GIS/EM4 - Modeling and potential use of hydrologic contributing areas for environmental applications

Modeling and potential use of hydrologic contributing areas for environmental applications

GIS/EM4 No. 234

Yuri Gorokhovich

Abstract

A method of calculation and visualization of hydrologic contributing areas using vector-based GIS is described. Selection of contributing areas around stream segments assists to define lateral and channel flow for each stream segment during the rainfall. During each time interval only certain area of the catchment contributes to the surface runoff. Developed GIS-based method helped to extract contributing areas and simultaneously calculate runoff values for each stream segment along the stream channel. It produced discharge values used in flow routing algorithm and simultaneously stored images of the expanding contributing areas for future animation. The basis of the presented methodology is a spatial, hydrologic response unit (HRU) matrix consisted of three GIS vector data sets: landuse, soil type and digital elevation model represented as a Triangulated Irregular Network (TIN). In this matrix each element has uniform surface runoff value. By overlaying stream outline vector data set and HRU matrix (which is a polygon), stream data set was broken into segments with different length, depending on the size of each element of HRU matrix. As a result, each stream segment acquired all HRU parameters from the matrix's polygon attribute table, i.e. landuse code, hydrologic group of the soil, slope, runoff coefficient and runoff discharge for each time interval during simulation. Algorithms and corresponding equations were developed to estimate discharge within the stream channel for each segment and time interval using contributing areas. Contributing areas, produced via GIS spatial iterations, were stored as graphic elements, annotated and exported into image files for animation. The method was tested for the small watershed in southern part of the New York state. The flow modeling predicted results within a range of 25% to 100% of observed values for 8 selected storm events. Predictions were generally better when storm durations exceeded 2 to 4 hours and time intervals were greater than 15 minutes. The method can be used for environmental applications dealing with the soil erosion and water quality modeling. Potential interest for environmental applications can be analysis of the pollutograph and dynamically changing contributing areas to define the location of the source of pollution.

Keywords

GIS, hydrologic modeling, contributing areas, spatial analysis

Introduction

GIS has been extensively used for the environmental modeling purposes (NCGIA, 1996). General datasets which are widely utilized for environmental modeling consist of soils data (polygons or raster data), stream networks (lines or connected cells), landuse (polygons or raster data) and digital elevation models (raster data). These data have various attributes which can be utilised in calculations, such as landuse areas, soil moisture coefficients, slope, stream length. Coupled with environmental models, GIS has proved to be an excellent tool to help manage a watershed and assist in solving water quality related problems for the drinking water supply, ( Gorokhovich and Janus, 1996). General models help to identify the source of pollution and associated concentrations or loads. Among specific questions addressed by models is a question of travel time or how fast pollution can reach a water supply. Storm events play important role in this process as a strong dynamic force. They move pollutants very fast and decrease travel time considerably. Therefore models should provide user with ability to predict travel time and quantity of pollutants. To accomplish this task models should be able to predict contributing areas around streams and water bodies. Several soil erosion studies have indicated importance of contributed areas. In soil erosion processes described by Lal et. al. (1994), sediment transport models show that detachement by shear forces occurs mainly in areas where water is concentrated (e.g., rills) rather than over a broad areas. Also they indicated that re-entrainment of sediments will take place mainly from the base of the rill. Novotny and Chesters (1989) mention the study of Dickinson and Pall (1982) which showed that sources of stream sediments do not necessarily coincide with major soil erosion areas because of the differences in capacity of different parts of a watershed to transport sediments. The authors underline that a source with a high soil erodibility located far from established channels may not contribute as much pollution to a stream as a less erodible source near stream. Walling (1996) mentions importance of contributing areas and sediment sources for calculating phosphorus loads. He uses such terms as "minimum active contributing areas", defined as a percentage of the total catchment area, in accord with the variable source area concept of storm runoff production (Kirkby, 1978). Therefore, contributing areas are important tool in quantifying pollution loads at specific time intervals during storm events.

