Click for next page ( 158


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 157
- TCRP [1-19 Final Report 6.0 RESEARCH AND DEVELOPMENT NEEDS 6.! Introduction This chapter presents a research and development (R&D) program for the short- to medium-term (e.g.5 over the next five years) which would result in a significantly- improved state-of-practice in the use of integrated urban models (as well as considerably more widespread use of these models in operational urban planning applications) by the end of this time period. The proposed R&D program follows directly from the specification of an "ideal" or "ultimate" integrated modeling system in Chapter 4, along with the assessment in Chapter 5 of operational or semi-operational models relative to this ideal model. The R&D program is also designed to be maximally consistent with and supportive ofthe guidelines for model implementation and use which are presented in the document Guidelines for Implementation and Use of integrated transportation land-use models [Miller' et al. ~ 1 998]. Section 6.2 presents several basic assumptions which underlie and guide the development of the proposed R&D program. Section 6.3 briefly describes our vision of hove integrated urban models should be implemented and/or upgraded within operational planning departments. This topic is dealt with in detail In the Guidelines report mentioned above, but is introduced here as a primary motivation for several key components of the proposed R&D pro~rarn. Section 6.4 summarizes the recommendations for land-use modeling R&D which were generated at the 1 995 TMIP Land Use Modeling Conference and which provide an important starting point for our recommendations. Section 6.5 then presents the R&D program In detail. 6.2 Starting Premise The R&D program outlined in Section 4 is based on a number of assumptions, both about integrated urban models and about desirable features of an R&D program. These assumptions are listed in point form in Table 6.1 and are discussed in greater detail belong-. 1. Operanona1 integrated urban models are essential to the urban planning process if the spirit and the letter ofthe legislated planning process (TEA-21, C.AAA, etc.) are to be addressed. Over and above legislative requirements. it simply makes practical/theoretical sense: transportation and land-use are fundamentally intertwined and should be analyzed in a holistic fashion. . 2. The transportation - land-use interaction is complex, but vie have achieved reasonable insight into it over more than forty years of research. Rather tears use this complexity as an excuse not to engage in integrated urban modeling, this complexity is a major rationale for employing such models, for at least two reasons: 1 - - - 1 ~D

OCR for page 157
TCRP H-17 Final Report Table 6.! R&D Program Assumptions I. Operational integrated urban models are essential to the urban planning process. 2. The transportation - land-use interaction is complex. This complexity is a major rationale for employing integrated urban models. The current and historical relative lack of use of integrated models is due to many factors. A properly designed R&D program can overcome these factors. 4. 6. Given our current technical capabilities, it is possible to describe an "ideal" or "desired" integrated urban modeling system which should be achievable with a concerted but reasonable R&D effort. An evolutionary approach is generally a sound one, both for an overall, national R&D program and for an Individual MPO designing its own mode] Improvement program. At He same time, some desired end states can not be achieved by following a purely evolutionary track. Rather, Hey require a "leap" off the trend line in a new . . Erection. The combination of the last few points Implies the need for at least a two-path R&D process, with one path involving the evolutionary development of current and emerging models, while the over, parallel effort involves the development of substantively new methods and procedures. S. The quality of the transportation sub-mode} within He overall Integrated urban mode} is critical to the policy sensitivity of the modeling system. 9. The R&D program should be "strategically focussed" but `' tactically broad." 10. R&D activities should be subject to rigorous professional scrutiny, including detailed documentation of results; use of peer review panels; conformance to agreed upon minimum performance standards and capabilities; and, above all else, ~ sound R&D program management process. - I:4

OCR for page 157
TCRP H-1 ~ Final Report i) - ~c; , ~d~ Integrated models provide We only mecharusm for properly exploring the full ramifications of any given policy (transportation or land-use) on urbar1 systems. Without comprehensive integrated models longer-r~ and/or secondary responses `~-ill inevitably be ignored. These may well be larger in magnitude and/or of opposite sign to shorter-run/initial responses captured by more partial models. ii) Many gaps in our understanding of urban processes derive from difficulties in adopting a holistic research framework. that is. most research is very partial in nature due to data limitations, etc. Comprehensive. integrated urban models hold the promise of providing "experimental platforms" or "computer laboratories" within which wee can experiment with hypotheses about urban processes in much more extensive ways tMiller and Salving, ~ 9983. While computer simulations can never totally replace real-world empirical observations. they can provide considerable detailed testing of assumptions. the results of which can then be "grounded" against real-~~orld experience. 3. The current and historical relative lack of use of integrated models (particularly in the United States) is due to many factors. Arnona the more important of these obstacles to the use of integrated urban models in operational planning contexts are the following: i) Disappointing experiences with earIv integrated modeling attempts. Pioneering land-use modeling efforts in the U.S. in the 1960s suffered many setbacks and resulted in a widespread disenchantment with these models Wee, ~ 9733. Dom which, to some degree' the profession has never fib- recovered. ii) Fragmentation of the planning process. Different departments. types of professionals. etc. typically deal with "land-use problems" versus "transportation problems," resulting in a failure to think holistically- about urbar~ systems. iii) Failure to acknowledge the transportation - Iand-use connection. Over and above the institutional fragmentation issue listed above. many analysts, planners and decision-makers either do not believe that a significant transportation - land-use interaction exists, or that it is so Tong-term that it is irrele~ art to normal transportation planning problems, anchor that land-use is not amenable to significant policy control (usually based on ideological principles). iv) Tendency to treat land-use forecasts as inputs to travel demand modeling. There are difficulties inherent in generating demographic and socio-economic forecasts for small zones. These include: reliability of existing data sources difficulty in allocating fixture land- use according to existing zoning. reliability of available data sources, the statistical reliability of forecasting arid allocating numbers among very small areas, and difficulty in conducting mear~ingfi~l sensitivi~v analyses with very small numbers. In some cases, such difficulties 1,,,

