Sie sind auf Seite 1von 8

Environmental Informatics Archives, Volume 3 (2005), 315 - 322 EIA05-040 ISEIS Publication series Number P002 2005 ISEIS

S - International Society for Environmental Information Sciences

Investigation of GIS-based Surface Hydrological Modelling for Identifying Infiltration Zones in an Urban Watershed
Nawshin Rumman, Grace Lin, and Jonathan Li* Department of Civil Engineering, Ryerson University, Toronto, Ontario, M5B 2K3 Canada Abstract. Infiltration derived from rainfall on ground surfaces is one of the important components in the hydrologic cycle process. Infiltration is a complex process and depends on the conditions of land use, soil type, and slope of the surface, evaporation, and precipitation. Urbanization increases impervious surface areas in a watershed and conveys the surface runoff to the storm sewer system which discharges to watercourses. This results less infiltration and groundwater recharge in built-up areas, higher surface flow velocity, and higher point discharge to the stream which leads to stream erosion and bank stability problems, etc. This paper presents a GIS-based hydrological modeling approach to identification of infiltration zones in an urban watershed. The comparative analysis results of the drainage network derived from digital elevation models (DEMs) by using HEC-GeoHMS in a GIS environment and the drainage extracted from surveyed topographical maps are demonstrated. Keywords: Surface water, hydrological modeling, GIS, urban watershed, infiltration zone. 1. Introduction

Many hydrological models have been developed to simulate and help us to understand hydrologic processes. The hydrological models are used as a watershed stormwater management tool to provide a direction to utilize natural water resources effectively and beneficially. According to Moor et al. (1991), the period from about 1960 to 1975 was the era of hydrologic modeling, in which mathematical descriptions of fluvial processes were developed and incorporated into hydrological models. Most of these models were concerned with predicting water quantities (e.g., runoff volumes and discharge) at a catchment or subcatchment outlet. These models were described as lumped parameter models, in which little or no consideration for spatially variable processes and catchments characteristics was involved. The emphasis of hydrologic modeling changed during 1975-1985. The growing concern with the environment, including management of pollution, resulted in the development of what have been commonly known as transport models. These models, using the hydrological models developed in the 1960s as the flow component, were perceived as the best way to predict water pollution. Just like their counterparts, transport models also poorly account for the effects of space and topography on catchment hydrology. Since mid-1980s, however, there has been an increasing recognition of the need to predict spatially variable hydrological processes at a fine resolution. This has led to the era of spatial modeling in hydrology. Digital terrain models (DTMs) and remote sensing data have been used to characterize catchment (e.g., vegetation cover) and are now considered as crucial data input to the new generation of hydrological and water quality models. Different forms of digital terrain data such as digital elevations models (DEMs), triangularly irregular networks (TIN) and contours) are used in different models to provide the spatial component of the analysis. The objective of the study is to delineate watershed using a surface hydrological modeling approach and compare with the surveyed map which was generated manually. We also investigate the applicability of the drainage comparative analysis approach developed by Saraf et al. (2004) to identify the potential infiltration zones in a typical urban watershed, Toronto, Ontario, Canada. Visualizing the distribution of infiltration zones will help us provide more meaningful infiltration parameters into hydrological models, analyze and interpret the results from hydrological models, and assist in watershed stormwater management practices and planning. Implementation of infiltration drainage features will provide benefit in stormwater quality and quantity control, groundwater recharge, and erosion control.
* Corresponding author: junli@ryerson.ca

