APPLICATIONS OF GIS IN HYDROLOGIC MODELINGByJun LuoB.Sc. Beijing Agricultural University, 1984M.Sc. Beijing Agricultural University, 1987A THESIS SUBMITTED IN PARTIAL FULFILMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF SCIENCEinTHE FACULTY OF GRADUATE STUDIESTHE DEPARTMENT OF BIO-RESOURCE ENGINEERINGWe accept this thesis as conformingto the required standardTHE UNIVERSITY OF BRITISH COLUMBIAOctober 1992©Jun Luo, 1992In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.(SignatureDepartment of ^io^ eSaboc,e r7-47.The University of British ColumbiaVancouver, Canada/otiizoibe)/^7, DateDE-6 (2/88)AbstractLiterature on applications of geographic information systems (GIS) to hydrologic modeling,stormflow generation in steep watersheds of humid regions, and automated configurationsof watersheds was reviewed. A hydrologic model in conjunction with GIS technology wasconstructed to simulate stormflow hydrographs from a watershed. The model consists ofthree components: stormflow generation, translation and detention. Two major aspectswere emphasised in the model: one is taking advantage of standard GIS to extract, overlayand delineate land and water-related characteristics required for stormflow modeling;another is integrating GIS with hydrologic modeling to simulate both spatial and temporaltransformation of rainfall into stormflow. A good agreement between simulated andobserved stormflow hydrographs was achieved when the model was tested using real datafrom Jamieson Creek, a well gaged and forested watershed in North Vancouver.Applications of this model in an ungaged and forested watershed, Nitinat watershed onVancouver Island, show that the model is capable of handling the effects of complexitiesof soil, land use, topography and rainfall intensity on stormflow simulation. It wasrevealed that for a given return period the maximum peak flow for a design storm wouldincrease with the duration of the storm and the Rational Method would not be suitable inestimating the maximum peak flow from a watershed with large storage capacity. The GIStechnology was highly recommended for water resources management because of itsability of managing and modeling spatial information. Further improvement and testing ofthe model would result in a comprehensive GIS based model, simulating stormflow, soilerosion and non-point source water pollution.iiTable of ContentsAbstract^ iiList of Tables viList of Figures^ viiAcknowledgement xiChapter I INTRODUCTION^ 11.1 Water Resource Management and Hydrologic Modeling ^ 11.2 GIS^ 31.3 GIS and Hydrologic Modeling ^ 41.3.1 GIS and parameterization 41.3.2 GIS and digital elevation model ^ 51.3.3 The interface between GIS and hydrologic models ^ 61.4 Objectives ^ 71.5 Organization of the Thesis^ 8Chapter 2 STUDY WATERSHEDS^ 102.1 Topography^ 122.2 Climate and Streamflow^ 172.3 Soil^ 202.4 Forest 222.5 Instrumentation and Data Collection^ 22iii2.5.1 Precipitation^ 222.5.2 Stormflow 232.5.3 Other data collection^ 24Chapter III WATERSHED CONFIGURATION^ 253.1 Stormflow Generation^ 253.1.1 Overland flow 263.1.2 Partial area^ 273.1.3 Variable source area^ 283.1.4 Comparison of the models of stormflow generation^ 293.2 Stormflow Translation^ 313.2.1 Assumptions of the simulation of stormflow translation ^ 323.2.2 Simulation of the pathways of stormflow^ 333.2.3 Configuration of stormflow time-area 413.3 Digital Derivations of Other Watershed Characteristics^ 523.3.1 Drainage density^ 523.3.2 Divides of watershed and its subwatersheds^ 543.3.3 Generalized stream channels^ 54Chapter IV MODEL ESTABLISHMENT AND TESTING^ 614.1 Delineation of Hydrologic Response Units 614.2 Simulation of Translation Hydrograph^ 654.3 Stormflow Detention^ 674.4 Model Testing 70iv4.4.1 Hydrograph separation^ 714.4.2 Hydrograph simulation 724.4.3 Comparison of parameters of the model^ 75Chapter V MODEL APPLICATION^ 775.1 Uniform Storm^ 785.2 Non-uniform Storm 825.3 Land Use Changes^ 91Chapter VI CONCLUSIONS AND RECOMMENDATIONS^ 104REFERENCES^ 109VList of Tables3.1 Comparison of the models of stormflow generation^ 303.2 Approximate average velocity in meters per second of runoff^ 464.1a Runoff coefficient in forest^ 644.1b Runoff coefficient in grass land 644.1c Runoff coefficient in urban area, 70% of area impervious^ 644.1d Runoff coefficient in forest under low antecedent soil moisture^ 654.2 Comparison of the parameters for the model^ 765.1 Regional R/(R + T) and areas of individual land use for the four scenarios ^ 1025.2 Parameters and results of the model for the four land use scenarios^ 102viList of Figures2.1 The locations of Jamieson Creek and Nitinat watersheds^ 112.2 A contour map of Jamieson Creek watershed^ 122.3 A 3-dimensional perspective of Jamieson Creek watershed^ 132.4 A contour map of Nitinat watershed^ 142.5 A 3-dimensional perspective of Nitinat watershed^ 152.6 Distributions of aspects of land slope in Jamieson Creek and Nitinat watersheds. 152.7 Distribution of land slope in Jamieson Creek watershed (after Cheng, 1975) ^ 162.8 Distribution of land slope in Nitinat watershed^ 162.9 Longitudinal stream channel profile of jamieson Creek watershed^ 182.10 Longitudinal stream channel profile of Nitinat watershed^ 182.11 Distribution of monthly precipitation in Jamieson Creek watershed^ 192.12 Distribution of monthly precipitation in Nitinat watershed^ 192.13 Distribution of daily average streamflow for each month from Jamieson Creek ^ 212.14 Distribution of daily average water stage measured at the outlet ofNitinat watershed for each month^ 213.1 Typical single-storm hydrograph 263.2 Flow chart of the downhill searching program (DSID) ^ 353.3 The direction of stormflow^ 353.4 The pathways of stornnflow 383.5 Simulated pathways of stormflow from each point on Nitinat watershed ^ 393.6 The accumulated numbers of upstream points ^ 40vii3.7 Derived and mapped stream channels in Jamieson Creek watershed ^ 423.8 Derived and mapped stream channels in Nitinat watershed^ 433.9 Linear regression lines for flow velocity as related to land slope with differentland uses^ 463.10 15-minute flow time-area for Jamieson Creek watershed^ 483.11 2-hour flow time-area for Nitinat watershed^ 493.12 10-minute flow time-area for Jamieson Creek watershed^ 503.13 1-hour flow time-area for Nitinat watershed^ 513.14 Raster derived divides on Nitinat watershed 553.15 Vector derived divides on Nitinat watershed^ 563.16 3-dimensional plot of Nitinat watershed, draped with derived divide lines ^ 573.17 Generalization of stream channels in Nitinat watershed^ 593.18 The derived variable area source of Jamieson Creek 604.1 Schematic of the model^ 624.2 Diagrammatic illustration of the discrete convolution equation for a linearwatershed^ 684.3 An observed hydrograph for the storm occurring in Jamieson Creekon July 16, 1977^ 734.4 An observed hydrograph for the storm occurring in Jamieson Creekon August 22, 1977^ 734.5 The simulated hydrograph for the storm occurring in Jamieson Creekon July 16, 1977^ 754.6 The simulated hydrograph for the storm occurring in Jamieson Creekon August 22, 1977 ^ 75VIII5.1 Rainfall intensity curve for 200-year return period in Nitinat watershed ^ 795.2 The simulated hydrographs at the outlet of Nitinat watershedfor the 200-year return period of rainfall with different durations ^ 815.3 The simulated peak flow at the outlet of Nitinat watershedfor the 200-year return period of rainfall with different durations ^ 815.4a Assumed storm hyetos on the north of Nitinat watershed 845.4b 3-dimensional appearance of the assumed storm on the north ofNitinat watershed^ 855.5a Assumed storm hyetos on the centre of Nitinat watershed^ 865.5b 3-dimensional appearance of the assumed storm on the centre ofNitinat watershed^ 875.6a Assumed storm hyetos on the south of Nitinat watershed^ 885.6b 3-dimensional appearance of the assumed storm on the south ofNitinat watershed^ 895.7 Stormflow hydrographs simulated for the four patterns ofassumed storms passing over Nitinat watershed^ 905.8a Land use scenario 1 ^ 935.8b Two-hour time-area map for the land use scenario 1 ^ 945.9a Land use scenario 2^ 955.9b Two-hour time-area map for the land use scenario 2^ 965.10a Land use scenario 3^ 975.10b Two-hour time-area map for the land use scenario 3^ 985.11a Land use scenario 4^ 995.11b Two-hour time-area map for the land use scenario 4^ 100ix5.12 Areas of 2-hour time-area for the four land use scenarios^ 1015.13 Distribution of runoff coefficient on the time-area maps for each scenario ^ 1015.14 Stormflow hydrographs simulated for the four scenarios of land uses^ 103xAcknowledgementI wish to express my deep appreciation to Dr. S.T. Chieng for his guidance,encouragement and great help throughout my whole graduate study. I extend my verysincere thanks to Dr. John C.W. Keng for financially supporting me with Dr. S.T. Chiengfor this M.Sc. programme and for his valuable advices for my study in Canada.I would like to gratefully appreciate Dr. D. Golding for generously providing me withlong term hydrologic data measured in Jamieson Creek watershed, and for his carefulreview and constructive comments on this thesis. Appreciation is also extended to Dr. A.Lau for being the committee member reviewing this thesis and giving me good advices.My special appreciation goes to Mr. K.C. Rai for his dedicated work in helping meobtain high quality hydrologic data and valuable references, and for his frank andconstructive advices. Special thanks are due to Mr. A. Loukas and Mr. G. Wu for theirsuggestive discussions regarding to my study. I also appreciate Mr. J. Huang for hisassistance in printing this thesis. Mr. R. Cheng, who read through the whole thesis forediting, deserves special thanks.There are so many professors, students and friends, who directly or indirectlycontributed to this thesis. I would like to express my deep appreciations to them.I am particularly indebted to my parents for their care, love, understanding andwarm encouragement.And finally, my greatest debt is to my wife, Saiping, who gave me tremendoussupport and whole-hearted love over the years. This thesis would never have beencompleted without her encouragement.xiChapter IINTRODUCTION1.1 Water Resource Management and Hydrologic ModellingWater resource is one of the most critical and dynamic natural resources in the world. Itis essential to plants, animals and human beings. Moreover, water is a renewable resourcethat is continuously in transit through various stages of the hydrologic cycle.The hydrologic cycle implies that water is in constant movement from place toplace and from one state to another. Although the overall amount of water is nearlyconstant for a certain place in a long period, the distributions of various components, suchas evapotranspiration, interception, infiltration, overland flow, interflow and channel floware rarely uniform in space and steady in time, which frequently causes flood or drought.The need of watershed management thus arises from the mismatch between thespatial and seasonal distribution of water as determined by the hydrologic cycle and thespatial and seasonal dimensions of the human need structure.In British Columbia, flooding resulting from long duration rains or a combination ofrain and snowmelt has caused extreme damage locally and in some cases over large area.In British Columbia, there has been widespread concern about changes to stormflowcharacteristics of streams following logging, particularly clearcutting on forestedwatersheds. Fishery managers are also concerned about the impacts of peak flow and thechange caused by logging on the environment of the streams where a large proportion ofChapter I Introduction^ 2coho, pink and chum salmon, and steelhead and cutthroat trout are produced.Commercial clearcutting, grazing, transportation and the increasing demands foragricultural land, industrial area and urban settlement have caused large scale change tovegetative cover and soil characteristics. The public has become increasingly concernedwith the subsequent impact on the hydrological regime in a watershed.Aside from land uses, the climate, soil, and topographic characteristics of awatershed also affect the hydrologic regime in general, and stormflow in particular.A reliable base of knowledge of the interactions between the aforementionedfactors and the hydrologic regime is therefore crucial for watershed management.However, a central problem for water resources managers and engineers to conductproper management is the insufficient measurement and study of hydrologic regime inmost of watersheds, especially in remote area.For this reason, a large number of models estimating peak flow and runoff volume,such as Stanford Watershed Model (Crawford and Linsley, 1966), HEC-1 (HEC, 1981),WATERSHED (Band, 1986a) and so forth, have been developed. Most of these models aredeterministic models, which can be divided into two categories: lumped models anddistributed models. In a lumped model, the hydrological processes are spatially averaged,or regarded as a single point in space without dimensions. For example, many models ofthe rainfall-runoff process treat the precipitation as uniform over a watershed and ignorethe internal spatial variation of the entire watershed. In contrast, a distributed modelconsiders the hydrologic processes taking place at various points in space and defines themodel variable as functions of the space dimensions.Since the hydrologic regime in a watershed varies in all three dimensions of space,Chapter I Introduction^ 3the estimation of average stormflow may not be sufficient to reflect the mechanism ofstormflow movement and satisfactory for the uses in large watershed management. Onthe other hand, however, explicitly accounting for all of this variation may make the modeltoo cumbersome for practical application, or even impractical for ungaged watershedswhere hydrologic information is limited.Apparently, the applications of many deterministic hydrologic models have beenlimited by the lack of ability to adequately represent spatial phenomena, a serious problemconsidering the basic spatial character of hydrologic processes. A solution or improvementon it may be found through the geographic information system (GIS) technology.1.2 GISThe term GIS has been defined in several ways. To some GIS means only the softwareused to analyze geographically referenced data; to others, the term includes the hardwareutilized by the system; yet others would include all processes from data acquisition to datapresentation, even the organizations operating the system. There are many definitions ofGIS (Tomlinson, 1984; Goodchild, 1985; Burrough, 1986; Aronoff, 1989 and Taylor,1991). The most obvious feature of GIS is the capability of spatial analysis. This capabilitydistinguishes GIS itself from other systems, such as computer-assisted mapping (CAM),computer-aided design (CAD), land information system (LIS), data base managementsystem (DBMS). One common definition is: GIS is a computerized system for capturing,storing, retrieving, analyzing and displaying data that are spatially referenced to the Earth.It can be said that GIS is an information system primarily concerned with spatial andChapter I Introduction^ 4temporal phenomena ranging in scale from the entire Earth down to a land parcel (Chieng,1990).A GIS combines two computer software technologies: data base management anddigital mapping. Data base management is a systematic way of organizing and accessingattribute data. Digital mapping represents map elements as points, lines, polygons, or gridcells. The key feature of a GIS is that the digital map elements are linked with the attributeinformation in