P38a complete paper |
| | Comments? | ©1996,7 William James | updated 97/03/03 | |
|
An indication that design practices are changing from event to continuous models
Abstract This paper indicates that attitudes and practices of engineers who use deterministic models in design, may be changing away from event models towards continuous models. During a recent Toronto conference on modeling the management of the impacts of urban stormwater, a survey was distributed to 124 engineers on this issue. For the group of 57 replies, which were probably biased towards design rather than planning or operational hydrology, their experience, preferences and current needs were identified. Although the sample is not large, the results indicate that, contrary to recent, official design manuals that proclaim that continuous models are neither feasible nor desirable, these practitioners clearly prefer using continuous models. Furthermore, a surprising number are experienced in both continuous simulation and in managing long-term precipitation records, the latter being considered in the literature to be an impediment to changing practice. Most replies did however indicate a need for further software utilities for data management for continuous models. These results are thought by the present writers to be important to writers of design manuals, several of which are currently known to be in production.
When computing runoff, design engineers can choose to use either event or continuous hydrological models. Event models only model discrete wet events. They are often distinguished by output that includes a single, single-valued (meaning no display of uncertainty) runoff hydrograph computed in response to a single input rainfall hyetograph, usually with a constant timestep. Frequencies of the input hyetograph are calculated from a single, simple parameter representing highly variable observed rainfall, and the frequency of the computed hydrograph is assumed to be the same. This approach ignores the dependence of computed responses on input assumptions of antecedent conditions. Since the real inter-event dry periods are not generally related mathematically to the input or design storm, the assumed start-up conditions are arbitary, and often conservative (James and Robinson, 1986). Continuous models, however, by including process code for drying, such as recovery of storages, and the soilwater deficit, model the hydrological processes between rainfall events. Continuous models typically use a dry-weather timestep of one hour or more, because of the slow speed of dry processes, and are applied to meteorological datasets up to many decades in duration (James and Robinson, 1986). Conventional hydrological design has been restricted to discrete events, because of unattractive computing costs in earlier times. However, it was argued more than a decade ago that the then emerging computer technology made it possible to supplant event modelling, and its associated design storm methodology, with more rational continuous modelling (James and Unal, 1984). This is not a new idea: thirty years ago, the Stanford Watershed Model, incorporated continuous simulation (Crawford and Linsley, 1966). Event-based hydrology and design storm methods, which are based on statistical simplifications of rainfall, have been widely criticized (Litwin and Donigian, 1978; Adams and Howard, 1985; James and Robinson, 1986; Huber et al., 1986). Only continuous simulation of long-term, historical rainfall records computes reliable statistics of flowrates and volumes. Additionally, continuous simulation is useful for the following design objectives (James and Robinson, 1986): · Compute the number and duration of flow or volume exceedances for evaluation of erosion and channel stability. · Evaluate water quality impacts on receiving waters. · Evaluate the impact of stormwater management alternatives. · Compute the performance of storage facilities. · Compute the number, duration and magnitude of diversions such as combined sewer overflows. Popular empirical, event-based models, such as those based on SCS and Rational methods, might in a specific application yield acceptable peak flowrate estimates for certain extreme storm events. But when design objectives include flow volumes, water quality, environmental or ecosystem considerations, such methods are fundamentally flawed. In event models, design storm volumes are arbitrarily linked to assumed input event durations, generally assumed to be commensurate with an estimated response time for the catchment in one of the as-was, as-is or to-be states of development. As a result, the artificial input hyetograph is composed of many combinations of rainfall intensities and time intervals (Huber et al., 1986). James and Robinson (1986) collected published arguments against continuous models, including: · Coarse