Problem statement

One of the main powers of GIS analytical functions is its ability to superimpose data layers in geographical space. This function allows scientists to understand better a spatial relationship between various data and their spatial patterns. In case of environmental modeling two data sets has to be superimposed: pollution source and environmental conditions affecting pollution dynamics. The source of pollution can be either point or non-point. Environmental conditions include surface description (surface of terrain or hydrogeological layer) and dynamic force (rainfall, storm event, flow, etc.). In this study an attempt was done to create a hydrological surface that would provide user with information on expansion of contributing areas, delivering surface runoff to the stream channel. Geographically located contributing areas can be superimposed with pollution source to predict pollutant concentration or time when pollutant enters a stream channel. Therefore, two models have to be created: model of hydrologic surface and model of contributing area expansion.

Approach

Hillel (1986) identified four principles that should guide model development: 1. Parsimony: "A model should not be any more complex than it needs to be and should include only the smallest number of parameters whose values must be obtained from data"; 2. Modesty: "A model should not pretend to do too much", "there is no such thing as THE model"; 3. Accuracy: "We need not have our model depict a phenomenon much more accurately than our ability to measure it"; 4. Testability:"A model must be testable" and we need to know "if it is valid or not, and what are the limits of its validity". Two models were created for the study: hydrologic model of terrain and the model of expansion of contributing areas. Hydrologic model of terrain was represented by hydrologic response units (HRU) matrix developed as a result of land use, slope and soil data integration via GIS. Various land uses, slopes and soil types create unique patterns responsible for the amount of surface runoff. The HRU matrix enabled user to estimate the runoff from various areas of the studied terrain. The selection of contributing areas for each stream segment can be accomplished through various methods used in hydrology to calculate time of concentration. There are several basic formulae available, including Kirpich's (Kirpich, 1940), Hathaway's (Hathaway, 1945) and FAA (Federal Aviation Agency, 1970).There are two essential spatial parameters presented in all methods: slope and distance from the outlet to the most distant point on the watershed boundary. Distance can be defined by using coordinates of points on a given stream segment and farthermost point on the watershed. Slope can be obtained from the attribute tables of the HRU matrix elements as an average value. After obtaining the time of concentration, it is possible to interpolate the buffer distance for each time interval to display and use contributing areas. For example, in case of linear interpolation, if the most distant point in the watershed is 800 meters from the given segment on the stream and the estimated time of concentration is 40 minutes, then for the first 10 minute interval of the storm, water will contribute to the studied segment from areas within the distance of 200 meters (assuming the same rate of precipitation). This is a simple approximation that produces distances necessary for spatial selection of HRU matrix elements. Linear interpolation was tried at the beginning of modeling exercise as a simple solution, but after several model runs it was replaced by exponential function of the distance vs.time.

Methods

According to Burrough, P.A. (1993), there are only two options for linking environmental models with GIS. First option is to run the model outside GIS framework and then use GIS for the data storage and display. Second option is to integrate model in the GIS by utilizing GIS functions, command or SQL language, or map algebra. First option has advantages when the modeled process is complex enough to use specific computer language or software. The disadvantage of this is a need to convert data formats and constantly move between compiled code and GIS system. Examples of such modeling are WASP4 model (DePinto J.V. et. al., 1993) and TR-20 model (Cahill T.H., et. al., 1993). Second option has advantage that the model and GIS use single database and functions. The disadvantage is that if the model is too complicated in computations, then GIS may not have enough powerfull tools to accomplish the goal of the modeling. It is only true though for the commercial GIS which are based on a proprietary codes and have limited number of functions. In case of public domain GIS the code is usually open and this allows user to implement any changes or use full capabilities of the programming language functions. Perfect example of such GIS is GRASS software, developed by US Army Corp. of Engineers. It is written in C language and excellent examples of model integrations within such GIS can be found in Proceedings of the Seventh Annual GRASS Users Conference (Waggoner G.S., editor, 1993). Since contributing area modeling will be based on some empirical methods which don't require complicated computations the second option is more practical and attractive. Another consideration which favors integration of GIS and modeling algorithm is that the result of modeling will be cartographical display of the final data. Modeler in this case will be able to manipulate with both display options and data processing from one single place. The time which will be gained by such integration can convert modeling process from tedious data/format exchange into the real dialog between scientist and a model.

