Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
TCRP H-12 Final Report 4~0 6`FUNCTIONAL DESIGN,, OF AN INTEGRATED URBAN MODEL 4.! Introduction Having built the case In Me preceding chapters for We need for integrated urban models, this chapter develops a general design of art 'ideal" integrated urban model. This is viewed as a critical step in the development of improved operational models for two reasons. First' the "ideal" model provides a consistent benchmark against existing models can be compared (Chapter 51. Second. it provides an explicit target towards which research and development (R&~) efforts can be directed (Chapter 69. The development of the ideal mode} design proceeds in several steps. Section 4.2 provides a high level overview ofthe ideal system. Section 4.3 defines a set of design issues which need to be addressed in turning this conceptual overview- into an operational model, while Section 4.4 discusses the closely associated issue of criteria for assessing a model's performance. Section 4.5 develops a more elaborated descup~on of the ideal mode] by addressing each of the design criteria developed in Section 4.2 in some detail. While far from an operational specification (something which is well beyond the scope of this project), the mode] description presented in Section 4.5 is sufficient for current purposes of assessing the current state of modeling practice and of defining a sensible R&D program in this area. 4.2 Overview of An Idealized, Integrated Modeling Process Figure 4. ~ presents a highly idealized representation of a comprehensive transportation land-use modeling system. The "behavioral core" of this system (shaded area of Figure 4.~) consists of four inter-relatec! components: [anct development: this models the evolution of the built environment, and includes both the initial development of previously "vacant" land and the redevelopment over time of existing land-uses. This component could also be labelled "building supply" since Bulldog StOCK Supply Actions (cons~ct1on, demolition, renovation. etc.) me included. · , .,, ~ , , ~ . - . . . 2. Location choice: ibis includes the location choices of households (for residential dwellings), firms (for commercial locations), arid workers (for jobs). 3. Activi~y/traveZ~-hether performed by traditional four-stage methods or by emerging activit:~-based models, this component involves predicting the tup-making behavior of the population. ultimately expressed in terms of ongin-destination flows by mode by time of day. - 49
TCR~ H-12 Final Report . Figure 4el Idealized Integrated Urban Modeling System Demographics ~1- ~ - ; . ~ Regional Economics ~ ~ :~ ~ ~ ~-~;~ - i. :: : -::: ~ ~ Government Policies 1 - ~ ~ If.: ~ : : : : : :: .. .. .. . Auto Ownership -, j Transport.System *- A ! . ~ ~ .. ~ ~ .... ~ .. ~ .. ... . . .... .. ~ ..... Land Use ~... : ~ .~ . . I. ~ .. ..... ... .. . .. - ' '~;:~ '1 . . ~ -:: ~ -; ' 1 : :: ~ .. . ~ -.:~ ~ ~1 2: it- ~ ~! e - ~ ~ ~ - .~ I. ~ ':::::':'2~'':1 ~ ~ ~::--1 Activity / Travel and .- ~ -I ~- (-J`,ods Movement ~ ~:~ :~ I: it: :: Location Choice ~. . -. .... - . - ~ ~ . ... ::: · ;. · . .. _~._ - _ _._ Flows, Times, etc. | | ExternalImpacts - sO
TCRP H-12 Final Report Auto ownership: this component models household auto ownership levels -- an important dete~minar~t of household travel behavior. Points to note concerning these four "behavioral core" components include the following: In speaking about "land-use" it is common for transportation planners to blur the ist~nchons among these four components, especially between the concepts of 1~d development and location choice. A properly specified model, however, must clearly distinguish among these components since they involve very different actors. decision processes and timeframes. As is discussed further below- they also represent distinctly different "degrees of freedom" for the system to respond to exogenous inputs (such as construction of new transit infrastructure). Each component involves a complex set of sub-models. In particular, market-based supply-demand relationships tend to dominate aggregate behavior in each cased (buyers and sellers of houses interact within the housing market; workers and employers interact within the labor market; etc.), with prices) both being endogenously dete~nined and playing a major role in determining the outcome of these supply-demand interactions. Models which ignore these major supply-demar~d interactions may fall to properly capture the dynamical evolution of We urban system over tone. . . A simple flowchart such as Figure 4.] never properly captures the temporal complexities of a dynamic system. The vertical hierarchy is chosen to represent shorts conditioning effects. That is, In the short run, most location choices are made within a "fixed" building stock supply. Similarly, in the short run, most activity/~avel decisions are made given a "fixed" distnbution of activity locations (and a fixed number of household autos). In the longer run, all four components evolve' at least partially In response to "feedback" Tom lower levels in the hierarchy (land-use evolves in response to location needs of households and firms, people relocate Weir homes armor jobs at least partially in response to accessibility factors; etc.~. The inclusion of auto ownership as a separate box within the "behavioral core" is somewhat unconventional. Auto ownership is often treated as simply one more (ofren exogenously determined) input to We travel model. As Ben-Akiva tI 974] has 4 With the possible exception of the auto ownership component, although even here a "supply side" clearly exists, even if we usually choose not to model it explicitly. Or, in the case of trip-making, travel times - 31
