Morris, G. L., Fan, J. (1997), Reservoir Atkinson, E. (1996), The feasibility of flushing sedimentation handbook, McGraw-Hill sediments from reservoirs, Report OD Pub. 137, HR Wallingford, Howbery Park, Oldeman, L. R., Hakkeling, R. T. A. & Sombroek, Wallingford, Oxon. W. G. (1991), World Map of Status of Fournier, F. (1960), Climat et al. Erosion, PUF, Human-Induced soil degradation: An Paris, (cf. Walling & Web, 1996). explanatory Note. ISRIC, Wageningen. Goklany I. M. (2002), Comparing 20th century Palmieri, A., Shah, F., Annandale, G.W. & Dina, trends in U.S. and Global Agricultural A, (2003), Reservoir Conservation Vol I: water and land use, Water International, The RESCON Approach, World Bank volume 27, Number 3, pp 321-329. Washington DC. Hotchkiss, R.H., Xi Huang. (1995). Hydrosuction Seierstad, A. and K Sydsaeter (1987), Optimal Sediment-Removal Systems (HSRS): Control Theory with Economic Principles and Field Test, Journal of Applications, New York: North Holland. Hydraulic Research, June: 479-489. Silvio G.D. & Hotchkiss, R.H. (1995), Two Years Activity of ICCORES (International Coordinating Committee on Reservoir Sedimentation), Sixth International Symposium on River Sedimentation, 7-11 November, 1995, New Delhi, India. Strakhov, N.M. (1967), Principles of Lithogensis, Oliver & Boyd. Edinbergh. Walling, D.E. & Webb. B.W. (1983), Patterns of sediments yield. In: Background to Palaeohydrology (ed. By K.J. Gregory), 69-100. Wiley, Chichester, UK. Walling, D.E. & Webb. B.W. (1996), Erosion and sediment yield: a global overview, Global and regional perspective, Proceedings of the Exeter Symposium, IAHS Publication no. 236, 3-17. Wapda (2006), Water & Power Development Authority: Water Sector Projects, Hydro- power, Summary, on http://202.38.50.35/ vision2025/default.asp. (accessed on 20.08.2006) White D.R. (2005), Review of current knowledge: World water storage In man-made reservoirs, Foundation for water research, Marlow, U.K. White, R. (2001), Evacuation of sediments from reservoirs, Thomas Telford Publishing, London. World Commission on Dams (WCD), (2000), Dams and Development a new Framework, Earthscan Publications ltd., London and Sterling VA, USA. WWAP (2006), The 2nd UN World Water Development Report: 'Water, a shared responsibility' March 2006. Yoon, Y.N, (1992), The state and the perspective of the direct sediment removal methods from reservoirs, Int. J. Sed. Res. 7(2). 178 Impact of 2004-tsunami Natural Disaster on Water District-10 Banda Aceh of Krueng Aceh River Basin Development Masimin1* and Zouhrawaty A. Ariff1 'Civil Engineering Department, Syiah Kuala University, Banda Aceh, Indonesia 23111 Corresponding author* E-mail: masimin@,plasa.com ABSTRACT Krueng Aceh river basin consisted of ten water districts (WD) and the city of Banda was covered in WD-10. Based on the study of water balance, this WD had already experienced in water shortages for three months in the year of 2000. By the construction of two water storages in the upper basin and rubber dam in lower reach river, WD-10 would have surplus waters till the year 2025. A 9M earthquake followed by a 14-minute tsunami attack on December 26, 2004 changed the face of the city of Banda Aceh and creating serious problem in Krueng Aceh river basin management. About one-third of the city population became victims and tsunami wave washed out their properties. Water supply systems were also destroyed with about 4.0 km wide of coastal strip become a toxic muddy field with nothing left behind The rehabilitation and reconstruction program of water supply system was a priority, since die shortage of water to the year 2025 of WD-10 was not to be expected. Keywords: river basin, management, tsunami attacks. 1. INTRODUCTION as urbanized area especially in me city of Banda The study of river basin management was Aceh and some irrigation schemes in Aceh Besar trying to match of natural resources capacity with district. The total population was about 625,499 some constraints and demands with some quality peoples in 2003 tiiat more man 50% of them were levels by considering a non destructed exploration living in the city of Banda Aceh. The population of the environment. The capacity of natural number was predicted to be 1.24 million peoples in resources was assumed to be in constant value that the year of 2025 tiiat the calculation was based on the extreme condition, such as flooding and drought, the annual population growth of 2.9% for the region. possibly influenced the availability and quality of The rapid population growth in the region and the the resources. The demand was vary in time that die development of some irrigation schemes tended to population tended to increase annually including the increase their water demands, land for agricultural increased of their level of quality life. practices, and the changing function of land used. Three important aspects in river basin Krueng Aceh river basin covered an area of management were die national strategy as die 1,775 km2, the length of river was 145 km, and the law guidance, institution which governed the average of river slope was 0,001064. There were implementation programs, and physical hydrology some tributaries in the river system and based on it, condition as me natural sources was to be exploited. the basin was divided into ten water district (WD) The national strategy (Water Laws) and institution that the most populated WD was WD-10 covering (Local Government) were not technical terms, so the city of Banda Aceh. The position of Banda Aceh that the focus of discussion in the paper was only was in between 05°30' - 05°35' N latitude and covering the physical hydrology condition 95°15' - 95°22' E longitude covering an area of especially for domestic municipal and industry 71.53 km2. Due to the position in the basin, WD-10 (DMI) water demands. Banda Aceh has experienced in annual flooding and The sample case was the Krueng Aceh river high tide that generating water management basin management mat last two years experienced in problem. tsunami natural disaster. Krueng Aceh river basin The WD-10 Banda Aceh has already in water was located in Aceh Province, Sumatra, Indonesia. shortage for three months in the year of 2000 and It covers two districts in die province; district of the government tried to solve the problem by Aceh Besar and me city of Banda Aceh, the capitol constructing two storage dams; Keuliling and Lubok city of the province. The basin has been developed Dam in WD-5 and WD-8 respectively. Sea water 179 problem due to high tide was protected by the 25% mountainous area. Most of the flat area was construction of rubber dam and the dam was also located in the lowland part of the basin where the providing the long storage in the upper reach of the city of Banda Aceh was located. The agricultural river. With the presence if three structures, the practice in the basin was in the hilly area WD-10 Banda Aceh would fulfill the water where some irrigation schemes were located and demands including water for DMI to the year 2025 mountainous area was mostly for forest and without shortage of water, so the DMI's problem conservation. The information of the land use in the was detected to be the water exploitation and basin consisted of forest (58.17%), paddy fields distribution networks. (8.77%), plantation fields (6.52%), scrubs (19.11%), The tsunami natural disaster on 26th urban area (3.72%), shrimp ponds (0.47%), and December 2004 had a bad impact to the water other function (3.24%). exploitation and supply distribution in WD-10 Krueng Aceh is the main river in me basin Banda Aceh caused of the damaged of destroyed of with 145 km river length and 0.001064 river slope the structures. The damage level of the water supply and the outfall of the river was in Malacca Strait. systems mat classified in four damage levels was as Therewere some tributaries in the main river totally damage (61%), heavily damage (11 (%), including Krueng Daroy, Lueng Paga, Krueng Jreue, slightly damage (18%), and no damage (10%). Since Krueng Keumireue, Krueng Boga, and Krueng some of the customers lost their life and the villages Inong. Itwas a tidal river and the effect of sea water were destroyed, the recovery program was done by reached the chaînage of km-18 extenting from the using water tank mobile for water distribution to die river mouth. The water intake for water supply for homeless and permanent rehabilitation and the consumption of the city of Banda Aceh was reconstruction of the water system network needed located in the chainage of km-15 and it was already some detail study. protected by the construction of rubber dam from sea water intrusion during dry season. 