such a way that, when either the map or the attribute data are manipulated,both sets of data are updated and adjusted to maintain the relationship between them(Lanfear, 1990). From this linkage perspective, the differences between GIS and LIS arenot great. However, LIS are used primarily for the storage and retrieval of spatial datawhile GIS are used essentially for more complex spatial analysis.1.3 GIS and Hydrologic Modelling1.3.1 GIS and parameterizationSince many hydrologic models have parameters defined in terms of land use, soil,precipitation and topography which vary greatly on a watershed, the potential for applyingGIS to hydrologic modeling is considerable. Several GIS applications in hydrologic modelinghave achieved significant improvements in the quality and efficiency of analysis in waterresources management (Muzik, 1990; Vieux, et al., 1988 and Jett, et al., 1979).GIS is in essence a spatial data base management system. Therefore, it has beenused to concentrate on providing computerized abilities to input, store, edit, query andoutput land and water-related information important in stormflow modeling, which makesChapter I Introduction^ 5spatial data collection become less model-oriented like data base management system(DBMS). With the flexible means of storing data representing the physical system, itbecomes possible to use a variety of alternative stormflow models with a single data baseselecting the most appropriate model for different phases of hydrologic modeling. Muzikand Pomeroy (1990) described a hydrologically oriented geographic information system.This system is a raster (i.e. grid cells) based GIS which stored hydrological parameterssuch as land use, soil type, rainfall intensity-frequency-duration statistics, runoff curvenumbers (CN), regional dimensionless unit hydrograph, and regional lag-time relationship,required for stormflow prediction. It was concluded that the relatively laborious task ofdata input for permanent storage in GIS is more than compensated for by the speed andefficiency achieved in subsequent hydrologic simulation. Sasowsky and Gardner (1991)developed a set of GIS techniques that provides many of the relevant topographic and soilparameters in hydrology modeling. It was pointed out that GIS allows for (1) rapidparameterization of relevant topographic parameters from grid cell digital elevation models,and (2) computation of weighted averages for appropriate topographic and soil parametersin each watershed configuration.1.3.2 Watershed configuration and digital elevation model of GISDigital elevation model (DEM) now has become a major feature of GIS, which gives GISthe capability of converting elevation contours or points into 3 dimensional graphs. Moreimportantly, the elevation data output from DEM is frequently used for watershedconfiguration. Since the hydrologic response of watershed is governed, in part, by thecharacteristics of watershed such as the shape, size, slope, the length of main stream andChapter I Introduction^ 6so forth, the degree of complexity presenting these hydrologic characteristics maysignificantly affect the simulation results of hydrologic models. The usefulness of DEM instormflow modelling has been recognized in several recent studies. Jett, Weeks andGrayman (1979) applied the Triangulated Irregular Network (TIN), a terrain network inwhich terrain is represented as a faceted surface with each facet being a triangular plane,to derive the stream network of 24 county area of the Ketucky river basin in the UnitedStates. Sasowsky and Gardner (1991) obtained the contributing areas and streamsegments in a watershed by using the DEM derived aspect data layer and x, y coordinatepair defining the location of the basin outlet cell. The same technique was used by Stuebeand Johnston (1990) to delineate the watershed boundary.1.3.3 The interface between GIS and hydrologic modelsGIS technology has evolved for 20 years since the first design concept was proposed byTomlinson in 1972. GIS application in hydrologic modeling has only about 10-year history.Many GISs only have some ad-hoc functions of spatial analysis, which make theminconvenient to be used for stormflow modeling. Most of GISs still lack of the powerfulspatial analysis capabilities specially for stormflow modeling. Alternatively, some GISapplications in stormflow modelling have to develop an interface between GIS andavailable hydrologic models. Hodge et al (1988) described the linkage of a hydrographicprogram and watershed process model with the raster GIS (GRASS). The hydrographicprogram, "Watershed" (Version 3.0), was designed to find watershed boundaries, stormdrainage channels and sub-basins for the watershed in GRASS. Wolfe and Neale (1988)used GRASS to provide limited data input to a finite element model. Fisher (1989)Chapter I Introduction^ 7developed an interface to automate many spatial display and analysis tasks with an easilyunderstandable screen selection.It has been felt that most existing GISs have many capabilities that may only be ofmarginal use in hydrologic modeling while there are many parameters and proceduresunique to stormflow modeling that are not included in standard GISs. Considerableresearch in the GIS applications to stormflow modeling has been conducted in the spatialvariability of rainfall and watershed characteristics such as terrain, land use, soil, etc.There has not been, however, commensurate level of effort spent on other spatialconsiderations in the actual transformations of rainfall into stormflow.The GIS applications to stormflow modeling in present study will be emphasisedin two major aspects: one is taking advantage of standard GIS to extract, overlay anddelineate land and water-related characteristics required for stormflow modelling; anotheris integrating GIS with hydrologic modeling to simulate both spatial and temporaltransformation of rainfall into stormf low.1.4 ObjectivesThe main objectives of this study are:a. to use GIS technology to establish a hydrologic model for simulating stormflowhydrograph,b. to estimate the peak flow of a watershed for a given rainfall return period,c. to simulate hydrographs of stormflow caused by spatially non-uniform andtemporally unsteady storms passing over a watershed,Chapter I Introduction^ 8d. to evaluate the impacts of land use changes on the discharge of a watershedunder a design storm, and,e. to explore the use of digital elevation model (DEM) in GIS for watershedconfiguration.1.5 Organization of the ThesisThis chapter (Chapter One) is an introduction of the study. The applications of GIS tohydrologic modeling have been reviewed and the objectives were specified.In Chapter Two, the background information of study watersheds is provided,including topography, climate, soil, forest, as well as instrumentation and data collection.Both study watersheds are steep watersheds with shallow soil and presently covered withdense coniferous forest. The physiographic similarities existing between the two studywatersheds have been analyzed in this chapter, which gives the confidence for the modelestablished in one watershed to be applied in another.The theories of stormflow mechanism and automated techniques in watershedconfiguration are reviewed in Chapter Three. A downhill searching program is described.This program is designed to simulate the direction and path of stormflow movement in awatershed. As the program searches through the grid formatted elevation matrix of awatershed, the flow length, flow time and contributing area of each grid are calculated,and the flow path is recorded as well. Consequently, many watershed characteristics suchas stormflow time-area, watershed boundary and drainage network can be figured outwhen this program is interfaced with GIS.Chapter I Introduction^ 9Subsequently, the establishment and verification of a stormflow model is discussedin Chapter Four. The synthetic Clark's Instantaneous Unit Hydrograph Time-Area Methodis modified and integrated with GIS in this model. Stormflow is considered to be generatedfrom hydrologic response units determined by soil, land use and topography of awatershed. The stormflow translation is simulated with the distributed method while thestormflow attenuation is presented using the lumped method. The two most importantfactors for the translation (time of concentration, 7) and attenuation (storage factor, R)are verified with the real data from a small, steep and forested watershed, JamiesonCreek, located at the North Vancouver of British Columbia.Chapter Five gives several applications of the verified stornnflow model. The effectsof land use and storm movement on stormflow are evaluated by changing the spatialdistribution of rainfall and land use within GIS. These evaluations are conducted for anungaged watershed, Nitinat watershed in Vancouver Island, where DFO (Department ofFisheries and Oceans) Nitinat River Hatchery is located. In addition, the Rational Methodis reexamined for design rainfalls with different durations in a large watershed by using theestablished model.It is concluded in Chapter Six that GIS is very useful in stormflow modeling. Thisstudy has demonstrated several uses of GIS in this area. The results obtained from thisstudy can be further improved when the antecedent soil moisture spatial distribution iscombined with GIS. Based on the watershed configuration derived from elevation data, themodel will be able to be expanded to include the simulation of soil erosion and non-pointsource water pollution on a watershed.Chapter IISTUDY WATERSHEDSTwo watersheds in southwest British Columbia were selected for this study. One is asmall but well gaged watershed, Jamieson Creek watershed, used for the model testing;another is a large but ungaged watershed, Nitinat watershed, used for the modelapplication.Jamieson Creek watershed has an area of 2.99 km 2 and is situated at the NorthVancouver of British Columbia (see Figure 2.1). Since 1969, it has been involved in aseries of extensive research programs conducted by the Faculty of Forest, University ofBritish Columbia. In addition to a long period of systematic rainfall and streamflow records,many results obtained from previous research in this watershed are available for furtherhydrologic study in this region.Nitinat watershed is a large watershed with an area of 426.90 km 2 , located on thesouthwest side of Vancouver Island (see Figure 2.1). There is no gage station installed tomeasure discharge for this watershed. Instead, some justifications of discharge peak flowcan be roughly made based on the data of flood stage provided by a pumphouse at theoutlet of the watershed. It is such a large watershed that the effects of the complexitiesof storm, soil, land use and topography on discharge prediction cannot be ignored for ahydrologic model.This chapter will describe the physiographic characteristics and data collections forboth watersheds in order to provide the background information for the interpretation of/ 0British ColumbiaChapter II Study Watersheds^ 11Figure 2.1 The locations (E) of Jamieson Creek and Nitinat watersheds.Chapter II Study Watersheds^ 12the assumptions, parameterizations, simulations and applications of the model in thesubsequent chapters.2.1 TopographyAs shown in the topographic map given in Figure 2.2, Jamieson Creek watershed haselevations ranging from 1,000 feet (305 m) at the mouth of the watershed to 4,000 feet(1,310 m) at the highest point on the divide. It can be seen from the 3-dimensional? 1900 2000mI Figure 2.2 A contour map of Jamieson Creek watershed. The contours weredrawn at intervals of 500 feet.Chapter II Study Watersheds^ 13perspective of the watershed (see Figure 2.3), this is a small and simple watershed interms of the relief of the watershed. Northeastly, eastly, southwestly and southly facingland slopes are dominant on this watershed, while the main channel is oriented to thesoutheast.Figure 2.3 A 3-dimensional perspective of Jamieson Creek watershed.The elevation of Nitinat watershed ranges from 100 m to 1400 m (see Figure 2.4).As indicated in Figure 2.5, the watershed main channel has a general orientation to thesouth. Compared with Jamieson Creek watershed, Nitinat watershed is more complex inthe relief of elevation. The topographic complexity comparative to Jamieson Creekwatershed is also reflected on a variety of aspects for land slopes (see Figure 2.6).Both of the study watersheds are steep watersheds. Figure 2.7 and 2.8 show theland slope distribution curves for Jamieson Creek and Nitinat watersheds respectively.Jamieson Creek watershed has an average slope of 48% with 87% of the area havingChapter II Study Watersheds^ 14Figure 2.4 A contour map of Nitinat watershed. The contours were drawn atintervals of 100 meters.Chapter II Study Watersheds^ 15Figure 2.5 A 3-dimensional perspective of Nitinat watershed.40s 304o20io F^N^NE^E^SE^S^SW^w^NWAspect of Land Slopega NItnat watershed Ng Jamieson CreekFigure 2.6 Distributions of aspects of land slope in Jamieson Creek and Nitinatwatersheds.a.20 40 60Percentage of Area with Slope Greater than hclicated Vdue10DChapter II Study Watersheds^ 16Figure 2.7 Distribution of land slope in Jamieson Creek watershed (after Cheng,1975).Figure 2.8 Distribution of land slope in Nitinat watershed.Chapter II Study Watersheds^ 17slope more than 30%. Similarly, the average slope of Nitinat watershed is 40% and thereis 79% of the area with slopes more than 30%.The longitudinal main channel profiles for Jamieson Creek and Nitinat watershedsare given in Figure 2.9 and 2.10 respectively. The gradient of the main channel at thelower portion in Jamieson Creek is 10%, but the upper portion of the watershed reachesa slope of 100%. The gradient of the main channel in Nitinat watershed is 5% at the lowerportion and 105% at the upper portion of the main stream.2.2 Climate and StreamflowThe weather of southwestern British Columbia is dominated by low pressure systems inthe winter and high pressure systems in the summer. Prevailing winds are predominantlyfrom the southeast in the winter, while northwest winds predominate in the summer.Extremes of temperature are rare. Both summer and winter temperature are mild in thisregion (Jungen, 1985).The easterly moving moisture-laden masses bring a large amount of precipitationto the watersheds. The annual average precipitation in Jamieson Creek from 1982 to1988 is 3245.2 mm with a range from 3023.0 to 4025.0 mm. In Nitinat watershed forthe same period, the annual average precipitation is 3649.1 mm ranging from 2165.6 to4371.3 mm.For both watersheds, as shown in Figure 2.11 and 2.12, most of precipitation areconcentrated in the period of November to February. The period between June andSeptember is the driest period of the year, which is also the snow free period.Chapter II Study Watersheds^ 18Figure 2.9 Longitudinal stream channel profile of Jamieson Creek watershed.Figure 2.10 Longitudinal stream channel profile of Nitinat watershed.Chapter II Study Watersheds^ 19Figure 2.11 Distribution of monthly precipitation in Jamieson creek watershed.Figure 2.12 Distribution of monthly precipitation in Nitinat watershed.Chapter II Study Watersheds^ 20In Jamieson Creek watershed, the average annual streamf low ranges from 0.27 to0.31 m3/s. Streamflows in Jamieson Creek watershed are usually quick responses toprecipitation. Unlike the monthly distribution of precipitation, however, hydrographs formonthly streamflow have two peaks in Jamieson Creek watershed (Figure 2.13). One peakoccurs in November mostly due to high rainfall intensity, while another peak occurs as aresult of snowmelt in May. As the annually maximum