TCRT H-17 Final Report observed, however auto ownership is an integral part of the '"mobility bundle'' (which, in terms of Figure 4. I, Ben-Akiva would define as the combination of the location choice, auto ownership and activity/trave} components) in that it is fundamentally interconnected with residential location arid work trip commuting decision-making. This point is strongly reinforced within the empincal literature discussed in Chapter 3, in which auto ownership is consistently found to be art important "intermediate variables connecting urban form (as measured by residential density, etc.) and Ravel behavior (as measured by transit usage, VMT, etch As shown in Figure 4. I, there are at least four major "drivers" of urban systems: 1. Demographics: evolution of the resident population in terms of its age-sex distribution, population size, education level, etc.; 2. Regional economics: evolution of the urban region economy in terms of its size, industrial distribution, etc.; 3. Government policies: zoning, taxation, interest rates, etc.; and 4. The transportation system: road, transit' etc. The extent to which these venous drivers are treated as being exogenous or endoger~ous to the model will vary from one modeling system to another. Government policies and changes to the transportation system are almost exclusively treated as exogenous inputs; demographic and regional economic processes are almost always Heated as at least partially endogenous. The key point is that the full range of "drivers" of land-use/location/travel decision-making should be included in the modeling system to ensure that the impact of any one policy (such as a change in the transit system) can be properly represented and evaluated by the model. In particular, it was often the case with early land-use models that they over-emphasized transportation system effects on land-use/location processes and hence were biased towards over-predicting the impact of transportation system improvements on these processes. As Knight and Trygg [1977] clearly demonstrate. however, transportation improvements are only one among many determinants of land development decisions -(see Figure 3.39. It is fair to say that no existing transportation - lard-use model Filly captures all aspects of the comprehensive modeling system described above. Such a system, however, provides the starting point for assessing both the state of the practice and the state of the art in this field. It also defines the goal towards which all such modeling systems should be striving, in that such a system would provide the analytical means for assessing the short and long run impacts of both transit and road alternatives in a balanced and comprehensive way. In particular, the long run impacts of major improvements in transit access on land development, residential and employment location choice - 52
TChP H-12 Final Report auto ownership levels and activity/travel could be assessed in a logical and defensible mater. It is also fair to say that, although not incompatible with a comprehensive transportation - land-use modeling process, TMIP is not actively developing such a process, despite the existence of the land-use Pack ("Track E"). Research to date within TMIP has focussed on act~vityl~ravel modeling (and associated emissions modeling), with the majority of the effort being placed on the development of the TRANSIMS network modeling system [Barrett' et al. ~ 1 995~. Figure 4. ~ can be expanded and reformulated ad infinitum. Figures 4.2 and 4.3 present two independently developed designs for "next generation" navel demand modeling systems which were generated as part of an FHWA "think piece" at the beginrung of the TMIP process. Without discussing these designs in detail, the key point to note is that land-use, location choice and demographics are all integral components of the proposed overall modeling systems. This inherent integration of land-use and transportation processes is further emphasized in Wegener's conceptualization of the process (Figure 4.4), in which he emphasizes the interactive, cyclical, feed- back nature of Me urban system. Cornrnenting on this figure, he notes: "The two-way interaction between land-use and transportation may be less commonplace for transportation modelers who are trained to take the land-use forecasts provided by planning departments as something beyond doubt. Now transportation planners, obliged to think about the land-use impacts of their proposals, call for land-use models as add-one to their ... models. Nothing could be more shortsighted. The land-use transportation feedback cycle needs to work its way through several iterations to equ~libnum or dynamic disequilibnum... The conclusion is that if transportation planners want larld-use forecasts, they have to integrate land- use models Into their models, or vice versa." [Wegener, ~ 99S, p. 27-28] Figure 4.5 provides one final representation of a transportation - land-use modeling system. The key point of this figure is to illustrate the potential sensitivity of such a system to a wide range of land-use and transportation policies. The figure shows a representative (although not necessarily ~ - 1 exhaustive) list of policies of interest. The arrows indicate the "entry points" for each of these policies within the modeling system, i.e., the point at which the given policy directly enters model calculations. For example, congestion pricing, gasoline taxes, etc. directly affect road user costs and so "enter" the modeling system at the road assignment stage, where these costs are calculated on a link by link and O-D pair by O-D pair basis, and where these costs presumably affect auto route choice decisions. The impact of such pricing policies, however, feeds back through the system, potentially affecting modal choices (and perhaps other aspects of travel choice), auto ownership, and possibly even residential and/or employment location choice. Thus, the modeling system, in principle, is capable of capturing both short- and long-run responses to a given policy (or combinations of policies) and, thereby, provide a more accurate estimate of system response, relative to more partial treatments of the problem.
