UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Applications of geographic information system data in the UBC Watershed Model Lee, Jeannie Mei Ling 1996

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-ubc_1996-0401.pdf [ 3.64MB ]
JSON: 831-1.0050351.json
JSON-LD: 831-1.0050351-ld.json
RDF/XML (Pretty): 831-1.0050351-rdf.xml
RDF/JSON: 831-1.0050351-rdf.json
Turtle: 831-1.0050351-turtle.txt
N-Triples: 831-1.0050351-rdf-ntriples.txt
Original Record: 831-1.0050351-source.json
Full Text

Full Text

APPLICATIONS OF GEOGRAPHIC INFORMATION SYSTEM DATA IN THE UBC WATERSHED MODEL by JEANNIE MEI LING L E E B. A. Sc., The University of British Columbia, 1994 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF APPLIED SCIENCE in THE FACULTY OF GRADUATE STUDIES Department of Civil Engineering We accept this ^ hesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA May 1996 ©Jeannie Mei Ling Lee, 1996 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department The University of British Columbia Vancouver, Canada DE-6 (2/88) 11 ABSTRACT The suitability of using geographic information system (GIS) data to describe a watershed for the U B C Watershed Model is investigated. A GIS combines the ability of a database management system to store, retrieve, and analyse information with the capacity to produce and manipulate graphical elements on a map. In hydrologic modeling, information from a GIS is commonly used to describe the physical characteristics of a watershed such as terrain, forest cover, and soil type. By using a GIS, possible sources of human error and subjectivity present when manually measuring attributes from maps or aerial photographs can be removed. 1 The Seymour Watershed located north of Vancouver, B.C. is used as an example in assessing the value of GIS data for the watershed model and to illustrate the complexities of calibrating the model. Terrain, ecological, and timber GIS databases for the Seymour Watershed are used to create a watershed description file (WAT) which is implemented in the model to produce synthetic watershed hydrographs. These hydrographs are then compared and calibrated against recorded historical streamflows, resulting in a calibrated Seymour Watershed model. This model is then used to forecast future streamflows when meteorological forecasts are also provided. Although GIS data is not perfect, it is valuable in describing a watershed as input for the U B C Watershed Model, which is then used to produce calculated hydrographs. The main Ill drawback of utilizing a GIS for the Seymour Watershed is the absence of adequate documentation of some GIS characteristics. Despite the removal of subjectivity and human error by using GIS data, inaccurate model streamflows remain due to errors in other non-GIS data such as improper reservoir elevation readings, missing historical streamflow or meteorological data, and inherent errors in the watershed modeling process. TABLE OF CONTENTS Abstract i i Table of Contents iv List of Tables vi List of Figures vii Acknowledgment ix 1.0 Introduction and Purpose 1 2.0 Literature Review 3 2.1 Perm state runoff model 3 2.2 Ward creek 6 2.3 FIPR hydrologic model 7 2.4 Summary of the past uses of GIS 9 3.0 Overview of the U B C Watershed Model 10 3.1 Required data 11 4.0 Geographical Information Systems 15 4.1 Raster or grid based system 17 4.2 Vector based system 18 4.3 Triangular integrated network 18 4.4 Comparison of raster based and vector based GIS methods 18 4.5 Input of GIS data 20 4.6 GIS applications in hydrology 21 5.0 Outline of Available GIS Data 24 5.1 Overview of the Seymour Watershed GIS data 24 5.2 Detailed description of Seymour Watershed GIS data 26 5.2a terrain database 27 5.2b ecological database 29 5.3 Creating the description file for the Seymour Watershed 30 Figures V . l through V.4 35 6.0 Applying the GIS Watershed File to the U B C Watershed Model :..39 6.1 Abstractions and other adjustments to the original recorded streamflow 41 6.1a storage flow 43 6.1b intake flow 43 6.2 Compilation of abstracted flows and combination with observed flows 43 Figures VI. 1 through VI. 8 47 7.0 Explanation of the Calibration Process 55 7.1 Stage 1 calibration 56 7.2 Stage 2 calibration 59 7.3 Stage 3 calibration 60 7.4 Optimization routine 60 7.5 Statistics option 61 8.0 Calibration and Discussion of the Seymour Watershed Model 64 8.1 Calibration of the Seymour Watershed model 64 8.2 Assess the meteorological accuracy of the Seymour Hatchery station 67 8.3 Investigate the accuracy of the determined form of precipitation 70 Figures V i l l i through Vni.12 73 9.0 Problems Associated with GIS Data, Abstraction Data, Historical Flow, and Meteorological Data 85 9.1 GIS data 85 9.2 Storage reservoir elevation data 87 9.3 AES and WSC data 88 10.0 Results and Conclusions 90 10.1 Results 90 10.2 Conclusions 94 References 99 Appendix 1 Statistics Report for Initial Seymour Watershed Model 101 Appendix 2 Statistics Report from Initial Seymour Watershed Model using Seymour Hatchery AES Station and Grouse Mountain AES Station..'. 104 Appendix 3 Sample Precipitation and Temperature Graphs for 1990 107 Appendix 4 Statistics Report from Seymour Watershed Model using Adjusted Precipitation Temperature 110 vi LIST OF TABLES Table V. 1 Summary of Initial Watershed Description File 33 Table VI. 1 Summary of Modified Watershed Description File 40 Table Vffl. 1 Statistical Report Summary for the Initial Calibrated Model 66 Table VIII.2 Statistical Report Summary for the Initial Calibrated Model Using Both Seymour Hatchery and Grouse Mountain AES Stations 68 Table VIII.3 Statistical Report Summary for the Final Calibrated Model Using Adjusted Precipitation Temperature 72 Table X . 1 Summary of Final Calibration Parameter Values 90 Table X.2 Final Description of the Seymour Watershed 92 vii LIST OF FIGURES Figure V . l initial uncalibrated annual model hydrograph 1989-1990 35 Figure V.2 initial uncalibrated annual model hydrograph 1990-1991 36 Figure V.3 initial uncalibrated annual model hydrograph 1991-1992 37 Figure V.4 initial uncalibrated annual model hydrograph 1992-1993 38 Figure VI. 1 modified uncalibrated annual model hydrograph 1989-1990 47 Figure VI.2 modified uncalibrated annual model hydrograph 1990-1991 48 Figure VI.3 modified uncalibrated annual model hydrograph 1991-1992 49 Figure VI.4 modified uncalibrated annual model hydrograph 1992-1993 50 Figure VI.5 modified uncalibrated annual model hydrograph with adjusted observed flows 1989-1990 51 Figure VI.6 modified uncalibrated annual model hydrograph with adjusted observed flows 1990-1991 52 Figure VI.7 modified uncalibrated annual model hydrograph with adjusted observed flows 1991-1992 53 Figure VI.8 modified uncalibrated annual model hydrograph with adjusted observed flows 1992-1993 54 Figure Vffl. 1 initial calibrated annual model hydrograph 1989-1990 73 Figure VTJI.2 initial calibrated annual model hydrograph 1990-1991 74 Figure Vffl.3 initial calibrated annual model hydrograph 1991-1992 75 Figure VTJI.4 initial calibrated annual model hydrograph 1992-1993 76 Figure Vffi.5 calibrated annual model hydrograph using both Seymour Hatchery and Grouse Mountain AES stations 1989-1990 77 Figure VtII.6 calibrated annual model hydrograph using both Seymour Hatchery and Grouse Mountain AES stations 1990-1991 78 viii Figure VIII. 7 calibrated annual model hydrograph using both Seymour Hatchery and Grouse Mountain AES stations 1991-1992 79 Figure VHI.8 calibrated annual model hydrograph using both Seymour Hatchery and Grouse Mountain AES stations 1992-1993 80 Figure VHI.9 final calibrated annual model hydrograph with adjusted precipitation temperature 1989-1990 81 Figure VIII. 10 final calibrated annual model hydrograph with adjusted precipitation temperature 1990-1991 82 Figure VIII. 11 final calibrated annual model hydrograph with adjusted precipitation temperature 1991-1992 83 Figure VIII. 12 final calibrated annual model hydrograph with adjusted precipitation temperature 1992-1993 84 ix ACKNOWLEDGEMENT I would like to thank Dr. M . C. Quick for his supervision and advice throughout this research. His invaluable assistance in the computing and research aspects of this project and his expertise with the U B C Watershed Model is much appreciated. I would also like to thank Lome Gilmour for his work in gathering the necessary GIS information. Finally, I would like to thank my parents and friends for their support and encouragement. 1 INTRODUCTION AND PURPOSE The purpose of this research is to investigate the suitability of Geographical Information System (GIS) data to describe a watershed whose hydrologic response will then be simulated by a hydrologic model. The U B C Watershed Model will be used to test the applications of GIS in watershed modeling. In addition to the investigation of the uses of GIS in hydrologic modeling, the complexities in calibrating the U B C Watershed Model will be illustrated and examined. GIS information is commonly used to describe the physical characteristics of a watershed such as terrain, soil, and ground cover properties. The accuracy of these properties is important because most computerized hydrologic models require details about the watershed to calibrate the model and to create flow forecasts. Improvements in estimating these parameters should increase the accuracy in forecasting watershed flows and may decrease the time needed to calibrate the hydrologic model. Since GIS information is digitally entered into a database, the measurements contained in a GIS are devoid of human subjectivity and error. Thus, it is believed that these measurements are more precise than those obtained by manually estimating characteristics from maps. 2 The UBC Watershed Model is developed to describe and forecast watershed behavior in mountainous areas. It requires historical streamflow and weather data plus an accurate description of the watershed including area, elevation, and forest cover properties. Presently, the values of these parameters are estimated by physically examining topographic maps and aerial photographs. The compatibility of GIS data with the U B C Watershed Model will be appraised using the Seymour River Watershed. This watershed is selected as a case study because sufficient GIS data is available from the Greater Vancouver Regional District watershed management office. The Seymour Watershed is about 126 km and is located north of Vancouver, B.C. This basin is used as a source of water for domestic consumption in the Greater Vancouver area. Historical meteorological and streamflow data are also readily available for this watershed area and river system. 3 II. LITERATURE REVIEW Information from a geographical information system has been previously applied to other hydrologic models. Generally, GIS information is used to estimate spatial data for runoff models and its use eliminates the subjectivity of estimating certain model input parameters. The following example cases illustrate the functions of GIS in hydrological analyses. 1. Penn State Runoff Model Geographical information system data is used in estimating input parameters for the Penn State Runoff Model (PSR model) (Shamsi, 1993). This model is used to simulate runoff hydrographs for various durations and frequencies. The Bull Run Watershed in Union County, Pennsylvania is a rural catchment covering 8.4 square miles and is used to illustrate the PSR model. Hydrographs generated by the PSR model are manipulated to create peak flow presentation and release rate tables which are utilized in the development of a watershed storm water management plan. A peak flow presentation table is the sum of individual flow contributions from all sub-basins draining to any given point. The release rate is defined as the ratio of 'sub-basin pre-development peak flow 4 contribution to watershed peak flow' to 'sub-basin peak flow.' The required physical input values of the PSR model include area, overland flow width, mean overland flow slope, percent imperviousness, and stream capacity and travel time. These input values are estimated using an assortment of basic GIS data layers. Information in a GIS is divided into separate layers each describing one specific characteristic. For example, elevation contours are stored as one data layer and slope segments are stored as another layer. To create a GIS database of elevation and slope, the layer of elevation contours is superimposed over the data layer for the slope segments. The final GIS data file contains an array of polygons covering the entire area, with each polygon representing a constant elevation and slope. A complete explanation of how a GIS functions is detailed in chapter IV of this report. The following vector coverages are created for the Bull Run Watershed: subbasins, streams, roads, soil types, and land use. Coverages of subbasins, streams, and roads are digitized from topographic maps. Soil types are established from survey maps and land use coverage is created from aerial photographs. A raster coverage of elevation is also generated. A vector coverage divides the data layer into an irregular pattern of polygons and a raster coverage uses a uniform grid of squares to describe as area. These GIS layers are then manipulated to establish sub-basin area, overland flow width, overland flow slope, stream length, stream slope, and centroids. Stream length, slope, and cross section dimensions are used to compute stream capacity and 5 travel time. Other remaining physical parameters, such as imperviousness and runoff curve numbers, are determined by subsequent processing of the basic layers. Data compilation is followed by model calibration. The GIS based physical input parameters are not intended to be altered during the calibration process. Rather, calibration is to be applied to the adjustment of the most difficult to define hydraulic parameters. Logically, the watershed parameters can be sub-divided into several groups. One group depends on physically measurable characteristics and are definable from GIS data. Other characteristics need to be calibrated from the hydrological response of the system. Since there is no continuous type rain gauge in the Bull Run Watershed, only three observed hydrographs are available. The nearest rain gauge is located approximately 13 miles south of the study area. The modeled peak flows compare favourably to available Federal Emergency Management Agency peak flow estimates based on the regional flood-frequency method developed by the US Army Corps of Engineers (1984). According to Shamsi (1993), experiences with other similar projects indicate that the GIS based models calibrate faster than non-GIS based models. This implies that GIS based parameters are more accurate than those computed from traditional manual measurement techniques. 