flood stages at the outlet of Nitinatwatershed were observed between November and February (Figure 2.14), the rainfallintensity can be reasonably considered as a major factor of the maximum design flood inNitinat watershed.2.3 SoilCheng (1975) classified the soils in Jamieson Creek watershed as two types: steepmountain soils and valley bottom soils. The steep mountain soils, mainly ablation andcolluvium, are shallow and very permeable. The Valley bottom soils, consisting of glacio-alluvial and lacustrine soils, are thicker and have varying permeabilities. In general, thesoils of Jamieson Creek are mostly coarse-textured sands and gravelly sand barns,underlain by mostly granitic impermeable bedrock.Field observations have indicated that soil channels in the form of old root holes,structural channels or cracks widely exist in the profiles of the watersheds (Cheng, 1975).Because of the porous soils distributed over the entire watershed, overland flow is rarelyobserved except near stream channels or on bedrock.Most of soils in Nitinat watershed are gravelly sandy loam or loamy sand (Jungen,-0.cU00.5_>.;t1 0.1 -a _0Chapter II Study Watersheds^ 21JAN FEB 1.1.AR APR MAY LN AL AUG SEP OCT NOV DECMonthFigure 2.13 Distribution of daily average streamflow for each month from Jamieson creek.6^5.5 -44.5 -JAN FEB MAR APR MAY JJN JUL. AUG SEP OCT NOV DECMonthFigure 2.14 Distribution of daily average water stage measured at the outlet of Nitinatwatershed for each month.Chapter II Study Watersheds^ 221985). The soils become more coarse in the area approaching to stream channels. Thereare stony soils sparsely distributed on steep slopes and bedrock exposures are occasionallyfound on the upper part of the watershed.2.4 ForestBoth study watersheds are covered with mature and over-mature coniferous forest. Onone hand, the combined interception loss and evapotranspiration from the forest make thewatersheds consume more water than the watersheds with other types of vegetation; onthe other hand, the decaying roots of forest create soil channels through which stormflowis quickly conducted. In addition, litters, mainly tree leaves, on the soil surface increasethe storage capacity of watersheds, which will greatly attenuate the watershed peak flow.In Jamieson Creek watershed, western red cedar, western hemlock and douglasfir occupy most of area below the elevation of 900 m. Above the 900 m level, thedominant species are mountain hemlock, yellow cedar and amabilis fir.In Nitinat watershed, the forest species have the similar changes with elevation asin Jamieson Creek watershed. At the lower portion of the watershed, red cedar andwestern hemlock are dominant, while yellow cedar and mountain hemlock again becomecommon at the upper portion of the watershed.2.5 Instrumentation and Data Collection2.5.1 PrecipitationChapter II Study Watersheds^ 23a. Jamieson Creek watershedFive recording rain gages (Belford weighing-type precipitation gage) were installedalong the contours trails of Jamieson Creek in 1970. The gage nearest to the place wherestormf low is measured was selected for this study. The gage is serviced weekly. For themodel uses, the recorded rainfall data were sampled at a time interval of 15 minutes.b. Nitinat watershedThere are daily rainfall data available for nine years collected by the Department ofFisheries and Oceans (DFO) Nitinat River Hatchery at the outlet of Nitinat watershed. Sincethe length of these data is not long enough to give the confidence to determine the rainfallintensity for design storms, rainfall data from nearby weather stations, Carnation Creek(1977-1988), Port Alberni Airport (1969-1988) and Cowichan (1960-1988) were obtainedin order to estimate intensity-duration-frequency (IDF) curves for Nitinat watershed.2.5.2 Stormflowa. Jamieson Creek watershedA 120 ° V-notch weir and water stage recorder was installed at the mouth ofJamieson Creek watershed in 1970. The water stage is monitored by a Leupold andStevens water level recorder. A calibrated water stage-discharge relationship is used tocalculate discharge based on the measured water stage. For this model, the discharge datawere also sampled with a time interval of 15 minutes.b. Nitinat watershedThere is no gage installed for measuring discharges of Nitinat watershed. Someestimations of discharges for the watershed can be roughly made using the flood stageChapter II Study Watersheds^ 24data provided by the pumphouse at the outlet of Nitinat watershed. The stage datarecorded from 1982 to 1988 was used for this study. Since the water stage readingswere taken twice a day in the morning and afternoon, they did not necessarily reflect thestages for instantaneous peak flood discharges. The discharge of the watershed wasmeasured once on November 9, 1989 when a peak flow occurred (McFarlane, 1990).2.5.3 Other data collectionTopographic maps for both Jamieson Creek and Nitinat watershed, 1:50,000, wereobtained from the Survey and Mapping Branch: Energy, Mines and Resources, Canada.Land use and soil maps of Nitinat watershed, 1:100,000, were obtained from theSurveys and Resources Mapping Branch, B.C. Ministry of Environment.Land use and soil maps of Jamieson Creek watershed, 1:15,840, were obtainedfrom the Faculty of Forest, University of British Columbia.These maps were digitized into GIS. The information related to these maps werelinked with the digitized maps and processed for the model in GIS.Chapter IIIWATERSHED CONFIGURATIONA physically distributed hydrological model is established on the understanding ofhydrological process. Overland, partial area and variable source area flow are consideredas three typical models of flow generation of watershed, describing the sources wherestormflow is generated in the process. The generated stormflow is subsequently translatedto the outlet of watershed in the forms of overland flow, interflow or channel flow. Thepathway through which the stormflow is translated to the outlet from each point on thewatershed and the travelling time that stormflow takes to a point of interest are simulatedusing the digital elevation model of GIS. This chapter will discuss the mechanism ofstormflow generation and translation, and describe a computer program developed by theauthor to automatically delineate some hydrologically important geomorphic characteristicsof watershed.3.1 Stormflow GenerationThe time distribution of stormflow at a point of interest is usually expressed as ahydrograph, which essentially reflects the hydrologic nature of stormflow. A typicalhydrograph for a single storm is conventionally divided into two components: quick flowand slow flow (see Figure 3.1). The quick flow, or direct runoff, is produced by a volumeof water derived from the storm event, which usually occurs soon after the beginning ofChapter III Watershed Configuration^ 26the storm. The slow flow, or base flow, is contributed from groundwater, the rate ofwhich changes relatively slowly.The portion of rainfall contributing to the quick flow is called excess rainfall.Theoretically, the amount of excess rainfall is equal to that of quick flow. However, theway that quick flow is generated from excess rainfall may change the time distribution ofquick flow at the outlet of watershed, which is more important for land managers andengineers. Overland flow, partial area and variable source area are three major modelsdescribing how the quick flow is generated from a watershed.ImTimeFigure 3.1 Typical single-storm hydrograph.3.1.1 Overland flowHorton (1933) considered that stormflow is generated in the form of overland flow. InChapter III Watershed Configuration^ 27Horton's theory, the overland flow is produced only after the rainfall rate exceeds theinfiltration rate of the surface soil. The infiltration rate decreases with time during a stormuntil a capacity value of soil, the saturated conductivity of surface soil, is reached. Whenthe rainfall rate is greater than the rate of soil infiltration, the rainwater collects on the soilsurface. As the surface storage is filled up, overland flow occurs. Since this modelconsiders overland flow occurs uniformly over the watershed, the source area isconsidered equal to that of the entire watershed. Horton proposed that overland flow iscommon and areally widespread.The Hortonian concept of stormflow generation prevailed for many years, and isstill valid when applied to land surfaces with low soil infiltration capacity. Much of rangeland and urban area fall in this category. However, overland flow is rarely observed inforested watersheds of humid regions, where infiltration capacities are usually greater thanexpected rainfall intensities (Cheng, 1975; Hibbert and Troendle, 1988).3.1.2 Partial areaIn fact, not all parts of a forested watershed contribute equally to the runoff process inhumid regions. Overland flow is rare except as "overland flow" in small ephemeralwaterways. Under these conditions, interflow becomes a primary mechanism forgenerating stormflow.Freeze (1974) illustrated the process of interflow by using the results of hismathematical simulation. This model emphasizes that stormf low results from overland flowgenerated only from small but relatively limited areas near the stream (riparian or partialareas) during a storm event. According to this model, the rainfall falling on channels is theChapter III Watershed Configuration^ 28first contributor to the stormflow. The second to contribute are areas of shallow watertable close to the channels where saturation occurs from below because of rising shallowwater table fed by vertical infiltration of rainwater from above.For the partial area model, the conditions necessary for interflow to be a significantcontributor to stormflow are quite stringent. The interflow is assumed to move laterallydue to the difference of hydraulic head before it seeps from the hillside above the watertable. The saturated conductivity values used as the highest limits of the velocity ofinterflow are usually inadequate to explain the flashy response of storm in a forestedwatershed of humid regions (Loukas, 1991).3.1.3 Variable source areaThe explanation for the quick response of interflow to a storm in a forested watershed canbe given partly by the fact that soil channels (macropores, or soil pipes) exist in mostforest soils and provide pathways of low resistance to interflow (Whipkey, 1965;Aubertin, 1971). Some soil channels may exist between surface soil and a relativelyimpermeable soil layer beneath the surface soil because of the frequent passing of lateralsubsurface flow. Activities of small animals, insects and decaying roots partly account forthe existence of the soil channels.Soil channels begin to fill when rainwater ponds on the soil surface in localdepressions or when the surrounding soil matrix becomes saturated. Water conducted intothese channels moves downward in response to the force of gravity. The water even canbe delivered in the soil channels through the unsaturated soil zones during its movementdownslope, without significant losses (Mosley, 1982). In other words, stormf low may beChapter III Watershed Configuration^ 29generated before soil is completely saturated in a forested watershed of humid regions.Since the surrounding soil is saturated when the channel flow starts to deliver waterquickly downslope, the velocities of channel flow are comparable to those of overland flow(Mosley, 1979).Hursh (1936) presented the concept of interflow as the primary source ofstormflow from forest lands of humid regions. As pointed out by Hewlett (1974), theorigin of the "variable source area concept" of stornnflow generation has been the subjectof reports by Hursh (1936). Based on the observations in the field that steep watershedswith shallow soils produce more stormflow than the low slope watersheds with deep soils,Hewlett and his co-workers further developed this concept to give more precise answerfor the quick response caused by the interflow. In essence, the concept assumes thatcertain regions within a watershed contribute runoff to the streamflow while other areasact as recharge or storage zones. It suggests that the stream channel system expands intoand shrinks from intermittent and ephemeral source areas during and following rainfall.3.1.4 Comparison of the models of stormflow generationThe stormflow generation models discussed above described three possible sourcesproducing stormflow, the forces conducting stormflow to the outlet, the ways throughwhich rainwater is conducted, as well as the physiographic conditions under which thesemodels can be applied (see Table 3.1). They may serve either as separated or integratedpathways for stormflow concentration. Hewlett and Troendle (1975) concluded that thesemodels are different ways of looking at the same complex process in the first-order basin,and it will come as no surprise that the others are merely special cases of the variableChapter III Watershed Configuration^ 30source area concept, which attempt to explain the entire range of hillslope processes ofstormflow generation. Like overland flow, there seems little question that overland sourcewill also vary with increasing precipitation. Oka (1990) also noted that additional wateris drained into the lateral channels from soil matrix and increases the contribution ofchannel flow in the total runoff.Table 3.1 Comparison of the models of storm flow generationModels Sources Forces Pathways Soils Land Use ClimateOverlandflowThe wholewatershedGravitationalpotentialSoilsurfaceFine soil,deep orshallowRange,cropped,urban areaArid andhumidPartialareaRiparianareas andchannelsHydraulicheadSoil matrix Coarse,deep soilForestedlandHumidVariable From Gravitational Soil Very Forested Humidsourceareariparian tothe wholewatershedpotential channels coarse,shallowsoillandAs described in Chapter II, both Jamieson Creek and Nitinat watershed are steep,presently forested watersheds with shallow, thin and well drained soils in humid area. Thewide existence of interconnected soil channels has played a very important role inproducing stormflow in these areas (Cheng, 1975; Loukas, 1991). The expanding andshrinking nature of the stream network observed by Cheng (1975) indicates thatstormflow generation mechanisms in forested watersheds of these areas are similar to themodel of interflow from a variable source area of the watersheds.Chapter III Watershed Configuration^ 313.2 Stormflow TranslationIn theory, all "effective" raindrops striking on a watershed will eventually find their waysto the outlet of the watershed. The sources in which the stormflow is generated fromthese raindrops have been discussed in the previous section. The question is: how is thestormflow translated from where it is generated to the outlet of a watershed? This sectiondescribes a computer program designed for simulating the pathway and time of stormflow,two important aspects of stormflow translation.Manual interpretation of the pathway and time of stormflow from topographic mapsor aerial photographs and the subsequent measurement or digitization of geometric ortopologic properties may be quite tedious, time-consuming and error-prone for any but thesmallest data sets. The development of automated techniques to extract, store andprovide measurements of the hydrological parameters related to geomorphology ofwatershed has recently received increased attention (Puecker and Douglas, 1975; Mark,1984; O'Callaghan and Mark, 1984; Marks et al., 1984; Band, 1986; Hodge et al., 1988).The growing availability of digital elevation model (DEM) facilitates the applicability ofthese techniques to a variety of hydrologic research.The most common form of