OCR for page 157
TCRP H-17 Final Report have resulted in a lack of credibility in the generated numbers -- a problem complicated by the potential political implications of fixture land-use. Minimal legal or institutional incentives to develop integrated models. In contrast, Gavel demand forecasting models have been de facto requirements for MPO planning for several ~ ~ ~ ~ . ~ ~ . ~ ! ~ _ ~ ~ - decades by virtue of me mandated JO transportation planning process etc. ISTE^N TEA 21 and COCA have since fo~alized this requirement. ~ _ , No such treatment exists for integrated models. Although TEA-21 and CAAA both require the coordination of transportation plans with land-use plans considerable leeway is provided in the implementation of this requirement -- in part one expects because of the lack of Federal jurisdiction in local land-use planning. vi) Lack of resources within MPOs to support integrated urban mode! development and use. This includes appropriately trained staff. computational resources and appropriate databases for model development. Ultimately, these constraints reflect a lack of funding of arid institutional support for integrated urban modeling at the local, state and federal levels, most probably due to a combination of the factors listed above. vii) Lack of support for integrated modeling R&D. Integrated urban modeling has received little to no support over the years from hinders of either basic or applied research. again undoubtedly due to the factors listed above. The result is a current modeling state of practice/art which is not wholly adequate for current and emergent planning needs and which is senousIx underdeveloped relative to what our current technical capabilities could support. The first five of these issues have been dealt with at length in Canons waYs in previous chanters. , ~. . . . . . . . ~ , ~ 1 ne purpose OI INS Cnapler IS lo explore in some aem1 what can be done about items vi and vii. 4. Given our current technical capabilities (computer hardware and software, datasets and data collection capabilities; modeling techniques; theoretical understanding of behavioral processes; etc.) as shown in Chapter 4, it is possible to describe an "ideal" or "desired" integrated urban modeling system which should be achievable with a concerted but reasonable R&D effort. Drawing from the comparison of currently operational models with this ideal model (Chapter 5~: i) All currently; operational models fall considerably show of the ideal model. Areas of significant shortfall include: - ii) At the same time current models individually and collectively display many strengths and generally provide a solid basis for further evolutionary improvements. Strengths include: - 136

OCR for page 157
TCRP H-19 Final Report iii) Despite the scope for significant evolutionary development among existing models, a "new generation" of integrated models w~11 need to be developed in order to fiully achieve the ideal model. . We move into the future, more often than not, in evolutionary steps. by which new, improved methods for doing things are developed as incremental improx ements of existing methods. This evolutionary approach is a sound one. both for an overall, national R&D program and for an individual MPO designing its own model improvement program. 6. At the same time. some desired end states can not be achieved by following a purely evolutionary track. Rather, they require a "leap" off the trend line in a new- direction. As noted above, such as the case appears to exist with the desired "ideals' modeling system: it will require the development of some significantly new- model strucn~res which Will not necessariT "organically emerge" out oftoday's models. 7. The combination of the last few- points implies the need for at least a two-path R&D process, with one path involving the evolutionary development of current and emerging models, while the other, parallel effort involves the development of substantively new- methods and procedures. The first involves working closely with a range of local planning agencies and model developers. typically within operational planning settings to implement and improve current moclels. The second includes a broader group of researchers, interacting with planners and mode] developers, to investigate fundamental issues in integrated urbar~ modeling. in the first instance. this second ~. .. . . .. ^. .. .. . . stream ot activity must be oti-llne , that Is, not tied to Implementation within operational planning settings. In the longer run. the two streams are expected to merge as the results of the off-line, more basic research activities are brought into operational practice. 8. Not surpns~ngly, the quality of the transportation sub-mode} within the overall integrated urban mode! is critical to the policy sensitivity of the modeling ss stem. This leads to two, related implications. First at the local planning agency level, the first step towards the development of a strong integrated modeling capability is to develop a strong travel demand modeling capability; indeed, there is little point in considering integrated models if the local travel demand mode] is not "in good shape". Second. at the national level while TMIP has an explicit (and currently underfunded!) "Track E" dealing explicitly with land-use modeling the entire TMIP program is generally supportive of the development of improved integrated models. Further. in designing "Track E" (or other land-use/integrated model research activities)- a holistic view of the overall modeling R&D effort and needs must be maintained. Any integrated urban model R&D program should be "strategically focussed" on achieving in ^^ . ~ a cost-ei~ect~ve way maximum Improvement In our operational modeling capabilities. At the same time, however, it should also be "tactically broad" in its implementation. That is. it should not begin by picking "winners and losers" among existing models in terms of which are - I :7