6 2. Ward Creek The Ward Creek Watershed located in Baton Rouge, Louisiana is a highly developed urban basin (Cruise and Greene, 1995). The 12.12 km 2 area is mostly mixed residential with some industrial and commercial areas. Local engineers are interested in developing an interface that further incorporates the spatial analysis capabilities of a GIS within the hydrologic analysis of an urban watershed. The researchers intend to use a GIS to create a spatial database that most accurately represents the hydrologic characteristics of the watershed. No previous GIS work has been done on this area and none of the needed physical data is available in digital form. The data acquired from various sources are of different scales and resolutions and must be manually geo-coded. Topography, soils, land use, pervious and impervious areas, storm drain system, stream channel, and street network of the urban watershed are coded into separate layers of data. The relational functions within the GIS are used to create a table of attributes corresponding to each information layer. Conclusions drawn from this project indicate that the coordinate values defining the locations and boundaries of features in the GIS can be used for spatial analysis. The coordinate values are used to identify which hydrologic response areas contribute to the flow at a particular inlet, the dimensions of the response areas, the overland distance from these areas to the inlets, and the storm drain that connects the particular 7 inlet to the next downstream inlet. The investigation of the effects that spatial location has on the discharge show that when changes to individual lots are made, even though there is little or no impact on the flow at the outlet, there is a larger impact at the nearest inlet receiving the flow from the lot surface. These results are obtained without any calibration of the hydrologic model. 3. FIPR Hydrologic Model The Florida Institute of Phosphate Research (FIPR) funded the development of an integrated hydrologic model to more adequately represent mine reclamation hydrology ( Burdge, Ross, Tara, 1993). The impetus of this model is the need for more accurate, reliable, and standardized quantitative hydrologic assessments of long term surface and groundwater effects associated with large scale mining impacts. A typical application for the FIPR Hydrologic Model (FH model) consists of five tasks: digital data gathering, GIS operations for model input, model input data processing, hydrologic simulation, and output post-processing such as statistical analysis and producing graphics. The F H model is ideally run on watersheds that are less than 10 km with sub-basins less than 1 km 2. It is considered a small scale application and requires fine GIS resolution namely, topographic resolution of 1 meter contour intervals. Soils are 8 broken down by hydrologic classifications and land use conditions are resolved into a simplified classification scheme. The GIS is contained within the model to perform the spatial data referencing and analysis function for generating model input. Hydrologic codes perform the calculations for time-dependant hydrologic simulation. There are four principal functions of the GIS in the F H model: perform the complex map overlaying to develop input data for the hydrologic models; provide the linkage mechanism between models with different spatial representations; provide the conversion of maps into common projections and scales; and provide limited post-simulation graphical output display. The significant points of the FEPR Hydrologic model are numerous. It is claimed that the GIS implementation of the FH model provides a means to standardize model parameter selection by omitting any user subjectivity. This model is also equipped with a complete user interface within the GIS to integrate the input data for modeling. Finally, the GIS application is able to display, interpret, and perform further analysis on the extensive output data generated by the model. 9 4. Summary of the Past Uses of GIS Of the three models reviewed, GIS is used most extensively in the Florida Institute Hydrologic Model where the geographic information system is integrated into the model. GIS is utilized to estimate input parameters, provide a link between models of different representations, convert various maps into common scales and projections, and create a graphical display. The Perm State Model uses GIS data to estimate input characteristics only. The model of Ward Creek uses GIS to represent the hydrologic characteristics of an urban watershed and is primarily used to forecast the effect of land development on storm drain flows. In all cases, GIS data is used to accurately describe the physical characteristics of a watershed. This research on the use of GIS in the U B C Watershed Model will be similar to the Perm State Runoff Model in that GIS data will be used to estimate input parameters for a hydrologic model. It is hoped that the use of GIS for the Seymour Watershed will conclude with encouraging results similar to those of the Perm State Model, namely model flows comparable to observed flows. 10 OVERVIEW OF THE UBC WATERSHED MODEL As described in the user manual, the U B C Watershed Model creates a computational representation of watershed behavior. This computer model calculates daily watershed outflow due to snowmelt and rainfall using maximum and minimum daily temperatures and precipitation data. In addition to streamflow values, the model also provides information on the accumulation and depletion of snowpack, soil moisture budget, soil and groundwater storage values, contributions to runoff from various portions of the watershed, and surface and sub-surface components of runoff. Given continuous meteorological input data, the U B C Watershed Model can operate continuously, accumulating and depleting the snowpack and producing estimates of streamflow. For calibration and verification purposes, the model uses historical meteorological and streamflow records as reference data and calculates performance statistics on total flow and hydrograph shape reconstitution. The watershed model is essentially designed for short term river flow forecasting. The accuracy of the forecast is dependant on how representative the meteorological data is to date, the accuracy of the meteorological forecast, and the current assessment of the snowpack, soil, and groundwater storage. Since the model operates continuously when given continuous meteorological input, it is possible to operate the model for longer term forecasting using projected future weather patterns. Once 11 an initial watershed status is specified, this long term forecasting capability assesses the range of possible outcomes for a whole season of snowmelt and rain runoff. Such seasonal forecasts are updated with recorded data as the season progresses thus gradually narrowing the range of possible outcomes. An extension of this mode of operation is used to estimate complete years of data when the streamflow records are non-existent but meteorological data is available. 1. Required Data The U B C Watershed Model requires a watershed to be represented as several consecutive horizontal bands of increasing elevation. Although up to twelve bands may be used, the operational manual suggests that four to eight bands are sufficient for most watersheds. The altitude intervals for each elevation band can be freely selected, though it is suggested to select bands which coincide with changes in the natural features of the watershed such as lake elevations, rock bluffs, or forest type. Each band is then distinguished by its physical characteristics: mean elevation, area, forested area, forest density, fraction of north/south orientation, glaciated area, glaciated orientation, and impermeable fraction of the soil. In addition, each band is assigned a meteorological station where temperature and precipitation data will be retrieved and used in calculations to estimate snowpack accumulation, snowmelt, evaporation, soil moisture status, and finally runoff which is calculated in terms of fast, medium, and slow contributions. 12 Mean band elevation is measured in meters and band area is measured in square kilometers. The forest properties of the watershed are depicted by the forested area, represented as a decimal fraction of the total band area, and the density of the forest. Forest density is a measure of the fraction of band area that is shaded by the forest. Relative north/south orientation of the band is measured on a scale between zero and one. Zero denotes a northerly orientation and one, a southerly orientation. For example, a value of 0.2 indicates a 20% southerly orientation. Glacier characteristics of the watershed are described by the glaciated area, measured in square kilometers, and the glacier orientation which is measured on the same scale as band orientation. Impermeablility is represented as the decimal fraction of the band that is impermeable. Also included in the description of the watershed are indexes indicating which meteorological station a band will use for temperature, precipitation, and evapotranspiration calculations. A fourth index adjusts the precipitation at each band, either increasing or decreasing the precipitation. In addition to the physical description of the watershed, historical data is needed. Streamflow and meteorological data of at least one full year are required for calibration and operation of the U B C model. The U B C Watershed Model uses meteorological information to generate synthetic streamflows which are compared with the historically observed flows to calibrate the hydrologic model. The accuracy and representativeness of the recorded data, both flow and meteorological, are crucial 13 to the performance of the model. The recorded observations should represent typical annual flow and meteorological trends and events to correctly simulate the response of a real watershed by the computer model. Generally, when more years of data are available, the calibration and forecasting performance of the model improves. Flow data must be measured in cubic meters per second and read either daily or hourly. Generally, daily data is preferred over hourly data. Meteorological data consists of a maximum and minimum temperature in degrees Celsius and precipitation in millimeters. These readings are also be recorded either hourly or daily. Another important stipulation of the watershed model is the recorded meteorological data be multiplied by a factor of 10. This includes both temperatures and precipitation measurements. For this research, daily stream flow data is obtained from the Water Survey of Canada (WSC) and meteorological data is obtained from Environment Canada which was previously named Atmospheric Environment Services (AES). In order for a meteorological or streamflow data file to be recognized as such by the U B C Watershed Model, it must be identified by a name and the extension ' .AES ' for meteorological data or ' . W S C for flow data. For each watershed, the U B C Watershed Model allows up to 5 AES stations but only one streamflow gauge station. The multiple meteorological stations are necessary for 14 large watersheds or watersheds which are particularly mountainous. Precipitation in mountainous watersheds can vary, sometimes greatly, between valley and summit areas. By using various AES stations, lower elevation bands can be represented by a different station than higher elevation bands. Each meteorological station must have a unique name and a measured elevation in meters. 15 GEOGRAPHICAL INFORMATION SYSTEMS A geographical information system is a computer system which links a database management system to a number of spatially distributed features that can be represented on a map. It combines the power of a database management system to store, retrieve, and analyse information with the ability to produce and manipulate graphical elements on a map. Geographical information system technology was first used in the 1960's to perform spatial operations. One of the earliest applications of this emerging technology in water resource engineering is reported by Solomon et al in 1968 who used the "square grid system" for computer estimation of precipitation, temperature, and runoff in Newfoundland (Muzik, Pomeroy, 1990). The use of GIS flourished in the late 1970's and in recent years has rapidly grown in number and complexity of applications. In the area of water resource planning and design, a major effort is focused on the development and utilization of microcomputer-based GIS systems. The modern GIS is an evolutionary result of other computer systems and is a merger between database and graphics software. As stated before, GIS programs are capable of producing maps and have developed this ability from computer-aided drafting (CAD) and thematic mapping. Computer-aided drafting is an automated drafting 16 program with the capability of dividing a drawing into a number of trait layers and printing only selected layers. The major limitation of C A D is the inability to interrelate distinct layers of information beyond a visual basis. In other words, an icon or characteristic on a C A D map cannot be spatially related to other icons or characteristics beyond a visual relationship on a map. In comparison, thematic mapping shades areas of a map based on the values of a single variable and is also unable to relate the values of different variables beyond a visual basis. Since thematic mapping is purely a graphic system, it can only be used as a display system to show results of a spreadsheet or database. A relational database management system allows the storage and retrieval of information from text records. Attributes of each record can be grouped into subsets to meet a specified criteria. Geographically referenced databases associate each record to a specific spatial location. The geographical database describes each object in terms of their position in a specified coordinate system, their associated attributes which are unrelated to position, and their spatial or topological interrelations which describe how the objects are linked together in the greater system. The final geographical information system product combines the analytic capabilities of a database management system with a high resolution computer graphics system. Generally, a GIS divides a map into distinct information layers where each layer is subdivided into an array of discrete units. Each unit contains a characteristic value or trait, which is unique from neighbouring units. Also, the location of each GIS unit 17 and its relation to other units are contained within the unit information. When creating a graphical output map of GIS data, attribute polygons are either coloured in coordination with characteristic values or values are marked on the polygons themselves. Over time, the GIS has developed three distinct forms to represent a database: raster or grid-based; vector or contour-based line networks; and triangular integrated networks. 