time-steps. This argument evidently arose from experience with rural watershed models whose minimum timestep was one hour, a resolution time too coarse for urban catchments. [Rebuttal: HSP-F can handle timesteps down to one minute, and SWMM to one second.] · Lack of available data. [Reality: new media, such as CD-ROMs and the Internet have greatly increased access to digital data.] · Costly computing time. [This argument becomes weaker as faster, less expensive computers reach the marketplace (Kuch and James, 1993).] Also, since different rainfalls cause different pollutant washoff, transport, and kinetic interactions, design storms are inappropriate for use with water quality standards (Medina, 1987). There is also a notion that calibration of continuous models is complicated, and fundamentally different from the calibration of event models (Nix, 1994). One of the present authors has shown that this is certainly not the case for models structured in the manner of SWMM, since the dry-weather processes are precisely the same, and the interaction of dry and wet processes may be easily handled (James, 1993). Event models do not permit their users to consider environmental issues, but long-term, continuous models are necessary, for instance, to address the long-term problems associated with nonpoint source pollution (Pitt, 1994b). Litwin and Donigian (1978) state ....only continuous simulation over extended time periods can provide a reliable means for quantification and statistical representation of nonpoint pollution effects..... In further support of the case for using continuous models for attacking problems involving ecosystem concerns, Pitt (1994a) states ....the long-term aquatic life effects of urban runoff are probably more important than short-term effects associated with specific events..... Legal precedents also exist: in Scarboro Golf and Country Club vs the City of Scarborough et al., the Canadian courts preferred continuous modelling to event-based hydrology. The decision affects land developers in areas where downstream conditions are sensitive to alterations in stream channel morphology (James, 1991).
Flood hydrology and drainage system design manuals, and stormwater management planning guides, seem to be decidedly reticent about, or even biased against, continuous models. Two examples are cited below. The first is Canadian, the recent (1989) Hydrology of Floods in Canada - A Guide to Planning and Design, written under the auspices of the Canadian National Research Council Associate Committeee on Hydrology, by five editors with the help of a committee of 27; of the total only four were not associated with academics or research. Thus the guide bears a stamp of official acceptance. Divided among eleven chapters and four appendices, the 260 pages devotes about 70 pages (27%) to computational hydrology, not counting the review of statistics. Continuous simulation is covered in only about 1/4 page, in a decidedly negative manner: ....for most cases of flood analysis and design, use of continuous models is either not feasible or prohibitively expensive or both.... (Associate Committee on Hydrology, 1989). In fact it expressly warns against continuous models because of : ....1. the extensive data needed to calibrate, validate and run the model, 2. the cost of computer time required, and 3. the specialized hydrologic expertise required to calibrate them..... The second is from the United States, the Water Environment Federation and the American Society of Civil Engineers Manual of Practice: Design and Construction of Urban Stormwater Management Systems published only two years ago (1992). Written, edited and reviewed by about 125 people, the manual comprises about 750 pages divided into sixteen chapters and one appendix. Quite different in scope from the above guide, only about 90 pages (slightly more than the Canadian guide) are devoted to computational hydrology, as opposed to hydraulics, and of these, almost one page is assigned to continuous models. This manual recommends a compromise use of both types of model: ....continuous simulation is most useful for planning....single-event simulation may then be used for detailed design..... Current textbooks support this proposition (Nix, 1994). The primary reason why large official committees should project a negative view of continuous modelling apparently results from the fear of use of hydrological models that are deemed to be complicated or to demand lengthy input data (Marsalek and Sztruhar, 1994). But among users, there is a growing body of evidence that continous models are to be preferred in almost all applications. Cao et al. (1994) offer a slightly more positive view than do the above manuals: ....At present, the ever-growing computing power of desktop units does not seriously limit the use of continuous simulation...., however, ....the