OCR for page 157
TCRP H-17 Final Report deserving of evolutionary support and which are not. As discussed in detail in Chapter 5. there are many models currently available, each of which brings various strengths arid weaknesses to the problem. and none of which necessani`- clearly dominates the others. TUeaTiy, the "marketplace" (both commercial and intellectual) should guide the evolutionary selection and adaptation of these models. rather than a strictly bureaucratic process. Thus, an R&D program should encourage substantive progress in our generic modeling capabilities, regardless of the particular modeling packable employed in a specific case study. fit the same time. it is recognized that government most likely is the only realistic body that is capable of sponsoring promoting. coordinating and sustaining the effort that ~ ould be required. Similarly. longer-run, more basic research activities should be broad-based involving a range of researchers, methods and concepts so as to maxirruze the possibility- ofthe "best" new methods being discovered and, eventually. adopted into practice. O. R&D activities should be subject to rigorous professional scrutiny both to maintain the highest technical standards and to ensure that products are being generated in a timers and useful fashion. This implies among other actions. detailed documentation of models, datasets. etc.. use of peer review- panels (both for selection of contract/grant winners and for on-going monitoring of R&D activities); conformance to agreed upon minimum performance standards and capabilities. and, above all else, a sold R&D program management process. 6.3 Evolutionary Development As is discussed in more detail in Miller, er al. tI 9983. the state of a planning agency's capabilities with respect to transportation and land-use modeling can be summarized in a two-dimensional ma~x such as shown in Figure 6. I. In this matrix, row s correspond to different levels of land-use modeling capability. While a continuum of levels obviously exists. five significant land-use modeling "states" or capability levels have been explicitly identified in Figure 6.~: it. L, ' . r ~ L: . None. The planning agency does not in any way model or forecast land-use. Zonal population and employment data. if required for travel demand modeling Purposes. are obtained from other agencies. ~ .L ~. Acti, ity + judgement. Activity- levels are estimated. perhaps on a scenario basis and systematically allocated to zones. Methods are often spreadsheet-based and usually involve considerable professional judgement. Non-market-based land allocation model. A formal land-use model is used, but this model is not market-based (i.e.. it does not include endogenous price signals or art explicit supply process). ITEUP is an example of this type of model. - I:8

OCR for page 157
TCRP H-12 Final Report Figure 6.1 A Taxonomv of Transportation - Land-Use ModeUng Capabilities \ Travel ~(~) |(~) |~) |(~) \ Demand No transit / Transit / Logit / Landde! mode s lit n logit (24 hr) peak-period A tivltvv-based Use~ | P | | assignment | Model_ \ - | ~_j ~: 9.S .~ef-- ~:~:~ ~:l ~fr~, ~ = l l None l l I ~;/ | . | I I , ~ a ~' ~' - Activity-+ I I ~ | ~| Judgement ~ - ~ ~.~ ;~ - >~; - ~-. ~ < f ~-~ :~-~x' -: ~.4;~ :.~i;.'--~. . ~.:-.~;. ~. ~.-: : f ~1 1 1 - 1 ~1 Non-market-based | l l . ~^: | ~: | Land Allocation I ~ ~ ~ -~. . 1 ~ ~ ~ .1 ~| ~l 1 ~) l I = ~| _ :; ag - t ~ ~ ~ ~ 1 -Logit allocation I I I ~ ~I 11 - with price signals | I I ~: Sh~ l ......... . ~N,, ~,., : -~:- . Full, ... . 1^ id M del j integrated market- | I I - A~ Lol term go~d~ based model l l l |~. : ~~ : :~. :~-; - ;-- FirstPath ~ CAdvanced' Path - 139

OCR for page 157
TCRP H-l ~ Final Report L4. Land allocation with price signals. A formal model is used. which includes endogenous price signals, but does not include a full demand-supply market process representation. This type of mode] is discmsed further below L:. Fully integrated market-based model. A filet system of market-based supply-demand relationships with explicit prices is used. MEPLAN, TRANUS, MUSSA, NYMTC- LUM and UrbanS~m are all current examples ofthis modeling approach as is the "ideal" model. Similarly columns In Figure 6.~ represent different levels oftravel demand modeling capability, of which four are explicitly shown: TI. No transit or mode split model. Only roads and auto travel are modeled. T2. TT4. Transit with simplified (non-Iogit) mode split Transit is represented in the modeling system, but modal split is performed using simplified (non-Iogit-based) methods. Assignments are usually based on daily rather there peak-penod volumes. usually using some form of capacity-restrained assignment. The modeling system is usually not iterated to achieve internal consistency. Lo~it mode split; peak-period assignment. A disaggregate To~it or nested logit mode choice model is used. Peak-perioc} equilibnum assignment is used. The system is iterated to achieve internal consistency. This level of travel demand modeling capability essentially defines the current "best practice" for medium to large cities. Activitv-based methods. This is an emerging approach -- i.e.. it goes beyond current best practice. The traditional four-stage process is replaced to varying degrees by activity-based (as opposed to trip-based) models. Portiar~d, Oregon is the most advanced along this path of model development in the U.S., by far' with a few other cities experimenting to varying degrees. The thrust of "Track D" of TMIP is to move U.S. modeling practice towards this 'next generation. modeling approach. Each cell in the Figure 6.! matrix therefore represents a land-use/transportation modeling combination. Virtually all cities can be categorized as being currently contained within one of the 20 cells in this matrix. Points to note about this matrix include the following: There are six desirable incremental capability levels as indicated by the arrowheads on Figure 6. I: I. Logit / peak-period assignment mode! / no land-use mode! (T3 Ll). - 140