1. Raster or Grid Based System The raster method separates different geographical attributes into distinct layers and divides each layer into a regularly sized grid pattern. For instance, one grid layer contains soil type, another grid layer contains vegetation type, and so on. Each enclosed square is described by the characteristics of its centre coordinates. The location and interrelation of each grid square unit is indicated by the position of the grid unit in the overall grid sequence. If the surface is thought of as a visual image with the dots having various colours and intensities similar to a computer monitor screen, the use of the term 'raster image' is revealed. The size of the grid pattern may vary for each attribute layer depending on the detail required to adequately represent the characteristic. 18 2. Vector Based System Instead of a grid, the vector based format uses a topological data structure of points, lines, and polygons defined by x, y coordinates to describe a map area. A topological data structure defines the elements of a map such that the spatial relationship between line segments, points, and areas are known. If a point on a map is randomly selected, it is possible to determine i f that point is within a particular polygon and i f that polygon is adjacent to a particular line segment. As with the raster method, distinct attributes are separated into different map layers. 3. Triangular Integrated Network A triangular integrated network (TIN) is a subset of a more general polygonal description of attribute regions. It is mainly used for topography where a network of irregularly spaced points are connected by lines to produce a triangular patchwork to indicate topographical traits. Each triangle is treated as a planar facet indicating significant peaks and valleys. Map attributes are also separated into distinct layers. 4. Comparison of Raster Based and Vector Based GIS Methods The most commonly used forms of GIS are grid based and vector based formats. In the past when comparing grid and vector based systems, a grid based system was preferred at times because it is more efficient in data storage than a vector based 19 system (Zhang et al., 1990). A case in point: the digital representation of a soil map in a vector based file contains both soil series codes and the corresponding Cartesian coordinates. In contrast, a raster based coverage contains soil series codes only, with the Cartesian coordinates implicitly expressed by the order of the grid pixels. The larger the coverage, the more storage space a vector based GIS requires compared to a raster based GIS. Most older raster based geographic information systems fail to adequately include multiple attributes capability so only one type of thematic attribute can be stored at a time. The same data layer is stored as many times as the number of modeling parameters thus reducing the advantage in storage efficiency. Modern GIS systems no longer suffer these problems and inefficiencies. The vector method is more complicated but generally gives higher resolution than the raster method. In the raster format, the resolution is a function of the size of the grid cell and the rate of change of the data. Usually, the smaller the grid cell, the higher the resolution but at the cost of increased data to be handled. To illustrate the difference between raster and vector data storage systems, an elevation contour map is examined. Elevation is a spatial parameter with a continuous function. Contours define points of uniform level or form discrete areas of uniform elevation. The boundaries and contours are usually complex and curvilinear in form. The vector approach stores the boundaries either as sets of coordinates or as polygons. The alternative raster method applies an orthogonal grid 20 to the elevation data and each grid cell is attributed with an elevation value. This value is assumed to be constant throughout the grid cell. The vector method is more complicated, but generally gives higher resolution. In contrast, the raster method is simpler and often computationally faster. 5. Input of GIS Data The method of entering data into the GIS varies and affects the cost of database creation and sometimes the data structure. In the past for example, the semi-automatic digitizer usually imposed a vector or chain structure, while scan digitizers imposed a grid structure (Clarke, 1986). Nowadays, systems are increasingly able to handle data from both vector and grid data structures by automatic raster-to-vector or vector-to-raster conversion. Other forms of data input include individual grid cell encoding and the use of remote sensing imagery. Stored data can be map and non-map data. Non-map data consists of characteristics such as soil type, land cover and use, etc. While the underlying assumption of any GIS application is that the database of physical information is available, the acquisition and compilation of this information is not a trivial task. Often, appropriate data is only available in map form, so that even with modern digitizing hardware the process is highly labour intensive. The eventual payoff comes from the multiple ways in which the data can be used once it 21 is made digitally accessible in a GIS. Consequently, a geographical information system is only cost efficient and effective i f there are demands for several different analyses on the same database. 6. GIS Applications in Hydrology Since so much of hydrology is linked to processes at the surface of the earth, the connection to the topographic, computer-based methodology of a geographic information system is a predictable step in the evolution of hydrologic engineering. Hydrologic applications of GIS systems range from synthesis and characterization of hydrologic tendencies to the prediction of response to hydrologic events. Several previous attempts have been made to incorporate GIS in hydrologic modeling. These attempts can be summarized in four categories (Cruise and Greene): 1. calculation of input parameters for existing hydrologic models 2. mapping and display of hydrologic variables 3. watershed surface representation 4. identification of hydrologic response units Presently, the majority of GIS applications in hydrologic analyses fall into the first two categories. 22 To exemplify of the amount of GIS data required to run a hydrologic model, the integrated hydrologic model sponsored by the Florida Institute of Phosphate Research is used (Ross, Tara). This model is developed to aid in the design of phosphate mine reclamation. It is a combination of a commercial geographical information system, an evapotranspiration code, and surface and ground water models. For the spatial data requirements, eight map layers are obtained or created in digital form for the GIS. The necessary layers are: 1. land use and vegetative cover 2. soil type 3. surface topography 4. hydrological routing elements such as streams and rivers 5. watershed and sub-basin delineation 6. water table contours 7. potentiometric surface contours 8. ground water confining layer elevations In a typical application of this hydrologic model, individual map layers are converted to a common scale and projection before any analysis. Then the first task of the GIS is to determine average conditions over each sub-basin. Next, attributes or characteristics tagged to a map feature on the land form maps are compared. The results of these comparisons are numerical values such as physical or empirical infiltration coefficients for the surface water model. Attributes are then averaged 23 over the smallest grid or parcel description in the model and the results are written to a data file to be read prior to hydrologic simulation. 24 OUTLINE OF AVAILABLE GIS DATA The Greater Vancouver Regional District (GVRD) is in the process of developing an extensive geographical information system for its three local watersheds: Seymour, Capilano, and Coquitlam. A l l the water for Vancouver and its surrounding suburbs is withdrawn from these three river systems. At present, the Seymour Watershed is the only G V R D watershed that has a complete GIS database which includes terrain and ecological information. Therefore, the Seymour Watershed is selected as the research area. In March 1993, a team of engineering consultants headed by Acres International Ltd. published an ecological inventory pilot study on the Jamieson-Orchid-Elbow Drainage area, located in the northern portion of the Seymour Watershed. The pilot study covers a wide range of aspects: climate, forest hydrology, sediment transport, terrain and terrain stability, biogeoclimatic classification, forest fire hazards, forest health, fisheries, and wildlife. From this pilot study and discussions with the G V R D mapping technologist, the required data from the Seymour Watershed is identified and extracted from the plethora of available GIS information. 1. Overview of the Seymour Watershed GIS Data The requested GIS data constitutes snow course or accumulation data, terrain maps, a 25 drainage network map, and biogeoclimatic data. Terrain maps consist of many layers of which the selected required layers are surficial materials, texture, material thickness, and slope angle characteristics. Biogeoclimatic classification at the site-series level is based on vegetation, soil, and other physiographic data. This group of data is divided into two sub-sets: site type and forest ecosystem. Information considered as site type are land form, elevation, and average slope and aspect. Forest ecosystem data include soil depth, water table depths, and soil moisture. Other useful biogeoclimatic information are percent coverage of all trees and understorey vegetation, stand composition, successional stage, and disturbance regime. The supplied GIS data is collected by using the ARCMNFO Geographical Information System. The necessary information is simply selected from the large GIS database via a Windows interface. To combine the data onto a few large spreadsheets, a selection of map layers has to be projected and scaled to a uniform GIS polygon size. For example, in a particular area the forest density polygon area is 10 but the soil is classified as two 5 m 2 polygon areas. The final spreadsheet to describe this area constitutes two polygons of 5 m each. The final collection of data is stored in three Excel spreadsheets however, some of the originally requested information is not provided. The absent data is not supplied due to the inordinate amount of time and effort needed to assemble this data which is not specifically required for the operation of the U B C Watershed Model. A l l the 26 essential information for the hydrological model is contained in a set of vast spreadsheets: one for terrain, one for ecosystems, and one for timber. The terrain and ecosystem databases describe the ground surface and the ecological properties^ respectively. The timber database is an appendage of the ecosystem spreadsheet and details the specifics of the forest properties. 2. Detailed Description of Seymour Watershed GIS Data The terrain and ecosystem databases are quite similar to each other. Both data spreadsheets are at least 20 columns long with thousands of entries. The first three columns of the terrain and ecosystem spreadsheets identify the GIS polygon, and indicate area in square meters and perimeter in meters. The polygon identification value is unique throughout each spreadsheet however, identification numbers are not compatible between any of the three GIS data sheets. Both spreadsheets also indicate the elevation span of each polygon. The terrain database gives the mean elevation and the elevation range of the polygon. Comparatively, the ecosystem database gives the mean, minimum, and maximum elevations of the polygon. In addition, the terrain spreadsheet indicates slope for each polygon; the ecosystem spreadsheet denotes the maximum, minimum, and mean slopes of each polygon 27 a. terrain database The terrain database describes the terrain, geological activities, aspect, and sediment transport in the watershed. Terrain is depicted by using terrain unit symbols to indicate which materials are present in a polygonal area and which materials are more abundant than others. Stratigraphic units are also utilized to show where surficial materials overlie a different material layer. Texture is included in this database and indicates the coarseness of the material and ranges from clay and silt, 62.5 um and less, to diamicton, a mass of pebbles and larger clasts in a matrix of fines. Materials and surface expression are divided into three sections to denote the dominant, secondary, and tertiary layers. Not all polygons require three layers of description but when more than one material is present, a blank, 7', or 7/' is used to separate each layer. A blank indicates an equal amount of dominant and secondary materials. A 7' indicates that the dominant is more extensive than the secondary and 7/' means that the dominant is much more extensive than the secondary. These dividers are also present between secondary and tertiary materials. Materials are distinguished into six categories: colluvium, fluvial sediments, 'active' fluvial sediments, till, organic sediments, and bedrock. Colluvium are products of gravitational slope movements including talus, landslide and debris flow deposits. Fluvial sediments are sands and gravels transported and deposited by streams and 28 rivers. Active fluvial sediments are in active deposition zones. Till is material deposited by glaciers without modification by flowing water and typically consists of pebbles, cobbles, and boulders in a matrix of silty sand. Organic sediments are materials resulting from the accumulation of decaying vegetative matter. Materials classed as bedrock are either outcrops of rock or bedrock within a few centimeters of the surface. Surface expression is specified in 14 classes: moderate slope, blanket, cone, fan, hummocky, gentle slope, moderately steep slope, rolling topography, plain, ridges, steep slope, undulating topography, veneer, and mantle of variable thickness. These categories are too numerous to be described here but, in general, each class pertains to the range of slope, amount of hillocks and hollows, and the overall shape of an area. Geological processes are expressed as the occurrence of avalanche, failing, rapid mass movement, and gullying. Failing includes sliding or slumping and is defined as a slope experiencing slow mass movement. Rapid mass movement is exemplified by debris flows, debris slides, avalanches, and rockfalls. Gullying and avalanches are self-explanatory. Information on slide count and slide density failure per hectare are also provided. Slide count is the number of slide initiation points in each GIS polygon and slide density failure is the slide area expressed per hectare for each polygon. 