2. KRUENG ACEH RIVER BASIN DEVELOPMENT STUDY 2.2 Analysis of Water Balance Based on me river system, the basin was 2.1 Climate and Physical Data divided into 10 (ten) sub-basins called the water The basin had tropical monsoon climate with district (WD) distributed in Krueng Aceh Basin. extreme condition was marked by strong winds from Based on the physical and hydrological data of the West direction especially during rainy season mat basin, the availability of water was determined. The occurred in October to May. The mean daily water demands consist of water for municipal uses, temperature and humidity were in the value of water for irrigation, water for industry, and water for 26.4°C and 79% respectively with the mean annual river maintenance. Water availability and demands rainfall was 1617 mm. The mean daily sunshine and for every WD was analyzed to see the monthly wind velocity are 44% and 233 km/day respectively water balance mat calculated using a 5-year time The distribution of wind direction was calm step with me starting point was the year of 2000 and (41.86%), N (5.76%), NE (7.20%), E (8.15%), SE the last prediction was the year of 2025. Since there (4.12%), S (8.18%), SW (2.34%), W (18.19%), and ware limited of water storages and the two storage NW (4.22%). The distribution of wind velocity was dams were under construction, the seasonal water calm <10 knots (41.86%), 10-14 knots (35.91%), availability in the sub-basins had a main key in 14-17 knots (15.84%), 17-21 knots (4.75%), and water balance analysis. The monthly water for 80% >21 knots (1.64%). level degree of availability was assumed in constant The area coverage of Krueng Aceh river condition and the critical values occurred during the basin was about 1,775 km2 and topographically the months of June to September as seen on Table 1. basin consisted of 40% flat area, 35% hilly area, and 180 Table 1, Monthly water availability for WD-1 to WD-10 in Krueng Aceh river basin WD Area Water Discharge (m3/s) # km2 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 1 247.0 2.86 1.64 2.60 2.63 2.85 1.4 0.75 1.00 1.70 3.35 2.80 2.80 2 44.8 0.71 0.41 0.65 0.65 0.71 1 0.19 0.25 0.42 0.84 0.70 0.70 3 408.0 9.37 9.75 10.1 9.02 9.15 0.3 2.65 2.99 2.57 5.45 9.64 7.17 4 108.0 3.14 2.91 7 2.93 3.05 5 0.86 0.96 0.98 2.15 3.17 2.55 5 308.0 7.07 7.36 3.21 6.81 6.91 4.0 2.00 2.26 1.94 4.12 7.28 5.41 6 84.5 1.94 2.02 7.67 1.87 1.89 1 0.55 0.62 0.53 1.13 2.00 1.49 7 241.0 6.13 6.01 2.11 5.54 5.84 1.3 1.66 1.68 1.38 3.47 6.15 4.82 8 72.5 12.80 11.9 6.22 11.0 12.0 7 3.37 3.14 2.52 7.19 12.6 10.2 9 102.0 13.48 8 12.3 9 1 3.0 0.66 0.75 0.64 1.36 2 4 10 91.3 10.32 2.44 9 2.26 2.29 3 2.23 1.31 1.56 2.23 2.41 1.79 6.65 2.54 4.80 8.03 0.8 8.85 8.68 5.94 3 2.5 9 5.3 8 1.0 0 3.8 0 The water demands were distributed for water from WD-7 could be discharged to lower WD water irrigation, domestic municipal water use and by some measure to eliminate the shortage of water industry, river flow maintenance, cooling process problem in WD-10. The WD-10 was also had a for power generation, waterways for fishing boats, problem of water intrusion during the low discharge and shrimp ponds fresh water supply. The future in dry season that this salt water tidal reached the water demands analysis was also considered for the intake of water supply in Lambaro. The construction 2.90% of population or livestock annual growth of rubber dam was not only stopped the sea water in using die 5-year time step witíi the starting point the downstream but also created a long storage of was the year of 2000. For the illustration, the fresh water in the upstream river reach of the dam. WD-10 die coverage the city of Banda Aceh was With the presences of three storage dams, the already been facing die water shortage in the starting WD-10 would be predicted to have surplus of water year of 2000 when there were no counter measures till the year of 2025 as seen on Fig. 2. Actually there to avoid the problem as seen on Fig. 1. was enough water in the WD-100 in the year of 6 3 When looking at me total availability of 2000 with surplus of water about 18.83xl0 m but water in the basin, there was still enough water to it was not well monthly distributed, so that three fulfill me demands so that me interconnection water months in that year die WD-10 had experienced in distribution looked like the solution. The surplus shortage of waters. AVAIL 2000 2005 2010 Month Fig. 2, Performance of WD-10 without any counter measures c O E ?5 E CO c (0 O *— 2020 (+129.08) •—2025 (+73.70 Fig. 3, Performance of water balance in WD-10 with the presence of three storage dams 3. AN OVERVIEW OF TSUNAMI ATTACK IN The west part of the city had two experienced of WD-10 BANDA ACEH tsunami attack, it meant that tsunami attacks was coming twice with the high of inundation was in the 3.1 Determination of Tsunami Wave Attack range of 10 to 17 m. In the west part of the city, the The region facing to Malacca Straits received number of tsunami attacks was three times with die the least magnitude of tsunami attacks with the wave height of inundation water was in the range of 7 to high was about 5.0-10.0 m and the number of wave 10 m. attacks were three times. The effect of tsunami Theoretically, there were two kinds of waves attacks in the region was letting the debris around in tsunami attacks; impulsive wave and standing the demolished structures. The coverage area of the wave. The first wave was quite high compared to the disaster was about 20.0 km length along the coast second one but the time of attack was too short. The and 4.0 km wide extending from the original coast standing wave was less high than the first wave, but lines. There were two kinds of tsunami attacks in the it occurred in the long time compared to the first one city of Banda Aceh when examining the location. and moved with massive of water. The second wave 182 was very danger and destroyed everything included 3.2 Effect of Tsunami Attack to WD-10 the coastal protection, beach forest, fish Banda Aceh ponds, roads, houses, public building, etc. City It was to understand that tsunami attack took facilities including the electric power system, a huge victim including peoples, properties, telecommunication system, drainage system, and infra structures, city facilities, and environment that water supply system were also destroyed and most of them were related to water resources everything became not in function that most of them management. With the total lost of life about 90,000 needed to be reconstructed again. peoples and uncountable livestock in Banda Aceh, it After the tsunami attack and water inunda had a significance influence to the total water tion, the backwash water was also creating problem demands in the near future. The infrastructures were since this flow water carried all things of floating also destroyed by tsunami wave attack and in bodies and some light debris going to the sea. About WD-10 Banda Aceh, the water supply and pipe two third of city became an open area covered by network system had in effect of tsunami disaster toxic mud and heavy debris that the original villages with some level of damages. The damage level was and roads could not be located. The survivals classified into four conditions as totally damage became homeless, no power, no food, no water, and (TD), heavily damage (HD), slightly damage (SD), some of them lost their family members. It was said and no damage (ND). The result of the survey, it that tsunami attack took a huge of victims, lost of was found that the damage level of the water supply lives and properties, and also destroyed the system was; (a) totally damage (61%), (b) heavily infrastructures of the city facilities including the damage (11 (%), slightly damage (18%), and non water supply systems. damage (10%). For sub-regions, the results were presented in Fig. 4. Fig.4, Damage level for infrastructures of water supply 183 3.3 Recovery, Rehabilitation, and Reconstruction and inter-connected operation of WD's was to be Programs conducted that the shortage of WD would get addi In the recovery program, the homeless who tional water from the water surplus of WD's. live in the gathering places were freely supplied of c. WD-10 Banda Aceh was suffering from the 9M fresh water distributed using mobile tanks. Upon the earthquake and 2004 - tsunami wave attack that temporary housing provided by government and most of the water resources facilities included water NGO, die rehabilitation of water supply distribution supply system had no function and destroyed that was rehabilitated to get the system to be in function. the recovery, rehabilitation, and reconstruction This was done by stopping the water supply going to programs should be conducted. unnecessary location with no people in the village d. The water distribution for homeless using mobile and doing minor works of the system, such as tanks were implemented during recovery program improved the water intake, stopped the water and minor works of the water distribution system leaking, etc. was implemented during the rehabilitation program. The reconstruction program was the hard e. Before going through the reconstruction program, work since the face of WD-10 Banda Aceh was the detail further study was mandatory since the changing compared to those of it before tsunami water demand and the water exploitation and the attack. The number and distribution of the peoples distribution were suspected to be different compared that proportional to the water demands were of those of previous study and the effect of the different for the time being compared to those based 2004-tsunami disaster. on the study in 2003. It meant that the detail study of water demand for reconstruction program was the Acknowledgment priority to be implemented especially for the WD-10 The writers want to thanks going to JICA for Banda Aceh. giving chance to participate in the survey for tsunami damage in Banda Aceh and to the 4. SUMMARY AND CONCLUSION Committee for giving chance to present the paper in Krueng Aceh river basin was divided into ten the International Symposium held in Bangkok, water districts that WD-10 was covering the city of Thailand, 16-20 October 2006. Banda Aceh. The city of Banda Aceh had classical water problems, an annual flood from upper basin REFERENCES and high tide that flow trough the down reach of the river disturbing the water quality in the water intake. Ahsan, M.R. & A. Das Gupta, (1999) Water re In the year of 2000, the WD-10 had already experi sources management - A Comprehensive enced the water shortage and by the presence of Approach. Proceedings of Civil and three dams in the basin this WD-10 would have no Environmental Engineering Conference, water shortage to the year of 2025. By the end of the Bangkok, Thailand. year of 2004, tsunami natural disaster destroyed the Ariff, Z.A., (2005), Damage level of the structures water supply system in WD-10 Banda Aceh creating due to tsunami attack in Banda Aceh, the water exploitation and water distribution. The Proceeding of Annual Civil Engineering recovery, rehabilitation, and reconstruction were set Seminar, Banda Aceh. to solve the problems. BCP (PT), (1999) Design Note for Ulee Lheue Ferry After gathering some information and brief Port, Public Works - City of Banda Aceh, discussion or summaries about the water study in the Banda Aceh. (Indonesian). basin and impact of 2004 - tsunami attack on CAE (PT), (2002) Design Note for Syiah Kuala WD-10 Banda Aceh, the conclusion could be Coastal Protection Works in Banda Aceh, withdrawal from mis study as follows: Dinas Pengairan (Water resources Bureau) a. The water global problem that water too short to Aceh, Banda Aceh. (Indonesian). become fresh water was also applied to the Krueng CAE (PT), (2003) Masterplan Development for Aceh basin by indicating that annual cumulative of Water Resources ofSWS-01.01 Krueng Aceh. water was greater than the demands but tiiere were Aceh Public Works, Banda Aceh. still shortage of water for some months of water Masimin & U. Budiman, (1996) Conceptual distribution. approach for Krueng Aceh river basin b. Some counter measures to avoid the problems management. Proceedings of Public Works could be done by constructing some storage dams and JICA Joint Seminar, Jakarta. 