OCR for page 157
TCRP H-12 Final Report . 2. Logit / peak-per~od assignment model / activity ~ Amusement land-use model (T3, L2). 3. Logit / peak-penod assignment model ~ posit allocation land-use mode} with price signals (T3, L41. This represents a short-term goal. 4. Logit / peak-penod assiar~ment mode] / fully integrated market-based land-use mode] (T3, L51. S. Activity-based travel demand model / posit allocation land-use mode] with price-signals (T4, L41. 6. Activity-based travel debark mode} / Shiv integrated market-based land-use model (T4. L51. This represents the long-tern~ goal of the ideal integrated model. "Appropnate" combinations of transportation and land-use modeling capabilities tend to lie along the major diagonal (i.e. from upper left to loner rights. That is, it makes little sense to combine a very complex land-use mode! with a very crude travel demand model. or vice versa (although there are cities that currently have sophisticated travel demand models with little or no land-use modeling capability). The ' appropriate' cell for a given city- obviously depends on a number of factors. including the city size. the nature and extent of its transit system, the extent to which it is interested in pursuing land-use as a policy tool, etc. The arrows in Figure 6.! depict logical / recommended ;;development paths.' or trajectories for urban areas desinng to upgrade their modeling capabilities. where the base of the arrow- represents a current capability (e.g., Ll,TI: no Iand-use model. no transit representation) and the tip of the arrow represents a logical incremental upgrade on that capability (e.g., Ll,T3: no land-use model. 'best practicer travel demand model). In specifying these development paths, several principlesiassumptions were invoked. These include the following: It is practically not possible to develop a credible land-use mode] (at least beyond the spreadsheet accounting / jud~ement level) without having a well developed travel demand model. Thus, an urbar1 area s first pnont~, should be to ensure that it has an operational 'best practice' travel debark mode] in place before taking on any significant land-use modeling tasks. ~ 8 8 Caveats may exist here with respect to small urban areas, within which transit services may be minimal or even non-existent. In such instances, integrated models without a significant transit / mode split modeling component may be used. - 141

OCR for page 157
TCRP H-12 Final Report Given the comments in the previous section concerning the attractions of an incremental, evolutionary approach to model development, onion Short jumps of one or two cells are generally shown in Figure 6. T. Thus even if the Tong-term objective is to move from the upper left corner ofthe matrix (i.e., virally; no modeling capability) all the way to the Tower right (i.e.. the ideal models one should accomplish this in several stages. Each stage is defined by a clear operational product which represents a significant improvement over the previous modeling state. ~ r ,~ It also provides a solid base (in terms of data, operational expenence, etc.: tor the next step forward. Figure 6. ~ shows a short-term goal of the L4 T3 model combination. This is seen as a realistic objective for advancing the state-of-the-practice In Integrated modeling in advance of the Tong-term realization of the ideal model, for two reasons: It reflects a travel demand mode! structure (Iogit / peak-penod assignment) that is relatively well advanced which, therefore can draw- from a wide boa>; of literature and practical installations. Therefore, it is readily achievable. The class of land-use mode} (land allocation with price signals) represents an improvement over non-market based land allocation models, but does not require a complete, filthy integrated market-based model. Although. as discussed below this class of land-use model does not yet exist operationally, it represents an attractive short-term advancement towards the ideal On-term model. The ordering of movement is important, since -- as noted above -- advancements in both types of models are linked and therefore must be coordinated. Two advanced paths are noted: Finally, Figure 6.! shows 'first' and 'advanced' paths of development. Movement from minimal travel demand / no land-use modeling Aft. TI or L1, T2) capabilities is recommendedfirst towards improvements in travel demand modeling (T3), then in land-use modeling (i.e., horizontally then vertically). But it also is practical to augment this at the same time with a correspor~ding improvement towards minimal land-use modeling capability (~21. More important is the sequence of movements from the short-tenn goal (~4, T3) to the Tong- term goal (ES, Ted. Here, the recommendation is to advanceirst towards fully integrated market-based land-use mode] (L S), then towards ~ activit:~-based travel dem=d mode] (T41. This sequence reflects the more advanced operational status of market-based land-use models, compared with activity-based travel demand models. How-ever, an acceptable alternate treatment is the reverse order: T4. then ES. - 149