29 Aspect is given for each GIS area in azimuth, clockwise from north. Sediment transport is indicated as the volume of fine sediment in cubic meters per hectare per year. A category for drain class is also given which relates to the permeability of the soil. b. ecological database The ecological spreadsheet characterizes the vegetative cover and assigns a timber identification number for each polygon area. Identification numbers in the ecological database do not coincide with the timber identification numbers. Area and perimeter are also supplied for each polygon. Site series, site type, variants, and successional stage are given to describe the type of vegetation, soil, and climate characteristics present. A site series is a group of sites with uniform soil moisture and nutrients. Site type is a division of a site series that is uniform in soil or landform characteristics and variants describe the usual local climate. Successional stage is the average age of forest growth and is depicted in seven categories: initial, shrub herb, pole sapling, young forest, mature forest, old growth, and no vegetation. Initial is growth two years and younger. Shrub herb is between 2 to 15 years; pole saplings are between 15 to 40 years. Young forest is 40 to 100 years, mature forest is 100 to 250 years, and old growth is over 250 years old. No vegetation is obviously no growth. Site disturbance is also provided which indicates fire, landslide, harvesting, etc. 30 Each ecological GIS polygon is assigned a timber identification number used in the timber spreadsheet to describe the forest cover of the area. The timber database illustrates the type, age, and size of trees in great detail and is an extremely large spreadsheet. The important data columns from this spreadsheet are crown closure, crown closure class, and density. Crown closure is the percent of ground area covered by the vertically projected tree crown area. The tree crown is commonly known as the limbed portion of the tree. Crown closure class is a measurement of the crown closure. Classes range from zero to ten and represent percent intervals from 0 to 100%. Class zero is from 0 to 5%, class 10 is from 95% to 100%, and the consecutive classes in between represent successive increments of 9%. Density is the quantitative measure of tree cover in an area expressed as the number of tree stems per hectare. 3. Creating the Description File for the Seymour Watershed To create the watershed description file (WAT) for the U B C Watershed Model, the first step is to sort the terrain spreadsheet data by elevation. After much consideration, it is decided that perimeter, geological processes, slide values, and sediment transport information are not needed to model the watershed. Furthermore, secondary and tertiary materials and their corresponding surface expressions are omitted because the U B C model does not benefit from such an in-depth soil 31 description. The terrain database is reduced to dominant material, surface expression, slope, aspect, elevation, area, and polygon identification numbers. Then, the information is arranged and summarized into 50 m intervals from 250 m to 1650 m. Characteristics of each elevation interval are a result of a weighted average of each polygon trait according to its area. The same procedure of data reduction and sorting is done to the ecosystem spreadsheet. The only practical information in this database is the successional stage, area, timber identification numbers, elevation, and slope. Forest characteristics are eventually added to the ecosystem spreadsheet in accordance to their timber identification values. These characteristics are then extracted from the timber spreadsheet. Next, the area of each 50 m elevation band is compared between the ecosystem and terrain databases. Most of the band areas compare well between the two spreadsheets, so another process of elevation band amalgamation is performed. These new elevation bands will result in a wider range of elevation for each interval. The selection of new elevation intervals is done by examing the successional stage and percent of total watershed area of consecutive 50 m elevation bands. A change in forest pattern is often a good indication where the final watershed bands should fall. Finally, the watershed bands are selected as roughly equal area bands combined with any obvious changes in ecosystem characteristics. The final elevation bands are 250 m to 450 m. 500 m to 750 m. 800 m to 900 m, 950 m to 1050 m, 1100 m to 1200 m, 32 and 1250 m to 1650 m, inclusive. Both terrain and ecosystem spreadsheets are divided into these elevation intervals. Discrepancies in the final elevation band areas between the two different databases are corrected by slicing GIS entries with very large areas into smaller portions of different elevations. In other words, the area of a large polygon is partitioned into smaller horizontal strips according to the elevation range of the polygon. The relationship between area and elevation in a polygon is assumed to be linear hence, the slope from maximum to minimum elevation is assumed constant. These segments are then added to other elevation bands that are low in comparative areas. With the final watershed bands defined, the characteristics from both terrain and ecosystem databases are averaged by weighted area. Successional stage and all identification numbers are deleted since they are no longer required. The completed WAT file consists of mean elevation, total area, forested fraction, shaded fraction, relative orientation, and drain class for each band. Eventually, drain class is discarded because no documentation was found to explain the class values and there is no apparent correlation between soil type and drain class. Percent of bedrock or rock replaces drain class as a measure of the impermeability in the elevation band. 33 The initial watershed description file (WAT) is as follows: band number band elevation band area forested fraction shaded fraction relative orientation fraction of rock 1 308 m 19.3 km 2 0.99 0.50 0.99 0.015 2 606 m 22.8 km 2 0.96 0.55 0.80 0.046 3 828 m 24.3 km 2 0.95 0.52 0.99 0.058 4 978 m 19.5 km 2 0.97 0.50 0.94 0.227 5 1127 m 20.3 km 2 0.94 0.41 0.95 0.306 6 1306 m 19.6 km 2 0.86 0.23 0.98 0.627 Table V . l In addition to GIS information, AES and streamflow data are necessary to operate the watershed model. The meteorological station data consists of maximum and minimum daily temperature, daily precipitation, and the station elevation and is easily attained from Environment Canada. Of the numerous stations in the Seymour Watershed area, the North Vancouver Seymour Hatchery station, station number 110N666, is considered the most suitable for this research area. The North Vancouver Grouse Mountain Resort, station number 1105658, is eventually used in this report as a supplemental AES station to examine the accuracy of the Seymour Watershed calibration. 34 Streamflow data is obtained from the Water Survey of Canada office for the Seymour River North Vancouver station. This station is located in Lynn Valley and measures the Seymour River a few kilometers north of the Burrard Inlet, where the river finally discharges. Figures V . l through V.4 are the synthetic hydrographs produced by the U B C Watershed Model using the initial Seymour Watershed description file. Each graph displays the calculated and observed streamflow for each hydrologic year, from 1989 to 1993. Although the model is yet to be calibrated, these calculated hydrographs are useful to illustrate the first step in the progression of refining the accuracy of the watershed model. 35 36 37 38 39 APPLYING THE GIS WATERSHED FILE TO THE UBC WATERSHED MODEL The official drainage basin for the Seymour River North Vancouver streamflow gauge station is 176 km 2. From the GIS analysis which measures only to the Seymour Dam, the Seymour Watershed is 126 km 2 . The 50 km 2 drainage area difference is the area from the dam downstream to the streamflow gauge station. In order to accurately recreate the observed streamflow with the synthetic model flow, the total area of the Seymour Watershed description (WAT) file must be increased. By examining topographical maps of the Seymour Watershed basin, most of the area variance is from elevations in the lower watershed bands. After much consideration and discussion, the 50 km 2 of additional area is added to the model in the following increments: 15 km 2 for band 1,13 km 2 for band 2, 10 km 2 for band 3, 7 km 2 for band 4, and 5 km 2 for band 5. The highest band, six, is not increased. The impermeable fractions of all six watershed bands are much lower than expected. Since these values are estimated by using the percent of bedrock at the surface, they are not necessarily representative of the actual impermeability. Areas consisting of materials other than rock can also be rather impervious. After performing an initial run of the watershed model, it is observed that insufficient flow is entering the fast 40 runoff component. The modeled hydrograph seems to be generally low compared to the observed graphs and too much flow is entering the groundwater component. To compensate for this problem, a blanket value of 0.2 is added to the impermeable fraction of each band. These alterations result in a modified WAT file: band number band elevation band area forested fraction shaded fraction relative orientation impermeable fraction 1 308 m 34.3 km 2 0.99 0.50 0.99 0.215 2 606 m 35.8 km 2 0.96 0.55 0.80 0.246 3 828 m 34.3 km 2 0.95 0.52 0.99 0.258 4 978 m 26.5 km 2 0.97 0.50 0.94 0.427 5 1127 m 25.3 km 2 0.94 0.41 0.95 0.506 6 1306 m 19.6 km 2 0.86 0.23 0.98 0.827 Table VT.1 Each watershed hydrograph is plotted from October 1 to September 30. Observed data is available from 1989 to 1993 so a total of four hydrological years are used for calibration. Found at the end of this chapter are figures VI. 1 through VI.4 which are the uncalibrated model hydrographs from the U B C Watershed Model implementing the newly modified WAT file. These graphs are very similar in appearance to the previous streamflow graphs which used the initial watershed description file. 41 Generally, the modified WAT file produces streamflows slightly larger than the flows from the initial WAT file. This is the desired result of increasing the drainage area and increasing the impermeability of each watershed description band. When compared to observed streamflow graphs, the calculated hydrographs generally follow the overall shape of the historical hydrographs. Significant peaks in the. observed flow coincide with peaks in the calculated flow. These similarities indicate that the preliminary watershed hydrographs are on the right track in modeling the Seymour Watershed. The most noticeable discrepancy between the model and the observed data is the amount of streamflow at any point in the hydrograph. In general, the calculated flow is substantially more than the observed flow. Some of this variance is caused by abstractions for domestic consumption and storage in the reservoir. In the following sections, calculations will be described which correct the historical flow data for these influences. 1. Abstractions and Other Adjustments to the Original Recorded Streamflow The Greater Vancouver Regional District regulates flow from the Seymour Lake reservoir by storing flow from the Seymour River and controlling discharges exiting the Seymour Dam, located at the south end of the lake. Flows are regularly removed from the river and used for domestic consumption or held as lake storage. The 42 observed streamflow from the Water Survey of Canada is measured downstream from this dam and is not a true measure of the natural river flow. The U B C Watershed Model uses meteorological data to calculate natural, unregulated flows which occur upstream from the dam. To compare the modeled flow with the observed flow, the amount of removed flow must be added to the original observed flow. Discharge files obtained from the G V R D are in either text or Excel spreadsheet format and cover the years 1989 to 1994. Each annual data file contains the date, reservoir elevation, weir elevation, head over crest, spillway flow, 24" hollow cone valve, 37" HB valve east and west, and intake flow. Elevations are measured in feet and flows are in either million gallons per day or cubic feet per second. The hollow cone and HB valves are used to release water into the river for fishery requirements. The information that is needed from the discharge files are date, reservoir elevation, and intake flow. The reservoir elevation is recorded manually on a daily basis but is not measured at a specific, recurrent time of the day. The measurement reading may be taken in the morning one day and in the late afternoon, the next day. Since there is no method to account for this randomness, there is no correction technique and it is hoped that it has little affect on the accuracy of the hydrologic model. 