Choice of the GIS type.

The choice of GIS type should depend on specific application that scientist keeps in mind. While working with stream networks and tracing water flow through channels, vector based GIS can be most effectively utilized. This kind of GIS will also has advantage for hydrologist who will use the resulting model because in hydrology rivers are traditionaly represented as vectors or segments. Most of the widely utilized hydrological models, such as HEC-1, HEC-2, TR-55 use the concept of segments or set of segments which are usually oriented toward the direction of outflow. In these models all segments are numbered and organized in a tabular form. This organization allows to calculate hydrological parameters for each segment and then build a linear graph or bar diagram representing distribution of a modeled parameter (usually discharge) along the selected network. Data associated with segments are stored in the table and may contain also information about areas surrounding stream network. Usually these are various landuse data, soil data, etc. More often these data are represented by grid-based or raster GIS because spatial analysis can be performed much faster. In case of integrated within GIS models, map algebra can be conveniently utilized. Raster data type is just a part of available GIS data structures which can be utilized for the modeling. Among main types of GIS models are hydrologic response units (HRUs), triangulated irregular networks (TIN), grid-based and contour-based models (Moore, I. et. al., 1991, M. F. Goodchild et. al., 1993). These types have one thing in common - they represent certain areas with unique continuous properties. While TINs, grid-based and contour models deal with elevations which is their only one unique attribute, HRUs are hydrologically similar areas derived by overlaying land use properties (land use classes) and soil properties (hydric groups). Hydrological property in this case is rainfall excess and infiltration, derived from combinations of land use classes and hydric soil groups. HRU data structure therefore fits most hydrological purposes and can be modified according to the modeling ideas. In presented study HRU represent main GIS structure for the proposed modeling.

GIS spatial analysis

Contributing areas increase during storm events until the whole watershed area reaches saturation. Contributing area for each time step is responsible for the certain amount of water delivered through the surface runoff to the stream channel. This means that during each time step for each segment on the stream there is both channel flow from upstream and from contributing areas that yields a certain amount of runoff in the form of lateral flow. During the first time step, the amount of runoff, delivered to the stream channel is equal to lateral flow, the total runoff value for the given contributing area. In the next time steps, the contributing area becomes larger and delivers runoff developed during the previous time step. Therefore, this runoff is taken into account as well. Figure 1 shows calculation routine for this process.

Figure 1.Relationship between travel time and expansion of contributoing areas was used in calculations of runoff (Q, L/sec) during a storm event modeling.

The general form for the previously described routine of calculating lateral flow is:

Figure 2.General form of equation linking expansion of contributing areas and water discharge.

The discharge Q was calculated from the HRU matrix using rational method and area A was calculated at each time step for each segment on the stream (Gorokhovich et. al., 2000). The algorithm was programmed using ARCINFO macro- language (AML). Figure 3 shows general structure of the developed algorithm.

Figure 3. General GIS algorithm that allows to calculate contributing areas, discharge and produce graphic images of contributing areas for each time step of the modeled time interval.