OCR for page 157
TCRP H-12 Final Report HOW to undertake these venous model development paths is discussed in detail in Miller et al. [1 998]. The primary purpose for introducing this figure here is to identify important R&D activities associated web this process. There are at least three such important activities implicit in this figure: T. Financial and technical support for cities wishing to undertake these types of mode! upgrades, especially with respect to improving their land-use modeling capabilities. Financial arid technical resource constraints have been identified above as major obstacles to improving integrated urban modeling capabilities. Mechanisms for overcoming these obstacles are discussed in detail in Section 6.5. - Considerable basic and applied research is required to turn the ideal mode! described in Chapter 4 into an operational system. This R&D effort is described in detail in Section 6.5. 3. The "~4: land allocation with price signals" type of mode} does not currently exist as an operational model. It represents a "~apt' between current non-market models such as IThUP arid "fully elaborated" market models such as MEp~T. MUSSA et al. We believe that this is a significant Pap in that it represents ~ important and attractive intermediate canabilitv. which mat well Drove useful in many urban areas" either as an end point for 1 ~ t ~ . ~ - t~ 1 t t~ r 1 1 _ _ x ~ it_ _ model development (l.e. good enough for local ourDoses) or as an 1mDonant sten In tne . . . ~ ,. evolution towards a ruti market-based system. This is elaborated in Appendix B. which provides a more detailed discussion of the proposed model. Section 6.5 discusses its place within the overall R&D program. 6.4 Summary Of 1995 Dallas TMIP Conference Recommendations As noted in Chapter 2, TMIP sponsored a conference in Dallas, Texas on land-use modeling in February. ~ 993. This three-dax- workshop-based conference was attended by 80 professionals drawn ~ , ~ , ~ . . . ~ ~ . ~ ~ . - ~ .~ ~ .1 ,t ~ ~ ~ 1 ^~ . 1 ~ _ 1 _ _ widely trom academia. government anct consulting, primarily Irom v~trun one umlea Stales OUI alSO from abroad. It generated a wide range of recommended actions to improve integrated urban modeling. These are summarized in Table 6.2.~9 Points to note from this table include the following: I. The Dallas Conference recommendations identify the need for both evolutionary improvement in existing methods arid the development of new models. ~9 For full documentation of the conference, including workshop reports and plenary papers, see Shunk, et al. [1 995]. - ]4:

OCR for page 157
TCRP H-19 Final Report 1 1 Table 6.3 Criteria For Selection of Case Study Sites/Models Site Selection She of metropolitan area (population) Economic/population growth rate, for example: mature, sIow-growing O newer, fast-growing CBD focus, for example: O single, well defined central business district o multiple central business districts Urban form, for example: o compact/constrained urban form o uncons~rainedisprawl Land-use controls, for example: o o well defined zoning re~lations/ordnances minimal zoning reaulations/ordnances Level of transit service, for example: o o o .- minimal/policy service only well-established/no rapid transit well-es~blished/rapid transit Position Within the Land-Use/Transportation Mode! Taxonomy . Conditions Travel demand modeling activity (between T! and T4 (per Figure 3. 1~) Land-use modeling activity (between Let and L5 (per Figure 3. l)) Data availability: o o o Travel demand profiles (recent origin-destination surveys, etc.) Land-use data (population, employment, zoning, development. etc.) Land market data (real estate transaction value, development activity, etc.) Platfonns/tools: 0 Type of integrated model used (if any) 0 Data platform (for example, GIS, data base manager, etc.) 0 Level of disaggregation of data In-house modeling capabilities (resources within the MPO) - ~>o

OCR for page 157
TCRP H-19 Final Report An important pan of each case study must be the development of new data bases or the updating of existing sources. These must include the "standard socio- economic/demo~craphic and travel data bases (such as population land-use, travel O-D, etch. However, special attention also must be given to the development of other, traditionally problematic data that are essential to the development ofthe more advanced Arouse models, notably, employment price data firm location choice. etc. Taking into account the need to ensure adequate applicability of the results. as defined by the criteria listed In Table 6.3 case studies also must be awarded to modeler-MPO partnerships on an open competitive basis. Thus, modeler-MPO partnerships must meet the following . . momma requirements: o demonstrated capabilities of the modeler to undertake the proposed implementation; o demonstrated capability and interest of the MPO to support the mode] implementation and to maintain and use the model; o demonstrated capability of the modeler and MPO to work together; o the technical quality' of the proposed mode] implementation (the proposed model should meet or, preferably exceed art explicit set of minimum technical standards); and o the potential for innovation within the mode] implementation (extensions of capabilities of existing models; introduction of new modeling methods Chain an operational setting; etc.~. Case studies should be selected based on recommendations from an independent peer review panel. This panel would be directed by, or be part of, the program management group. The pane] should also review- the mode! implementation and testing results in each case. specify the "starld=d tests" to be performed, and undertake the cross-comparisons of standard test results obtained Mom the case studies. The role and structure of the peer review- panel. and its relationship to the management ofthe R&D program. are elaborated in Sub-section 6.5.7. Advantages of the proposed case study pro Pram include: it directly improves the state of practice in many locations it provides operational experience which can be extrapolated elsewhere; it is a cost-effective means for controlled experimentation; and it provides a practical. direct way to improve databases for bow operations and research. 6.5.3 Development of "Next Generation" Models The "ideal" or "next generation" model described in Chapter 4 wall not simply "organically evolve" out of the current set of operational models. As with TRANSIMS (and other "next generation", activity-based travel modeling systems currently emerging), the next generation of Is}