184 Masimin & Z.A. Ariff, (2003) Coastal Protection Works in the City of Banda Aceh Indonesia, Proceedings of COPEDEC VI International Conference, Colombo, Sri Lanka. Masimin & Ariff, Z.A., (2005) Report on Tsunami Natural Disaster in Aceh Indonesia (Victims for Know Nothing)^ Workshop on Tsunami Disaster and Mitigation, PARI Yokosuka, Japan. Meidiatama Indoconsult (PT), Design Note for Kuala Cangkoy (Ulee Lheue) Protection Works in Banda Ace^ Water resources Bureau for Aceh, Banda Aceh.(Indonesian). Verhaeghe, R.J., (1999) Integration of river basin and spatial planning Proceedings of Civil and Environmental Engineering Conference, Bangkok, Thailand. Wahana (PT), (2002) Identification study for raw water condition on SWS-01.01. Aceh Public Works, Banda Aceh. 185 Surface Water Treatment with Microfiltration at The Village of Pranon, Nakhon Sawan Province, Thailand William R. Sellerberg, P.E. Voravuthi RakTae-ngan Pall Corporation Pall Corporation Filtration & Separations (Thailand )Ltd. 25 Harbor Park Drive Unit 2501 Rasa Tower 25* Floor Port Washington, New York, U.S.A. 555 Phaholyothin Rd., Chatuchak Bangkok 10900 Thailand ABSTRACT The use of low-pressure microfiltration membrane systems to produce potable water has become more common over the last 15 years due to their abilities to meet increasingly stringent regulatory requirements. In contrast to conventional treatment processes where contaminants are typically removed by multiple barriers, membrane processes provide a simple and absolute barrier to many microbial pathogens. Microfiltration membranes have a fixed engineered pore size of 0.1 micron and consistently provide high quality filtrate regardless of the incoming raw water quality. As a result, water treatment becomes less complicated and more reliable. These technological advantages make low-pressure membrane filtration an excellent treatment option to ensure public health. A pilot study was conducted at the Village of Pranon in the Nakhon Sawan Province of Thailand using microfiltration to treat surface water for potable use. The objective was to demonstrate microfiltration as a feasible technology for treating the highly variable raw water source. This paper presents the operating and performance data from the pilot study. Raw water turbidity averaged 50 NTU with sustained spikes greater than 100 NTU while the filtrate water turbidity consistently averaged 0.03 NTU. Design conditions were determined and operating parameters were developed for the design of the full-scale AX-2 Pall MF system donated to the Village of Pranon. Introduction Pall Corporation has donated a new water The Village of Pranon has operated a treatment plant using membrane filtration in me conventional water treatment plant since June 2000 name of His Majesty King Bhumibol Adulyadej of using coagulation & sedimentation followed by sand Thailand. Pall Corporation conducted a pilot study filtration. The chemical addition system consisted from April 2006 to July 2006 to verify the viability of running a small side stream of the raw water over of a Pall Microfiltration system for this application. solid blocks of alum. More blocks of alum are added during periods of high turbidity in the raw Specific objectives of this pilot test were as follows: water (>1000 NTU). The resulting coagulant dose was extremely variable and was estimated to range • Demonstrate the design membrane flux and between 150 mg/L and 250 mg/L. Filtrate water operating parameters for the full-scale AX-2 quality was also extremely variable. High color and system. high turbidity in the filtrate were problematic, • Demonstrate the ability of the system to especially during periods of high turbidity in the raw consistently produce high quality filtrate water. Short filter runs were also a problem, (average 0.03 NTU) regardless of raw water resulting in bypassing a portion of the settled water turbidity without chemical addition. around the sand filters direct to the clearwell in • Demonstrate and optimize different pretreatment order to keep up with demand. options to further enhance the filtrate quality • Demonstrate the ability of the system to remove particulate matter, bacteria, Giardia lamblia cysts, and Cryptosporidium oocysts from the feed water 187 Test Methods & Equipment microfiltration pilot unit. The plumbing was A Pall automated pilot system was installed configured to bypass the pretreatment pilot unit and for this evaluation. The system was equipped with a backwashable disc strainer. An isometric drawing single hollow-fiber MF membrane. The treatment of the MF pilot unit is shown in Figure 1 and a train consisted of a pretreatment pilot unit process flow diagram is shown in Figure 2. Photos followed by a backwashable disk strainer and the are included in Appendix A. Figure 1 CHoms Tank and Doeing Punp Figure 2 Pt.r&cn fmwßu PII 3faltar MQdUteJ'frPmw * •Ht!. J«WÉ«fl.a>Ma* &SMML Tïï ¡MmTa***ic*SSTiS» itCLUEniiiaiiaaHe lTTlaSLträ!» Mots- «irm&S'iAF ï« AP ?irw*íC*K>n¡v»* 188 There are four basic modes of operation for the An on-board computer and PLC controlled membrane unit: the operation of the system. The system was 1. Forward Filtration: The feed pump draws water monitored and controlled remotely using satellite from the feed tank and pumps it through the connection and remote access software. Critical membrane filter through the feed port at the bottom operational parameters were logged continuously of the module. The filtrate exits the filtrate port at and recorded automatically on the system top end of die module. computer's hard drive. The real-time data was used 2. Air Scrub (AS) is an effective way to clean the for operation optimization and troubleshooting. membrane hydraulically. Air is injected into the module, on the feed side of the fibers while filtrate Test Results & Discussion is pumped in the reverse direction through the The pilot testing took place from April 2006 module. All discharge during the AS is sent out the to July 2006. Over the course of the study, the pilot upper drain. The combined water-air flow creates was run on pretreated water, raw water and clarified strong turbulent and shearing force to dislodge dirt water from the existing system. Different deposits on the membrane surface. pretreatment chemicals were tested on-site as well 3. Feed Flush or Reverse Flush (FL or RF): This as in the laboratory to test additional color process follows an AS to flush waste out of the reduction. Grab samples were collected monthly module. Either feed water or filtrate can be used. and sample analysis was conducted by Intertek During a FL, the feed pump is used to pump feed Testing Services (Thailand) Ltd., Bangkok, water into the bottom of module and out the upper Thailand. drain to waste. During a RF, the RF pump is used to Results obtained during May 2006 are pump filtrate water into the top of the module and presented in die report since the raw water quality out the upper and lower drains to waste. Both FL during this period was the most challenging. and RF were tested during this pilot study. Standard operating parameters are shown in Table 1. 4. Enhanced Flux Maintenance (EFM): At a user Process data are presented in Figure 3. The system defined time interval, the system will stop while in was operated at a constant flux of 45.4 LMH (26.8 forward filtration mode. During the EFM process, GFD). Recovery ranged from 89.2% to 94.5%. the feed side of the system is drained and filtrate is During periods of high turbidity, operational then pumped from the RF tank off skid to the water conditions were adjusted to manage the heater. Heated water is displaced out of the heater transmembrane pressure (TMP) increase. For and a chemical (typically chlorine) is injected as the example, the filtration duration interval was reduced filtrate returns to the test skid. This solution then from 30 minutes to 15 minutes and daily chlorine flows into the feed tank until a sufficient quantity EFMs were enabled. has been transferred. The solution is then recircu lated for 30 minutes through the system on the feed Table 1 side of the filter and back to the feed tank. The solution is drained and the system is flushed using a Flow 2.3 m5/h = 10 gpm standard AS and FL. The EFM procedure functions Flux 46.4 LMH = 26.8 GFD to extend the interval between the full chemical Recovery 94.5% cleaning (CIP). An unheated chlorine EFM was tested during this pilot study. The EFM on the AX- Filtration Duration 30 minutas 2 full-scale system is performed manually. Air Scrub Ak-Flew Rate DSCrM The pretreatment test rig consisted of two Air Scrub Filtrate How Rate 1.8irrVh = 8gpm 250 gallon tanks. Each tank had an under-baffle, Air Scrub Duration 60 seconds dividing the tank into two compartments. Reverse Filtration Flow Rate 3,6 irrVh = 16 gpm A variable