DEM used for deriving the hydrological characteristicsof watershed is the elevation matrix usually obtained from quantitative measurementsfrom stereoscopic aerial photographs. The elevation matrix can also be converted fromelevation contours digitized or scanned from quadrangle topographic maps using GIStechniques.Because of the ease with which matrices are processed in the computer,Chapter III Watershed Configuration^ 32particularly in raster-based geographical information systems, two 1:50,000 topographicmaps were digitized and then converted into the elevation matrices using DEM of GIS inthis study. The computer program designed for searching the stormflow pathway andcalculating the stormflow time is called downhill searching program (DSP), implying thatthe program traces the pathways of stormflow in the downslope direction throughout theelevation matrix of watershed.3.2.1 Assumptions of the simulation of stormflow translationBased on the models of stormflow generation as discussed in Section 3.1 and thephysiographical characteristics of both Jamieson Creek and Nitinat watersheds, thefollowing assumptions are made:Assumption 1. If the two watersheds are kept forested, soil channels are mainpathways for both of them to conduct rainwater to stream channels; and, if theland use is changed to grass land, urban area or a combination of them, overlandflow will be dominant where such changes occur.Since both interflow through soil channels and overflow on the soil surface are theresponse of stormflow to the gradient of gravitational potential, another assumption is:Assumption 2. The stormf low always goes in the direction of steepest slope.For forested watersheds in humid regions, rainwater is usually delivered throughChapter III Watershed Configuration^ 33vertical soil channels to a relative impermeable soil layer beneath the surface soil andbecomes lateral flow above the layer. Because the surface soil is shallow and coarse, itcan be assumed:Assumption 3. The impervious soil layer and soil channels conducting the lateralflow are basically parallel to the soil surface.The above three assumptions constitute the basis on which the pathways ofstormflow are simulated over the study watersheds. The last assumption gives theconvenience for modeling the pathways of interflow with the available surface elevationdata.3.2.2 Simulation of the pathways of stormflowPuecker and Douglas (1975) employed a simple local algorithm that uses a kernel of fourcells to detect stream lines and ridges. As this kernel is moved over the elevation matrixfor each set of four cells at a time, the concave and convex pixels are flagged as potentialstream and ridge points, which, however, are not well connected. Band (1986) improvedthis algorithm with a kernel of nine cells to refine the potential stream and ridge points intogeomorphologically reasonable, connected graph structures representing drainage channelsand divide networks. Although this algorithm works well, it is not based on anyunderstanding of fluvial process.Mark (1984) proposed a more realistic algorithm to detect the drainage lines on awatershed. For each point in the elevation matrix of the watershed, its elevation isChapter III Watershed Configuration^ 34compared with its eight neighbours within the 3 x 3 kernel. The lowest neighbour isflagged and the kernel is moved to the lowest neighbour. When this process is repeated,this algorithm accumulates upstream drainage area to successively lower pixels,delineating the major drainage lines.Sasowsky and Gardner (1991) applied a set algorithm to configure a watershed intodiscrete, connected channel segments and corresponding contributing areas. Thisalgorithm delineates the topographic boundary and drainage area of a basin by using theDEM derived aspect data layer and an x,y coordinate pair defining the location of the basinoutlet cell. The aspects of neighbours are recursively evaluated to determine the numberof cells accumulated and if a cell can be added to the channel network.By reference to the previous research, this study attempts to make animprovement for delineating stream channels and divides in steep watersheds. The contourlines were processed using a GIS program (TerraSoft registered by Digital ResourceSystems Ltd.) to produce the elevation matrices with a square-grid of 40 m x 40 m forJamieson Creek and 400 m x 400 m for Nitinat watershed. The downhill searchingprogram (see Figure 3.2), written in FORTRAN language, searches through the matricesto find the flow path from each point to the outlet and accumulates the number ofupstream points drained to each point from its upstream drainage area. Finally, the flowpathways and accumulated numbers of points are stored in two separated files, and inputback to TerraSoft as a new layer in GIS and a new GIS theme respectively. GIS displaysthe drainage network on the new layer, and outlines stream channels and divides of thewatershed with the new GIS theme.The program searches for the pathways of points on the watershed through theChapter III Watershed Configuration^ 35Figure 3.2 Flow chart of the downhill searching program (DSP)Digital Elevation DataA Kernel of 3 x 3 PointsChapter III Watershed Configuration^ 36elevation matrix with a kernel of 3 x 3 points. The direction in which the programconducts the search is determined by detecting the steepest downwards slope existingbetween the central point and its eight neighbour points within the kernel of 3 x 3 points(see Figure 3.3).Land Surface^Flow DirectionFlow DirectionFigure 3.3 The direction of storm flow. It is simulated by detecting the steepestslope between the central cell and its eight neighbour cells.The biggest problem for algorithms in using elevation matrices for continuouslydetecting the pathways of stormflow is that of "pits" in the digital surface caused eitherby data error or complex topography with natural or artificial depressions. Although thenumerical smoothing method, i.e., moving a kernel of 3 x 3 points over the elevationmatrix to recalculate the elevation data, can remove a large number of pits, it has beenChapter III Watershed Configuration^ 37noted that it may oversmooth the topography, leading to a loss of significant terraininformation (O'callaghan and Mark, 1984). To avoid the significant loss of terraininformation, the downhill searching program gives a small increase of elevation and a delayof travelling time of stormflow when a pit is met. Such an increase and delay will continueuntil the program finds its way out of the pit. In other words, the rainwater has to fill upthe pit before it flows out of it. If the pit is too big to be filled up within a certain of time,the area drained to the pit will be considered as a local drainage area excluded from thewatershed.For the areas, usually flat areas, within which there exist two or more points withequally steep downward slopes from the central point in the kernel, the program willarbitrarily designate the direction to the point closest to the outlet.If the program meets a point on the edge of the elevation matrix rather than theoutlet while searching for the flow pathways for a point of interest, the point of interestwill be flagged, indicating that the point of interest does not belong to the watershed,otherwise segments from the point of interest will be connected to the outlet as its flowpathway (see Figure 3.4).Figure 3.5 shows the pathways of stormflow from each point on Nitinat watershed.This graph was stored in the vector file with TerraSoft compatible format. Clearly, theflow pathway from any point on the watershed can be traced on the graph.As the flow pathway of each point is traced to the outlet, the points traversed bythe flow pathway automatically accumulate the numbers of upstream points from whichthey receive rainwater until all points have been evaluated. Figure 3.6 shows theaccumulated numbers derived from the elevation matrix presented in Figure 3.4.Chapter III Watershed Configuration^ 381010^1106^1000^1027^1091^11231 1 NIZ 11001--> 998—> 990-0- 928^989^994I NZ1001^1087^1000 ---b• 945 ---1. 912^945Ii1035 —4464-4- 901--* 898—÷ Flow Path^110271 Elevation DataFigure 3.4 The pathways of storm flow. The pathways are connected in thedirection of the deepest slope calculated using the elevations between theconcerned point and its eight neighbour points. They are stored in a vector filecompatible with GIS.The results are recorded in two files with the GIS compatible formats. One is avector file, containing the pathways of stormflow for each point; another is a raster file,containing the numbers of accumulated points. Each file is then transferred to GIS withcorresponding UTM (Universal Transverse Mercator) coordinates. The vector data becomesa new layer of watershed map in GIS, while the raster data is treated as a GIS theme forfurther analysis.I9874---1020Chapter III Watershed Configuration^ 39tiMIALA=9=''lliAmffl!'''"'"Figure 3.5 Simulated pathways of stormflow from each point on Nitinat watershed.Chapter III Watershed Configuration^ 40Figure 3.6 The accumulated numbers of upstream points. It shows the number ofupstream points from which rainwater contributes to each point. Divides can belocated where the number is equal to one, while the stream channels can beidentified for the points with the larger accumulated numbers.Since each point has the same area of grid cell in the elevation matrix, theaccumulated number of upstream points, in essence, represents the area contributingrainwater to the point of interest, namely, the drainage area.Unlike the pathways of stormflow from each point, stream channels can only befound in the places where stormflow frequently reaches a certain amount. Because theChapter III Watershed Configuration^ 41derived contributing area indicates a relative amount of stormflow that a point receivesfrom its drainage area, the points with larger contributing areas are more likely to becomestream channels than those with smaller contributing areas. Therefore, a threshold ofcontributing area can be selected to identify the stream channels by using the contributingareas data. Points with their contributing areas larger than the threshold are consideredon the stream channels of the watershed, as shown in Figure 3.6.To match the derived stream channels to mapped ones, the threshold may varywith watersheds because of the problem of different map scales. For Jamieson Creekwatershed, the threshold is selected as a contributing area of 30 points (48,000 m 2) todefine the stream channels (see Figure 3.7), whereas a threshold contributing area of 5points (800,000 m 2) is used to identify the stream channels in Nitinat watershed (Figure3.8). The good consistency between the derived and natural channels shown in Figure 3.8for Nitinat watershed demonstrates that the downhill searching program can perform wellin automatically deriving the stream channels in forested watersheds with steep slope inhumid areas. For Jamieson Creek watershed as shown in Figure 3.7, the result could bebetter if more accurate elevation data could be provided, because the program becomesmore sensitive to the accuracy of elevation data as the grid resolution increases.3.2.3 Configuration of stormflow time-areaAlthough the quantitative analyses of drainage network have gone through dramaticadvance, there has not been a commensurate level of effort spent on the coupling of theseanalyses with the most important hydrologic variable, namely, the streamflow responseto rainfall in a watershed. Some researchers (Rodrigues-Iturbe, Valdes and Devoto, 1979)o^1000m1 1^ 1Chapter III Watershed Configuration^ 42Figure 3.7. Derived (x) and mapped (—) stream channels in Jamieson Creekwatershed.Chapter III Watershed Configuration^ 43Figure 3.8. Derived (x) and mapped (—) stream channels in Nitinat watershed.Chapter III Watershed Configuration^ 44attempted to link the instantaneous unit hydrograph (IUH) with some geomorphologicparameters described by Horton's well-known empirical laws (Horton, 1945). The derivedequations express the IUH as a function of Horton's numbers, a mean velocity ofstreamflow, but they do not give a good representation of the physical mechanism of thestormflow process.An interesting way to understand how excess rainfall is converted into stormflowis to use the concept of flow time-area. It assumes that the outflow hydrograph resultsfrom the pure translation of direct runoff to the outlet, ignoring any storage effects in thewatershed. If one unit of excess rainfall were spread uniformly over the watershedinstantaneously, rainwater first flows from areas immediately adjacent to the outlet, andother upstream areas will subsequently contribute to the outlet until rainwater from theremotest area reaches the outlet with the travelling time equal to the time of concentration(Tc). The flow time-areas are defined by dividing the area of a watershed into subareaswith distinct runoff travelling times to the outlet. These areas are delineated withisochrones of equal travelling time, spaced in equal travelling time increments numberedsequentially upstream from the outlet (Hoggan, 1989).The flow time-area concept provides useful insight into the runoff phenomena.However, its application for estimating stormflow is limited because of the difficulty ofconstructing isochronal lines and because the hydrograph must be further adjusted orrouted to represent storage effects in the watershed. An algorithm to construct isochronallines for a watershed using digital elevation data is described below. Discussion of thestorage effects on the time-area method is given in next chapter.The travelling time of stormflow from one point on a watershed to another can be2/3^1/2R. XS1„,V.=nm 3-2-2Chapter III Watershed Configuration^ 45deduced from the flow distance and velocity.3-2-1where tu is the flow travelling time from a point with the UTM coordinates of i,jtothe outlet;m=1,2,...,M, the steps from the point (i,j) to the outlet;A/m is the length of segment m on the pathway; and,Vm is the flow velocity with which rainwater goes through the segment m.In Manning's equation, the flow velocity of segment m is expressed as:where R. is the hydraulic radius;nm is the Manning roughness coefficient; and,Sim is the friction slope, equal to the slope of land surface or channel'sbottom for a steady uniform flow.To simplify the estimation of the hydraulic radius R„„ R.213/n. in Manning's equationis considered as a constant variable, P, for each pattern of land use. So the Manning'sequation can be rewritten as:V.=PXS.112^3-2-3where P is regressively derived by reference to data in Table 3.2.As shown in Figure 3.9, the velocities, in meters per minute, for different land usepatterns described by the regression equations are:Fora*YE•16.98*Sul R0.9687.^.^;Slope (X)IIE10Natund ChannelacYin2823*S'a^R-011529.•w----"'"------------i-^.---.--2--- a ma•••■■ 300 -LAChapter III Watershed Configuration^ 46Table 3.2 Approximate average velocity in metres per minute of runoffSlope in PercentLand use 0-3 4-7 8-11 12-Forest 0-28 28-46 46-59 59-Grass land 0-46 46-64 64-78 78-Urban area 0-155 155-247 247-311 311-Natural channel 0-37 37-73 73-128 128-(Source: Applied Hydrology, Chow et al., 1988, p.165)1^5Slope (%).^10^. 