Discussion

GIS allows one to model contributing areas and lateral flow as an input to the hydrological flow routing model. Spatial modeling took into account different watershed characteristics, as well as their location and potential influence on runoff, depending on travel time. A linear relationship was replaced by an exponential one to obtain the radius of contributing areas after extremly high volumes of runoff were predicted during several of the first time steps. Figure 4, 5 and 6 show sequence of the GIS algorithm end results. All three images refer to the first time interval of the modeling sequence. Figure 4 shows contributing area for the first segment in cyan color. Figure 5 shows contributing area for the following stream segment and Figure 6 shows contributing area for the last segment on the stream. White area around the stream channel is a contributing area that already supplied surface runoff into the channel. Green color refers to the HRU matrix, consisted of combination of soil, landuse and elevation (TINs) data.

Figure 4. Visualization of runoff and contributing area for the first segment on the stream for the first time interval.

Figure 5. Visualization of runoff and contributing area for the second segment on the stream for the first time interval.

Figure 6. Visualization of runoff and contributing area for the last segment on the stream for the first time interval.

The hydrological model presented here conceptualizes how different variables on the watershed influence flow rate and surficial runoff (i.e., channel and lateral flow to the stream). The variables analyzed are coefficients of the surface runoff, contributing areas and slope. This model works on the time scale of a single storm event that may last for a few days using time intervals during the storm ranging from 10 to 50 minutes. Therefore, some large variability in RMSE can be expected as the result of a dynamically changing environment (e.g. the sharp transition from antecedent conditions, high winds, rain gusts, and channel blockage by falling trees, etc.). Most of the modeled storm simulations overpredict flow. Underprediction can be possibly explained by the short time interval of storms, low discharge values and antecedent moisture conditions. Most of the simulated storms differ considerably from underpredicted ones in terms of duration, seasonality and rain intensity. To account for all different conditions, runoff coefficients were adjusted to satisfy all of them. This could explain the fact that some storms became underpredicted. Figures 7 and 8 show examples of the storm event modeling. More detailed description of the hydrologic model and its applications described in (Gorokhovich et. al., 2000).

Figure 7. Flow modeling results for the storm event, November 9, 1996

Figure 8. Flow modeling results for the storm event, July 24, 1997

Conclusion

Spatial distribution results in increasing numbers of computer commands iterations within the modeling code and complex GIS structure of the input data. This complexity slowed down the simulation process. ARC/INFO GIS software, used for the main part of the programming, like most other commercial GIS packages, utilizes macro language. Macro language differs from the regular programming language because it is not compiled. Therefore, performance of the programming code is slow and an average model run takes 7 to 12 hours of computer time. Nevertheless, a spatially distributed model allows the analysis and study of particular portions of the watershed which might be affected by the point source pollution, or implementation of specific engineering projects. By using this distributed modeling approach, it became possible for the first time to visualize boundaries of areas, contributing surface runoff to stream segments. Results were animated, showing expansions of contributing areas with each time interval. This is one of the first attempts to visualize boundaries of contributing areas - isochrones and convert them into spatial objects. File formats used for the visual frames are GIF's (Graphics Interchange Format) and can be used by most of the commercial and shareware animation software packages.

Recommendations for future research

Contributing areas in this study were estimated through empirical equation, which was experimentally derived. However, there is a need for the field experiments, which would allow the better understanding of contributing areas as a spatial phenomenon. Such a study would also support coefficients, used in non-linear interpolation of contributing areas. These experiments should allow the observation of relations between the rainfall depth, duration and growth of contributing areas. The distance between point of observation and most distant point on the watershed is an important parameter used by the time travel formula. The most distant point on the watershed is usually selected arbitrarily from a map, showing watershed basin, or through computer by simply calculating the length of the watershed basin. Nevertheless, this length can vary, depending on topographical complexities, such as height of hills, number of depressions and valley shapes. Future enhancement of the travel time computation should include a better estimation of the distance between point of observation and most distant point on the watershed, using GIS capabilities.

Acknowledgements

Special appreciation is due to New York State Electric Research and Development Authority (NYSERDA) grant which allowed to fulfill field data collection and purchase of some monitoring equipment.

References

Burrough, P.A. 1986. Principles of geographical information systems for land resources assessment. Clarendon Press, Oxford. 193 pp.