Environmental Informatics Archives, Volume 3 (2005), 315 - 322

2. Background and Related Work One of the basic tasks in hydrological analysis is to delineate drainage basins and stream networks. The resulting stream networks can then be used in various applications, such as studies of stream flow hydraulics, prediction of flooding, and modeling of chemical transportation and deposition of pollutants in surface waterways. Traditionally, watershed delineation was mainly conducted by the manual delineation method. Although DEMs were introduced in the late 1950s (Miller and Laflamme, 1958), their application potential was not fully realized until the late 1980s when DEM s become widely available. With the advent of geographic information systems (GIS), DEMs have been used to delineate drainage networks and watershed boundaries, to calculate slope characteristics, to enhance distributed hydrologic models and to produce flow paths of surface runoff (Moore et al., 1991; USACE, 2003; Saraf et al., 2004). Recent studies have demonstrated that the accuracy of parameters extracted from DEMs is comparable to those obtained by manual methods while the processing time is much less (Wang et al. 1998, Islam, 2004; Saraf et al. 2004). These parameters include the basin size, basin slope, main channel length, and stream length. For example, Garbrecht et al. (1993) extracted network parameters from 1:24,000 DEMs for a small basin (84km2). They compared the DEM generated network parameters with those based on contour lines of topographic maps and found that the DEM generated parameters were in general within 5 % of those measured manually from the maps. Islam examined the differences between manual delineation and GIS-based delineation using two types of DEMs. One DEM was generated from traditional stereo images and the other was provided by Australian National University (ANU) DEM. The latter incorporated the knowledge of actual river network and a different algorithm for filling sinks. Two watersheds of 26 km2 and 70 km2 in Southern Manitoba Region of Canada were used in his study. It was found out that the ANU DEM provided more accurate delineation than the traditional DEM. It was concluded that although hand-delineation was cumbersome, its accuracy was not lower than computer generated delineation, and the watershed areas obtained from different methods did not vary significantly. Two major factors can affect the accuracy of stream networks derived from DEMs: the scale of DEMs and the drainage density. Wang et al. (1998) compared the parameters derived from the 1:250,000 DEMs with those from the 1:24,000 DEMs in 20 basins ranging from 150 to 1000 km2 and showed that the goodness-of-fit between parameters estimates based on the DEMs varies. Results clearly showed that superior estimations are produced from the 1:24,000 DEMs. However, the 1:250,000 DEMs provided reasonably good estimates to some geometric and topological parameters for lower order streams such as stream length and frequencies. They also pointed out that small scale DEMs may not necessarily result in improved accuracy for higher-order stream networks. The current trends in using DEMs and GIS to perform hydrological analyses are beyond the step of preparation of hydrological inputs and focus on the development of GIS based distributed rainfall-runoff modeling, in which one attempts to establish a linkage between GIS and hydrological models. These models are called object-oriented model (Maidment, 1993). For example, Jain et al. (2004 and 2005) developed and tested a DEM-based overland flow routing model for computation of surface runoff from isolated storm events using the diffusion wave approximation. Spatially distributed information for model inputs, such as topography, soil, land use, etc. for each of the discretized cells of the catchment was provided through a GIS. The catchment DEM was utilized to derive the flow direction and the computational sequencing for flow routing for each of the discretized cells of the catchment represented as a proper hydrologic cascading system. A finite volume-based numerical solution of the diffusion wave approximation of the St Venant equations for DEM-derived overland flow and channel flow was developed. Using Philip (1969)s infiltration model, infiltration was computed on a cell basis. The model produces in spatial and temporal domain the flow discharge, depth, and velocity due to isolated rainfall events on a catchment. The proposed model was calibrated and verified, using rainfall and runoff data collected on the Banha catchment in India. The results of model application indicated that the model satisfactorily predicted the runoff hydrograph. The U.S. Army Corps of Engineers Hydrologic Engineering Center (HEC) has developed HEC-GeoHMS and HEC-HMS. HEC-GeoHMS is currently available to the public, and can be used for stream and watershed delineation. Similar to the ANU DEM, the depressionless DEM in HEC-GeoHMS is created by filling the depressions or pits by increasing the elevation of the pit cells to the level of the surrounding terrain in order to determine flow directions, USACE (2003). The HEC-HMS, the Hydrologic Modeling System, was designed to simulate rainfall-runoff processes of watershed systems. It can be used to assess the performance of delineated watershed in rainfall-runoff simulation, and generate outflow hydrographs. In regard of identifying infiltration zones by GIS tools, the focus of the Sullivan et al. (1996)s study was to determine which data collection techniques were best suited for quantifying and delineating spatial variations in surface infiltration and for estimating infiltration parameters used in computer models for simulation of runoff