speed mixer was located in die first Reverse Filtration Duration 30 seconds compartment of each tank. Hand valves were adjusted to use 0, 1, or 2 tanks. Three chemical dosing pumps were available for chemical addition. The pretreatment test rig was used for a portion of this pilot study 189 Figure 3 Nakonsawan Met, Thailand May ZOOS 5 S S 3 S S S S S | g § Ü i i § g i g g § g g g S 5 The raw water turbidity exceeded the upper turbidity ranged from 1 to greater than 100 NTU. range (100 NTU) of the on-line turbidimeter for A high range turbidimeter capable of measuring up more than 10 days during May for a total of to 9999 NTU was installed on July 12, 2006. The approximately 23 days during the pilot. The filtrate filtrate turbidity ranged between 0.02 NTU to 0.05 turbidity consistently averaged 0.03 NTU regardless NTU. Data due to clogged sensors and during shut of raw water turbidity. Feed turbidity and filtrate downs has been removed. The monthly grab sample turbidity for May are shown in Figure 4. The feed recorded the turbidity at 494 NTU. Analytical results are presented in Table 2. Figure 4 Nakonaawan Pllat, Thailand TurtldKy May 2006 5 S S 3 S S S S S g g § § i | g S § i g § g g § § | | §§ § g 5 190 Table 2 Teat Parameter UnH Raw water Treated Water TISI2S7 Regulation* Sampling date /time IO/May/2006 /18:14 10/May/2006/18:58 - Taste/ Odour - Not objectionable Not objectionable - pH pH unit 7.41 7.65 6.5-8.5 Color mg/I R-Co 21332 54.50 5.00 Turbidity NTU 494 0.03** 500 Total Solids mo/l 466 240 500.00 Chloride mo/1 36.73 38.90 250.00 Fluoride ma/I 0.40 0.41 0.70 Total Hardness mg/I 55.40 46.95 300.00 Nirats mg/I 53.68 1.55 45.00 Sulfate mpyi 195.14 31.91 200.00 IronfFe) mg/I 17.40 0.01 0.50 ManganesefMn) mart 0.76 O.001 0.30 Copper (Cu) mg/1 0.02 0.01 1.00 2nc Summary Low-pressure microfiltration membrane systems are a feasible technology to treat extremely variable raw sources for potable use. High filtrate quality is consistently produced regardless of raw water turbidity. The following conclusions are drawn based upon the pilot study conducted at the Village of Pranon in the Nakhon Sawan Province of Thailand: • The Pall MF system was successfully operated at a flux rate of 45.4 LMH (26.8 GFD). • The filtrate turbidity consistently averaged 0.03 NTU regardless of raw water turbidity. • During periods of high turbidity, operational con ditions were adjusted to successfully manage the TMP increase. • The chemical cleaning process demonstrated the ability to restore membrane permeability. • Membrane integrity was verified during the pilot study using a pressure-hold test. 191 Appendix A: Photos Photo 1: Raw Water Supply Pump Photo 4: MF Pilot Unit and Backwashable Disc Strainer Photo 2: Pilot Site Overview Photo 5: Pretreatment Pilot Unit Photo 3: Satellite Dish, Existing alum dosing & flocculation Photo 6: Raw Water Sample vs. MF Filtrate Sample 192 Overview of United States Drinking Water Regulations Anthony M. Wachinski, PhD, PE Pall Corporation 25 Harbor Park Drive Port Washington, New York, USA ABSTRACT The principal law governing drinking water safety in the United States is the Safe Drinking Water Act (SDWA). The SDWA directs the US Environmental Protection Agency (USEPA) to promulgate and enforce national regulations - drinking water standards -necessary to ensure safe drinking water for the consumer and to protect public health. The United States compliments its strict drinking water standards with three programs. The National Sanitation Foundation, International is a not-for-profit, non-governmental organization that develops standards for products related to public health. It has offices in the USA, Brazil, the United Kingdom, and China. It is a World health Organization Center for drinking water safety. Two NSF certifications are critical to the production of safe drinking water. NSF 61 ensures that products and materials in contact with drinking water are safe. NSF 60 ensures that chemicals added to drinking water are safe at their maximum use levels. The Environmental Technology Verification Program provides verification testing of drinking water system performance so the private utilities and municipalities are assured that drinking water systems they procure will perform to the manufacturer's claims. The water treatment plant certification program is a state -run program. It ensures that all water treatment operators are properly trained and tested. This program is vital to the delivery of safe drinking water to all in the United States. Introduction Since that event, drinking water regulations in the United States have undergone major and dramatic The production of drinking water in the changes. Current trends suggest that they will United States is governed by Federal Regulations continue to become more stringent and complicated. promulgated by the United States Environmental The principal law governing drinking water Protection Agency (USEPA) and ultimately State safety in the United States is the Safe Drinking Regulations. A number of water treatment programs Water Act (SDWA). The SDWA was passed by further ensure that all drinking water treated in the Congress and signed into law by the President in United States meets the high standards of safety, 1974. The SDWA directs the US Environmental quality, and taste. Each and every water treatment Protection Agency (USEPA) to promulgate and plant producing water must meet standards set forth enforce national regulations - drinking water by the National Sanitation Foundation and State standards -necessary to ensure safe drinking water certification requirements for its water plant for the consumer and to protect public health. operators set forth by each state. The law also delegates primary enforcement This paper provides a brief overview of the responsibility to the states that adopt drinking Safe Drinking Water Act (SDWA) and federal water regulations that are as stringent as those drinking water regulations in effect at this time. promulgated by the USEPA. The regulations cover It also describes the role of National Sanitation microbial, chemical, and radionuclide microbial Foundation International, The Environmental contaminants and establish specific roles for the Technology Verification Program, and the training state governments, and public water. and certification of water treatment plants operators. Public Water Systems Safe Drinking Water Act USEPA has further divided public In April 1983, the largest waterbome water systems that are covered by SDWA require disease outbreak in the United States occurred in ments into three categories based on the type of cus Milwaukee, Wisconsin, when 400,00 people were tomers served, as follows exposed to the protozoan parasite Cryptosporidium Parvum. This event attracted national attention to 193 National Primary Drinking Water Regulations • Community public water systems serve year- The National Primary Drinking Water around residents and include municipal systems, Regulations (NPDWRs) specify MCLs or treatment mobile home parks, and apartment buildings techniques for contaminants that may have an with their own water system serving 25 or more adverse health effect on humans. The Maximum people. Contaminant Level (MCL) is the highest level of a contaminant allowed in drinking water. MCL's are • Non transient, non community public water enforceable standards. The SWDA attempts to systems are entities with their own water supply establish a maximum contaminant level (MCL) and serving an average of at least 25 persons who do an MCLG for each drinking water contaminant. not live at the location, but who use the water for The MCL is set at a level as close as possible to the more than six months per year. These systems MCLG but at a concentration that is reasonable and include churches, schools, and office buildings. economically achievable with Best Available Technology (BAT). When it is impossible or • Transient, non community public water sys impractical to establish an MCL, me USEPA can tems are establishments that have their own establish a treatment technique (TT) and specify water system, where an average of at least treatment methods that must be used to minimize 25 people per day visit and use the water exposure of the public. In some cases the MCLG is occasionally or for only short periods of time. economically achievable and in other instances it Examples include restaurants, hotels, motels, and is not. For non-carcinogens, the MCLG is a finite parks. number. For known or suspected human A public water system covered under the carcinogens, the MCLG is zero. Table 1 lists the provisions of the SWDA supplies piped water for primary contaminants and MCLs. human consumption and has at least 15 service The primary regulations are mandatory and connections; or serves 25 or more persons 60 or must be complied with by all public water systems more days each year. to which tiiey apply. If analysis of the water Approximately 220,000 public water systems produced by a water system indicates that an MCL in the United States are regulated under USEPA for a contaminant is being exceeded, the system and SDWA rules. About 60,000 are classed as must take steps to stop providing the water to community systems, and 160,000 fall under one of the public or initiate treatment to reduce the the two non-community systems. contaminant concentration to below the MCL. The Water Quality Association categorizes community public water systems according to num National Secondary Drinking Water Regulations ber of customers and percent of population: The National Secondary Drinking Water 1. Very small systems serve less man 500 people, Regulations basically apply to drinking water constitute 64% of the community water systems, and contaminants that may adversely affect die aesthetic serve 2% of the community water system qualities of water, such as odor and appearance. population. These qualities have no known adverse healtíi 2. Small systems serve 501 to 3,300 customers, effects, and thus secondary regulations are not constitute 24% of the community water systems, and mandatory. However, diese qualities do seriously serve 8% of the community water system affect acceptance of water by die public, and for this population. reason compliance with die limits is very strongly 3. Medium systems serve 3301 to 10,000 recommended. Table 2 lists die secondary MCLs, customers, constitute 7% of the community water and the adverse effects of secondary contaminants. systems, and serve 11% of the community water system population. Variances and Exemptions 4. Large systems serve 10,001 to 100,000 Each drinking water regulation includes customers, constitute 4.5% of the community water provisions for variances and exemptions to provide systems, and serve 35% of the community water relief from full compliance to water systems tiiat system population. have a legitimate problem meeting the requirements. 