OCR for page 157
TCRP H-12 Final Report integrated urban models must be developed through an explicit dedicated and aggressive R&D effort, if their are to be developed in a timely fashion. Indeed, given their experimental nature and the level of effort required to develop and implement them. it ~7ill only be through a concerted, funded R&D program that such models can be developed at all. since this effort clearly lies well beyond the capabilities and resources of virtually all local (and even state) planning agencies. Such a R&D effort should consist of at least the following three sub-components: 1. basic'-esearch addressing fundamental methodological and behavioral questions which need to be resolved if next generation models are to be successfully developed and deployed; 2. development of mo&teiing system prototypes, and 3. implementation and testing of models in operational settings. Each of these steps is critical to the eventual successful implementation of next generation models in operational planning applications. Of these, basic research component is often the most difficult for agencies to support, given that they are anxious to get on' with the development of working models. As a result, there has been a chronic tendency; in the transportation field to rush into model development. The inevitable result is that we make assumptions very early in the model design process about what modeling methods to use (usually the "tried and true"), about how- to represent behavior (usually in a simplified manner to reduce model complexity). and about what data to use (usually what are available, which are not necessarily what are needed). The net result is usually a very conservative approach to model development, focussed on generating a working model within a fixed period of time. Such an approach is not conducive to innovation, to testing he potheses and to developing fundamentally new approaches to modeling urban processes. A potentially long list of basic research topics exists. Example topics include: microsimulation of land markets; demographic synthesis and updating (especially household formation and evolution); supply-side modeling (e.g., developer behavior); _ ~ a, micro-scale urban modeling; linkages to activity/trave] models; ~.. . . . . . . goods movement; linkages hetvv-een regional macroeconomics =d scatiallv micro processes: deallOg With dynamlCS (Short- VS. long-mn processes; feedback; reaming, "optimal", time steps); and determination of an "optimal" level of spatial resolution. - I32

OCR for page 157
TCRP H-19 Final Report The division between model prototyping and mode} implementation can be somewhat arbitrary . The argument however is that fundamentally new- models do need a carefi~x staged implementation process. This begins with smaller-scale testings quite possibly of mode] sub- syTstems, before filll-scale implementation is attempted. In particular, prototyping allows experience to be gained with the model. sub-systems to be tested. and mode! weaknesses to be discovered and corrected. prior to "locking into" the fi~-scale, implemented version of the model. The TRANSIMS model development process provides a good example of this approach. Here. both the Dallas and Portiar~d Oregon case studies are providing important intermediate tests ofthe TRANSIMS concept prior to the anticipated hill deployment' of the model. This R&D pro gram should be strategically- focussed. it should have as its explicit coal the development of next generation models which approach the ideal mode] described in Chapter 4 as closely as possible and it should only support projects which clearIx and substantively contribute in a coordinated may towards achieving this coal. At the same time. the pro gram should be "tactical!`; broad" in that it does not prejudge the methods or theories which will "best" SUDDort next . . . .. . generation models and that it Is open to a diversity of researchers and research organizations. 6.5.4 Other Research end Analysis In addition to the formal modeling tasks associated with development and implementation of current and next generation models discussed in the previous two sub-sections a number of other research tasks which are directly supportive of improved transportation - land-use analysis and planning have been identified, both within this project and by the Berkeley Research Needs conference. Examples include the following: Case studies documenting the evolutionary development of improved transportation - iand- use modeling methods within MPOs. As MPOs follow- an evolutionary path through Figure 6. I, their experiences can be documented in order to develop best practice guidelines for other urban areas wishing to undertake similar model improvements. As opposed to the actions discussed in Section 6.5.2, which focussed on individual models and their implementation/improv-ement, this component focuses on the process of model capability upgrading within an urban area, regardless of the mode} used. Development and implementation of the logit, price-signal-based land allocation model described in Appendix B. Development of various simple tools and procedures for use in situations in which formal integrated urban models are not available (e.~., small urban areas or as an interim measure in larger urban areas which currently do not have a significant integrated urban modeling capability). I,,

OCR for page 157
TCRP H-19 Final Report Development of recommended procedures for developing employment databases in urban areas where such data currently do not exist (a major obstacle in many urban areas to model development -- and planning analysis in general). Review of lard-use and transportation policies, Heir effectiveness and the processes used to analyze them, as well as the relationship between policies and modeling (i.e. between qualitative and quantitative analysis). For example, policies defining urban growth boundanes, road pacing etc., clearly provide the context within which integrated models are developed and applied. As clearly, integrated models would have important roles in simulating the impacts of alternate policies. etc. 6.5.5 Training and Information Dissemination A major barrier to the implementation of integrated urban models is a lack of trained staff in MPOs who can properly use these complex models. In this case, "trained" not only- means experience with the mechanics of running a given piece of software, but' much more important, having sufficient technical understanding of the behavioral and methodological foundations of the modeling system so that they apply appropriate jud~ement in operating the mode! and interpreting its results. Such training can be supported by a number of actions, including: . . . Mounting short training courses and seminars. These could focus in the first instance on macroeconomic theory of market behavior random utility models of human decision-making, transportation - land-use interactions etc. They could also provide an overview of integrated urban modeling practice (current and emerging), as well as the relevance and benefits of applying integrated models to practical situations. Development of "best practices" manuals case study reports arid other aids for self-learning and reference. Development of many on sketch planning methods for integrated land-use - transportation analysis, perhaps along the approach of a ; quick-response' system. Literature reviews on selected topics not dealt with elsewhere. Disseminating the documented results of the implementation case studies and other R&D efforts described in the previous sub-sections. Creating an "integrated modelers' users group" for the exchange of information. problem solutions, etc. - 154