316

Environmental Informatics Archives, Volume 3 (2005), 315 - 322

quantity. Field data were combined with coverages in an existing GIS database and field observations of surface attributes (soil type and vegetation) which are known to contribute to variation in infiltration rates. The coverages in the GIS database were used to develop a map of the watershed which delineates areas of low potential infiltration from areas of high potential infiltration. In Saraf (2000)s study, the identification of groundwater recharge zones was based on nearest neighbour analysis on an intersection theme of two drainage networks. Using simulated and surveyed drainage networks, nearest neighbour analysis was done using an ArcView extension, Nearest Neighbour Analyst Extension developed by Saraf (2000). The analysis applied a simple test of significance for deviation from randomness, using the standard error of the expected difference. In his study, the watersheds showed a R-value < 0.5 on two-drainage networks intersection theme and illustrated clustering tendency. In relatively flat areas where rocks and soils were suitable for groundwater recharge, a considerable mismatch was observed between two drainage networks. High correlation indicates high surface runoff, whereas, in flat areas, it could be inferred from the misfit that in these areas water was infiltrating more and these may be the recharge areas. The groundwater recharge maps of the two study areas showed that the areas where groundwater recharge was high, the degree of mismatch between the surveyed and simulated drainage was also high. Such mismatch between simulated and surveyed drainage networks can be exploited to delineate the groundwater recharge zones. The basis of such groundwater recharge zone identification is that the water which is coming through rainfall is not completely flowing as surface runoff; rather some part of it is going into groundwater and hence mismatch between two drainage networks are depicted. It is important to note that in the surface hydrological modeling it is assumed that there is no loss to water when it flows on the ground. However, whenever there is mismatch between surveyed and simulated drainage network it automatically means that there is some loss of water to groundwater, assuming that evapotranspiration losses remain the same throughout the watershed. Saraf et al. (2004) also compared their findings with the groundwater recharge zones identified using remote sensing and GIS based integrated methodology as developed by Saraf and Choudhury (1998). This visual comparison showed good degree of match between groundwater recharge zones. Up to today, little study has been done for the use of DEMs to delineate drainage networks and watershed boundaries in built-up areas watersheds. This paper presents our investigation on the use of HEC-GeoHMS to perform the watershed delineation and the drainage comparative analysis in urban built-up areas.

3. A Case Study in the City of Toronto 3.1 Study Area and Data sets. Flowing through the heart of Toronto, the Don River is one of the Canada's most degraded urban rivers. Over 80% of its 360-km2 land is over urbanized and is home to over 800,000 people. Figure1 presents the Don River watershed which consists of seven subwatersheds.

Figure 1 Study area the Don River watershed in the City of Toronto, Ontario, Canada

317

Environmental Informatics Archives, Volume 3 (2005), 315 - 322

The data sets we used in this study include a 10-m DEM from Ontario Ministry Natural Resources, land use shapefiles from the City of Toronto, and the Ontario Soil Series system shapefiles from Toronto and Region Conservation Authority (TRCA). By selecting the theme property it is found that there are 3486 rows and 2890 columns at 10 m resolution that cover a rectangular area of 1007.45 km2, which encompasses the Don River watershed of 358 km2. The free-downloaded HEC-GeoHMS Extension of ArcGIS software was used for stream and watershed delineation. This extension uses 8-Point Pour down approach to determine flow paths. Using the terrain data as input, a series of terrain preprocessing steps were performed to derive the drainage networks. The steps consist of computing the flow direction, flow accumulation, stream definition, watershed delineation, watershed polygon processing, stream processing, and watershed aggregation. According to Saraf et al. (2004), depression areas surrounded by neighbouring pixels of higher values are considered as hindrance for the determination of hydrological flow direction using DEMs. Some depressions may be data errors while some exist in reality. The extension HEC-GEOHMS to ArcView can remove depressions by filling and raising them to the lowest elevation value on the rim of depression. 3. 2 Methodology The approach used in the study is GIS-based surface hydrological modeling method (see Figure2). DEMs are extremely valuable in understanding the properties of the terrain (e.g. slope, aspect, curvature, flow accumulation, stream ordering, etc.). Further, DEMs can provide a wealth of information about the geomorphical and hydrological properties of the area. The extension HEC-GeoHMS to ArcGIS was used to determine the flow direction and flow accumulation. Flow is routed through the path of the steepest slope of eight neighbouring pixels, assuming the surface to be insulated. The flow direction grid represents the direction of flow out of each cell into one of its eight neighbours. Flow direction data are used to create the flow accumulation grid, where each cell is assigned the value of the number of cells flowing to it. Cells having flow accumulation value of zero generally correspond to local topographic highs, whereas cells with high flow accumulation correspond to stream channels. The flow accumulation is utilized to produce a drainage network applying a suitable threshold value. Threshold value indicates the minimum upstream drainage area (threshold area) necessary to maintain a stream. The areas with the flow accumulation value less than the threshold value is considered relatively flat and the flow has a tendency to infiltrate depending on the land cover and soil type at those locations. Different threshold values would result in stream networks with different total stream lengths, and consequently, different drainage density values. Determination of the threshold value is subjective and could be based on the calibration of stream network from the spatial model matching with the one extracted from surveyed topographical maps. In the same procedure drainage network for the subwatershed also determined. A comparative analysis of drainage network derived from DEMs by ArcGIS and that extracted from surveyed topographic maps was performed. However, whenever there is mismatch between surveyed and simulated drainage network it automatically means that there is some loss of water to groundwater that means water is infiltrate in that area. We assume that evapotranspiration losses remain the same throughout the watershed. Combining the calibrated stream network with the land use and soil type spatial data, the areas with high potential infiltration are identified. For example, a flat pervious area with sandy soil will have high potential infiltration.
DEM Surveyed map Soil & Landuse maps