5. Very large systems serve greater 100,001 Variances and exemptions serve only as time delays; customers, constitute 0.5% of the community water water systems are expected to comply widi all systems, and serve 44% of the community water regulations eventually. system population. 194 Public Notification The National Sanitation Foundation, In writing the SDWA, Congress expressed International the belief that public water systems have a The materials of construction used to treat responsibility to keep their customers informed and produce drinking water that come in direct about the quality of their water. Therefore, non- contact with the water must not leach into me water complying systems are required to provide public any harmful organic or inorganic chemical. notification (PN). Systems that violate operating, These materials are tested and certified by the monitoring, or reporting requirements or briefly National Sanitation Foundation International exceed an MCL must inform the public of the (NSFI). Chemicals used in the production of problem. Even though the problem may have drinking water to include inorganic coagulants such already been corrected, an explanation must be as alum and ferric salts and organic polyelectrolytes, provided in the news media describing the public and disinfectants must also not exceed maximum health significance of the violation. concentrations set forth by NSFI. The current requirements went into effect in The National Sanitation Foundation, 1989. The regulation requires PN if anyone of six International, founded in 1944, is a not-for-profit, conditions occurs: non-governmental, World Health Organization • failure of a system to comply with an applicable Collaborating Center for food and water safety. The MCL NSFI develops standards for products related to • failure to comply with a prescribed treatment public health. It has offices in the USA, Brazil, the technique United Kingdom, and China. In August 2006, NSF • failure to perform water quality monitoring as International entered into an agreement with the required by the regulations Philippine National Government to use NSF • failure to comply with testing procedures as standards as the basis for all for all drinking water prescribed by a primary drinking water units tested, certified, and evaluated in the regulation Philippines. The agreement allows NSF technical • issuance of a variance or exemption by the experts to work directly with the Philippine Bureau primacy agency of Health Devices and Technology. • failure to comply with the requirements of any schedule that has been set under the terms of a The Environmental Technology Verification variance or exemption Program Water treatment plant system manufacturers Monitoring and Reporting Requirements must demonstrate to each State regulator that its To ensure that drinking water meets federal system meets or exceeds the state's performance and state requirements, all water systems are requirements. Today over 90 percent of new water required to regularly sample and test the water they plants and water plant upgrades in the United States, supply to consumers. The regulations specify Canada, and Europe are membrane based systems, minimum sampling frequencies, sampling locations, either microfiltration or ultrafiltration hollow fiber testing procedures, requirements for record keeping, systems. This verification is accomplished by and routine reporting to the state. The regulations piloting the smallest scalable system for 30 to 360 also cover special reporting procedures to be days on the target water source—usually a surface followed if a contaminant exceeds an MCL. water. For membrane plants, a pilot facility is most often a single module system. The purpose of the Monitoring pilot is to demonstrate that effluent water quality, The federal regulations specify minimum turbidity< 0.1 ntu over 95% of the time meets monitoring frequencies, which in many cases are a federal and state requirements and to establish function of the type of water source being used, the operating parameters such as flux (flow rate per unit type of treatment used, and the size of the water area of membrane) and flux maintenance system. All systems must provide periodic testing techniques—operations such air scrub and reverse for microbiological contamination and analysis for flow that clean the membrane. some chemical contaminants. 195 To standardize manufacturer's claims and Summary minimize the number and amount of pilot tests The principal law governing drinking water required, the USEPA, in the late 90's implemented safety in the United States is the Safe Drinking the Environmental Technology Verification (ETV) Water Act (SDWA). The SDWA directs the US program. The ETV program provides verification Environmental Protection Agency (USEPA) testing of drinking water system performance so the to promulgate and enforce national regulations - private utilities and municipalities are assured that drinking water standards -necessary to ensure safe drinking water systems they procure will perform drinking water for the consumer and to protect to the manufacturer's claims. To encourage public health. The United States compliments its manufacturers to participate in the program, the strict drinking water standards with three programs. USEPA co-funded the extensive studies required The National Sanitation Foundation, International is to validate performance claims of new water a not-for-profit, non-governmental organization that technologies. Although no longer in existence, the develops standards for products related to ETV program's test reports are available. More than public health. It has offices in the USA, Brazil, the 60% of the states accept ETV report claims in lieu United Kingdom, and China. It is a World health of or to compliment addition piloting of new Organization Center for drinking water safety. Two technologies. NSF certifications are critical to the production of safe drinking water. NSF 61 ensures that products Water Treatment Plant Operator Program and materials in contact with drinking water are The water treatment plant certification safe. NSF 60 ensures that chemicals added to program is vital to the delivery of safe drinking drinking water are safe at their maximum use levels. water to all in the United States. The program is The Environmental Technology Verification state-run. It ensures that all water treatment Program provides verification testing of drinking operators are properly trained and tested. water system performance so the private utilities and The program involves education, testing, and municipalities are assured that drinking water certification. Although each state administers its systems they procure will perform to the own program, in general operators are classified I, manufacturer's claims. The water treatment plant II, III, or IV. The certification classification is based certification program is a state-run program. upon the production flowrate of the water treatment It ensures that all water treatment operators are plant and the complexity of the storage and distribu properly trained and tested. This program is vital to tion system. the delivery of safe drinking water to all in the United States. Antimonv 0.006 0.006 Increase in blood choles Discharge from petroleum refineries; terol; decrease in blood fire retardants; ceramics; electronics; sugar solder Arsenic 07 0.010 Skin damage or problems Erosion of natural deposits; runoff as of with circulatory systems, from orchards, runoff from glass & 01/23/ and may have increased risk electronics production wastes 06 of getting cancer Asbestos 7 million 7MFL Increased risk of developing Decay of asbestos cement in water (fiber >10 fibers benign intestinal polyps mains; erosion of natural deposits micrometers) per liter Barium 2 2 Increase in blood pressure Discharge of drilling wastes; discharge from metal refineries; erosion of natural deposits Beryllium 0.004 0.004 Intestinal lesions Discharge from metal refineries and coal-burning factories; discharge from electrical, aerospace, and defense industries Cadmium 0.005 0.005 Kidney damage Corrosion of galvanized pipes; erosion of natural deposits; discharge from metal refineries; runoff from waste batteries and paints Chromium 0.1 0.1 Allergic dermatitis Discharge from steel and pulp mills; (total) erosion of natural deposits Copper 1.3 TT8; Short term exposure: Gas Corrosion of household plumbing Action trointestinal distress systems; erosion of natural deposits Level= Long term exposure: Liver 1.3 or kidney damage People with Wilson's Disease should consult then- personal doctor if the amount of copper in then- water exceeds the action level 197 Contaminant MCLGI MCL Potential Health Effects Sources of Contaminant in Drinking (mg/L)2 or from Ingestion of Water Water TTl (mg/L) 2 Cyanide (as 0.2 0.2 Nerve damage or thyroid Discharge from steel/metal factories; free