43 a. storage flow Flow in or out of storage is calculated by multiplying the change in reservoir elevation by the lake area. Since no elevation-storage curve of the reservoir is available, a static value of lake area is implemented. In reality, when the surface elevation of the lake changes, the area also changes. This incremental area change is a small percentage of the entire lake area and is assumed to be insignificant. Lake area is estimated from the GIS database of the Seymour Watershed by extracting polygons which are identified as lake segments. The total surface area of the reservoir is 3.8 km 2 or 41 million square feet. b. intake flow Intake flow is the water used for domestic consumption purposes such as drinking water. It is generally a uniform amount of water withdrawn from the reservoir and varies from 5 to 10 ft3/s. Both storage and intake flow must be converted into cubic meters per second in order to add these flows to the observed WSC flow. 2. Compilation of Abstracted Flows and Combination with Observed Flows Storage and intake flows are combined into one spreadsheet file to represent the total flow removed from the Seymour River. This flow file is then converted into a format compatible with the WSC file format. Two different computer programs are created 44 to perform this task and to combine the removed streamflow values with the observed values. Streamflows entered in a WSC file are arranged in 9 columns of data. The first column is an identification number for the specific row of data. It consists of 8 characters: three for the last three digits of the year, two for the month, one for the row number, and two for the total number of days in the month. For example, 989011031 represents the first row of data for the month of January, which has 31 days, in the year 1989. Flow values are displayed to one tenth of a unit. An important feature for the format of the WSC file in the U B C Watershed Model is that the end of the ID number must be 16 spaces from the left edge, the first data value in the row must run from space 17 to 24, and the following data values must be contained within 8 spaces starting from space 25 to 33. A l l data values are right justified within each input space. If the data file is not in this format, the watershed model will not read the historical streamflow values properly resulting in an erratic observed hydrograph. The first computer program, written in Visual Basic 3.0, is created to transform abstraction data from a text file into a WSC compatible format. Before this transformation program begins, the file of removed flows must be manually transformed from a spreadsheet format to two columns of continuous text: one for the day, from 1 to 365 or 366, and the other for the flow value. Also, a change in year 45 must be indicated as a data entry. In other words, data for 1989 would start with 1989 as the first row followed by the two columns of data beneath it. Only one year of data is manually converted from spreadsheet to text format at a time, then consecutive years are joined together as two long columns of text. A dummy data entry, such as 'xxxx', is required at the end of the completed text file to indicate the end of file. When these modifications are complete, the program is employed to convert the text file into a WSC format. The final file containing all flow abstractions is named 'removed.wsc.' The second computer program is used to add the removed flows to the original WSC file, 'seynvan.wsc' This program creates the file 'add.wsc' which is the corrected Seymour River flow. Before applying this addition program, the first line of 'seynvan.wsc' which is used as a gauge station descriptor or identification line must be deleted. This addition program is also written in Visual Basic 3.0. A new run of the watershed model produces hydrographs with flows that are more similar to the observed historical flows. Figures VI.5 through VI. 8 display the hydrographs produced by the watershed model and the hydrographs of the new observed streamflows which include the flows removed or added for storage or intake. The model flows are not altered from the previous graphs of figures VI. 1 to V I A 46 By taking into consideration abstracted river flows, the overall observed flow increases. However, there are also new irregularities which arise with the addition of removed flows. There are instances where the observed streamflow drops below zero which are due to sudden changes in reservoir elevation. These unexpected fluctuations in surface elevation are likely due to effects of wind on the lake surface or errors in reservoir level records which will be investigated in chapter 9. In addition, there are remaining occurrences where the model flow does not follow the observed graph. Thus far, all of these observations are from an uncalibrated watershed model. An uncalibrated model retains the default watershed parameters and compares the initial, unaltered model hydrographs to the observed hydrographs. The following chapter will detail the calibration process where the synthetic hydrographs are adjusted for total yearly or monthly volumes and graph shape in an attempt to resemble the historical hydrographs. 47 48 49 50 51 52 53 55 VII. EXPLANATION OF THE CALIBRATION PROCESS Once the description of the watershed is complete and the WSC file has been adjusted for storage and other abstractions, calibration of the model can commence. According to the U B C Watershed Model manual, the calibration of the watershed involves adjusting parameter values in a WAT file until the annual and monthly volumes and patterns of calculated runoff approximate historical records. The WAT file contains a large array of parameters but the adjustment of only a few of these parameters is necessary for the calibration of the model. A majority of these parameters have default values which have been pre-calibrated and do not require further alteration. Calibration parameters can be adjusted manually or they can be tuned using the optimization routine provided in the U B C Watershed Model. When using the optimization routine, it is best to alter a limited number of parameters at one time. Some variables are dependant on other variables, so calibrating many parameters at once is not effective. The general calibration procedure is iterative and consists of three steps: 1. adjust particular parameter values in the WAT file 2. run the U B C Watershed Model 56 3. evaluate the results of the model using the 'graphics' and 'statistics' options to determine the degree to which the estimated hydrograph agrees with recorded streamflow behavior This process is repeated until suitable parameter values are established. The full calibration procedure is done in three stages, each dealing with different groups of parameters but following the same iterative pattern of modification and evaluation. Stage one is concerned with the meteorological distribution parameters and is initially carried out in a very simplified manner to determine a working basis for further calibration. Stage two deals with the time distribution of runoff and stage three assesses gradients of behavior in the watershed. Stages two and three are refining processes which should not be attempted until the annual and monthly volumes of runoff, calculated in stage one, correspond closely to the historical streamflow. After completing stages two and three, the user may return to the first stage to achieve greater parameter refinement. Visual comparison of estimated and observed flows are useful especially when calibrating the model for runoff timing. 1. Stage 1 Calibration The parameters manipulated in stage one calibration are EOLMID, EOLHI, POGRADL, P O G R A D M , POGRADU, POSREP, and PORREP. The first five variables are found in the section 'distribution of meteorological variables' in the 57 WAT file and the last two variables are within the 'AES station elevations and parameters' section. POGRADL is the precipitation gradient for elevations below EOLMTD. POGPvADM is the precipitation gradient for elevations above EOLMTD but below EOLHI. POGRADU is the precipitation gradient for elevations above EOLHT. EOLMTD is often taken as the elevation of the middle of the barrier height and EOLHI is estimated as two-thirds the barrier height. Stage one calibration begins by setting EOLMTD and EOLHI to values greater than the maximum elevation of the watershed. The other calibration parameters initially remain at their default values and should not be changed until a preliminary run of the model has been performed. After running the U B C Watershed Model, the 'statistics' function is used to provide statistical analysis on how well the model hydrographs estimate the total flow, shape, and timing of the observed hydrographs. Next, examine the statistics file, comparing the total observed flow and the total estimated flow values. These values represent the monthly and annual runoff figures taken from the WSC file and the U B C Model, respectively. Their comparison indicates how accurate the precipitation gradients distribute precipitation over the watershed. When necessary, adjust annual and monthly runoff statistics by modifying the value of POGRADL. This parameter is increased if the calculated flow is too low and decreased i f the flow is too high. Re-run the watershed model with the new parameter values and compare annual runoff values again. This process is repeated until observed and calculated flows are similar. 58 As stage one calibration proceeds, evidence may indicate the necessary refinement of POSREP and PORREP. These two parameters are AES adjustment factors for snowfall and rainfall data, respectively. Normally, in watersheds where the available data for the basin is reliable and elevation is accurately represented, these parameters remain unchanged at zero. This indicates that the measured precipitation data is not changed in the watershed model. In some situations, there may be strong evidence that the measured precipitation should be increased or decreased to match the calculated watershed values. At a later stage in calibration it may become apparent from the hydrologic response, snow course information, or a combination of the two, that precipitation gradients are higher or lower at various elevations. To account for this, P O G R A D M can be activated by setting EOLMID to an elevation value below which POGRADL will govern and above which P O G R A D M will govern. Once annual and monthly volumes are close to the desired levels, the next step is to determine i f the runoff time distribution is similar between the results generated by the U B C model and those from historical data. The 'Graphics' option on the main menu provides a visual representation of the shape of the runoff. Using the display options, the user can observe the characteristics of the estimated and observed flows. 59 2. Stage 2 Calibration The important factors in stage two calibration are COIMPA, POPERC, PODZSH, POUGTK, and PODZTK. COIMPA is the fraction of impermeable area in a band and controls the amount of water entering the sub-surface. POPERC allocates how much of this volume of water entering the sub-surface can be stored in the groundwater, PODZSH divides the groundwater component into an upper and deep zone component, and POUGTK and PODZTK are the routing time constants for these two zones, respectively. The volumes of runoff and the shapes of the groundwater recession flows are then examined, particularly at the end of the summer for POUGTK and through the winter for the longer recession PODZTK. The procedure for stage two calibration starts with determining the routing time constant for the deep zone groundwater reservoir from recorded winter flows. Then, the required seasonal allocation to deep zone groundwater storage is determined. Next, determine the upper groundwater routing time constant and adjust the groundwater percolation to calibrate allocation to upper groundwater storage. The fraction of impermeable area in a band, the soil moisture deficit production, and impermeable area recessional parameter may also require adjustment. Any of these steps can be separately repeated to improve the calibration of the model. 60 3. Stage 3 Calibration Stage three of the calibration process deals with gradients of behavior in the watershed, identified by the fraction of impermeable area in a band. It is a fine tuning process that can only be carried out when stages one and two achieve a high level of model performance. 4. Optimization Routine Calibration of the Seymour Watershed is performed using the 'optimization method'. The optimization method is chosen over manual calibration of the watershed because it removes the tedium of having to vary parameters and then re-running the watershed model. It is helpful i f the user has some knowledge of the hydrological processes involved in watershed modeling. Through experience and judgment, the user can define reasonable value ranges for the parameters and let the computer do the work of finding the best values for them. The optimization module investigates three groups of parameters separately. The first group, precipitation distribution, adjusts precipitation gradients. This optimization routine is performed until the estimated volumes begin to converge with the observed volumes. The next group to be optimized is the water distribution group which distributes the rain and snowmelt to groundwater, interflow, and surface runoff 61 through the soil moisture budget. The third group of parameters, routing constants, adjusts the time constants for each component of flow, controlling the length of time taken to pass through the watershed. Once a value range is defined for each parameter, the optimization routine randomly selects values from the range for the set of tagged parameters in each group, executes the model with these values, and calculates the coefficient of efficiency, the coefficient of efficiency compensated for volume error, the coefficient of determination, and the estimated total flow for the period. This procedure can be run for a number of iterations, at the end of which the ten best efficiencies and their corresponding parameters are saved to a file. The WAT file is then updated with the most efficient parameter values. This updated WAT file is then used to optimize the next optimization group. When all the selected groups are processed, the resultant WAT file contains the parameter values that give the best efficiency for the watershed. The entire optimization process can be repeated to obtain better calibration efficiencies. 