1^5^ inMGM (%)Figure 3.9 Linear regression lines for flow velocity as related to land slope withdifferent land uses. The experimental data are referred to Table 3.2.Chapter III Watershed Configuration^ 47Forest, V,,, = 16.98xS„,1'2Range Land, V,,, = 23.81xS„,1/2Urban Area, Vm = 91 .18xS,,,1/2Natural Channels, V,,,= 28.23xSmv2The flow travelling time is given by:titif L1'n =i1 Alm = 1 if A insS„-,1/20=1 Vnt m=1 M112 P m=i 3-2-4where tu is calculated by an application program written in FORTRAN language.This application program is quite similar to the downhill searching program (seeFigure 3.2). The difference is that it sums the travelling time to the outlet for each pointwhile tracing its flow pathway. In addition to the elevation matrix data, the derivedcontributing area data are processed to locate the natural channels where the flow velocityis different than other land uses. The travelling time of stormflow for other land usepatterns will be discussed in Chapter V.The travelling time data are then input back to TerraSoft as another GIS theme forthe configuration of time-area maps. Figure 3.10 and Figure 3.11 show the derived 15-minute and 2-hour time-area maps for Jamieson Creek and Nitinat watersheds respectivelywhen both watersheds are covered by forest and drained by natural channels.Because the flow travelling time data are stored in GIS with grid-based format foreach point, the flow time-areas can be spaced at any selected time intervals by simplychanging the color scheme for the representation on computer. As a result, GIS can easilyproduce a flow time-area map with selected flow time interval suitable for hydrologicalmodeling. Figure 3.12 and Figure 3.13 give another example of time-area maps at intervals0^1000m^2000mI I IChapter III Watershed Configuration^ 48Figure 3.10 15-minute flow time-area map for Jamieson CreekChapter III Watershed Configuration^ 49Rintat-WEEME2"immli111.11=1" ""Figure 3.11 2-hour flow time-area map for Nitinat watershed0^1000m^2000m1 I 150Chapter III Watershed ConfigurationFigure 3.12 10-minute flow time-area map for Jamieson Creek watershed.Chapter III Watershed Configuration^ 51NIT I Ntiftagffiligg111.0•1^gitlifill 11110•■•Figure 3.13 1-hour flow time-area map for Nitinat watershedChapter III Watershed Configuration^ 52of 10 minutes and 1 hour time for Jamieson Creek and Nitinat forested watershedsrespectively.3.3 Digital Derivations of Other Watershed CharacteristicsThe derived contributing area data have considerable potential for automaticallyconstructing the hydrological characteristics of watershed. The capabilities of GIS indisplaying and analyzing these spatially distributed data have made such constructionsmuch easier. The followings are some examples.3.3.1 Drainage densityThe drainage density is the length of streams per unit area within a watershed. In thesame climate region, a watershed with a higher value for drainage density will usually havea fine-textured topography and short, generally steep, slopes. Conversely, a watershedwith a lower drainage density will have longer, gentle slopes and larger distance betweenstream channels. To some extent, the similarity of drainage density between watershedsindicates the similarity of their geomorphological characteristics and hydrologicalresponses.The drainage density (Do) may be expressed as:3-3-1where IL. is the total length of streams; and,A is the drainage area.Chapter III Watershed Configuration^ 53Since each point in the contributing area matrix represents a square cell with samearea, the drainage density can be rewritten as:1xEn^nDd=^ = E12xEN 1xEN 3-3-2where In is the total number of the points whose contributing areas are greaterthan the selected threshold of stream channels;IA/ is the total number of the points within the watershed boundary; and,/ is the length of one side of the square cell for which a point represents.For Jamieson Creek, the selected threshold of accumulated numbers is 500 withan area of 800,000 rre, equivalent to the contributing area designated as the threshold of5 accumulated numbers for stream channels in Nitinat watershed. In, IN and / are 44,1711 and 40 m respectively. The drainage density is calculated as:Dd= En^44 NEN 40 X 1711 =0.0257(m/m2)3-3-3For Nitinat watershed, the threshold of 5 accumulated points is selected. In, INand / are 652, 2668 and 400 m respectively. The calculated density is:Dd= En . 652 =0.0244(m/m2)/xE N 400x2668 3-3-4The similar values of drainage density in Jamieson Creek and Nitinat watershedsgive another evidence that the hydrological similarity does exist between the twowatersheds, which supports the application of the hydrological model derived fromJamieson Creek to Nitinat watershed.Chapter III Watershed Configuration^ 543.3.2 Divides of watershed and its subwatershedsThe contributing area data indicate the total number of points from which a point ofinterest will receive rainwater if the rainfall continues indefinitely. Theoretically, a pointright on a hydrological divide will only receive rainwater from itself no matter how long therainfall lasts. The points with their accumulated numbers equal to one, therefore, can beautomatically distinguished as the divides of the watershed and its subwatersheds, asshown in Figure 3.14. The vector connections of these points can be used to delineate thesubwatersheds (see Figure 3.15).Having draped these derived vector-based divide lines on the 3-dimensional plot ofthe watershed, the GIS shows the good agreement existing between the derived dividesand the ones in the real world (see Figure 3.16).3.3.3 Generalized stream channelsThe channel network is a concept of both theoretical and practical importance in drainagewatershed analysis. As defined by Shreve (1966), a channel network consists of allchannels upstream from a given (arbitrarily chosen) point in the drainage system.Most ofexcess rainfall on the watershed is conducted to tributary channels through land surfaceor soil channels and then converged to the main stream which carries the water to theoutlet where the stormflow hydrograph is measured.In contrast to the tree-shaped drainage network that gives a perspective of thepathways of stormflow translation, the generalization of the drainage network has at leasttwo implications: one is that it represents the relative importance of stream channels inconducting stormflow; another is that it gives a rough idea of variable source area of theChapter III Watershed Configuration^ 55Figure 3.14 Raster derived divides on Nitinat watershed.Chapter III Watershed Configuration^ 56Figure 3.15 Vector derived divides on Nitinat watershed.Chapter III Watershed Configuration^ 57Figure 3.16 3-dimensional plot of Nitinat watershed, draped with derived dividelines.Chapter III Watershed Configuration^ 58watershed.Because the contributing area represents the area contributing to a concerned point,the points with larger values of contributing areas will be more important than the pointswith small values in the terms of producing and conducting stormflow. Therefore, thestream channels can be generalized to distinguish the tributary channels and main streamby flagging the points whose contributing areas are greater than the given thresholds.Figure 3.17 gives the results of the generalization of stream channels on Nitinatwatershed.Also, by adjusting the thresholds of contributing areas, the variable source area canbe roughly mapped out. Figure 3.18 shows that the sources of stormflow in JamiesonCreek expands from the places with larger contributing areas to the places with smallercontributing areas during rainfall and contracts thereafter. The principle applied here is thatthe more water a point receives the more easily it produces streamflow. The numericalsimulation of variable source area of Jamieson Creek is based on this principle andsuggests that the downhill searching program is very useful in enhancing the automatedconfigurations of hydrological characteristics of watershed. 7.1=11ii.M..MmImMEMAIIINIPChapter III Watershed Configuration^ 59MminallImie••••• Sio.1111.../...(a) (b)(c)(d)Figure 3.17 Generalization of stream channels on Nitinat watershed. The thresholds ofaccumulated numbers are 5 for (a), 30 for (b), 100 for (c) and 300 for (d).o^ImolaI (a)Chapter III Watershed Configuration^ 60(b)(c) (d)Figure 3.18 The derived variable area source of Jamieson Creek. The variable sourceareas are indicated by changing the thresholds of accumulated numbers. The flow sourceexpands from the areas with larger accumulated numbers to the areas with smalleraccumulated areas during rainfall, and shrinks in the reverse direction after rainfall. Thethresholds selected in this graph are, 300 for (a), 100 for (b), 30 for (c) and 10 for (d).Chapter IVMODEL ESTABLISHMENT AND TESTINGThe synthetic Clark's Instantaneous Unit Hydrograph Time-Area Method (Clark 1943),considering the discharge at any point in time as a function of the translation and storagecharacteristics of the watershed, was modified for this study. A model has beenestablished based on the integration of the modified Clark's method and GIS technology.By taking advantage of GIS technology in spatial information analysis, an attempt in thismodel has been made to simulate the effects of the complexities of rainfall, soil, land useand topography of watersheds on the prediction of stormflow. This is a storm event basismodel, consisting of three main components as depicted in Figure 4.1. The firstcomponent calculates runoff coefficients and generates excess rainfall for each discretearea of watershed. The second component simulates the translation of the generatedexcess rainfall from the discrete areas to the outlet and produces a pure translationhydrograph. The third component then routs the translation hydrograph to a hypotheticlinear reservoir and outputs the final simulated hydrograph. These three components ofthis model are run in sequence. The results of simulation were compared with themeasurements conducted in Jamieson Creek watershed.4.1 Delineation of Hydrologic Response Units6 /To describe the spatial variances of parameters affecting the runoff generation, theChapter IV Model Establishment and Testing^ 62Figure 4.1 Schematic of the model. It consists of three components, describingstorm flow generation, translation and detention.Chapter IV Model Establishment and Testing^ 63entire watershed is partitioned into small subareas, each of which is small enough to behydrologically considered as homogeneous area. These small subareas are calledhydrologic response units (HRU), which are the unique combination of land use, soil andtopography. The physiographic features of each identified unit in the watershed arerepresented by one set of parameters for each hydrologic process. These parameters areconsidered to be uniformly distributed with their average values within each unit.The delineation of the boundaries of HRU can be effectively conducted using GIStechnology. The land use, soil and topographic data are widely available in the forms ofmaps, satellite imageries or even digital files, which can be digitized, scanned ortransferred directly to GIS. In GIS, these digital information are topologically organized asindividual themes. The topologic structure for each theme not only puts the intelligenceto the graphic features of the maps to find their relationships, but also establishes thelinkage between the graphic features and their textural attributes. When these themes areoverlaid together, the topographic structure is reorganized to construct a new theme withits attributes extracted from the original themes.In this study, the land use, soil, and topographic maps of studied watersheds weredigitized and organized as themes using the GIS software, TerraSoft. The hydrologicresponse units (HRU) were identified as a GIS theme by overlaying these themes withtime-area theme in GIS. Each unit in the HRU theme was automatically related to a recordin the accompanying attribute data base. The attribute data base, therefore, contains landuse patterns, soil texture, flow travelling time and land slope originally existing in theseparated data base related to each individual theme. By referring to Table 4.1a, 4.1 b,4.1c and 4.1d, each HRU were assigned a runoff coefficient based on the combination ofChapter IV Model Establishment and Testing^ 64its land use pattern, soil texture and land slope. The runoff coefficient for each HRU wasthen multiplied by rainfall data extracted from the layer containing the information of apassing storm to yield excess rainfall.Table 4.1a Runoff Coefficient (Rc) in ForestSlope in Percent Open Sandy Loam Clay & Silt Loam Tight Clay0-5 0.10 0.30 0.405-10 0.25 0.35 0.5010-30 0.30 0.50 0.6030-50 0.40 0.60 0.7050-80 0.50 0.70 0.80>80 0.60 0.80 0.90Table 4.1b Runoff Coefficient (Rc) in Grass LandSlope in Percent Open Sandy Loam Clay & Silt Loam Tight Clay0-5 0.22 0.42 0.55-10 0.29 0.49 0.5710-30 0.35 0.55 0.6330-50 0.44 0.64 0.7250-80 0.53 0.73 0.81>80 0.62 0.82 0.92Table 4.1c Runoff Coefficient (Rc) in Urban Area, 70% of Area ImperviousSlope (%) 0-5 5-10 10-30 30-50 50-80 >80Rc 0.65 0.70 0.80 0.86 0.90 0.95Chapter IV Model Establishment and Testing^ 65Table 4.1d Runoff Coefficient (Rd) in Forest under low antecedent soil moistureSlope in Percent Open Sandy Loam Clay & Silt Loam Tight Clay0-5 0.07 0.25 0.345-10 0.10 0.30 0.4010-30 0.16 0.36 0.5530-50 0.22 0.42 0.6050-80 0.29 0.48 0.66>80 0.35 0.54 0.72(Compiled from Van Der Guilk, et al. (1986), McFarlane (1990), Chow, et al.(1988) and Viessman, et al. (1989). For rainfall intensity with 200 year return period,these runoff coefficients will be increased by 1.5 times. The maximum runoff coefficientis one.)4.2 Simulation of Translation HydrographThe translation of excess rainfall from its drop falling to the watershed outlet isaccomplished using the time-area curve for the watershed. The time-area curve is a formof unit hydrograph. If one unit of instantaneous excess rainfall is uniformly placed on thewatershed at t = 0, the runoff from each time-area will pass the outlet in the upstreamtime-area sequence. Accordingly, the ordinates of the one unit translation hydrograph isquantitatively equal to the dimensions of the time-area curve.Assuming that discharge is proportional to excess rainfall in a watershed, called alinear watershed, the discharges produced by excess rainfall separated with the sameselected time interval as the time-area can be added and convoluted to yield a stormChapter IV Model Establishment and Testing^ 66translation hydrograph for a given storm event. The governing convolution equation for thestorm translation hydrograph in discrete form is given as:nsMQ=, PmxA„,„+,-, 4-2-1where Q„ is the discharge at the outlet during the nth time interval;Pm is the depth of instantaneous excess rainfall uniformly falling on thewatershed at the beginning of mth time interval;A„,, i is the area of (n-m+ 1)th time -area.M is the number of time intervals of rainfall;m is 1 ,2,...,M.The summation is conducted to m = 1 ,2,...,n for n< =M, but for n> M, it is limitedto m = 1 ,2,...,M.This equation allows us to determine an instantaneous storm translation hydrographproduced by a temporally varying rainfall, but it does not account for the effects of thespatial distribution of rainfall and land use on the hydrograph, which may prevent themodel from being applied to large watersheds.To describe the spatial distribution of rainfall and land use in the model, each time-area is further divided into small hydrologically homogeneous sub-areas, i.e., HRUs. Anychanges of rainfall or land uses are, therefore, immediately reflected in the excess rainfallgenerated in each HRU. The excess rainfall falling on each HRU is summed for every time-area and then convoluted to yield the instantaneous stormflow translation hydrograph.The convolution equation is modified as:Chapter IV Model Establishment and Testing^ 67n5A4Qn'E E^x A,,m., 4-2-2where an is the discharge at the outlet during the nth time interval;Pm.n_m+t, is the depth of instantaneous excess rainfall falling on An.