OCR for page 157
TCRP H-12 Final Report Promotion of 'special interest groups' among parallel/related professional organizations, such as the American Planning Association etc. In addition to developing arid mounting the training coursers), an important component of this activity should be the development arid maintenance of a well advertized web site which would provide planners with access to a centralized library of information concerning integrated urban modeling and which would also provide a central contact point for the proposed users group. Information contained within the web site could include: all reports and other documentation generated by the R&D program described above; all manuals and other training materials, documentation of existing models, and literature reviews, annotated bibliographies etc. 6.5.6 Database AssembIv and Management Data limitations may well be the single most often cited obstacle to the development of operational integrated urban models. as well as to research efforts in this area. One of the single biggest contributions which a coordinated R&D effort can make is to improve the national database for transportation - Arouse analysis and modeling. This can be achieved through three major . . . activities. The first involves the development of a centralized data library containing well documented databases from a number of urban areas. A standard condition for funding support for any of the R&D efforts described above should be that the complete mode] database (land-use and transportation) must be provided. suitably documented, to the central data library. These documented databases would then be available to any researcher or planner who wishes to access them. The existence of such a data library would provide a tremendous boost to integrated urban modeling in terms of: supporting cross-cit ~ comparisons of transportation - land-use interactions which will provide deeper more generalized insights into these interactions. supporting tests of mode] transferability in which a given mode] is applied to multiple locations; ~nnn~inc, Retina of multinie models within one or more urban areas and supporting the development of national default parameters and relationships as generalized results emerge from the above analyses.

OCR for page 157
TCRP H-12 Final Report While the exact contents of these databases will inevitably vary from one urban area to another. it should prove possible to impose minimum standards upon their contents and structure to facilitate the sorts of comparative analyses described above. Second, data collection standards and procedures should be developed for urban areas to follow, especially with respect to cntical data items which currently are often not "well handled" in many urban areas. This might include developing recommended procedures for using tax assessment and/or real estate databases to support modeling activities, procedures for collecting and maintaining employment databases and, possibly even. procedures for collecting goods movement data. Third. assistance can be provided for selected data collection efforts. While aeneraiized, nation-wide support for data collection is probably beyond the budget of even the most extensively funded program. very focussed binding of special, high-return data collection efforts should prove to be very cost effective. One example of this is the development of a very high quality- micro-level database for portions of one or more cities to support R&D efforts with respect to micro-scare modeling tDeakin =d LapTop forthcoming]. Another example is to support pilot-testing of novel data collection me~ods. particularly if they address one of the traditionally problematic data items (employment, price data. firm location choice etc.~. 6.5.7 Program Management The development of He R&D program requires both a sponsor and a means of managing it. It is reasonable to expect that these would be one and the same agency. In this Brats, the sponsoring / managing agency wait be able to: maintain budgetar~v and administratix e control, keep the program on track, and move it forward; serve as a clearinghouse for the exchange and dissemination of products. information, etc. liaise with other. related efforts (notably, other TMIP tracks. but also TEA-211^ and provide a "voice" and a profile for the program. ., However. In addition to the administrative/budgetary managements a technical management function also will be required. Cle=ly all of these functions must be linked. However. wee propose that a separate technical management body be established, under the direction of the admin~strativelbud:,etary management bode;. with responsibility for the following: the independent peer review panels. which would oversee the case studies (see Section 6.5.2), - I 36

OCR for page 157
TCRP H- 1 ~ Final Report technical coordination and specification of all aspects of the R&D program, including: o development of technical specifications for each activity O development of ;goodness-of-fit enters for each activity; o ensure technical consistency among the disparate activities; o technical evaluation of contractor proposals for conducting each activin,~; o oversight and review of each activity; and o technical advisory services to the managing agency The independent peer review- pane] for the case studies arid the technical management function could be one and the same. Or. a separate peer review- panel could be established for each case study. perhaps under the direction of (or at least with the advice of) a member of the technical management function. This increases the central managerial requirement. but has the advantage of bringing more participants into the process awhile spreading the technical burden. It also is conceivable. and even desirable that the contractor (modeler) for one case study could serve on the peer review pane! for a second case study. 6.~.8 Budget and Time Frame Table 6.4 presents a recommended budget for the R&D program proposed above. Attempting to develop a budget based on this still Verdi general discussion is difficult. and requires a number of assumptions. Table 6.4 Recommended Budget, Fir e-Year R&D Program ..... .... .......................... -- ......................... .... ~. .................... ' ........................... . ~. ~............. , r ~................ . Y.ear- ~ ~ -.- . - :--.- ~ ~ ' ' " ' ' ': "' "':" " ,. . " . '" ": ' .................. ...... . .. - ' " " - -: ' ' ' " " ' ":' ,' ' ' ~- T 10 I 100 ~0.0 1 030 ~0.i0 ~3.30 2 ~1.0 ~1.00 ~o.0 ~020 ~0.10 ~3~0 _ i3 1 1. 0 1 1.50 1 0. 0 ~0.20 ~0. 0 ~ 3.60 . 4 1.25 1.50 0.20 0.20 0.20 3.3s t5 T 1 200 1 1 0.10 1 Go 1265 L Total 1 6. 0 1 7.00 I 1.1 ~1 1.00 1 1./~0 1 16.00 Nore: Allfigures are in millions of dollars - 157