input

HEC-GeoHMS

output
Watercourse Compare delineation of watercourse & watershed Identify Infiltration zones

Watershed

Figure 2 Flow chart of our methodology

318

Environmental Informatics Archives, Volume 3 (2005), 315 - 322

3.3 HEC-GeoHMS Modelling The operation of HEC-GeoHMS streamline and watershed delineation consists of fill sinks, flow direction, flow accumulation, stream definition, stream segmentation, watershed delineation watershed polygon processing, and watershed aggregation. Figure3 shows the results after flow direction calculation. Figure4 shows delineated subwatersheds in the study area. Figure4 shows the results of watershed aggregation of the study area.

Figure 3 Flow Direction

Figure 4 Delineated subwatershed in the study area.

4. Comparison of Modelled Results Comparison of the simulated watershed and streamlines were carried out based on the following aspects. (1) Threshold values. A wide range of threshold values from 1% to 10% of the largest drainage area of 342 km2 was tested for the Don River watershed delineation. It was found that the higher threshold value the less dense drainage networks were produced. As illustrated in Figure6, significant difference in streamlines and watershed size resulted from threshold values of 2% and 5% of the total Don River watershed area. (2) Comparing with Surveyed Map. The simulated drainage network was compared with the surveyed map. As shown in Figure7, five out of seven subwatersheds match with the surveyed map very well. However, mismatching was found in the Massey Creek and Lower Don subwatersheds. A discussion with the City of Toronto staff indicated that the streamlines are partially enclosed in sewer conduits due to urbanlization in these two subwatersheds. It is also shown in Figure7 that streamlines mismatch at the upstream of tributaries due to the limitation of the threshold value. Comparing with the streamlines in Upper East Don subwatershed in Figure8, the streamlines extended further and better match was obtained since smaller threshold value was used.

Figure 5 Results of watershed aggregation.

319

Environmental Informatics Archives, Volume 3 (2005), 315 - 322

Figure 6 Watershed delineation with threshold values of 5% (left) and 2% (right) of the Don River Watershed

Table 1 Comparison of modeled and surveyed watershed areas. Watershed Area of Watershed Delineated (sq. km) Manual delineation in surveyed map Don Watershed Upper East Don 358 65 Computerized delineation using HEC-GeoHMS 342 65

Figure 7 Comparison of Don watershed and streamline between modeled (green lines with a threshold value 2%) and surveyed map (yellow lines)

Figure 8 Land use map of the Upper East Don subwatershed

320

Environmental Informatics Archives, Volume 3 (2005), 315 - 322

5.

Identification of Infiltration Zones

An investigation of using Saraf et al. (2004) approach to identify the infiltration zones was carried out in the Upper East Don subwatershed, where the watershed delineation was identical by both hand and HEC-GeoHMS delineation. The Upper East Don subwatershed has a total area of 65 km2 and approximately 35% of the area is undeveloped and the rest of area is urbanized, as shown in Figure8. Mapping of surficial soil classification according to the Ontario Soil Series system was provided by TRCA as a GIS map layer. The soil classification can be seen for the Upper East Don watershed in Figure9. Mismatch between the surveyed and simulated streamlines are shown in Figure9 and 10, respectively. Most of the mismatches locate in the upstream end of the tributaries, where normally tends to be flat. It can be seen that the modeled streamlines extend beyond the surveyed streamlines. As concluded in Saraf et al. (2004), whenever there is mismatch between surveyed and simulated drainage network it automatically means that there is some loss of water to subsurface, assuming that evapotranspiration losses remain the same throughout the watershed. Combining with the soil information, it is confident to identify the high infiltration zone locations, where sandy soil with misfit surveyed and simulated streamlines locate.