Cahill, T.H., J.McGuire and C.Smith. 1993. Hydrologic and water quality modeling with geographic information systems. In: Proceedings of Symposium on Geographic Information Systems and Water Resources. J.M.Harlin (Ed.), 640 pp.

DePinto, J.V. , J.F.Atkinson, H.W.Calkins, P.J.Densham, W.Guan, H.Lin, F.Xia,P.W.Rodgers, T.Slawecki and W.L.Richardson. 1993. Development of GEO-WAMS: a modeling support system for integrating GIS with watershed analysis models. First International Conference/Workshop Integrating GIS and Environmental Modeling. Breckenridge, Colorado, 1993.

Dickinson, W.T., and R. Pall. 1982. Identification and control of soil erosion and fluvial sedimentation in agricultural areas of the Canadian Great Lakes basin. Contract No.2360. 0152-S-1-0433. Final Report to Supply and Services, Canada, Ottawa, Ontario.

Federal Aviation Administration. 1970. Airport drainage. AC 150/5320-5B.

Goodchild, M.F (Ed.).1993. GIS and Environmental Modeling. New York, Oxford University Press. pp.488

Gorokhovich,Y.,R.Khanbilvardi, L.Janus, V.Goldsmith,D.Stern.2000. Spatially distributed modeling of stream flow during storm events. Journal of the American Water Resources Association, June 2000.

Gorokhovich,Y. and L. Janus. 1996. The NYC water quality division geographical information system (GIS) and its applications for the watershed management. Proceedings, Watershed'96 Technical Conference, June 8-12, 1996, Baltimore, Maryland. pp.530-532.

Hathaway, G.A. 1945. Design of drainage facilities. Transactions, ASCE, (110):697-730.

Hillel, D. 1986. Modeling in soil physics: a critical review. Future Developments in Soil Science Research (A Collection of Soil Science Society of America Golden Anniversary Contributions). Presented at the Annual meeting, Soil Science Society of America, New Orleans, Louisiana, pp. 35-42

Kirkby M.J. 1978. Hillslope hydrology. Chichester, John Wiley and Sons.

Kirpich, Z.P. 1940. Time of concentration of small agricultural watersheds. Civil Engineering, 10:362

Lal R.(Ed.). 1994. Soil erosion research methods. Soil and Water Conservation Society.

Moore, I.D. and J.C. Gallant. 1991. Overview of hydrologic and water quality modeling. In I.D. Moore (ed), Modeling the Fate of Chemicals in the Environment, Canberra: Center for Resource and Environmental Studies, the Australian National University, pp. 1-8.

NCGIA. 1996. Proceedings of the Third International Conference/Workshop on Integrating Geographic Information Systems and Environmental Modeling, Santa Fe, New Mexico.

Novotny V. and G. Chesters.1989. Delivery of sediment and pollutants from nonpoint sources: a water quality perspective. Journal of Soil and Water Conservation, Nov.-Dec., 1989. pp. 568-576.

Waggoner, G.S. 1993. Geographic Information Systems: Proceedings of the Seventh Annual GRASS Users Conference. Technical Report NPS/NRGISD/NRTR-93/13. US Dept. of Interior.

Walling, D.E. 1996. Suspended sediment transport by rivers: a geomorphological and hydrological perspective, in: Advances in Limnology 47. Suspended Particulate Matter in Rivers and Estuaries, Proceedings of an International Symposium, Reinbek, Germany. E.Schweizerbart'sche Verlagsbuchhandlung, Stuttgart 1996. pp 1-27.

Authors

Yuri Gorokhovich, Ph.D., Head of GIS Group, Division of Water Quality and Control, New York City Department of Environmental Protection, 465 Columbus Avenue, Valhalla, NY 10595 E-mail: ygoro@valgis.dep.nyc.ny.us, Tel: (914)742-2085, FAX: (914)773-0365.