main limitation to wider use of this straightforward [continuous model] approach, which should otherwise be preferred, follows from the general lack of sufficiently long, continuous and detailed rainfall records..... The latter limitation is not borne out by the findings presented below in this present note. In fact it seems more likely that event models are still preferred by official committees because of their apparent simplicity. As if to emphasise this very point, in an invited address to the U.S Army Corps of Engineers Hydrologic Engineering Center, one of the present authors wrote: ....the principal argument in favour of design storm methods is design economy..... But, unfortunately, .... cheap stormwater drainage design is an avoidance of consideration of the inevitable long-term ecological impacts of that design, tantamount to a deliberate decision to remain ignorant of the impacts of the design.... (James, 1994). This section of this paper has used only two recent examples; they serve to show how soon the recommendations of a Manual or Guide may become out-dated, and sound a warning against their uninformed use. Indeed, modern rates of advance in both computational hardware and software indicate that the half-life of official guides and manuals will soon be less than the time required for the official committees to write them in the first place. Survey Of Stormwater Modellers The 1995 annual workshop and conference on Stormwater and Water Quality Management Modelling in Toronto provided an excellent opportunity to determine the practices and needs of engineer-modellers from across North America, since participants had diverse experience and professional backgrounds. Previous model-user surveys were conducted either at similar conferences or through the mail by graduate students (Ahmed, 1993, and Carvalho, 1992) and were deemed to be successful. For this survey, the purposes were twofold: 1. to provide a summary of a variety of on-going models-in-design practices and opinions, and 2. to assess the needs of users of continuous models. On February 27, 1995, 30 questionnaires were distributed at the three-day computer workshop that preceded the conference. An additional 94 questionnaires were distributed at the two-day conference on March 2, 1995. Origins of attendees of the conference were: USA 25% [NE: 3%, NW: 6%, MW: 6%, SE: 10%]; Canada 74% [ON 65%, PQ 3%, AB 3%, BC 2%]. Origins of attendees of the workshop were: USA 26% [all NE], and Canada 74% [ON 59%, PQ 12%, MB 12%, BC 6%, AB 6%]. Conference attendees were deemed on the average to probably be more experienced modellers than attendees at the workshop. Typically this annual series of conferences has attracted many participants already aware of continuous model environments, such as the Storm Water Management Model (SWMM). Because this specialty conference was promoted as a forum for the exchange of emerging technologies, responses were possibly biased towards individuals who apply, or at least are supportive of, state-of-the-art practices. In other words, since the conference was clearly not intended to support traditional, event-based hydrological models, the results might not represent all practitioners. Nevetheless, it is the absolute numbers of users of continuous models that is considered here to indicate a potentially significant trend. Additionally, results should be considered in light of recent media publicity promoting the capabilities of the Internet. As a result, potential bias may exist due to topical interests; these influences are, however, deemed to be genuinely important extraneous factors in changing design attitudes. In total, 57 of 124 surveys were completed and returned, a response rate of 46%; 33 respondents from the conference and 24 from the computer workshop. Respondents were first asked to describe their profession, area of specialization and computing environment. For this section of the survey, multiple responses were permitted, and so totals may exceed 100%. Results for this section did not vary significantly between respondents who attended the conference and those who attended the workshop. Responses indicated that, by far, engineers comprised the majority of participants; only six of fifty-seven respondents (11%) were not engineers. Of the 51 who classified themselves as engineers: · 29 were water resources engineers (57%). · 12 were environmental engineers (24%). · 7 were both water resources and environmental engineers (14%). · 3 described themselves as either municipal, development, or civil engineers (6%). Two-thirds of all respondents (67%) were employed as consultants, and nearly one-third (30%) were government employees. These figures clearly indicate that the results are biased towards engineers who are active in design, rather than exclusively planning (planners did not check off that category) or in operational hydrology. IBM-compatible personal computers were the overwhelming favourite of all participants (91%). 