5. Statistics Option The U B C Watershed Model provides a 'statistics' option to calculate various statistical data for each month of a calibrated WAT file. Statistics include the mean observed flow averaged over the specified period, mean estimated flow averaged over 62 the period, total observed flow, total estimated flow, difference between the observed flow and the estimate flow, coefficient of efficiency, and coefficient of determination. The coefficient of efficiency, e!, relates how well the estimated hydrograph compares in shape and total flow to the historical hydrograph. It is calculated as: e! = l - ( £ ( Q ^ - Q ^ ) 2 £(Qobs /' " Qobs avg ) where Q o b s a v g = ZQobs n n = number of days for daily runs or hours for hourly runs Qobs / = observed flow on day or hour / Qest / = estimated flow on day or hour / An efficiency of 1.0 indicates a perfect fit between the observed and the calculated hydrographs. An efficiency less than 1.0 is a result of imperfect shape, total flow, or timing. Negative values for the coefficient of efficiency can also be attained. The calculated efficiency is more sensitive to large peaks than small underlying flows. The coefficient of determination, d!, is a factor which relates how well the shape of the estimated hydrograph corresponds to the shape of the observed hydrograph. It is independant of total flow differences between the two hydrographs however, timing 63 does affect the value of this statistic. For example, i f two hydrographs are identical except one has more volume than the other, the coefficient of determination is 1.0. If one of these hydrographs is shifted slightly to the left, the coefficient of determination decreases. The coefficient of determination is calculated as follows: d! = l - £ ( Q ^ - ( b * C U ; + a))2 savg . ^XQobs ;' " Qobs avg ) where a = (SQ o b s/-b * Z Q ^ ) n b = S f C U , * O^fJ) - S Q ^ * ZQnhw * (1/n) £ ( Q e s u ) 2 - ( l / n ) * Z ( Q e s t / ) 2 n = number of days for daily runs or hours for hourly runs Qobs / = observed flow on day or hour i Qesi / = estimated flow on day or hour / 64 VIII. CALIBRATION AND DISCUSSION OF THE SEYMOUR WATERSHED MODEL 1. Initial Calibration of the Seymour Watershed Model The calibration of the Seymour Watershed requires the modification of some parameter values. EOLMTD is changed to 850 m and EOLHI is 1000 m. The precipitation gradients POGRADL and P O G R A D M are both calibrated to 1, but POGRADU remains at zero. POSREP is adjusted to -0.67 and PORREP is adjusted to 0.11. The low precipitation gradient factors are consistent with research done by Loukas and Quick. They found that the precipitation of the Seymour Valley is similar to the precipitation found in the neighbouring mountainous areas. In the Seymour River Watershed, precipitation was found to increase with elevation up to a height of 260m. Beyond this elevation, the rainfall dramatically drops and then a further slight increase in rainfall, leveling off at the upper elevations. This process is apparent in the values of POGRADL, POGRADM, and POGRADL. 65 In the second stage of calibration, only POPERC and PODZSH are altered. POPERC is changed to 9 and PODZSH is changed to 0.7. The other stage two calibration parameters are unaltered and remain at their default values. A statistical analysis is performed on the Seymour WAT file for each year from 1989 to 1993 and for the four years overall. The entire statistical output is found in appendix 1 located at the end of this report. For each month starting from October 1989 to September 1993, the mean observed and estimated flows, total observed and estimated flows, difference between observed and estimated flows, coefficient of efficiency, and coefficient of determination are calculated and displayed. At the end of the statistical report, the above values are calculated for each year from 1989 to 1993 and for the full four year period. The coefficient of determination for 1989 to 1993 is 0.5314 and the coefficient of efficiency is 0.5245. The difference between observed annual flow and estimated flow is 6.3% of Q o b s . The statistical results for each one year interval are as shown in table VIII. 1. Figures VIII. 1 to VIII.4, found at the end of this chapter, are the final calibrated model hydrographs of the Seymour Watershed. These graphs are created using the modified watershed description file, the observed streamflows including abstractions, and the calibrated watershed parameters. 66 period interval coefficient of determination coefficient of efficiency observed flow [m3/sl model flow fm3/sl observed -estimated fm3/sl percentage flow difference 1989 -1990 0.6422 0.6176 6137 5269 868 14% Qobs 1990-1991 0.4706 0.4596 8334 7291 1043 12.5% Q o b s 1991 -1992 0.5929 0.5914 5959 6254 -295 -5% Qobs 1992-1993 0.5168 0.4927 4983 5006 -23 -0.5% Q o b s 1989-1993 0.5314 0.5245 25412 23820 1592 6.3% Q o b s Table Vm.1 The graphical output of the U B C Watershed Model indicates that the calculated hydrograph is not consistent in accurately predicting the observed hydrograph and there are times when the observed flow is negative. The negative historical flows are most likely due to inaccurate reservoir elevation values which can be altered to avoid negative flow values. However, these negative flows remain in the observed data because they are infrequent and do not drastically affect the shape of the hydrograph. The modeled flow is less reactive than the observed data. It appears that the late-winter, early-spring months from January to April have the most discrepancies. During these problematic months, the estimated flow is low and flat while the observed flow shows numerous peaks. These disparities suggest that the model is not correctly responding to the precipitation data which may be due to an inappropriately used meteorological station. In other words, the streamflow station may be located in 67 an area that does not correlate on a meteorological basis to the information provided by the environmental station. A supplemental station will be used to evaluate this possibility. 2. Assess the Meteorological Accuracy of the Seymour Hatchery Station The data from the North Vancouver Grouse Mountain Resort station is used to ensure that the Seymour Hatchery station is a good representative for the meteorological events of the Seymour Watershed. The hatchery station is located in a valley at an elevation of 210 m but the Grouse Mountain station is near the summit at 1128 m. Sometimes valley stations do not accurately exemplify the weather outside of the local area and are heavily affected by orthographic lifting of the clouds. Since the Grouse station is 1128 m in elevation, it is used to represent the highest bands in the watershed, 5 and 6. The other bands are unchanged and continued to use the Seymour Hatchery station. IOTSTA, IOPSTA, and IOESTA are the parameters that require changing to implement the secondary meteorological station. The new WAT file, containing the Grouse Mountain station, is calibrated using the optimization routine. With the use of two AES stations, there are two snow adjustment factors and two rain adjustment factors; a pair each for the Seymour Hatchery and the Grouse Mountain stations. The final calibration results in a similar 68 WAT file as before but with altered snow adjustment factors, rain adjustment factors, and POGRADL. For the hatchery station, the snow adjustment factor is -0.67 and for the Grouse Mountain station, the snow adjustment factor is now 0.90. The rain factor for the hatchery is 0.2 and the rain factor for Grouse Mountain is -0.80. POGRADL is calibrated to 2. Statistics are also run for this new WAT file for the years 1989 to 1993. The full statistical report is found in appendix 2. This report uses the same calculations and time periods as the report produced for the previous watershed model which used only the Seymour Hatchery AES station. For the four year period, the coefficient of determination is 0.5432, the coefficient of efficiency is 0.5423, and the observed flow exceeds the estimated flow by 2.4% of Q o b s . A statistical summary is shown in the following table. period interval coefficient of determination coefficient of efficiency observed flow [m3/sl model flow [m3/sl observed -estimated [m3/sl percentage flow difference 1989 -1990 0.6447 0.6348 6137 5304 833 13.5% Q o b s 1990-1991 , 0.4984 0.4920 8334 7440 894 11% Qobs 1991 -1992 0.5964 0.5932 5959 6340 -381 -6% Q o b s 1992-1993 0.5011 0.4811 4983 5687 -704 -14% Qobs 1989-1993 0.5432 0.5423 25412 24804 608 2.4% Q o b s TableVin.2 69 The statistical results from the WAT file using the Grouse Mountain station does not indicate a significant improvement from the original WAT file which used only the Seymour Hatchery station. In general, the coefficient of determination and the coefficient of efficiency improve when the mountain station is added but the annual flow errors increase. By using the supplemental station, the greatest improvement in the coefficient of efficiency is 0.324 occurring in 1990 to 1991 and the highest increase in flow difference is 13.5% Q o b s in 1992 to 1993. After comparing the statistics of the new WAT file with the original WAT file over the 1989 to 1993 period, the lack of general improvement is evident: overall coefficient of determination improves by 0.0118, coefficient of efficiency increases by 0.0178, and total flow error for the four year period decreases from 1591.8 m3/s to 608.3 m3/s. Figures VIII. 5 through VIII. 8 display the calculated hydrographs using both the Seymour Hatchery and Grouse Mountain meteorological stations. The observed hydrographs are unchanged from previous graphs and include abstracted flows. A visual inspection of the new and original hydrographs indicate that the Grouse Mountain station does not improve the lack of reaction during the January to April period. The new model hydrographs do not indicate the peaks in flow that are present in the observed hydrographs but still generally resembles the original model hydrograph. This implies that the Grouse :Mountain station and the Seymour Hatchery station are showing the same precipitation response. There is no advantage 70 in implementing a supplemental AES station in this model. The diversion of the model hydrograph from the observed hydrograph may be due to a slight temperature inversion which is investigated in the following section. 3. Investigate the Accuracy of the Determined Form of Precipitation A temperature inversion is when a pocket of cold air is trapped in a topographical depression by lighter, warm air. Since this model is using one meteorological station located in the Seymour valley, the recorded station temperature is possibly lower than the temperature in the rest of the watershed which is higher in elevation. An unidentified inversion causes the model to underestimate the rain precipitation and over-estimate the amount of snowfall. To investigate this potential problem, the rainfall and snowfall calculated by the model is plotted with the maximum and minimum temperatures for each year of available data. Appendix 3 contains a sample plot of these parameters for January through March, 1990 and a plot of the observed flow, calculated flow, rainfall, and snowfall for the same period. From these graphs, a temperature inversion in the valley is discovered. Rainfall events are present during January to April which the model does not incorporate as streamflow but the observed data records an increase in flow. The temperatures during these rain events are close to zero and much of the precipitation is being miscalculated as snow by the hydrologic model. To correct this situation, the 71 parameter which represents the value added to the temperature before determining the form of precipitation, POTASR, is increased. With the modification of the temperature parameter, the calibration constants of POSREP and PORREP are altered again. Calibration runs using different POTASR values found that the optimal amount of added temperature is 3 °C. A new run of the optimization routine results in a POSREP of -0.44 and a PORREP of 0.014. Notice that the absolute value of both the snow and rain AES adjustment factors have decreased from the previous calibrated WAT file. This indicates that the historical precipitation data is incorporated by the model with less prior alterations. This new WAT file produces synthetic hydrographs which resemble the observed graphs more than the previous Seymour WAT file which did not adjust the POTASR factor. There is a clear improvement in the problematic months of January through April. Figures VIII.9 through VIII. 12 are the annual hydrographs produced by this final, improved WAT file. The statistical results have also improved by using the new Seymour WAT file and are displayed in full in appendix 4. A statistical summary is shown in table VIII.3 on the next page. 72 period interval coefficient of determination coefficient of efficiency observed flow [m3/sl model flow [m3/sl observed -estimated [m3/s] percentage flow difference 1989-1990 0.6697 0.6665 6137 5857 280 4 . 6 % Q o b s 1990-1991 0.5408 0.5342 8334 8449 -115 -1.4% Qobs 1991 -1992 0.5830 0.5668 5959 6713 -754 -12.6% Q o b s 1992-1993 0.5428 0.5153 4983 5322 -339 - 6 . 8 % Q o b s 1989-1993 0.5754 0.5672 25412 26341 -929 - 3 . 6 % Q o b s Table Vffl.3 73 74 75 o O N O N l O N 0 0 O N C o &0 3 o on 3 O u. o c o •s u-3 O D 00 -»-» O 00 .g 3 s-oT) 3 ^ •a ° % T O T S i * "8 0 0 CO 1 g 78 80 81 o O N O N i — i O N 0 0 O N 1) e o 03 O T 3 T3 M d. ca o >--o T3 O s T3 <D +-» a o O N 82 O N os o os O S c3 ex o O T 3 O T3 l-H O c 3 84 85 PROBLEMS ASSOCIATED WITH GIS DATA, ABSTRACTION DATA, HISTORICAL FLOW, AND METEOROLOGICAL DATA Much of the data used in this project requires some adjustments and assumptions before it is entered into the watershed model. The GIS, storage flow, streamflow, and meteorological databases each have particular problems, such as missing or inaccurate data, which must be addressed. 1. GIS Data The geographic information system data has proved to be very useful in describing the Seymour Watershed but requires some modifications to the area of large GIS polygons and clarification in definitions of some terrain characteristics. In the GIS database, there are some polygons with excessively large areas which negatively affect the accuracy of the description of the watershed. The mean elevation of a vast area does not give an accurate description of the polygon since it is unlikely that the area lies laterally along the mean elevation with little above or below this level. It is more likely that a sizable but equal amount of area lies above the mean elevation as there is below it. These large polygons skew the area within an elevation band, hence 86 all large GIS polygons were divided into smaller polygon slices of varying mean elevations. These new subdivided areas are utilized to supplement elevation bands which do not have equal terrain and ecosystem areas. Each polygon is visualized as a large rectangle where the area grows linearly with the increase in elevation from minimum to maximum. This is a necessary assumption since there is no information to indicate how area relates to a change in elevation for each polygon. Another problem with the GIS database is the lack of available information on some watershed characteristics, namely drain class. Unlike other classifications like crown closure and successional stage, drain class does not have any accessible documentation to explain its definition or numerical values. The only information obtained about drain class is from the mapping technologist who believed it was related to soil permeability. A direct measurement of imperviousness would be very useful for the U B C Watershed Model. Since the drain class definition is unavailable, the percent of impervious area in a band is approximated by the percent of bedrock present in an elevation band. As stated before, this severely underestimates the amount of impervious area and an additional percentage is required. Without previous experience working with the U B C Watershed Model, it is difficult to estimate how much additional imperviousness is needed and identify which bands require supplementation. 