OCR for page 157
TCRP H-12 Final Report The first and most basic of these is the assumption of ~ five-year time period for the program. The choice of five years is clearly arbitrary, and could be charmed without significantly altering the overall logic of the program. It is assumed, however, Cat five years is about as a long a period which government agencies are likely to be w~ling/able to commit to ~ individual program. At the same fume, it must be recogruzed Hat a concerted, multi-year effort is required if sigruficant progress is to be made in this long-neglected field. In particular, five years is actually quite a short period of time to accomplish the development of operational "next generation' models. Table 6. j lists the specific assumptions used to generate each of the budget items shown in Table 6.4. Given the high degree of variability and uncertainty which exist with respect to the cost of one project versus another (implement mode] X in city Y; test N sub-components of a yet-to-be-designed next generation prototype mode] Z) these dollar amounts represent "reasonable judgements" concerrung how much should be spent in order to have a significant impact on the integrated urban modeling state of practice and art over the next five years. Much less than these amounts will not generate the level of improvement in modeling capabilities which are desperately needed; while expenditures considerably in excess of these amounts most likely will not be cost-effective at this point in time. As detailed in Table 6.4 we recommend an expenditure of $l6 million over a five year period, for an average expenditure of $3.2 million per year. Over 80% of this ($ ~ 3 million) is allocated to current and next generation mode] development arid Implementation, vow the remaining $3 million (just under 20%) allocated to supporting R&D activities. We believe the recommended budget is appropriate to sustain and move forward the recommended R&D program towards the development of practical tools The proposed expenditure filthy complements TRANSIMS and the other TMIP tracks, thereby moving the TMIP "picture" to completion. It also expands significantly the current minimal level of research support, which is a necessary condition towards achieving the stated Track E objectives. Moreover, it is also an extremely small sum compared to the billions of dollars that are spent annually on transportation and land-use investments in the United States. - 158

OCR for page 157
TCRP H-19 Final Report Table 6.~ Assumptions Used to Construct Program Budget Task I: Current Mode! Implementation & Testing i.! 1.2 In each of Years ~ Trough 4, $~.0 - I.25 minion ($~.5 million In Year I) is allocated to mode! implementation case studies. The amount awarded per case study may vary significantly (perhaps from S0. ~ to 50.5 million per case study). Or, in a given year, funds can be concentrated on one or two particular case studies. On average, perhaps 34 case studies are initiated each year. In each of Years 2 through 5, S0.25 to $0.50 million is allocated to testing / validation of the models unplemented in the previous year. Task 2: Next Generation Mode} Development 2.! 2.2 In Years ~ and 2 $~.0 million per year is allocated to targeted "basic research" activities. Individual amounts per project will vary. Some projects may be funded over two years, others will be for shorter durations. In Years 3 arid 4 $~.5 million per year will be allocated to implementation and testing of selected next generation model ~rototvDes. This activity will be staged, with increasingly elaborated prototypes being tested as time goes on. 2.3 In Year ~ $2.0 million will be spent on the full-scale test implementation of the most promising prototypets). Task 3: Other R&D ~ 1 $0.2 Bullion will be spent In each blear to support other, related R&D activities. This should be sufficient to fund a wide range of useful small to medium sized projects of varying types and durations. - 1~9

OCR for page 157
TCRP H-12 Final Report Table 6.5 Assumptions Used to Construct Program Budget, continued Task 4: Training & Information Dissemination 4. ~Year ~ costs include "start-up" costs associated with hardware acquisition and web site development. 4.2 Years 2 through 4 represent "steady-state" activity levels associated with trainings development of resource documents and maintenance/expansion of the web site. 4.3 Year ~ assumes reduced funding requirements due to less development of new resource material and reduced demands for web site maintenance and updating. Task 5: Data Assembly and Management 5. I, 5.2, 5.3 Same assumptions as for Task 4. However, an important assumptions is that complementary funds will be made available from TEA-21 activities, since: o data assembly and management is a task that requires cooperation with other complementary activities, ir1 order to achieve efficiencies in cost; o the full costs of dam assembly and management should not be born by the proposed R&D effort (which otherwise would be overwheIrned by data activities); and o the possibility exists of partnerships with other sources (e.g., land developers, private carriers, etc.) for whom the data would provide an added bereft; and o similarly, technological advances in data management are occurring, independent of this effort; therefore, the R&D program should be in a position to benefit from these but cannot realistically direct these advances in a meaningful way. General 6. ~ Figures shown do nor include internal program management costs for the sponsoring agency. 6.2 Figures shown do not include costs associated with the peer review process. - 160