„,+u, atthe beginning of mth time interval;A,,„,,t; is the area of the ith HRU within time-areaL„,„+, is the total number of the HRUs in the time-area A,,„,,I.According to the Rational Method (Chow, 1988), the depth of excess rainfall isgiven as:Pm,n-m+1,I=Cn-m+1,iX m,n +1,14-2-3where /m,n_m+ij is the depth of instantaneous actual rainfall falling on^at thebeginning of mth time interval;Cn_m+t; is the runoff coefficient of the HRU,The term /m,,,m+ti represents the variance of rainfall in both space and time, whilethe term Cn_m+ t; reflects the effects of land uses, soils and topography of watershed onstormflow generation.An example of the application of the modified convolution equation to a passingstorm in a linear watershed is shown diagrammatically in Figure 4.2.4.3 Stormflow DetentionWhile stormflow is translated downstream over the land surface, through soil channels orIsochrone Watershedt+ tPulAzttChapter IV Model Establishment and Testing^ 68Storm at setFigure 4.2 Diagrammatic illustration of the discrete convolution equation for alinear watershed.Chapter IV Model Establishment and Testing^ 69along natural waterways, its peak flow is usually attenuated and delayed because ofstorage and resistance, i.e., stormflow detention of the watershed. The impact ofdetention on stormflow translation depends on the coverage, soil, slope of land, channelconditions and size of the watershed.Unlike stormflow translation, stormflow detention cannot be directly measured froma watershed. In synthetic Clark's Instantaneous Time-Area Method, a hypothetical linearreservoir with a storage coefficient, R, is assumed to describe the impact of stormflowdetention of the entire watershed on the translation hydrograph. The reservoir outflow,Qv and storage, St, are linearly related by:St=RxQ,4-3-1When a pure translation hydrograph is routed through the hypothetical linearreservoir, the relationship of inflow, 4, outflow, Qt, and storage, St, for the reservoir canbe given by the flow continuity equation as:4-0 dSt,=dt 4-3-2This equation can be approximated with Eq. (4-3-3) in the discrete form, where thesubscripts "t" and nt-1" denote the beginning and end, respectively, for At.0,1+0,=st-s,-,2^2^At^ 4-3-3Combining Eq. (4-3-1) with Eq. (4-3-3) and reorganizing the terms yield:= Atli +1)+(2R-At)at-,of^ -, (2R+At) 4-3-4Chapter IV Model Establishment and Testing^ 70Since 1,1 and 4 are known from the translation hydrograph for every time increment,the routed hydrograph is accomplished by solving Eq. (4-3-4) for successive timeincrements using each Of as 4_, for the next time increment.R is a constant and has the same time unit as At. For ungaged watersheds, it canbe determined from the ratio of RAT,- F R) which may only vary with land use patterns ina physiographically similar region, where 7; is the flow time of concentration. Thedimensionless term of RfiTe-FR) represents the storage characteristics for a kind ofwatershed regardless of the size of the watershed. Generally, the larger the value ofRfiTc+R), the larger the storage capacity of watershed. Not surprisingly, the watershedcovered with forest usually has larger value of RATc+13) than that covered with grass orurbanized.Because of the assumption of instantaneous rainfall in the Clark's method, the finalsimulated hydrograph produced by real continuous excess rainfall has to be computed byaveraging two same routed hydrographs spaced at an interval At, equivalent to thatselected for rainfall, apart withq= Qt+Qt-,2 4-3-5where 121 is the ordinate of the simulated hydrograph.4.4 Model TestingThis model includes two important parameters: stormflow travelling time, T, and thestorage coefficient, R. The stormflow travelling time can be determined based on the dataChapter IV Model Establishment and Testing^ 71of land use, land slope and flow travelling distance, which has been discussed in ChapterIII. To obtain the parameter, R, for this model, the simulated hydrograph was verified usingtrial and error technique to best fit the outflow hydrograph observed in Jamieson Creek.The verified model was then applied to another storm event in the same watershed tocheck for the agreement between the simulated hydrograph and the observed one.In general, any storm event with apparent discharge rise caused by pure rainfallintensity can be selected for the model testing. To determine the storage coefficient, R,for single land use, the storm events in Jamieson Creek occurring on July 16 and August22 of 1977 were selected, as the watershed was completely covered by forest at thattime. Both rainfall and stormflow data were sampled with the same interval of 15 minutesas that for time-area (see Figure 4.3 and 4.4).4.4.1 Hydrograph separationIn this study, no attempt has been made to partition the total stormflow hydrograph intooverland flow, interflow, channel flow and ground flow. Instead, the total stormflowhydrograph measured during a storm event is considered to be made up of twocomponents: quick flow and slow flow. The quick flow may be overland flow, interflowor channel flow, or a combination of them, which consists of the flashy rise part ofoutflow hydrograph. The slow flow results from ground flow, which comes gradually upand down during and after a storm event.There are many empirical techniques available for hydrograph separation. However,these techniques are more or less arbitrary since they do not have a sound physical basis.Fortunately, the slow flow is less important compared to the quick flow, which makesChapter IV Model Establishment and Testing^ 72most of the separation techniques acceptable in simulating stormflow hydrographs.A time-based separation technique proposed by Hewlett and Hibbert (1967) wasselected for this study. With this method, the hydrograph separation is made by drawinga line with a given slope from the point of initial hydrograph rise to the falling limb of thehydrograph. This technique has been widely applied to forest hydrology for many years.Cheng (1975) tested this technique with 20 stormflow hydrographs of Jamieson Creekwatershed and selected a slope of 0.55 litre s-' km-2h-1 for the separation line. After havingexamined the agreement between hydrographs separated by this technique and a modelusing 17 stormflow hydrographs observed in Jamieson Creek watershed, Loukas (1991)also suggested this technique can be used for the separation of hydrographs in JamiesonCreek watershed with the same slope of 0.55x10-3 m3s-1 km-2h-1.The same slope of the hydrograph separation line is used for this study. Since thearea of Jamieson Creek watershed is 2.99 km2, the slope is transformed to 0.41x10' m3s-1(15 minutes)-1 for the convenience of the model testing, as shown in Figure 4.3 and4.4).4.4.2 Hydrograph simulationWith the rainfall data of July 16, 1977 as input, the translation equation (4-2-2) anddetention equation (4-3-4) are run recursively using macros in a spreadsheet program,Quattro Pro. The simulated hydrograph was finalized with Eq. (4-3-4). After several timesof testing for the storage coefficient, R, the simulated hydrograph was obtained with thestorage coefficient of 180 minutes (see Figure 4.5).The obtained R is then applied to simulate an observed hydrograph for anotherChapter IV Model Establishment and Testing^ 73Figure 4.3 An observed hydrograph for the storm event occurring in Jamieson Creekwatershed on July 16, 1977.Figure 4.4 An observed hydrograph for the storm event occurring in Jamieson Creekwatershed on August 22, 1977.Chapter IV Model Establishment and Testing^ 74storm event occurring on August 22, 1977. Because the baseflow before the stormoccurred was very low, the runoff coefficient for dry condition in forest was used here toreduce the effect of antecedent soil moisture on stormflow simulation (see Table4.1d). Figure 4.6 shows the result of the hydrograph simulation. Obviously, a goodagreement has been achieved between the simulated hydrograph and the observedhydrograph. In addition, both the value of peak flow and the elapse time to peak flow forthis storm event are well approximated by the simulation.4.4.3 Comparison of parameters of the modelAs shown in Figure 4.3 and 4.4, there exists an inflection point on the falling limb ofhydrograph for every distinct storm. If a watershed is treated as a linear watershed witha linear reservoir at the outlet, the inflection point indicates the moment when the inflowfrom the linear watershed to the linear reservoir becomes zero. Therefore, dropping theinflow term at this point in Eq. (4-3-3) and combining Eq.(4-3-1) give:R= 10t-i +0d /2(0,-1-0,)At (4-4-1)Since the inflow terminates its influence on the outflow hydrograph at the outletat the inflection point, the flow time of concentration Tc, can be estimated as the timefrom the end of excess rainfall to the inflection point (Hoggan, 1989).Table 4.2 summarises the values of R, T, and the ratio of RAR+T) calculated forthe two selected storm events and used for the simulation model.The closeness of model parameters R and Tc, as well as the hydrographs, betweenthe observed and simulated results is the basis for confidence in applications of this modelChapter IV Model Establishment and Testing^ 75Figure 4.5. The simulation hydrograph for the storm event occurring in Jamieson Creekwatershed on July 16, 1977.Figure 4.6 The simulation hydrograph for the storm event occurring in Jamieson Creekwatershed on August 22, 1977.Chapter IV Model Establishment and Testing^ 76to ungaged or/and more complex watersheds, which will be discussed in next chapter.Table 4.2 Comparison of the parameters for the model.Sources R (minute) 7; (minute) RN! + TelModelStorm on July 16Storm on August 221801651806060750.750.720.74Chapter VMODEL APPLICATIONHundreds of hydrologic models have been developed for engineers and land managers topredict the behavior of hydrologic regime in watersheds since the first comprehensivehydrologic model was developed by a group of scientists at Stanford University (Crawfordand Linsley, 1966). A major advantage of simulation models is the insight intomechanisms of hydrologic processes gained by the practices of simulating the responsesof the overall model system and the interactions between various components of thesystem. Another advantage is that models can be nondestructively and repeatedly testedwith a little cost.However, most applications of these hydrologic models have been limited by thefacts that there exist the inherent variability in natural processes and the shortage ofhydrologic data required as input to the models. Lumped models can only be used in smallwatersheds due to their inability to deal with the complexities of hydrologic processes ineither space or time. Distributed models, on the other hand, are limited to well-gaugedwatersheds where long term systematic measurements have been conducted.Apparently, to be applied in a large but ungaged watershed, a hydrologic modelmay have to have the advantages of both lumped and distributed models. This chapter willdemonstrate the usefulness of the model described in last chapter. This model isessentially a combination of lumped and distributed models. With the power of GIS inhandling the variability of hydrologic processes, the model can even simulate stormflow17Chapter V Model Application^ 78hydrographs caused by complex storms in a watershed where soil, land use, rainfall andtopographic maps may be considered as only source of input data to the model.In this study, Nitinat watershed was selected for the applications of this model. Asdescribed in Chapter II, it is an ungaged watershed with an area of 426.90 km'. Only thepump house at the outlet of the watershed and nearby weather stations provide roughlyestimated peak flow and long term rainfall data for this watershed. Soil, land use andtopographic information are available in the forms of paper maps, which have beendigitized into GIS for this study.The DFO (Department of Fisheries and Oceans) Nitinat River Hatchery is located atthe outlet of this watershed. During the hatchery's nine years of operation, river floodlevels have come to within a few feet of the site elevation. Since a flood level in excessof 11.5 m site elevation would cause the immediate escape of the annual fish hatch,economic losses would be substantial, as hatchery returns thirty to fifty million dollars tothe wild fishery annually. In addition, a more substantial flood level could enable plannersto decide the viability of future hatchery expansions and upstream developments.According to McFarlane's study (1990), the substantial flood level is dependant onthe outflow, especially maximum peak flow caused by storm in this area. In this chapter,the effects of variability of storm and land use on the shape and timing of the responsehydrograph for Nitinat watershed will be examined.5.1 Uniform StormIt is usually the case that design storms are assumed to be distributed uniformly over aChapter V Model Application^ 79watershed. For the sake of safety, a return period of 200 years for these design stormswas used for this study. As shown in Figure 5.1, the relationships between rainfallintensity and duration has been found using the Gumbel Method (Gumbel, 1958) with therainfall data from nearby weather stations (McFarlane, 1990).Figure 5.1 Rainfall intensity curve for 200-year return period in Nitinat watershed.Because Nitinat watershed is currently covered with forest, the regional ratio,R/(R +Tc), has the constant:R =0.75R+T, 5-1-1The flow time of concentration, Tc, for Nitinat watershed can be obtained from theChapter V Model Application^ 80time-area map of the watershed, which is 16 hours (see Figure 3.11 and 3.13).Rearranging the above equation with the flow time of concentration, the storagecoefficient of Nitinat watershed is given as:R= 0.75T, =3 x 16 =48hours1-0.75 5-1-2Figure 5.2 shows the resultant hydrographs simulated by the model for differentdurations of the design storm with the 200-year return period. It should be noted thatwithin the same return period the maximum peak flow increases with the duration ofrainfall. Such increase becomes slow as the duration increases as shown in Figure 5.3,which suggests that the maximum peak flow can only be determined when the rainfallintensity with a comparable long duration is used for the model.This result seems to conflict with the concept of Rational Method, because, inRational Method, the maximum peak flow for a given return period is supposed to occurwhen the duration of rainfall reaches the flow time of concentration of the watershed.According to Rational Method, the maximum peak flow for the 200-year return period inNitinat watershed is only estimated as 666.8 m3/s if the duration of rainfall is 16 hours,the flow time of concentration in Nitinat watershed. In contrast, Figure 5.2 indicates thatthe maximum peak flow can expected to reach 1200 m3/s as the 200-year return periodof rainfall lasts longer than 56 hours. The low estimation of maximum peak flow obtainedby using Rational Method is partly because the Rational Method does not account for theeffects of storage in a watershed on the maximum peak flow.The watershed storage delays the occurrence time of peak flow and graduallyaccumulates the contribution of rainfall to the peak flow. Since the storage factor (R) forChapter V Model Application^ 81Figure 5.2 The simulated hydrographs at the outlet of Nitinat watershed for the200-year return period of rainfall with different durations.Figure 5.3 The simulated peak flows at the outlet of Nitinat watershed for the 200-year return period of rainfall with different durations.Chapter V Model Application^ 