47 of the 57 respondents (82%) used IBM-compatible machines exclusively, and only 5 (9%) used multiple computer environments. This was similar to the conclusion of a recent survey of 200 water quality agencies in the United States (Saito et al., 1994). Responses to the further detailed technical questions were as follows: Question 1. One-third of all respondents (35%) already had access to the Internet from their work computer, and half (54%) indicated that they would like to have access. Only 2 (one each from the workshop and the conference) had no desire for Internet access (4%). Question 2. This question was intended to indicate experience with continuous models, however, it is clear that results attested more to the length of the rainfall time-series used in their model practices. One complication in the interpretation of this result is that it is likely that some respondents obtained several months of rain data, for example, yet only used selected events for their models. Nonetheless, it is surprising that half (46%) have had experience managing long time-series of rainfall, especially among those that attended the conference (61%). Several respondents noted that they had used continuous rainfall records of twenty years or more. Question 3. As discussed earlier, numerous publications have ignored or downplayed the importance of continuous simulation. The purpose of this question was to assess practitioners’ opinion on this issue: nearly two-thirds (60%) believed that continuous models are important, gainsaying traditional arguments. Furthermore, of those that expressed an opinion, 81% felt that continuous modelling was necessary. Question 4. As in question 2, most respondents (70%) were experienced with obtaining long precipitation records, especially the conference attendees (82%). The results did not indicate popular retrieval modes, but it is assumed that most had obtained their data on floppy diskettes, directly from the source. Question 5. This question was divided into two parts: in the first part, the overwhelming majority of respondents (88%) felt that they would adopt continuous simulation if access to datasets were more readily available; in the second part, respondents selected their preferred medium on which to collect data, assuming required datasets were readily available. Those choosing the Internet (83%) outnumbered those who chose CD-ROM (17%) by about five to one. Question 6. Respondents were generally willing to wait for data to be downloaded from the Internet. Immediate retrieval (less than one hour) was important to only one-quarter of respondents (25%). Of those that replied, three-quarters (74%) were prepared to wait at least one day for the data to be retrieved. Question 7. This question was intended to indicate the respondent’s experience with continuous models. On the first page of the questionnaire, continuous modelling was defined in question 3 as the simulation of both dry and wet weather processes for periods of ten years or more. After interviewing a number of respondents, the present authors were not confident that continuous modelling was consistently defined by all participants. For example, some respondents may have regarded multiple wet events as continuous, rather than as just a sequence of separated wet events. Perhaps the definition should have been more explicit and also repeated again in question 7. Nevertheless, it is surprising that the majority (57%) indicated experience with calibrating a continuous model, especially among conference attendees (73%). Even among those that did not have experience with continuous model calibration, 54% believed that they were capable of successfully calibrating a continuous model. The last part of the question showed strong demand (85%) for software utilities for model calibration. All workshop respondents (100%) were in favour of the provision of such software. Question 8. The final question sought preferences for types of Internet services. Respondents were asked to assign rankings from most important (number one rank) to least important (number six). Results indicated support for a wide range of desirable features. There was a slight preference for technical user support (30%), which is also borne out by the current popularity of the SWMM-Users electronic mailing list operated by the University of Guelph. However, all listed Internet services were generally approximately equally desired. Significant Conclusions Results of this survey produced a useful number of significant findings of interest to writers of future manuals of practice: · Generally, respondents agree that continuous models are preferable, if not necessary. This contradicts numerous design manuals and planning guides that