cyanide) problems discharge from plastic and fertilizer factories Fluoride 4.0 4.0 Bone disease (pain and ten Water additive which promotes strong derness of the bones); Chil teeth; erosion of natural deposits; dis dren may get mottled teeth charge from fertilizer and aluminum factories Lead zero TT8; Infants and children: Delays Corrosion of household plumbing Action in physical or mental systems; erosion of natural deposits Level= development; children could 0.015 show slight deficits in attention span and learning abilities Adults: Kidney problems; high blood pressure Mercury 0.002 0.002 Kidney damage Erosion of natural deposits; discharge (inorganic) from refineries and factories; runoff from landfills and croplands Nitrate 10 10 Infants below the age of six Runoff from fertilizer use; leaching (measured as months who drink water from septic tanks, sewage; erosion of Nitrogen) containing nitrate in excess natural deposits of the MCL could become seriously ill and, if un treated, may die. Symptoms include shortness of breath and blue-baby syndrome. Nitrite 1 1 Infants below the age of six Runoff from fertilizer use; leaching (measured as months who drink water from septic tanks, sewage; erosion of Nitrogen) containing nitrite in excess natural deposits of the MCL could become seriously ill and, if un treated, may die. Symptoms include shortness of breath and blue-baby syndrome. Selenium 0.05 0.05 Hair or fingernail loss; Discharge from petroleum refineries; numbness in fingers or toes; erosion of natural deposits; discharge circulatory problems from mines Thallium 0.0005 0.002 Hair loss; changes in blood; Leaching from ore-processing sites; kidney, intestine, or liver discharge from electronics, glass, and problems drug factories 198 Summary of Drinking Water Regulations Regulation Promulgation In Effect National Interim Primary Drinking Water Regulations (NIPDWRs) 1975 Total Trihalomethanes Rule 1979 Fluoride rule 1986 1987 Phase I Volatile Organic Compounds Rule (Phase I VOCs) 1987 1989 Surface Water Treatment Rule (SWTR) 1989 1990 Total Coliform Rule (TCR) 1989 Phase II Synthetic Organic and Inorganic Contaminants Rule 1991 1993 Lead and Copper Rule (LCR) 1991 1994 Phase V Synthetic Organic and Inorganic Contaminants Rule 1992 1993 Stage 1 Disinfectants/Disinfection By-products Rule (Stage 1 D/DBPs) 1998 Interim Enhanced Surface Water Treatment Rule (IESWTR) 1998 2000 Radionuclides Rule 2000 2003 Arsenic Rule 2001 2004 Filter Backwash Recycling Rule (FBRR) 2001 Long-term 1 Enhanced Surface Water Treatment Rule (LTIESWTR) 2002 2002 Ground Water Disinfection Rule ? Radon Rule ? Stage 2 Disinfectants/Disinfection By-products Rule (Stage 2 D/DBPs) 2006 2006* Long-term 2 Enhanced Surface Water Treatment Rule (LT2ESWTR) 2006 2006* * Utilities serving population > 100,000 199 Effect of ENSO on Southeast Asian Rainfall Tsing-Chang (Mike) Chen Department of Geological and Atmospheric Sciences, Iowa State University, U.S.A. tmchen(a),iastate. edu 1. Introduction (Kalnay et al. 1996) for 2003-2005 will be used to The impact of ENSO on the regional portray the water-vapor budget and weather and hydrological cycle has been a very attractive subject climate systems concerned in this study. not only to climate researchers, but also to the general public. The research community measures 3. ENSO cycle the ENSO effect on the regional hydrological cycle The maximum rainfall over the South Asia in terms of the accumulation of precipitation. (SA)-western tropical Pacific (WTP) region In contrast, the general public observes this impact migrates from south of the Equator in winter (Fig. on its daily life, namely, the ENSO impact on the la) to north of the Equator (Fig. lc). Regardless of variation of daily precipitation which is related to this seasonal migration, the maximum rainfall zone the development of daily weather systems. How do in the SA-WTP region is modulated by the ENSO we fill the research gap between the interannual cycle. The sea surface temperature (SST) anomalies variation in the regional hydrological cycle and the over the NOAA NIN03.4 region (5°S-5°N, 170°- rain-producing weather system? In other words, we 120°W), DSST(NIN03.4), are superimposed on the cannot use only statistics from the regional x-t diagrams of winter DP(EQ-5°N) (Fig. lb) and hydrological cycle to present the ENSO impact. summer DP(10°N-15°N) (Fig. Id), where Instead, this impact should be depicted in terms of DP is rainfall departure from its seasonal-mean the possible change in the daily precipitation in the context of the interannual variation of precipitation p under the ENSO effect. With this understanding, value ( ) and averaged over a latitudinal zone the regional authority/government can develop a indicated inside the parenthesis. The extreme better policy for water management when the phases of ENSO cycles appear in winter. Warm possible extreme climate condition occurs. In order (W) and cold (C) extreme ENSO phases during the to accomplish this goal, we shall present the impact 1979-2004 period are marked in Fig. lb. Three of the ENSO cycle on the regional hydrological interesting features emerge from the comparison cycle coupled with the modulation of the ENSO between these two figures. cycle on daily precipitation events, based on two (1) The warm and cold extreme phases of summer sources of information: our previous studies and DSST(NIN03.4) marked in Fig. Id may not always current research of the Southeast Asian monsoon follow extreme winter ENSO phases. (2) The DSST rainfall. (NIN03.4) anomalies in summer are generally smaller than in winter. (3) Because DSST anomalies over the WTP are out-of-phase with 2. Data DSST(NIN03.4) anomalies, it is expected that The daily rainfall data with a l°xl° DP anomalies over the SA-WTP region are also resolution over the entire global was compiled by out-of-phase with DP anomalies in the the GPCP analysis (Huffinan et al. 1997) for the central-eastern tropical Pacific, as shown in both period of 1997-2005. Prior to 1997, the period of Figs, lb and Id. This out-of-phase relationship of 1980-1989 is covered by the MSU precipitation DP anomalies between these two regions gives us a dataset (Spencer 1993), while the rest of period is general view of how the ENSO cycle affects the analyzed by using the GPI dataset (Joyce and Arkin 1997). The time period covered by these dataseis is rainfall in the SA-WTP region. However, this long enough for us to deal with the rainfall variation overall view of the ENSO effect cannot provide us on the interannual to synoptic time scale. with the synoptic details of how the ENSO cycle Meteorological and hydrological variables affects daily rainfall events. assimilated by the ERA-40 reanalysis (Kâllberg 2004) with a 2.5°x2.5° resolution for the period of 1979-2002 and by the NCEP/NCAR reanalysis 201 Fig. 1 Long term-mean rainfall (GPCP) averaged in (a) winter (DJF) and (c) summer (JJA), and the longitude-time cross-section of rainfall departure from climatology averaged between (b) EQ-5°N in winter and (d) 10°-15°N summer. The index of DSST(NTN03.4) is superimposed in (b) and (d) as a dark-red line. 4. Winter As indicated by DSST(NIN03.4) in Fig. lb, the 1997/98 and 1999/00 winters are extreme warm and cold ENSO winters, respectively. In order to explore how the ENSO cycle affects the daily rainfall event, the x-t diagrams of P(5°N) of these two winters are displayed in Figs. 2a and 2b, respectively. The daily rainfall over the SA-WTP region is active during the cold ENSO year (because of warm DSST anomalies in the WTP region). The eastward propagation of rainfall with a period about 1-2 months can be clearly seen during the cold ENSO year. These low-frequency rainfall clusters Fig. 2 (a) Longitude-time cross-section of daily are formed by westward-propagating short-period rainfall (GPCP) along 5°N in a warm year (high-frequency) rainfall clusters. It becomes clear from 6/1/1997 to 5/31/1998. (b) Same as (a) that rainfall in the SA-WTP region is generated by except for a cold year from 6/1/1999 to 5/31/2000. Contour interval is 2 mm-d-1. multiple-scale processes. What are the time scales of the rain-producing weather systems? Based on the winter rainfall distribution (Fig.la), the rainfall approximately 10-24 days (Chen et al. 2002, 2004). center over Borneo (102.5°-117.5°E, 5°S-10°N) is It is inferred from Fig. 2b that the winter rainfall in selected to perform a power spectral analysis. The the tropical SA-WTP region is produced by power spectra of Borneo rainfall in every winter cold-surge vortices which are regulated by the 10-24 over the 1979-2004 period is shown in Fig. 3. Three day mode (i.e. cold surge) and the 30-60 day major peaks emerge from these rainfall spectra: 30- intraseasonal mode. To support our argument, the 60 day, 10-24 day, and 3-6 day. As indicated by the rainfall generated by cold-surge vortices over the x-t diagrams of daily rainfall in Fig. 2, rainfall is three most active regions (marked by red squares) of generated by the 3-6 day mode, which includes these vortices are shown in Fig. 4. Over 70% of synoptic vortices induced by cold surges. The rainfall in these regions is generated by cold-surge occurrence frequency of cold surges in East Asia is vortices. The impact of the ENSO cycle on the Fig. 3 Power spectral analysis of daily rainfall over Area 2 (indicated in Fig. 4) for 26 win ters (11/1-3/31) from 1979 to 2004. The data sources include GPCP (1997-2004), P,Pv PrPv P,P» P.P. GPI (1979, 1989-1996), and MSU (1980- 1988). Warm, cold, and normal years are Fig. 4 (a) Rainfall contributed from cold surge vortices indicated by red, blue, and yellow lines, (Pv), (b) winter-mean total rainfall (PT), and (c) respectively. contribution of the cold surge vortices to rainfall in terms of the ratio of PV/PT- The contour interval is given to the right of each figure. Three areas associated with maximum vortex activity are interannual variation of the SA-WTP rainfall is outlined by red dashed lines with numbers. The accomplished through the modulation of the cold domain of Area 1 is (72.5°~102.5°E, 5°S~10°N), surge and cold-surge vortex activity by the 30-60 Area 2 (102.5°~117.5°E, 5°S~10°N), and Area 3 day intraseasonal mode. (117.5°~170°E, 2.5°S~12.5°N). The amounts of PT and PV in each area and their average are displayed as histograms in (d). 