82the same land use pattern increases with the size of watershed, the use of RationalMethod will underestimate the maximum peak flow in larger watersheds due to thestorage effect. Because of this, the Rational Method should not be recommended for theselarger watersheds without modification.The lowest stage of water flow measured at the outlet of Nitinat watershed during1982-1988 is 4.6 m. It has been observed that a stage of 9.0 m was reached when totaldischarge was measured at 600 m 3/s at the same place on November 9, 1989(McFarlance, 1990). For an expected flood discharge of 1200 m 3 /s for the 200-year returnperiod, it is reasonable to expect the maximum flood stage at the peak flow to exceed thesite elevation of 11 m.5.2 Non-uniform StormAnother reason restricting the use of Rational Method is the assumption that rainfall fallsuniformly over the entire watershed and steadily continues from the beginning to the endof a storm event. Many other lumped models follow the same assumption because of theshortage of information in ungaged watersheds. Unfortunately, this assumption is rarelytrue in reality due to the large variability of rainfall intensity in both space and time.To estimate the effects of non-uniform storms on the time distribution of dischargeat the outlet of Nitinat watershed, four patterns of storms were examined with the model.The average rainfall intensity and duration are the same for the four patterns, as describedbelow:Chapter V Model Application^ 83Pattern 1. A storm moves from north to south toward the outlet of thewatershed for six hours. Figure 5.4a, 5.5a and 5.6a give three positions where thestorm moves by. The storm assumably spends two hours moving from one positionto another.Pattern 2. The storm moves along the same path and with the same time intervalas in pattern 1 but in the reverse direction.Pattern 3. Storms stay on the north, centre and south of the watershedsimultaneously for six hours.Pattern 4. An uniform and steady storm continuously stays over the entirewatershed for six hours.As seen in Figure 5.7, the storm moving toward watershed outlet results in thehighest peak flow and shortest lag time of peak flow compared with that in other patterns.The simulated hydrographs for these storm patterns also indicate that hydrograph shapecan be modified by spatial and temporal variations in rainfall intensity. Storms moving inopposite direction may produce significantly different hydrographs even though they bringthe same amount of rainfall to the watershed.To simulate the spatial and temporal variations of storm in the model, elevationcontours in DEM of GIS are replaced with storm isohyets so that rainfall information canbe interpolated to any point on the watershed (see Figure 5.4b, 5.5b and 5.6b). The stormChapter V Model Application^ 84Figure 5.4a Assumed storm hyetos on the north of Nitinat watershed.Chapter V Model Application^ 85Figure 5.4b 3-dimensional appearance of the assumed storm on the north of Nitinatwatershed.86Chapter V Model ApplicationFigure 5.5a Assumed storm hyetos on the centre of Nitinat watershed.Chapter V Model Application^87Figure 5.5b 3-dimensional appearance of the assumed storm on the centre of Nitinatwatershed.Chapter V Model Application^ 88Figure 5.6a Assumed storm hyetos on the south of Nitinat watershed.Chapter V Model Application^ 89Figure 5.6b 3-dimensional appearance of the assumed storm on the south of Nitinatwatershed.Chapter V Model Application^ 90Figure 5.7 Storm flow hydrographs simulated for four patterns of assumed stormspassing over Nitinat watershed.isohyets are organized on separate layers in GIS, representing the variations of rainfallintensity for each time interval. GIS sequently extracts these rainfall data on separatelayers into each hydrologic response unit (HRU), where excess rainfall is generated andtranslated to the outlet.It should be pointed out that it is the digital elevation model (DEM) that makes itpossible for the model to simulate discharge hydrographs caused by complex storms forungaged watersheds in real time if the isohyets maps of the watersheds can be timelyreceived from satellites, airplanes or weather stations.Chapter V Model Application^ 915.3 Land Use ChangesAt present, Nitinat watershed is completely covered with forest. The land use changesincluding clearcutting and urbanization on any part of this watershed in a large scale willdefinitely change the discharge pattern at the outlet, which may affect the developmentof fish hatchery downstream. Since some land use changes are irrevocable, the effectsof alternative schemes for such changes on hydrologic regime should be examined usingcomputer models before the mistakes are made to landscape.Four scenarios of land use changes for Nitinat watershed were evaluated in thisstudy:Scenario 1. The existing forest is kept untouched. The total area of forest is426.90 km2 equivalent to the area of the entire watershed (see Figure 5.8a).Scenario 2. The existing forest is altered to the combination with the urban areaof 76.44 km2 (17.91%), grass land of 212.61 km2 (49.80%) and forest of137.85 km2 (32.29%) (see Figure 5.9a).Scenario 3. The urban area as indicated in Figure 5.9a is changed to grass land(see Figure 5.10a), where the grass land is developed to 289.05 km2(67.71%) andthe area of forest is 137.85 km2 (32.29%).Scenario 4. The grass land as indicated in Figure 5.9a is returned to forest (seeChapter V Model Application^ 92Figure 5.11 a), where the forest occupies 350.46 km2(82.09%) and the urban areais 76.44 km2 (17.91%).The changes of land uses also change the travelling time of stormflow to the outletof the watershed. Consequently, each scenario has its own flow time-areas, which canbe calculated by using the flow velocity equations described in Chapter III and spaced withselected time intervals in GIS. Figure 5.8b, 5.9b, 5.10b and 5.11b show 2-hour time-areamaps for scenario 1, 2, 3 and 4 respectively . As compared in Figure 5.12, however, thedifference between the areas spaced for each time interval on these maps is small becausethe stormflow on large watersheds is mainly conducted through natural channels. Whatis significantly changed is the runoff coefficients for the areas of each time interval (seeFigure 5.13), which determines the generation of excess rainfall for each time-area in themodel.The ratios of R/(R + T) and storage coefficients for these scenarios in Nitinatwatershed depends on the regional ratios of R/(R + 7) and areal proportions for individualland use types involved in each scenario, which are given as:for scenario 1,R/(R + T) = 0.75R = 48.0 hoursfor scenario 2,RAR +^= 0.10 x 76.44/426.90 + 0.50 x 212.61/426.90 ++ 137.85/426.90= 0.51Chapter V Model Application 93Figure 5.8a Land use scenario 1. The land is completely covered with forest.Chapter V Model Application^ 94Figure 5.8b Two-hour time-area map for the land use scenario 1.95Chapter V Model ApplicationFigure 5.9a Land use scenario 2. The land is covered with forest, grass and urban area.Chapter V Model Application^ 96Figure 5.9b Two-hour time-area map for the land use scenario 2.97Chapter V Model ApplicationFigure 5.10a Land use scenario 3. The land is covered with forest and grass.Chapter V Model Application^ 98Figure 5.10 Two-hour time-area map for the land use scenario 3.99Chapter V Model ApplicationFigure 5.1 la Land use scenario 4. The land is covered with forest and urban area.Chapter V Model Application^ 100Figure 5.11b Two-hour time-area map for the land use scenario 4.Figure 5.12 Areas of 2-hour time-area for the four land use scenarios.0.200.35-0.50 --E-11" 0.75-U4I■4- 0.70 -0o0 0.65 -+-46 0.60 -CnLE 0.55 -0.50 -2^4^6^6^10Time (hr.)---1— smaria I -El- Scenario 2 -eli- Scenario 3 +Scenario 40.4512^14 16Chapter V Model Application^ 101^10000^2000-5000-4Ee 7000-047.e00 moo-4■13 4000-C3cr 3000-2000-^1000 ^2^4^6^6^lbTime (hr.)-+- Scenario 1 -ia- Scenario 2 -Xi- Scenario 3 +Scenario 4Figure 5.13 Runoff coefficients on each area of time-area maps for the fourscenarios.Chapter V Model Application^ 102R = 16.6 hoursfor scenario 3,R/(R + T) = 0.50 x 289.05/426.90 +0.75 x 137.85/426.90= 0.58R = 22.0 hoursfor scenario 4,R/(R+T) = 0.10 x 76.44/426.90 + 0.75 x 350.46/426.90= 0.63R = 27.2 hoursThe parameters related to these scenarios are summarised in Table 5.1 and 5.2.Table 5.1 Regional R/(R+T) and areas of individual land use for the four scenarios.Forest Grass land Urban areaArea(km2)% Area(km2)% Area(km2)Regional RAR +Tc) 0.75 0.50 0.10Scenario 1 426.90 100 0 0 0 0Scenario 2 137.85 32.29 212.61 49.80 76.44 17.91Scenario 3 137.85 32.29 289.05 67.71 0 0Scenario 4 350.46 82.09 0^. 0 76.44 17.91Table 5.2 Parameters and results of the model for the four land use scenarios.7", (hours) R (hours) R/(R + T) Peak flowDischarge (m3/s) Time (hours)Scenario 1 16 48.0 0.75 666.8 28Scenario 2 16 16.6 0.51 1715.1 22Scenario 3 16 22.0 0.58 1246.4 24Scenario 4 16 27.2 0.63 1214.9 2418001600 -•••-■...!) 140014.3 1200-E woo-cp0) 800--C 600-0.V) 400 -200 -32 40 48 56 64 72 BO 88 96 104Time (hr.)-11*- Scencrio 1^A Scenario 2 —)k— Scenario 3 -la- Scenario 4Chapter V Model Application^ 103The input rainfall intensity is given by using an uniform design storm with 200-yearreturn period and duration of 16 hours. As shown in Figure 5.14 and Table 5.2, the peakflow increases and the lag time to the peak flow decreases as the watershed is developed.The land use scenario 2, a combination of forest, grass land and urban area, will cause thehighest peak flow and shortest lag time to the peak flow, while forest tends to reducepeak flow and delays the time of arrival of peak flow. According to the relationshipbetween the flood stage and the peak flow discussed in Section 5.2, it has beeninterestingly noted that keeping the existing forest is the only scenario not causingoverbank flood for the storm with 200-year return period and 16-hour duration. Bycomparing the scenario 3 and 4, it can also be concluded that the watershed may respondto storms in the similar way though the land use scenarios may be different.Figure 5.14 Storm flow hydrographs simulated for the four scenarios of land uses.Chapter VICONCLUSIONS AND RECOMMENDATIONS6.1 ConclusionsThe GIS based model described in this thesis has demonstrated usefulness of applying GIStechnology to hydrologic modeling. With the proposed approach, the impact of land usepractices, such as logging, road construction and urbanization, on the hydrologic response(i.e., stormflow), can be quickly evaluated so that proper land use scenarios can beadopted before any irrevocable mistake has been made to the watershed itself.Time-area method is the key for the model. Unlike most of GIS based hydrologicmodels which mainly focus on model parameterizations, derivations of hydrologiccharacteristics of watershed, or interface between GIS and existing hydrologic models,this study took one step further to integrate the movement of stormflow with GIS by usingthe time-area method. Such integration provided the model with a greater power inhandling the effects of complexities of soils, land uses and rainfall on the simulation ofstormflow hydrogra ph.Spatial and temporal variations of rainfall intensity have substantial effects onstormflow in a watershed. Examining design storms on computer will aid in providing earlywarning of anticipated flood conditions. In this study, four patterns of storms passing overNitinat watershed were examined with respect to the stormflow responses to the storms.In fact, it is possible for the model to simulate the response to more complex stormstog--Chapter VI Conclusions and Recommendations^ 105moving at any velocities and in any directions. By processing rainfall intensity data in thedigital elevation model (DEM) of GIS, the model is capable of forecasting floods in real-timewith the rainfall information timely received from meteorologic radars, weather stations,satellites or airplanes.The model was established with the combined characteristics of distributed andlumped models. It can be used not only for modeling stormflow in a well gaged watershed,but also for giving a reasonable simulation of stormflow in an ungaged watershed. Soil,land use and topographic maps are commonly available information for most watersheds,but they may be only hydrologically related information available for ungaged watersheds.By taking advantages of GIS in spatial information processing, the model can help usersto achieve satisfactory simulations of stormflow hydrographs with these limitedinformation in ungaged watersheds.Because of the effect of watershed detention on stormflow, the assumption for theRational Method, that maximum peak flow caused by a storm occurs when the durationof the storm reaches the flow time of concentration, is rarely true except for small andsimple urban watersheds. In the watersheds with large storage capacity, peak flow isgreatly attenuated and the time to peak flow is delayed. As a result, for a design stormwith a given return period, peak flow increases with the duration of storm. Such increase,based on the results from the examination on the simulated stormflow in Nitinat watershedby using the model, may approach a stable level as the rainfall proceeds much longer thanthe flow time of concentration. Therefore, the maximum peak flow for a storm with agiven return period will be underestimated if the Rational Method is applied to thewatersheds with large storage capacity.Chapter VI Conclusions and Recommendations^ 106There are three models explaining the sources of stormflow generation: overlandsource, partial area and variable source area. It has been pointed out in Chapter III thatvariable area source is dominant in the study watersheds. Stormflow is conducted throughwidely existing soil channels in these watersheds and becomes overland flow near streamchannels. Though roughly, this study has made the first attempt to automatically simulatethe dynamic change of variable source area.The digital elevation model (DEM) is recognized as a useful tool to automaticallyderive hydrologic characteristics of watershed. Many previous researches have beencarried out for the automated delineations of these characteristics such as watershedboundaries and stream channels, but very few of them attempted to link the automaticallyderived hydrologic characteristics to hydrologic response. The delineation of flow time-areais an effort made in this study to establish such linkage so that GIS technology can bebetter integrated with hydrologic modeling.There are many parameters that have frequently been employed to characterizewatershed topography. These parameters, such as drainage density, main stream profile,slope, aspect, elevation of land surface, usually represent the long term interactionsbetween hydrologic regime and watershed surface. The similarities of the theseparameters between watersheds may determine the applicability of a hydrologic modelfrom one watershed to another. Manual interpretation of these parameters could be verycumbersome, time-consuming and error-prone. The automated techniques used for thisstudy provide a fast and more accurate way to characterize a watershed.The results of the model testing in Jamieson Creek are very encouraging.Applications of the model to Nitinat watershed indicates that the maximum flood level forChapter VI Conclusions and Recommendations^ 107the storm with 200 years return period would exceed the present hatchery site elevation.Logging and urbanization in the watershed would substantially increase peak flow for agiven return period.6.2 RecommendationsThe effects of soil, land use, rainfall intensity and topography on