87 2. Storage Reservoir Elevation Data Reservoir elevation measurements are required to calculate the flow that enters or exits lake storage. These daily measurements are read manually by a technician who records values at the Seymour Dam. Usually, the surface elevation rapidly increases due to rain events then gradually decreases as water is withdrawn from the reservoir during dry periods. Occasionally, the elevation significantly increases or decreases one day then, the next day, suddenly reverts back to levels similar to the days prior to the sudden change in elevation. In actuality, these elevation fluctuations are almost impossible because it indicates a large volume rapidly entering storage which exits the next day in the same abrupt manner. The result of these jumps is an erratic observed watershed flow. Regularly, the observed hydrograph will drop below zero or suddenly increase in a dramatic spike. The behavior of these unusual readings indicate an error in measurement or some behavior of the reservoir surface that has yet to be considered. Seiching or periodic oscillating of the water surface is a possible cause for these inaccuracies. These oscillations are due to a drag force exerted by the wind on the surface of the lake. The actual stress felt by the water surface is influenced by the wind strength, the stability of the meteorological boundary layer over the water surface, the variability of the wind speed over the lake, the length of fetch, the degree of wave development, and the amount of wave energy dissipation at the shores of the 88 lake. Of these influences, wind speed is the dominant factor since stress is related to the square of the wind velocity. This drag force pushes water towards the far-side of the lake resulting in a standing wave at the reservoir surface. When the wind stops, the surface oscillates until the water level is returned to a stable position. If elevation measurements are taken while the lake level is unstable, an inaccurate reading results. Some unexpected readings are due to these periodic surface oscillations, but either instrumental or human error is the most probable cause for most measurement irregularities. Since readings are done manually, it is conceivable that values are misread by the technician. The only way to eradicate this inconsistency is to automate the task of measurement taking which is considered too costly. In an attempt to correct these errors, lake elevations which are considered inconsistent with adjacent daily values are changed to represent an average of succeeding and preceding daily measurements. The elevation of Seymour Lake is plotted for all available years so that these significant variances can be identified and altered. Once this is complete, the observed flow is rarely negative. 3. AES and WSC Data The meteorological and streamflow data files contain one significant problem: missing data values. Sometimes weeks of precipitation, temperature, or flow data are 89 absent, indicated by a reading of '-9999'. If a substantial amount of data is missing from an annual file, the entire year of information is useless. But i f only a few sporadic days of data values are absent, they can be estimated and replaced. To replace missing data for maximum or minimum temperatures, an average temperature is taken from the available adjacent days of data. Usually, either the maximum or minimum temperature is absent per day so a sudden change in temperature is still indicated by the available values. Missing streamflow data is averaged in a similar manner as the temperature. Maximum and minimum flow values are not provided but it is assumed that daily streamflow is similar to its preceding and succeeding days. Absent precipitation data is substituted with a zero since an average of adjacent days is not an accurate replacement. Unlike temperature or streamflow, precipitation can widely vary each day and no supplemental information is available to indicate a change. 90 X. RESULTS AND CONCLUSIONS 1. Results The final calibrated parameters for the Seymour Watershed are: EOLMTD 850 m EOLHI 1000 m POGRADL F P O G R A D M F POGRADU 0 ^ snow adjustment factor -0.44 rainfall adjustment factor 0.014 impermeable fraction varies per band ground water percolation 9 mm deep zone share 0.7 upper groundwater runoff 30 days* deep zone share time 150 days* evapotranspiration 100 mm* impermeable area factor 100 mm1 temperature increase before determining form of precipitation 3.0 °C Table X.1 * indicate an unchanged default value Notice POGRADL and P O G R A D M both equal 1. This means that for elevations below 1000 m, EOLHI, the precipitation increases 1% for every 100 m increase in elevation. Above 1000 m the precipitation gradient is zero. The low gradient values are in agreement with research done by Loukas and Quick where it was found that 91 precipitation in the Seymour Valley is similar to the precipitation in the higher elevations of the adjacent mountains. The snow adjustment factor of -0.44 means that the recorded snowfall is decreased by 44 % to match the calculated watershed value. In contrast, the observed rain is increased by 1.4 %, the rainfall adjustment factor, to correspond with the estimated watershed value. If these two parameters remained unchanged at zero, the available data is considered reliable and representative for the basin. The rainfall adjustment factor of 0.014 is minimal enough to indicate that the historical data is accurate. The value of -0.44 for the snowfall factor indicates that the historical snowfall data is not perfect but still valid. The impermeable fraction for each band is not altered from the initial watershed description. The amount of groundwater percolation, indicating the maximum capacity of sub-surface storage, is 9 mm where any excess runoff enters interflow. Deep zone share value of 0.7 divides the groundwater into an upper component and deep zone component: 30 % upper zone and 70 % deep zone. The upper groundwater runoff time constant is 30 days and deep zone share time constant is 150 days. These variables indicate the volumes of runoff and the shapes of the groundwater recession flows at the end of the summer as upper groundwater and as deep zone share through the winter. 92 Evapotranspiration is measured as the actual evapotranspiration from the potential value. This is compared to how much moisture has satisfied the soil demands. The impermeable area modification factor is compared with how much moisture has satisfied the soil demands. These two parameters are used in an exponential decay function to reconstitute summer rainfall runoff. The final description of the Seymour Watershed is as follows: band 1 2 3 4 5 6 mid band elevation 308 m 606 m 828 m 978 m 1127 m 1306 m band area 34.3 km 2 35.8 km 2 34.3 km 2 26.5 km 2 25.3 km 2 19.6 km 2 forested fraction 0.99 0.96 0.95 0.97 0.94 0.86 density of canopy 0.50 0.55 0.52 0.50 0.41 0.23 orientation 0.99 0.80 0.99 0.94 0.95 0.98 glaciated area 0 0 0 0 0 0 glacier orientation 0 0 0 0 0 0 impermeable fraction 0.215 0.246 0.258 0.427 0.506 0.827 AES temp, station 1 1 1 1 1 1 AES ppt. station 1 1 1 1 1 1 ppt. adjustment 0 0 0 0 0 0 AES evap. station 1 1 r 1 1 1 AES station North Vancouver Seymour Hatchery WSC station North Vancouver Seymour Arm Table X.2 93 It is difficult to determine i f the GIS data improved the amount of time needed to calibrate the UBC Watershed Model since data from other sources were not used. However, it is reasonable to conclude that the GIS watershed information was very useful in describing some characteristics of the Seymour Watershed. Without it, all watershed characteristics must be measured manually. After examining the statistical analysis and resulting hydrograph of the Seymour Watershed WAT file, it is concluded that GIS data can adequately describe some of the physical characteristics of a watershed for the U B C Watershed Model. The calculated model flow follows the historical flow well. It is not a perfect correlation but the main characteristics of the measured hydrographs are replicated. Flow discrepancies from the observed data occurring between January and April of the initial calibrated watershed model are due to a temperature inversion present in the Seymour Valley. The meteorological station used by the model is located in the valley and occasionally records temperatures which are lower than those in the remaining higher elevations of the watershed. An attempt to correct this error with the use of a supplemental AES station was not successful. The inversion was finally amended by increasing the value added to the temperature before determining the form of precipitation, POTASR. 94 Despite the benefits of using GIS to describe the watershed, there are some inaccuracies and obstacles in utilizing the information. Where large GIS polygons are used to describe an area, an assumed linear relationship between area and elevation increase is adopted to divide the polygon into smaller slices which better represents the area. A more serious problem with using GIS data is the absence of adequate documentation for some of the characteristics and values. Without sufficient background explanation, some watershed characteristics are estimated from crude assumptions and measurements. It is also important to note that the use of GIS does not erase inaccuracies present in other data used in the watershed model such as meteorological data and reservoir elevation records. 2. Conclusions A Geographical Information System is very useful for the description of a watershed in the U B C Watershed Model. Without a GIS, watershed characteristics such as area and forest cover must be measured manually which is subjective and vulnerable to human error. There is a vast amount of information available in a GIS but the most practical databases for the watershed model are focused on terrain and ecology. In this investigation, GIS information is used to establish the elevation, area, forested fraction, shaded fraction, relative orientation, and the impermeable fraction of each descriptive watershed band in the Seymour Watershed. 95 Most of the GIS information requires some manipulation before it can be applied to the watershed model. Original data from the GIS is transferred to a spreadsheet and is then sorted into elevation bands. Band characteristics are summarized to accurately describe each elevation interval of the watershed. Since only the minimum, maximum, and mean elevations of each GIS polygon are available, very large polygons are divided into a number of smaller areas of varying elevations to accurately depict the watershed area. These smaller polygons are then distributed to the elevation band that it best represents. The U B C Watershed Model is able to simulate observed hydrographs in the Seymour Watershed with a coefficient of efficiency of 0.5672 and a coefficient of determination of 0.5754 for the period between 1989 to 1993. The discrepancy between the total observed flow and the total calculated flow is 929 m3/s or 3.6 % of the observed flow for the four year period. Upon visual comparison of the annual hydrographs, the calculated flow is quite similar to the observed flow. The main hydrograph peaks are present and the general shape and tendencies of the observed graph are reproduced. Although the coefficients for efficiency and determination are not extremely high, the similarities between the modeled hydrograph and the observed hydrograph lead to the conclusion that the data from a GIS can be used to correctly depict the physical characteristics of a watershed in the U B C Watershed Model. Normally, expected 96 values for these coefficients are greater than 0.75 for rain dominated watersheds and greater than 0.85 for snowmelt dominated basins. The discrepancies between estimated and observed flows and the low coefficients of efficiency and determination are not entirely due to errors within the GIS data. Inaccuracies and additional problems present in other data sources, such as impermeability values and storage flow, also affect the accuracy of the model. In general, the GIS data is plainly displayed and explanations of most characteristics and values are given or are available from noted sources. However, there is one trait with no available information which may be useful to the U B C Watershed Model, namely drain class. An explanation of drain class may provide direct information on the impermeability of an area. Instead, the impermeable fraction of the Seymour Watershed is derived from the fraction of rock in each elevation band. This greatly underestimates the actual amount of impermeable area and requires some correction. The amount of additional impermeability required by each band cannot be determined from calculations or measurements but is derived from the user's previous experience with the watershed model. A user with experience calibrating the UBC Watershed Model will be knowledgeable of the expected fraction of impermeability of the average watershed. Generally, the amount of impermeable area increases with elevation and is not less than 0.10 but can be up to 1.0, indicating complete impermeability. 