still argue that continuous modelling is neither feasible nor desirable. · A surprising number of modellers are experienced in both continuous simulation and in managing long precipitation records. · By far the majority of respondents would adopt continuous modelling if datasets were more readily accessible. · Most modellers believe that the Internet is an important resource. · Most modellers would prefer to access such data through the Internet, rather than by purchasing datasets on CD-ROM. · Most modellers would be willing to wait one day to retrieve available data. · Modellers expressed a need for many relevant Internet services. There is a slight preference for technical user support. · Development of utilities that assist continuous modelling tasks is desirable, and perhaps necessary. · Applications should be developed for IBM-compatible personal computer environments. References Adams, B.J., and Howard, C.D.D. (1985); “The Pathology Of Design Storms”; University of Toronto, Department of Civil Engineering, Publication No. 85-03; 32 pages. Ahmed, F.H. (1992); “The Utility Of An Object-Oriented Software Tool For Planning Management Practices For Urban Stormwater In Ontario”; M.Sc. Thesis, University of Guelph; 183 pages. Associate Committee on Hydrology (1989); “Hydrology Of Floods In Canada: A Guide To Planning And Design”; W.E. Watt et al. (Editors), National Research Council of Canada; ISBN 0-6660-12876-4; 245 pages. Cao, C., Piga, E., and Saba, A. (1994); “A Continuous Simulation Approach To Design Storm Calibration”; Water Science and Technology, Journal of the International Association on Water Quality, Vol. 29(1-2); ISSN 0273-1223; pages 11-20. Carvalho, L.M.J. (1992); “Computer-Integrated Crop-Management System (CICMS) With On-Line Weather Station”; M.Sc. Thesis, University of Guelph; 119 pages. Crawford, N.H., and Linsley, R.K. (1966); "Digital Simulation In Hydrology: Stanford Watershed Model IV"; Stanford University, Department of Civil Engineering, Technical Report No. 39; 210 pages. Huber, W.C., Cunningham, B.A., and Cavender, K.A. (1986); “Continuous SWMM Modelling For Selection Of Design Events”; In Urban Drainage Modelling, C. Maksimovic and M. Radojkovic (Editors), Pergamon Press; ISBN 0-08-032558-0; pages 379-390. James, W. (1991); “Channel And Habitat Change Downstream Of Urbanization”; Engineering Foundation Conference on Urban Runoff and Receiving Streams, Crested Butte, CO; in press. James, W. (1993.) Rules for responsible modeling. Booklet pub by CHI as report # R184. ca 100 pp. Available form the authors. James, W. (1994); “On Reasons Why Traditional Single-Valued, Single-Event Hydrology (Typical Design Storm Methodology) Has Become Simple-Minded, Dishonest, and Unethical”; US Army Corps of Engineers, Workshop on Urban Hydrology and Hydraulics, Davis, CA; in press. James, W., and Robinson, M.A. (1986); “Continuous Deterministic Urban Runoff Modelling”; In Urban Drainage Modelling, C. Maksimovic and M. Radojkovic (Editors), Pergamon Press; ISBN 0-08-032558-0; pages 347-378. James, W., and Unal, A. (1984); “Evolving Data Processing Environment For Computational Hydraulic Systems”; Canadian Journal of Civil Engineering, Vol. 11(2); ISSN 0315-1468; pages 187-195. Kuch, A.W., and James, W. (1993); “A Sensitivity Analysis Tool For Three Generation Continuous Urban Stormwater Quality Management Modelling”; American Society of Civil Engineers, Runoff Quantity and Quality Model Group Conference, Reno, NV; in press. Litwin, Y.J., and Donigian Jr., A.S. (1978); “Continuous Simulation Of Nonpoint Pollution”; Journal of the Water Pollution Control Federation, Vol. 50(10); pages 2348-2361. Marsalek, J., and Sztruhar, D. (1994); “Urban Drainage: Review of Contemporary Approaches”; Water Science and Technology, Journal of the International Association on Water Quality, Vol. 29(1-2); ISSN 0273-1223; pages 1-10. Medina Jr., M.A. (1987); “Water Quality Modelling And The Regulation Environment”; In Pollution Control Planning, W. James (Editor), Computational Hydraulics Inc.; pages 176-191. Nix, S. J. (1994). Urban stormwater modeling and simulation Lewis Publishers. ISBN 0-87371-527-6. ca 220pp. Pitt, R. (1986); “Runoff Controls In Wisconsin’s Priority Watersheds”; In Urban Runoff Quality: Impact And Quality Enhancement Technology, B. Urbonas and L.A. Roesner (Editors), American Society of Civil Engineers; ISBN 0-87262-577-X; pages 290-313. Pitt, R. (1994a); “Effects Of Urban Runoff Controls On Aquatic Biota”; In Handbook Of Ecotoxicology, Lewis Publishers Inc.; ISBN 0-873715-85-3; Chap. 30 Pitt, R. (1994b); seminar notes, Stormwater Quality Management Workshop, Guelph, ON; Lectures 2 and 9. Water Environment Federation, and American Society of Civil Engineers. (1992). Design and construction of urban stormwater management systems. Pub by ASCE and WEF. ISBN 0-87262-855-8. ca 750pp. |