Figure 9 Soil classification map (red sandy loam, green clay loam, blue loam) of the Upper East Don subwatershed overlaid with modeled streamlines (green lines with a threshold vale of 1%) and surveyed streamlines (yellow lines).

Figure 10 Identification of potential infiltration zones

321

Environmental Informatics Archives, Volume 3 (2005), 315 - 322

Concluding Remarks A comparative analysis of the drainage network derived from DEM by HEC-GeoHMS and drainage extracted from surveyed topographical maps has been carried out. In general, HEC-GeoHMS model produces agreeable results on the watershed and streamline delineation comparing with surveyed maps. Mismatching in watershed boundaries from both methods indicates streamlines would be enclosed in sewer conduits due to urbanlization. However, if the DEMs incorporate knowledge on existing network drainage, HEC-GeoHMS watershed delineation can be applied to an urbanized area. This also demonstrates that watershed and stream delineation from geospatial hydrology model provides results that can be verified with published information to detect possible errors in terrain model. The drainage comparative analysis approach developed by Saraf et al. (2004) in identifying groundwater recharge zones are adopted in the present study to identify potential infiltration zones. High correlation of stream delineation from HEC-GeoHMS and surveyed map indicates high surface runoff; whereas, in flat areas, it can be inferred from the misfit that in these areas water is infiltrating more and these may be the recharge areas. Combining HEC-GeoHMS hydrology model with soil type and land use information, identification of infiltration zone is reliable. Field investigation to confirm identified infiltration zones will be useful in determining the application of this approach. References City of Toronto, (2002). Wet Weather Flow Management Mast Plan. Garbrecht, J., and I. Martz, (1993). Network and subwatershed parameters extracted from digital elevation models the Bills Creek experience, Water Resources Bulletin, Vol. 29, pp. 909-916 Islam, M. D. R. (2004). A review on watershed delineation using GIS tools, Dept. of Civil Engineering, University of Manitoba, Winnipeg, www.wrrc.dpri.kyoto-u.ac.jp Jain, M. K., U. C. Kothyari, and K.G. Ranga Raju, (2004). A GIS based distributed rainfall-runoff model, Journal of Hydrology, Vol. 299, pp. 107-135 Jain, M. K., and V. P. Singh, (2005). DEM-based modelling of surface runoff using diffusion wave equation, Journal of Hydrology, Vol. 302, pp. 107-126 Johnson, C.A., A.C. Yung, K.R. Nixon, and D.R. Legates, (2001). The use of HEC-GeoHMS and HEC-HMS to perform grid-based hydrologic analysis of a watershed, www.dodson-hydro.com Lo, C.P. and A. Yeung. (2002). Concepts and Techniques of Geographic Information Systems, New Jersey: Prentice Hall. Maidment, D.R. (1993). GIS and hydrologic modelling, In Environmental Modeling with GIS, by Goodchild, M.F., Parks, B.O., and Steyaert, L.T. (eds.), New York: Oxford University Press, pp. 147-167. Moore, I. D., Grayson, R.B., and Ladson, A.R. (1991). Digital terrain modelling: A review of hydrological, geomorphological and biological applications, Hydrological Process, Vol. 5, pp. 3-30 Philip, J.R. (1957). Theory of infiltration, Soil Science, Vol. 83, pp. 345-357. Saraf, A.K, and Choudhury, P.R. (1998). Integrated remote sensing and GIS for groundwater exploration and identification of artificial recharge sites, International Journal of Remote Sensing, Vol. 19, pp. 1825-1841. Saraf, A.K, Choudhury, P.R., Roy, B. Sarma, B., Vijay, S. and Choudhury, S. (2004). GIS based surface hydrological modelling in identification of groundwater recharge zones, International Journal of Remote Sensing, Vol. 25, pp. 5779-5770 Sullivan,M., Warwick, J.J., and Tyler, S.W. (1996). Quantifying and delineating spatial variations of surface infiltration in a small watershed, Journal of Hydrology, Vol. 181, pp. 149-168 USACE (2003), Users Manual, Geospatial Hydrologic Modeling Extension, HEC-GeoHMS, Version 1.1, U.S. Army Corps of Engineers Hydrologic Engineering Center, California, USA, www.hec.usace.army.mil Wang, X., and Yin, Z. Y. (1998). A comparison of drainage networks derived from digital elevation models at two scales, Journal of Hydrology, Vol. 210, pp. 221-241

322

Das könnte Ihnen auch gefallen