stormflowgeneration are reflected on the runoff coefficients assigned to each hydrologic responseunit (HRU) in the model. Although the antecedent soil moisture also greatly affectsstormflow generation, its spatial distribution is not involved in the current considerationof the model. Because antecedent soil moisture may vary greatly with watershedtopography, a more detailed analysis of antecedent soil moisture distribution will benefitthe improvement of the model accuracy. Further research may use the digital elevationmodel of GIS to relate antecedent soil moisture to evapotranspiration and solar radiationthat are partly determined by aspect, slope and elevation of land surface.The downhill searching program is able to trace the flow pathway for every pointon a watershed. By reversing the searching direction of the program, this program willhave the capability of delineating the boundary of drainage area for any point on thewatershed. Potentially, the traced flow pathway for each point on the watershed providesthe boundary conditions important for the finite element analysis in a numerical hydrologicmodel.The derived contributing areas from DEM for points on a watershed are very usefuldata, from which many important hydrologic characteristics such as watershedChapter VI Conclusions and Recommendations^ 108boundaries, stream channels, drainage density and variable source area of stormflow canbe obtained. Together with flow pathways and time-area data, the contributing area canbe further used to simulate soil erosion and non-point source water pollution of thewatershed because these derived results have indicated where, when and how muchstormflow comes to a point of interest on the watershed.It appears that a more comprehensive hydrologic model in conjunction with GIS canbe established to simulate water quantity and quality in a watershed. Because both GIStechnology and its applications to water resources management are still in their infantstage, many parameters and procedures unique to stormflow modeling have not beenincluded in standard GISs. With the improvement of GIS technology, however, it is notimpossible to establish a hydrologically oriented GIS with the power in both spatial datamanagement and modeling specific for water resources management.i 0 yREFERENCESAronoff, S., 1989. Geographic information systems: a management perspective, WDLPublication, Ottawa.Aubertin, G.M., 1971. Nature and extent of macropores in forest soils and their influenceon subsurface water movement, U.S. Forest Service Research, Paper NE-191.Band, L.E., 1986a. Analysis and representation of drainage basin structure with digitalelevation data, Proceedings of the International Conference on Spatial Data Handling,Seattle, Washington, pp. 437-450.Band, L.E., 1986b. Topographic partition of watersheds with digital elevation models,Water Resources Research, vol. 22, no. 1, pp. 15-24.Ball, D., and R. Babbage, 1989. Geographic information system: defence application,Brassey's Australia, Toronto.Barry, R.G., and R.J. Chorley, 1982. Atmosphere, weather & climate, Methuen, London.Bedient, P.B., and W.C. Huber, 1988. Hydrology and floodplain analysis, Addison-Wesley,New York.Bonvoisin, N., R. Moore, and D. Morris, 1990. A water information system: case study,GIS'90 Symposium, Vancouver, Canada.Bruce, M.A., and R.H. Clark, 1969. Introduction to hydrometeorology, Pergamon Press,Oxford.Burrough, P.A., 1986. Principles of geographical information systems for land resourcesassessment, Claredon Press, Oxford.Carter, J.R., 1988. A topology of geographic information systems. Technical Papers ofASCM-ASPRS Annual Convention, Vol. 5, pp. 207-215.Chang, M., S.P. Watters, and A.K. Sayok, 1989. A comparison of methods of estimatingmean watershed slope, Water Resources Bulletin, Vol. 25, No.2, pp. 327-333.Cheng, J.D., 1975. A study of stream flow hydrology of a small forested watershed in thecoast mountains of southwestern British Columbia, Ph.D. Thesis, Faculty of Forestry,References^ 110University of British Columbia.Chieng, S.T., 1990. Integrating GIS technology and CADD techniques for agriculturaldrainage planning, system design and drafting, Fourth International Drainage Workshop,ICID, pp. 115-123.Chow, V.T., D.R. Maidment, and L.W. Mays, 1988. Applied hydrology, HcGraw-Hill, NewYork.Clark, C.O., 1943. Storage and the unit hydrograph, Proceedings of the American Societyof Civil Engineers, Vol. 9, pp. 1333-1360.Cline, T.J., A. Molinas, and P.Y. Julien, 1989. An Auto-CAD based watershed informationsystem for the hydrologic model HEC-1, Water resources Bulletin, Vol. 25, No. 3, pp. 641-652.Connors, K.F., T.W. Gardne, and R.L. Day, 1989. Deriving stream parameters forhydrologic modeling from digital elevation data. New Directions for Surface WaterModeling, International Association of Hydrological science (IAHS), No. 181, pp. 398-406.Crawford, N.H., and P.K. Linsley, 1966. Digital simulation in hydrology, StanfordWatershed Model IV, Technical Report 39, Civil Engineering Department, StanfordUniversity, Standford, California.Dunne, T., W. Zhang, and B.F. Aubry, 1991. Effects of rainfall, vegetation, andmicrotopography on infiltration and runoff, Water Resources Research, Vol. 27, No. 9, pp.2271-2285.Dutton, G., 1978. Harvard papers on geographic information system, Harvard University.El-Kadi, A.I., 1989. Watershed models and their applicability to conjunctive usemanagement, Water Resources Bulletin, Vol. 25, No. 1, pp. 125-137.Engman, T.E., and A.S. Rogowski, 1974. A partial Area model for storm flow synthesis,Water Resources Research, Vol. 10, No. 3, pp. 464-472.Fairfield, J., and P. Leymarie, 1991. Drainage networks from grid digital elevation models,Water Resources Research, Vol. 27, No. 5, pp. 709-717.Flint, A.L., 1987. The effect of surrounding topography on receipt of solar radiation,Forest Hydrology and Watershed Management, Proceedings of the Vancouver Symposium,August 1987, pp. 339-347.Freeze, R.A., 1972. Role of subsurface flow in generating surface runoff II. Upstreamsource areas, Water Resource Research, vol. 8, no. 5, pp. 1272-1283.Freeze, R.A., 1974. Streamflow generation, Review of Geophysics and Space Physics, vol.References^ 11112, no. 4, pp. 627-647.Goodchild, M.F., 1985. Geographic information systems in undergraduate geography: Acontemporary dilemma. The Operational Geographer, No.8, pp. 34-38.Gosselin, C.R., 1985. Effect of storm size on peak streamflow changes following loggingof a small costal watershed in southern B.C., A Bachelor Thesis in the Department ofForestry, UBC.Gray, D.M., 1973. Handbook on the principles of hydrology, Water Information Centre,Ottawa.Gumbel, E.J., 1958, Statistics of extremes, Columbia University Press, New York.Hetherington, E.D., 1987. The importance of forests in the hydrological regime, CanadianAquatic Resources, Canadian Bulletin of Fisheries and Aquatic Science 215, pp. 179-211.Hewlett, J.D., and J.D. HeIvey, 1970. Effects of forest clear-felling on the stormhydrograph, Water Resources Research, Vol. 6, No. 3, pp. 768-782.Hewlett, J.D., 1974. Comments on letters relating to 'Role of subsurface flow ingenerating surface runoff' by R.A. Freeze, Water Resources Research, vol. 10, no. 4, pp.605-608.Hewlett, J.D., and C.A. Troedle, 1975. Non-point and diffused water sources: a variablesource area problem, Proceedings of the symposium watershed management, ASCE, pp.21-46.Hibbert, A.R., and C.A. Troendle, 1988. Streamflow generation by variable source area,Forest Hydrology and Ecology at Coweeta, pp. 111-127Higgins, D.A., S.B. Maloney, A.R. Tiedemann, and T.M. Quigley, 1989. Storm runoffcharacteristics of Grazed watersheds in eastern Oregon, Water Resources Bulletin, Vol.25, No. 1, pp. 87-100.Hodge, W., 1988. Linking the ARMSED watershed process model with the grassgeographic information system, Proceedings ASAE International Symposium on ModelingAgricultural, Forest, and Rangeland Hydrology, Hyatt Regency Hotel, Chicago, IL, Dec. 12-13, pp. 501-510.Hoggan, D.H., 1989. Computer-assisted floodplain hydrology and hydraulics, McGraw-HillPublishing Company, New York.Horton, R.E., 1933. The role of infiltration in the hydrologic circle, American GeophysicsUnion Transactions, Vol. 14, pp. 446-460.References^ 112Horton, R.E., 1945. Erosional development of streams and their drainage basins:hydrological approach to quantitative morphology, Bulletin of Geology Society ofAmerican, pp. 275-370.Hursh, C.R., 1936. Reports, hydrology, American Geophysics Union Transactions, Vol. 15,pp. 301-302.Huxhold, W.E., 1991. An introduction to urban geographic information systems, OxfordUniversity Press, Oxford.Hydrologic Engineering Centre, 1981. HEC-1 flood hydrograph package: User's Manualand Programmer's Manual, updated 1985, U.S. Army Corps of Engineers, Davis,California.Jett, S.C., Weeks, A.D., and Grayman, W.M., 1979. Geographic information systems inhydrologic modeling, Hydrologic Transport Modeling Symposium, 1979, The AmericanSociety of Agricultural Engineering Conference Proceedings, pp. 127-137.Lanfear, K.J., 1989. Geographic information systems and water-resources applications,Water Resources Bulletin, Vol. 25, No. 3, p. v-vi.Linsley, R.K., M.A. Kohler, and J.P.L. Paulhus, 1982. Hydrology for engineers, McGraw-Hill, New York.Loukas, A., 1991. Analysis of the response and the rainfall distribution in a mountainouswatershed, M.A.S. Thesis, Faculty of Civil Engineering, University of British Columbia.Luo, J., and S.T. Chieng, 1992. GIS and watershed configuration, Proceedings ofCPGIS'92, Buffalo, New York.Luo, J., S.T. Chieng, and J. Keng, 1991. Application of GIS to design flow determination,Canadian Society of Agricultural Engineering at the Agricultural Institute of Canada AnnualConference July 29-31, Fredericton, New Brunswick, Paper no. 91-122.MacLaren, J.F., 1975. Review of canadian design practice and comparison of urbanhydrologic models. Canada-Ontario Agreement Research Report, No. 26, Ontario Ministryof Environment, Toronto, Ontario.Mark, D.M., 1984. Automated detection of drainage networks from digital elevationmodels, Cartographica, no. 21, pp. 168-178.Marks, D., J. Dozier, and J. Frew, 1984. Automated basin delineation from digitalelevation data, Geoprocessing, no. 2, pp. 299-311.McCuen, R.H., and W.M. Snyder, 1986. Hydrologic Modeling, Prentice-Hall, EnglewoodCliffs, New Jersey.References^ 113McFarlane, B., 1990. Nitinat river flood frequency study, A Bachelor Thesis in theDepartment of Bio-Resource Engineering, The University of British Columbia.McMillan, T., 1990. Water resource planning in Canada, Journal of Soil and WaterConservation, November-December 1990.Moore, I.D., J.C. Panuska, R.B. Grayson, and K.P. Srivastava, 1988. Application of digitaltopographic modeling in hydrology, Proceedings ASAE International Symposium onModeling Agricultural, Forest, and Rangeland Hydrology, Hyatt Regency Hotel, Chicago,IL, Dec. 12-13, pp. 447-461.Muzik, I. and S.J. Pomeroy, 1990. A geographic information system for prediction ofdesign flood hydrographs, Can. Journal of Civil Engineering, Vol. 17, p. 965-973.Mosley, M.P., 1979. Streamflow generation in a forested watershed, New Zealand, WaterResources Research, vol. 15, no. 4, pp. 795-806.Mosley, M.P., 1982. Subsurface flow velocities through selected forest soils south island,New Zealand, Journal of Hydrology, vol. 55, pp. 65-92.Oka, T., 1990. Rainfall infiltration and macropores in a hillside slope, Bulletin of DisasterPrevention Research Institute, Kyoto University, vol. 40, no. 344, pp. 1-13.O'Callaghan, J.F., and D.M. Mark, 1984. The extraction of drainage networks from digitalelevation data, Computer Vision Graphics Image Processing, Vol. 28, pp.323-344.Puecker, T.K., and D.H. Douglas, 1975. Detection of surface-specific points by localparallel processing of discrete terrain elevation data, Computer Graphics Image Processing,no. 4, pp. 375-387.Reed, C.N., 1990. Using CAD data in a GIS, GIS'90 Symposium, Vancouver, Canada.Robinson, V.B., A.U. Frank, and H.A. Karimi, 1987. Expert systems for geographicinformation systems in resource management, AL Applications, Vol. 1, No. 1, pp. 47-57.Rodriguez-lturbe, I and J.B. Valdes, 1979. The Geomorphologic structure of hydrologicresponse, Water Resources Research, vol. 15, no. 6, pp. 1409-1420.Rodriguez-lturbe, I., G. Devoto, and J.B. Valdes, 1979. Discharge response analysis andhydrologic similarity: the interrelation between the geomorphologic IUH and the stormcharacteristics, Water Resources Research, vol. 15, no. 6, pp. 1435-1443.RTAC, 1982. Drainage manual, Vol. 1, Roads and Transportation Association of Canada,Ottawa.Sasowsky, K.C., and T.W. Gardner, 1991. Watershed configuration and geographicReferences^ 114information system parameterization for SPUR model hydrologic simulations, WaterResources Bulletin, vol. 27, no. 1, pp. 7-18.Schoolmaster, F.A., and P.G. Marr, 1992. Geographic information systems as a tool inwater use data management, Water Resources Bulletin, Vol. 28, No. 2, pp. 331-336.See, R.B., D.L. Naftz, and C.L. Qualls, 1992. GIS-assisted regression analysis to identifysources of selenium in streams, Water Resources Bulletin, Vol. 28, No. 2, pp. 315-330.Shaw, E.M., 1984. Hydrology in practice, Van Nostrand Reinhold, Berkshire, England.Shreve, R.L., 1966. Statistical law of stream numbers, Journal of Geology, Vol. 74, pp.17-37.Smith, A.Y., and R.J. Blackwell, 1987. Development of an information data base forwatershed monitoring, GIS for resources management: A compendium, American Societyfor Photogrammetry and Remote Sensing and American Congress on Surveying andMapping.Stuebe, M.M., and D.M. Johnston, 1990. Runoff volume estimation using GIS techniques,Water Resources Bulletin, American Water Resources Association Vol. 26, No. 4, pp. 611-620.Surkan, A.J., 1974. Simulation of storm velocity effects on flow from distributed channelnetwork, Water Resources Research, Vol. 10, No. 6, pp. 1149-1161.Taylor, F., 1991. Geographic information system: the microcomputer and moderncartography, Pergamon Press, Oxford.Tomlinson, R.F., 1984. Geographic information system - a new frontier, Proceedings ofInternational Symposium on Spatial Data Handling, 20-24 August, Zurich, pp. 1-14.Valdes, J.B., Y. Fiallo, and I. Rodriguez-Iturbe, 1979. A rainfall-runoff analysis of thegeomorphologic IUH, Water Resources Research, vol. 15, no. 6, pp. 1421-1434.Van Der Gulik, T.W, S.T. Chieng, and F.J. Smith, 1986. B.C. agricultural drainage manual,B.C. Ministry of Agriculture and Food, Agricultural Engineering Branch.Viessman, W. T.E. Harbaugh, and J.W. Knapp, 1972. Introduction to hydrology, Intext,New York.Vieux, B.E., V.F. BraIts, and L.J. Segerlind, 1988. Finite element analysis of hydrologicresponse areas using geographic information systems, Proceedings ASAE InternationalSymposium on Modeling Agricultural, Forest, and Rangeland Hydrology, Hyatt RegencyHotel, Chicago, IL, Dec. 12-13, 1988, pp. 437-446.References^ 115Whipkey, R.Z., 1965. Subsurface stormflow from forested slopes, Bulletin of InternationalAssociation of Scientific Hydrology, vol. 10, no. 4, pp. 74-85.Wilson, E.M., 1984, Engineering Hydrology, MacMillan, London.Wolfe, M.L. and C.M.U.Neale, 1988. Input data development for a distributed parametershydrology (FESHM), Proceedings ASAE International Symposium on Modeling Agricultural,Forest, and Rangeland Hydrology, Hyatt Regency Hotel, Chicago, IL, Dec. 12-13, pp. 462-469.Worrall, L., 1990. Geographic information systems: development and applications,Belhaven Press, London.Zhang, H., C.T. Haan, and D.L. Nofziger, 1990. Hydrologic modeling with GIS: anoverview, Applied Engineering in Agriculture, Vol. 6, No. 4, pp. 453-458.