5. Summer The impact of the ENSO cycle on the summer rainfall over the SA-WTP region is this region. The effect of the ENSO cycle on the revealed from the x-t diagram of DP(10°N-15°N) interannual variation of rainfall in this region should superimposed with DSST(NIN03.4) in Fig. Id. The be accomplished through the variation of these interannual variation of rainfall in this region is weather systems' occurrence frequency. This argu out-of-phase with mat in the central-eastern tropical ment will be illustrated and substantiated in terms of Pacific and with summer DSST(NIN03.4) our previous studies and current research. anomalies. It was shown by Chen and Yoon (2000) that the global divergent circulation converges a. Tropical cyclones and other synoptic more (less) water vapor toward the Asian disturbances monsoon hemisphere in response to the east-west Monsoon depressions develop over the Bay differentiation of DSST anomalies. This argument of Bengal from westward-propagating residual lows is confirmed by the interannual variation of summer of different weather systems across Indochina. rainfall over Indochina (inside the box in Fig. 5) It was shown by Chen and Weng (1999) that indicated by its rainfall histograms (Fig. 5b-c). interannual variations in populations of these However, the increase or decrease of Indochina residual lows across Indochina follow those in summer rainfall may not be uniform over the entire populations of related weather systems in die South summer. Tropical cyclones and other synoptic China Sea-western tropical Pacific (Fig. 6). The disturbances are rain-producing weather systems in population changes are out-of-phase with DSST 203 Geography and Surface b. North-Pacific summer vortex It was demonstrated by Chen et al. (2001) that summer upper-level vortices can be generated by the instability of the North Pacific oceanic trough. These vortices may move to East Asia and Indochina, and suppress convective activity over these regions, as indicated by satellite observation (Fig. 8a). The heavy monsoon rainfall season in Indochina is occasionally marked by the passage of North-Pacific summer vortices (Figs. 8a-c). The population of these summer northern Pacific vortices is about 20 on average over the entire summer, and over six of them propagate across Indochina. ,0.NIV<850mb)^ 0^h \/ *-r. 9ÍTE 100"E M0°E lZQnE M or. s MU Dt'proiaton * monaoon depression (MD) K n u • c » u n i 9 trop*c*l cyclone (TC) g IÍ|(C)TC+( 12-24)day modes * 12-24 day manaocnlcMr tooTMD a tond gana* of M3 TO ai sa es ar 8» oi «s » Fig. 5 (a) Analysis domain (dark-blue box) and surface station, rainfall from (b) CMAP and (c) station averaged over this domain, and (d) occurrence frequencies of tropical cyclones and 12-24 day monsoon lows. (NIN03.4). Follwing Chen and Weng (1999), Chen ••nan and Yoon (2000) measured interannual variation of If lift tllu gemas:* Indochina summer rainfall (as shown in Fig. 5). Chen and Yoon's (2000) analysis was extended by Chen and Wang (2006) to explore the contribution Fig. 6 (Top) 850-mb streamfunction superimposed to Indochina rainfall from different weather systems with various disturbances contributing to genèses (Fig. 7). It is revealed from this new analysis that of monsoon depression. Interannual variations of about 50% of Indochina summer rainfall is produced the population in each type of synoptic by westward-propagating "residual lows". Both the total rainfall and the contribution from residual lows DSST(NIN03.4) anomalies (Fig. 9). In other words, undergo an interannual variation out-of-phase with some non-rainy days caused by passages of North- DSST(NIN03.4). Populations of North-Pacific Pacific vortices across Indochina are also regulated vortices in the North Pacific and Indochina undergo by the interannual variation of the large-scale an interannual variation out-of-phase with summer circulation following the DSST(NIN03.4). 204 -7 mm-d'1 Fig. 8 (a) X-T diagram of GMS IR images near 12.5°N from 5/25 to 7/13/2000 at 12Z. (b) 200-mb «¿BVBS'BS SO 93 04 »S'SB 00 0L2 04 year streamline superimposed with daily precipitation on »r P(R«adud La«) 6/5/00 at 12Z. The vortex over Indochina is marked by a red cross, (c) Time-height cross-section of 4 temperature departure from its average of 1 t - >,-3.4mrad- 5/20-6/15/00 superimposed with daily precipitation (50%) (histogram). The vortex of 6/5/00 is marked by fat-red arrow in (a) and (c). BO 82 64 80 86 90 92 94 99 88 00 02 04 Fig. 7 (a) Daily location of residual lows recorded during summers from 1979 to 2004. Interannual variations of (b) summer rainfall and (c) rainfall contributed from residual lows, averaged over the region marked in (a) (blue square). 6. Summary Climate researchers generally consider the impact of the ENSO cycle on the regional water cycle through the accumulation of rainfall. However, the management of water resources cannot consider only the overall long-term rainfall accumulation, but must consider daily rainfall Fig. 9 (a) The upper-level summer circulation depicted events as well. In this study, we illustrate the ENSO by 200-mb streamlines superimposed with daily impact on rainfall through multiple-scale processes locations of vortices (dots). Vortices which merged in both winter and summer. With such an approach, with the midlatitude trough are marked by blue dots, the regional authority/government can develop the while those propagated westward are marked by red proper management policy to handle disastrous dots, (b) and (c) are population histograms of total events when extreme rainfall events may occur North Pacific summer vortices and those moved following the modulation of the ENSO cycle on the across Indochina, respectively. regional weather systems through multiple-scale interaction. Acknowledgements This research task is supported by Cheney Research Fund, and accomplished with assistance provided by Simon Wang, Paul Tsay, Judy Huang, and Adam Clark. 205 References Kâllberg, P., A. Simmons, S. Uppala, and M. Chen, T.-C, and S.-P. Weng, 1999: Interannual and Fuentes, 2004: The ERA-40 archive. intraseasonal variations in monsoon de ERA-40 Project Report Series No. 17 pressions and their westward-propagating (available online at http://www.ecmwf.int/ predecessors. Mon. Wea. Rev., 127, 1005- publications/). 1020. Kalnay, E., and coauthors, 1996: The NCEP/NCAR , and J.-H. Yoon, 2000: Some remarks on the 40-year reanalysis project. Bull. Amer. westward propagation of the monsoon Meteor. Soc, 82, 247-267. depression. Tellus, 52A, 487-499. Huffman, G. J., R. F. Adler, P. A. Arkin, A. Chang, , and S.-Y. Wang, 2006: Transition of synoptic R. Ferraro, A. Gruber, J. Janowiak, A. residual lows across Indochina (in prepa McNab, B. Rudolf and U. Schneider, ration). 1997: The Global Precipitation Climatol , W.-R. Huang and J.-H. Yoon, 2004: Interan ogy Project (GPCP) combined precipita nual variation of the East Asian cold surge tion dataset. Bull. Amer. Meteor. Soc, 78, activity. J. Climate, 17,401-413. 5-20. , M.-C. Yen, G.-R. Liu and S.-Y. Wang. 2001: Joyce, R. and P. A. Arkin, 1997: Improved estimates Summer upper-level vortex over the North of tropical and subtropical precipitation Pacific. Bull. Amer. Meteor. Soc, 82, using the GOES Precipitation Index. J. 1991-2006. Atmos. and Oceanic Tech., 14, 997-1011. , , W.-R. Huang and W. A. Gallus Jr., Spencer, R.W., 1993: Global oceanic precipitation 2002: An East Asian cold surge: Case from the MSU during 1979-91 and comparisons to study. Mon. Wea. Rev., 130,2271-2290. other climatologies. J. Climate, 6, 1301-1326. 206 Tropical storms and associated flood risk on Grande Terre, New Caledonia Ray A. Kostaschuk Department of Geography, University of Guelph, Guelph, ON, N1G2W1, Canada James P. Terry School of Geography, The University of the South Pacific, PO Box 1168, Suva, Fiji Islands Geoffroy Wotling Observatoire de la Ressource en Eau, DAVAR, BP 256, Noumea, Nouvelle-Calédonie ABSTRACT The South Pacific island territory of New Caledonia is affected by tropical storms in the summer months that can cause intense rainfall and extensive flooding. Most of the largest historical flows in the Tontouta River were caused by tropical cyclones, with the largest floods caused by category 3 or 4 hurricanes that pass near the watershed. The log Pearson Type III distribution consistently provides the best fit to both annual maximum and partial duration flow series from the Tontouta River. Partial duration series of daily flows are higher than annual maximum series at lower probabilities and partial duration series of instantaneous flows are consistently higher than those of daily flows. Instantaneous flows should be used for constructing partial duration series for risk assessment in small, steep tropical basins affected by tropical cyclones. 