97 It is important to ensure that all the water produced in the watershed is accounted for in the recorded streamflow data. The Seymour Watershed contains a lake which is used as a storage reservoir and a source of domestic water for the Vancouver area. These abstracted flows must be added to the historical streamflow to accurately represent the natural flow produced by the watershed. Storage flow is calculated from the change in reservoir elevation which is manually measured at a random time of the day. Occasional errors in these elevation recordings are apparent when calculating storage flow. Expected normal storage flow should dramatically increase over a few days of rain as the reservoir fills with water and then gradually decrease as flow is released from storage. At times, the calculated storage flow suddenly increases or decreases one day but is followed by flow readings similar to days prior to the sharp change. This flow pattern is most likely due to an erroneous reservoir elevation reading which can be affected by human error, the time of the day, and lake surface oscillations as previously explained. In order to calibrate the model against observed streamflow, the model watershed area must equal the actual observed area. The area of each band may be altered to accurately reproduce the real watershed outflow. In the case of the Seymour Watershed, the total area according to the GIS data is less than the area accounted by 98 the Water Survey of Canada. The Water Survey of Canada is the source for historical streamflow of the Seymour River and measures the entire watershed area contributing to the Seymour River North Vancouver flow gauge station. The flow gauge is located downstream of the Seymour Dam; however, the GIS data measures the watershed area down to the dam only. The difference in area due to the distance between the dam and the gauge station must be accounted for to produce an accurate model calibration. From a topographical map, this additional area downstream of the dam is identified and added to the appropriate elevation bands. Presently, GIS data cannot be used as a direct input to the U B C Watershed Model without prior manipulation. The next developmental step in the evolution of GIS and the U B C Watershed Model is to create a detailed computer interface which systematically corrects or adjusts the GIS information in preparation for the watershed model. Ultimately, GIS information is very valuable in describing most of the physical attributes of a watershed. A Geographical Information System can completely eliminate any need for manual measurements from topographical or aerial maps. Although using a GIS diminishes the subjectivity of measuring physical watershed characteristics, the accuracy of the watershed model is still affected by the accuracy of other measurements. 99 REFERENCES Acres International Limited et al. G V W D Watershed Ecological Inventory Pilot Study (Jamieson-Orchid-Elbow Drainage). Final Report. March 18, 1993. Bhaskar, Nagashwar R., Wesley P. James, and Ravikumar S. Devulapalli. "Hydrologic Parameter Estimation Using Geographic Information System." Journal of Water Resources Planning and Management. Vol.18, No.5, September/October, 1992: 492-511. Brooks, Norman H. , et al. Mixing in Inland and Coastal Waters. San Diego: Academic Press, Inc., 1979. Burdge, Jeffrey, Mark A. Ross, and Patrick D. Tara. "New Directions in Integrated Hydrologic Modeling with GIS." Engineering Hydrology. Proceedings of the Symposium, Hydraulics Division on the American Society of Civil Engineers, July 25-30, 1993: 563-568. Edited by Chin Y . Kuo. New York: American Society of Civil Engineers, 1993. Cruise, J. F. And R. G. Greene. "Urban Watershed Modeling Using Geographic Information System." Journal of Water Resources Planning and Management. Vol.121, No.4, July/August, 1995: 318-325. DeVantier, Bruce A. And Arlen D. Feldman. "Review of GIS Applications in Hydrologic Modeling." Journal of Water Resources Planning and Management. Vol.119, No.2, March/April, 1993: 246-259. French, Steven P., and Lyna L. Wiggins. GIS: Assessing Your Needs and Choosing a System. Chicago: The American Planning Advisory Service, 1991. Haan, C. T., D. L. Nofziger, and H. Zhang. "Hydrologic Modeling with GIS: an Overview." American Society of Agricultural Engineers. Vol.6(4) July 1990: 453-457. Hendricks, L. A. , et al. " A Geographical Information System (GIS) User Interface for Delineating Wellhead Protection Areas." Ground Water. Vol.31, No.3, May-June, 1993: 480-488. Loukas, Athanasios, and Michael C. Quick. "24-H Design Storm for Coastal British Columbia." Journal of Hydraulic Engineering. Vol.121, No. 12, December, 1995: 889-899. 100 Ministry of Forests. Province of British Columbia. Forest Inventory Manual. Forest Classification/Sampling and Environmentally Sensitive Areas. Vol.2. 1992. Muzik, I. And S. J. Pomeroy. " A Geographic Information System for Prediction of Design Flood Hydrographs." Canadian Journal of Civil Engineering. Vol.17 (1990): 965-973. Ross, Mark A. And Patrick D. Tara. "Integrated Hydrologic Modeling with Geographic Information Systems." Journal of Water Resources Planning and Management. Vol.119, No.2, March/April, 1993: 129-137. Shamsi, Uzair M . " A GIS Application to Hydrology." Engineering Hydrology. Proceedings of the Symposium, Hydraulics Division on the American Society of Civil Engineers, July 25-30, 1993: 371-375. Edited by Chin Y. Kuo. New York: American Society of Civil Engineers, 1993. University of British Columbia Mountain Hydrology Group. U B C Watershed Model Manual. Version 4.0. Vancouver: Mountain Hydrology Group, Civil Engineering, U B C , 1995. APPENDIX 1 Statistics Report for Initial Seymour Watershed Model 11018D15Ea3LslOhlOV |s4B U.B.C. |s4B WATERSHED MODEL I I |s4B STATISTICS REPORT sl2h8V a95CDate: 07-12-1996 Time: 14:56:40 STATISTICS FOR THE OCT 1 , 1989 - SEP 30 , 1993 WATER YEAR(S) Mean Qobs cms/d Mean Qest cms/d Tot Qobs cms Tot Qest cms Tot Qobs -Tot Qest Coeff.of Eff Coeff.of Det 1989 OCT 28.2 29.5 874.3 914.6 -40.3 0.6262 0.6272 NOV 31.4 36.9 941.4 1107.1 -165.7 0.8069 0.8279 DEC 15.5 20.3 481.5 627.9 -146.4 0.6708 0.6940 1990 JAN 19.1 11.1 591.2 344.3 246.9 0.4185 0.7113 FEB 14.6 3.4 409.5 95.2 314.3 -0.3839 0.3149 MAR 18.1 3.9 560.8 119.7 441.1 -1.7588 0.2483 APR 22.9 22.3 687.6 670.2 17.4 -1.0350 0.1422 MAY 17.9 13.5 555.0 419.3 135.7 -0.9831 0.0945 JUN 23.5 19.8 705.1 593.8 111.3 0.6300 0.6910 JUL 3.8 3.0 119.0 92.0 27.0 0.2593 0.3811 AUG 3.0 4.5 93.6 140.7 -47.1 -0.3912 0.3871 SEP 3.9 4.8 117.6 144.4 -26.8 0.4266 0.5478 OCT 25.6 26.0 792.1 805.8 -13.7 0.2289 0.2367 NOV 59.4 69.5 1782.1 2086.2 -304.1 0.5626 0.5792 DEC 18.0 8.5 557.4 264.3 293.1 0.1184 0.4197 1991 JAN 19.1 4.5 592.3 139.8 452.5 -0.0868 0.6174 FEB 49.2 22.9 1377.1 640.2 736.9 -0.0062 0.2461 MAR 8.5 4.8 263.9 150.2 113.7 -0.2533 0.0001 APR 23.8 23.7 712.5 709.8 2.7 0.4171 0.4440 MAY 21.4 29.5 662.2 913.8 -251.6 0.2559 0.5306 JUN 15.5 8.3 464.7 248.4 216.3 -1.3685 0.1448 JUL 6.8 5.5 212.0 169.1 42.9 -0.1473 0.1681 AUG 27.6 28.1 856.2 872.6 -16.4 0.7169 0.7204 SEP 2.1 9.7 61.7 290.6 -228.9 -1.6898 0.7301 102 s0B| s0B| s0B| OCT 1.6 2.7 49.1 83.6 -34.5 -0.1279 o.oocfe 0 NOV 36.2 40.2 1086.2 1206.8 -120.6 0.4669 0.4783 DEC 22.7 22.9 705.1 710.8 -5.7 0.4693 0.4733 1992 JAN 43.8 31.2 1357.8 968.5 389.3 0.5848 0.6775 FEB 24.3 22.5 704.8 653.8 51.0 0.4097 0.4232 MAR 10.0 16.0 308.5 496.9 -188.4 -0.1364 0.6689 APR 24.9 28.4 747.7 851.5 -103.8 0.4794 0.5404 MAY 9.9 13.3 306.4 413.5 -107.1 -1.7485 0.2818 JUN 6.0 5.4 179.8 163.1 16.7 0.3290 0.3827 JUL 5.3 4.5 165.4 139.4 26.0 0.2576 0.2958 AUG 5.1 3.6 158.0 111.6 46.4 -0.0051 0.3563 SEP 6.3 15.2 189.8 454.6 -264.8 -0.7909 0.3146 OCT 20.9 25.9 649.4 801.8 -152.4 0.6859 0.7043 NOV 21.7 24.1 651.3 723.8 -72 .5 0.1491 0.2121 a59RFile DEC : SEYMOUR6.STAall5CPage 1 of 5.6 4.6 173.1 2 143.6 29.5 -0.4575 0.2676 1993 JAN 12.3 2.4 382.2 75.2 307.0 -0.0759 0.6907 FEB 8.7 2.0 242.4 55.6 186.8 -0.5930 0.6922 MAR 22.2 23.6 689.5 731.6 -42.1 0.5665 0.5727 APR 28.3 37.8 849.4 1135.0 -285.6 0.4406 0.6600 MAY 24.6 21.1 763.2 652.7 110.5 -0.8603 0.0673 JUN 11.8 10.2 353.5 305.0 48.5 0.3325 0.4204 JUL 4.7 6.4 146.7 198.6 -51.9 -0.1409 0.2768 AUG 0.6 4.2 19.3 131.1 -111.8 -1.3429 0.2520 SEP 2.1 1.7 62.5 52.2 10.3 -0.1298 0.4321 Defined Period Period 17.4 (891001-16.3 930930) 25411.9 23820.1 1591.8 0.5245 0.5314 Defined Period Period 16.8 (891001-14.4 900930) 6136.6 5269.2 867.4 0.6176 0.6422 Defined Period Period 22.8 (901001-20.0 910930) 8334.2 7290.8 1043.4 0.4596 0.4706 Defined Period Period 16.3 (911001-17.1 920930) 5958.6 6253.9 -295.3 0.5914 0.5929 Defined Period Period 13.7 (921001-13.7 930930) 4982.5 5006.2 -23.7 0.4927 0.5168 APPENDIX 2 Statistics Report from Initial Seymour Watershed Model Using Sey Hatchery AES Station and Grouse Mountain AES Station 11018D15Ea3LslOhlOV |s4B U.B.C. |s4B WATERSHED MODEL I I |s4B STATISTICS REPORT sl2h8V a95CDate: 07-12-1996 Time: 14:59:50 STATISTICS FOR THE OCT 1 , 1989 - SEP 30 , 1993 WATER YEAR(S) Mean Qobs cms/d Mean Qest cms/d Tot Qobs cms Tot Qest cms Tot Qobs -Tot Qest Coeff.of Ef f Coeff.of Det 1989 OCT 28.2 24.6 874.3 762.6 111.7 0.6027 0.6448 NOV 31.4 33.4 941.4 1002.7 -61.3 0.8062 0.8201 DEC 15.5 19.0 481.5 587.7 -106.2 0.6596 0.6767 1990 JAN 19.1 12.4 591.2 383.7 207.5 0.4930 0.6945 FEB 14.6 3.6 409.5 99.5 310.0 -0.3592 0.3159 MAR 18.1 5.0 560.8 154.3 406.5 -1.4636 0.1656 APR 22.9 22.9 687.6 685.8 1.8 -0.2755 0.1375 MAY 17.9 20.3 555.0 629.4 -74.4 -0.3144 0.0864 JUN 23.5 20.2 705.1 605.8 99.3 0.6246 0.6670 JUL 3.8 3.3 119.0 102.5 16.5 0.2750 0.4391 AUG 3.0 4.2 93.6 128.9 -35.3 0.1366 0.3681 SEP 3.9 5.4 117.6 160.9 -43.3 0.0472 0.2533 OCT 25.6 23.9 792.1 741.7 50.4 0.2187 0.2226 NOV 59.4 65.5 1782.1 1964.5 -182.4 0.6117 0.6255 DEC 18.0 8.9 557.4 275.7 281.7 0.1400 0.4206 1991 JAN 19.1 4.8 592.3 148.6 443.7 -0.0633 0.6188 FEB 49.2 25.2 1377.1 704.7 672.4 0.0354 0.2465 MAR 8.5 5.0 263.9 155.1 108.8 -0.1700 0.0232 APR 23.8 24.1 712.5 721.7 -9.2 0.4903 0.5277 MAY 21.4 27.6 662.2 855.3 -193.1 0.4895 0.6301 JUN 15.5 20.3 464.7 608.0 -143.3 -0.3997 0.2130 JUL 6.8 9.0 212.0 279.0 -67.0 0.3098 0.5023 AUG 27.6 22.9 856.2 710.1 146.1 0.6913 0.7251 SEP 2.1 9.2 61.7 275.2 -213.5 -1.1497 0.7136 105 s0B| s0B| sOBI 1 0 6 OCT 1.6 3.1 49.1 95. 0 -45 .9 -0. 0957 0. 0240 NOV 36.2 40.2 1086.2 1206. 3 -120 .1 0. 4377 0. 4502 DEC 22.7 23.8 705.1 739. 2 -34 .1 0. 4745 0. 4774 1992 JAN 43.8 33.3 1357.8 1032. 8 325 .0 0. 5927 0. 6591 FEB 24.3 25.5 704.8 740. 4 -35 .6 0. 4081 0. 4159 MAR 10.0 19.4 308.5 600. 4 -291 .9 -1. 1475 0. 6314 APR 24.9 24.3 747.7 729. 7 18 .0 0. 4578 0. 5675 MAY 9.9 14.5 306.4 450. 3 -143 .9 -1. 6688 0. 3854 JUN 6.0 4.9 179.8 148. 3 31 .5 0. 3744 0. 4273 JUL 5.3 4.1 165.4 125. 9 39 .5 0. 2486 0. 3702 AUG 5.1 3.2 158.0 100. 4 57 .6 0. 0448 0. 4169 SEP 6.3 12.4 189.8 372. 0 -182 .2 -0. 1099 0. 3234 OCT 20.9 23.8 649.4 736. 5 -87 . 1 0. 6756 0. 6893 NOV 21.7 25.0 651.3 749. 3 -98 .0 0. 1430 0. 2111 a59RFile DEC : SEYGROUS.STAall5CPage 1 of 5.6 4.9 173.1 2 150. 6 22 .5 -0. 5072 0. 2732 1993 JAN 12.3 2.5 382.2 78. 0 304 .2 -0. 0637 0. 6925 FEB 8.7 2.4 242.4 66. 6 175 .8 -0. 4734 0. 4773 MAR 22.2 23.8 689.5 738. 6 -49 .1 0. 6108 0. 6155 APR 28.3 35.3 849.4 1059. 1 -209 .7 0. 5380 0. 6860 MAY 24.6 27.3 763.2 846. 0 -82 .8 -0. 4030 0. 0005 JUN 11.8 24.3 353.5 728. 2 -374 .7 -2. 0642 0. 4317 JUL 4.7 8.9 146.7 277. 3 -130 .6 -0. 0658 0. 3684 AUG 0.6 5.7 19.3 176. 1 -156 .8 -2. 6602 0. 2657 SEP 2.1 2.7 62.5 80. 3 -17 .8 -0. 2805 0. 2852 Defined Period Period 17.4 (891001-17.0 930930) 25411.9 24770. 8 641 .1 0. 5424 0. 5434 Defined Period Period 16.8 (891001-14.5 900930) 6136.6 5303. 9 832 .7 0. 6348 0. 6447 Defined Period Period 22.8 (901001-20.4 910930) 8334.2 7439. 7 894 .5 0. 4920 0. 4984 Defined Period Period 16.3 (911001-17.3 920930) 5958.6 6340. 4 -381 .8 0. 5932 0. 5964 Defined Period Period 13.7 (921001-15.6 930930) 4982.5 5686. 8 -704 .3 0. 4811 0. 5011 107 APPENDIX 3 Sample Precipitation and Temperature Graphs for 1990 108 109 110 APPENDIX 4 Statistics Report from Seymour Watershed Model Using Adjusted Precipitation Temperature 11018D15Ea3LslOhlOV s4B U.B.C. sOB| S4B WATERSHED MODEL sOB| 111 I S4B STATISTICS REPORT sOBI sl2h8V a95CDate: 07-15-1996 Time: 11:50:10 STATISTICS FOR THE OCT 1 , 1989 - SEP 30 , 1993 WATER YEAR(S) Mean Qobs Mean Qest cms/d cms/d Tot Qobs Tot Qest Tot Qobs Coeff.of Coeff.of cms cms -Tot Qest Eff Det 1989 OCT NOV DEC 1990 JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC 1991 JAN FEB MAR APR MAY JUN JUL AUG 28.2 31.4 15.5 19.1 14.6 18.1 22.9 17.9 23.5 3.8 3.0 3.9 25.6 59.4 18.0 19.1 49.2 8.5 23.8 21.4 15.5 6.8 27.6 26.9 38.7 22.1 20.0 11.9 8.7 21.9 12.8 17.8 3.1 4.4 4.4 25.5 81.1 17.0 12.5 38.2 7.2 29.2 23.2 7.3 4.9 25.2 874.3 941.4 481.5 591.2 409.5 560.8 687.6 555.0 705.1 119.0 93.6 117.6 792.1 1782.1 557.4 592.3 1377.1 263.9 712.5 662.2 464.7 212.0 856.2 834.1 1161.5 684.7 620.3 334.5 270.1 656.2 396.3 534.1 96.0 136.1 133.3 792.0 2432.3 527.2 387.0 1068.8 224.4 875.4 720.6 218.2 150.5 780.2 40.2 -220.1 -203.2 -29.1 75.0 290.7 31.4 158.7 171.0 23.0 -42.5 -15.7 0.1 -650.2 30.2 205.3 308.3 39.5 •162.9 -58.4 246.5 61.5 76.0 0.6321 0.7913 0.6230 0.6686 0.3456 -0.5065 -0.9906 -0.8805 0.5722 0.2521 -0.1159 0.5826 0.1530 0.5451 0.4695 0.3642 0.2338 0.2560 0.3385 0.5494 -1.6792 -0.1141 0.7178 0.6405 0.8175 0.6630 0.6793 0.3823 0.3355 0.1210 0.0982 0.6946 0.4008 0.3911 0.6093 0.1611 0.6073 0.4852 0.6061 0.2651 0.3377 0.4868 0.5677 0.1664 0.1709 0.7210 112 OCT 1.6 2.4 49.1 74.8 -25.7 -0.0713 0.0003 NOV 36.2 42.9 1086.2 1287.4 -201.2 0.3512 0.4067 DEC 22.7 31.4 705.1 973.1 -268.0 0.1521 0.3676 1992 JAN 43.8 46.9 1357.8 1455.1 -97.3 0.6441 0.6495 FEB 24.3 27.2 704.8 788.7 -83.9 0.3314 0.4029 MAR 10.0 14.0 308.5 434.5 -126.0 0.2915 0.6396 APR 24.9 20.2 747. 7 604.7 143.0 0.4339 0.5430 MAY 9.9 10.8 306.4 335.2 -28.8 -0.8629 0.3175 JUN 6.0 4.6 179.8 138.3 41.5 0.2742 0.3556 JUL 5.3 3.8 165.4 117.7 47.7 0.1239 0.2492 AUG 5.1 3.0 158.0 93.9 64.1 -0.1796 0.3221 SEP 6.3 13.7 189.8 410.0 -220.2 -0.4332 0.3159 OCT 20.9 23.3 649.4 721.6 -72 .2 0.6889 0.6979 NOV 21.7 26.9 651.3 806.9 -155.6 0.1793 0.2842 a59RFile DEC : SEYMOUR7.STAall5CPage 1 of 5.6 7.8 173.1 2 242.7 -69.6 -0.6045 0.2007 1993 JAN 12.3 6.1 382.2 188.4 193.8 0.2961 0.7101 FEB 8.7 3.5 242.4 96.7 145.7 -0.1482 0.7475 MAR 22.2 31.3 689.5 969.6 -280.1 0.4726 0.6255 APR 28.3 38.5 849.4 1155.3 -305.9 0.4300 0.6778 MAY 24.6 16.7 763.2 517.2 246.0 -2.0215 0.0126 JUN 11.8 9.3 353.5 277. 7 75.8 0.3025 0.4157 JUL 4.7 5.8 146.7 178.7 -32.0 0.0604 0.2761 AUG 0.6 3.8 19.3 116.6 -97.3 -0.9105 0.2433 SEP 2.1 1.7 62.5 50.3 12.2 -0.1108 0.3827 Defined Period Period 17.4 (891001-18.0 930930) 25411.9 26340.8 -928.9 0.5672 0.5754 Defined Period Period 16.8 (891001-16.0 900930) 6136.6 5857.2 279.4 0.6665 0.6697 Defined Period Period 22.8 (901001-23.1 910930) 8334.2 8448.6 -114.4 0.5342 0.5408 Defined Period Period 16.3 (911001-18.3 920930) 5958.6 6713.3 -754.7 0.5668 0.5830 Defined Period Period 13.7 (921001-14. 6 930930) 4982.5 5321.8 -339.3 0.5153 0.5428 


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items