1 INTRODUCTION 1.1 Background produce annual rainfall exceeding 2,000 mm on Tropical cyclones (TCs), also known as much of the east coast while the west coast, in rain 'hurricanes' and 'typhoons', are a major cause of shadow, receives less man 1,000 mm. In the floods in the humid tropics (Gupta, 1988). summer months (November to April) recurrent There are an average of nine to ten TCs annually in depressions bring hot and wet conditions, and the South Pacific and the intense precipitation occasionally TCs moving from the north or associated with these storms generates river floods north east bring torrential rains that cause extensive that cause widespread damage to agriculture, homes river flooding (Bird et al., 1984). and businesses and play a major role in sediment Figure 1. Location of New Caledonia in the transport in channels (Kostaschuk et al., 2003) and South Pacific. The track and category of deposition on floodplains (Terry et al., 2002). Tropical Cyclone Beti in 1996 is illustrated. Despite their importance to South Pacific Islands, TC-induced floods are poorly understood 160° E 170° E (Kostaschuk et al., 2001). This investigation fc. 15° S examines the risk of floods in the Tontouta River in New Caledonia (Fig. 1) by comparing the annual . - - v* /* maxima and partial duration series of daily and l instantaneous flows. % Vanuatu m 1.2 Study Area New\ New Caledonia is a group of islands in the Caledonia « south west Pacific (Fig. 1) consisting of one large Category V^ mountainous island known as Grande Terre (16,750 - - - 1 2 km ) and many smaller islands. The east coast of % Grande Terre is generally steep and mountainous, % while the west coast has foothills and wider % coastal plains. Prevailing south easterly trade winds 25° S 207 The natural vegetation of New Caledonia was Figure 3. Total daily rainfall and mean daily originally largely forest, but extensive areas were discharge in the Tontouta River caused by Tropi cleared and burned by early Melanesian farmers. cal Cyclone Beti. See Fig. 1 for the storm track. There was further cleaning and burning associated with the introduction of grazing animals and the 400 1800 development of mining activities after the arrival of • 1600 350 —*— rainfall European colonists in the 1860s. The existing 300 —•— discharge - 1400 vegetation cover has also been extensively disrupted 250 • 1200 1 1000 200 by landslides, gullying, and associated fan 800 150 deposition (Bird et al, 1984). All of these land-use 600 s 100 changes undoubtedly exacerbate flood flows. 400 SO • The Tontouta River is located on the 200 0 southwestern side of Grande Terre (Fig. 2). The catchment is about 380 km2 with steep slopes and rugged terrain in its upper reaches. Bankfull SS/S/S///S/** discharge at the hydrometric station is approxi mately 600 m3/s. 3 RESULTS AND DISCUSSION 2 DATA SOURCES AND METHODS 3.1 Tropical cyclones and river flow A list of TCs affecting the south west Pacific Figure 3 illustrates the dramatic impact of a and maps of TC paths (e.g. Fig. 1) for 1969-2003 tropical cyclone on rainfall and mean daily was obtained from the Fiji Meteorological Service. discharge in the Tontouta River. In 1996, TC Beti Precipitation and flow data for the Tontouta approached Grande Terre as a Category 4 hurricane hydrometric station (Fig. 2) were supplied by the and passed directly over the Tontouta watershed as a Observatoire de la Ressource en Eau in New Category 3 (Fig. 1). Rainfall began on 24th March, Caledonia. Records of mean daily discharge extend peaked on 27th March at nearly 400 mm/day and from 1954-2003 but mere are many missing ended abruptly on 28th March. Mean daily observations from 1954-1969 so these were not used discharge began to rise on 25th March, peaked on in our analyses. In addition, some data are missing 27th March at nearly 1700 m3/s and then declined for the years 1987-88, 1993-95 and 1999-2000 so slowly until about 4th April. In comparison, peak these were not considered in the analyses of mean instantaneous discharge on 27th March was over daily flows. Total daily precipitation data are 3300 m3/s (Table 1). The lag between maximum available from 1990-2003. total daily rainfall and mean daily discharge was less The stage recorder at Tontouta was adapted than one day and flows exceeded bankfull from th to record large floods in the 1970s so that peak 26-29 March. instantaneous flows could be determined. Flood marks were used before that and in cases when the 3.2 Annual maximum and partial duration series recorder malfunctioned. Extrapolations for peak Two approaches are commonly used in flows were also estimated from the HEC-RAS hydrologie frequency modelling - the annual model calibrated with observed flows. maximum series (AM) and the partial duration Figure 2. Location of the Tontouta River system series (PD). Madsen et al (1997) note that the and hydrometric station on Grand Terre. AM approach, although widely used, has serious limitations. First, it uses the largest annual value, and does not consider secondary events that may 100 km exceed the annual maxima of other years. Second, the AM may include annual peaks from very dry years, which can bias the analysis of large values. The PD considers flows above a certain threshold and avoids these limitations. It is also more flexible in flood representation and provides a more complete description of flood-generating processes (Langet al, 1999). 208 A number of probability distributions have been Figure 4 shows that AM and PD series for daily proposed for both AM and PD, including the flows are identical above a Weibull probability of generalized Pareto, log Pearson Type III and around 0.6 and the PD values are higher at lower Gumbel Extremal. We compared these distributions probabilities. In addition, PD series of instantane for the data in this study and found that the log ous flows are consistently higher than the daily Pearson Type III generally provided the best fit. flows (Fig. 5), which is expected given that instanta Kostaschuk et al. (2001) found a similar result for neous flows exceed daily flows for specific storms. data from Fiji, a Pacific island nation similar to New Figure 6 illustrates PD series for daily and instanta Caledonia that is also affected by tropical cyclones. neous flows in the Tontouta expressed as return pe The log Pearson Type III distribution was thus used riods up to 200 years. The instantaneous flow curve for all of the data in this study. is much larger than the daily curve, reflecting the 'flashiness' of the hydrographs. The dramatic dif Table 1. Maximum instantaneous discharge (Qi) ference between these series illustrates the need to for the Tontouta River and named storms related use instantaneous flows in constructing PD predic to the Hows for 1969-2003. tions for small, steep tropical basins affected by TCs, including those in New Caledonia and other Date and time Qi Name South Pacific islands. (m3/s) 02/02/1969 04:30 551 02/01/1971 11:30 780 Rosie 06/02/1972 23:40 745 Daisy Figure 4. Annual maximum (AM) and partial 03/06/1972 15:00 613 Ida duration (PD) series for the Tontouta River from 04/02/1974 08:45 1744 Pam 1969-2003 based on mean daily discharge. 08/03/1975 04:00 2613 Alison 18/04/1975 01:30 318 17/01/1976 13:20 1055 David 3000 06/01/1978 23:00 514 12/02/1981 23:45 1553 Cliff 07/03/1981 06:00 355 Freda 24/12/1981 22:00 2420 Gyan 13/01/1988 00:00 4583 Anne 15/11/1988 06:45 330 17/12/1988 14:54 417 Eseta 02/01/1989 22:48 852 Delilah 21/01/1989 16:30 899 12/02/1989 04:36 400 Harry 0.2 0.4 0.6 0.8 11/04/1989 05:18 716 Lili Weibull Probability 23/01/1990 09:50 1076 02/02/1990 05:44 582 25/02/1990 19:05 986 Figure 5. Partial duration series for daily and 08/03/1990 05:05 478 instantaneous discharge for the Tontouta River for 17/02/1992 11:27 475 Daman 1969-2003. 05/03/1992 4:57 928 Esau 10/03/1992 20:39 471 Fran 24/03/1992 06:21 637 • dailyactual 4 08/04/1992 01:57 1783 daily predicted 06/01/1994 11:34 686 Rewa A instantaneous actual 27/02/1994 13:10 1010 Theodore - - - - instantaneous predicted 27/03/1996 19:36 3356 Beti 08/01/1997 12:36 1637 Drena 07/07/1997 12:06 593 Discharg e (m'/s ) r 25/01/1999 04:09 362 Dani i i il Ért/ á %V1 l 'a iV" ' 07/03/2002 00:16 909 Des ) 0.2 0.4 0.6 0.8 1 14/03/2003 11:25 3166 Erica Weibull Probability 16/07/2003 05:02 503 209 cause intense rainfall and extensive flooding. Most references of the largest historical flows in the Tontouta River Bird, E.C.F. Dubois, J.-P. & Itis, J.A. 1984 The were caused by tropical cyclones with the largest Impacts of Opencast Mining on the Rivers and floods are caused by category 3 or 4 hurricanes that Coasts of New Caledonia. The United Nations pass near the watershed. However, some extremes University, 64pp. floods (as in 1992) are not related to 'named' storms Gupta, A. 1988 Large floods as geomorphic events and some strong hurricanes (as defined by wind in the humid tropics. In: V.R. Baker, R.C. speed and/or low pressure) do not necessarily cause Kochel and P.C. Patton (eds), Flood more rainfall. Geomorphology, p.301-315. John Wiley & The log Pearson Type III distribution Sons, New York. consistently provides the best fit to both annual Kostaschuk, R.A., Terry, J.P. & Raj, R. 2001 maximum and partial duration flow series from the Tropical cyclone floods in Fiji. Hydrological Tontouta River. Partial duration series of daily Sciences Journal 46: 435-450. flows are higher than annual maximum series at Kostaschuk, R.A., Terry, J.P. & Raj, R. 2003 lower probabilities and partial duration series of Suspended sediment transport during instantaneous flows are consistently higher than the tropical-cyclone floods in Fiji. Hydrological daily lows. Instantaneous flows should be used Processes 17: 1149-1164. for constructing partial duration series for risk Lang, M., Ouarda, T.B.M.J. & Bobée, B. 1999 assessment in small, steep tropical basins affected Towards operational guidelines for by tropical cyclones. over-threshold modeling. Journal of Hydrology, 225: 103-117. Figure 6. Partial duration series predictions for Madsen, H., Rasmussen, P.F. & Rosbjerg, D. 1997 daily and instantaneous discharge for the Comparison of annual maximum series and Tontouta River for 1969-2003. partial duration series methods for modeling extreme hydrologie events. Water Resources Research, 33: 747-757. - - - - daily predicted instantaneous predicted Terry, J.P., Garimella, S. & Kostaschuk, R.A. 2002 Rates of floodplain accretion in a tropical island river system impacted by cyclones and large floods. Geomorphology 42: 171-182. • •• 3 s .-••"'" Discharg e (m /s ) ) 50 100 150 200 2 SO Return Period (years) 210