TCRP H-12 Final Report Figure 4.2 SAMS: Sequenced Activity-Mobil~ty System (Resource Decision Consultants, Inc., as reported in Spear ~19943) > ~ l ' PLANNING I PR( )CFRS ~;~~ \TDMs, , v ~ _ , ~ ~ , ~ /Command & Control . . ~ . ~ > 1 *Runs on GIS platform URBAN SYSTEM MICROSIMULATOR* I Land-use Re/development, price, densification Residential and Job Locations HIGHWAY/TRANSIT NETWORK SOCIO DEMOGRAPHIC MICRO SIMULATOR* Individuals Households Fi~ms/Developers t 11 - 1 1 '1 DYNAMIC HOUSEHOLD VEHICLE TRANSACTIONS MODULE ~ ~ T ~ ACTIVITY-BASED DAILY TRAVEL SIMULATOR* onginJdestination, time of day, mode · vehicle allocation & vehicle occupancy I t DYNAMIC NETWORK ASSIGNMENT* Link flour by clock time Level of service Cold/hot starts . AIR QUALITY | EMISSIONS MODULE I 1 Indicators for Mobility, Air Quality & Value of Travel Services - 34
TCRP H-12 Final Report Figllre 4.3 SMART: Simulation Mode} for Activities, Resources and Travel (Louisiana Transponahon Research Center, as reported in Spear t19943) - Land \~Prices - Site Schedules/ Constraints ~ ~'l' Land \~ Uses J :Infrastructure Investments by - >( Transpor~tion \` System ,J . .......... Linkage proposed Linkage suggested for later inclusion - Area ~ HH no; r ~ Characteristics J Needs/ ~ HH~ Activin- Pattern \~ Rind Constrain: - r M=datO~'~r Household Activity Simulator Routing Flexible ~l + - 1 Optlonal SpeedfConoestion ~ Net volumes by time of day ~ Net volumes by time of darer 1 ~ Net volumes by time of day 1 (E up by link)
TCRP H-12 Final Report Figure 4.4 The Land-Use/Transportation Feedback Cycle Source: Wegener L1995] Mode choice He Route choice a/ / Firm loads / Travel times/ distance/costs 1 Accessibility Attractiveness - Location decisions of investors - Destination choice Transport Land Use - - - ~~ Construction - 56 Tnp decision \ Car ownership Activities Location decisions of users
TCRP H-12 Final Report l Figure 4.5 Policy Inputs into an Integrated Urban Mode} Source: Miller and Hassour~ah t1993] ZoninglLand Use Policies e Interest Rates an. 9= Housing Policies Current Technology Changes ~ AdvancedlNew Technologies-_ . _ AVeh.Technology~AutoOwnershipt ~1 Inspection & Maintenance Programs ~ Vehicle Purchase Taxes A. Incentives for Scrapping"Bad" Cars Tax Policies Re. Auto Use DeductionsJ Carpooling Policies ' Parking Policies Telecommuting Policies Flexible Work Hours, etc. HOV Lanes Congestion Pricing ~ Traffic Control Measures ~ ~ ..~. Gasoline Taxes J ~ a..... Transit Improvements Transit Prionty Measures Transit User Subsidies J Employment Retail, etc. ==~_t Building Stock Residential Distributiont _ _ _ ~_ _ _ , .~'' ~ _ _ _ ~ 1 1 ., Non-Work Travel: _ _ · Generation ~ . ~ by time of day e Distribution w /weekend - Mode Split | eekday Auto Driver Trips (by time of day) (+ auto occ.) Transit Tnps (by time of day) Walk/Bicycle Trips (by time of day) 41 . . .. ~ ~ ~ Auto Transit ! As s i Elm ent A s s i gem ent _ ~ _ , I _ _ - ~ r ~ ! Auto Energy Use Transit Energy & Emissions Use & Emissions ~, r I I t=t+^t - 57
TCRP [I-17 Final Report - 4.3 Design Issues 4.3.! Introduction A large number of issues must be considered in the design of an operational integrated unbars mode} derived Mom the idealized modeling system presented In the previous sub-section. As many as possible of these are listed In Table 4. ~ and briefly discussed below. Different models, of course, wall address these issues In a variety of ways, ranging from ignoring them completely to dealing with them in a very computationally detailed and/or theoretically rigorous6 mariner. No "right' answer/approach necessarily exists with respect to any one of these issues. As with any design exercise, the "nght" or "best" design depends on the specific application context (data availability, computational and technical support capabilities, ar~alysis/forecasting needs, etc.~. In addition, no one issue or "dimension" of the problem can be "optimized" in isolation; it is the overall balance across design dimensions which is impor~t (e.~., very fine spatial resolutions may be difficult/~mpossible/ur necessary to m~nta~n within very Tong-range forecasting applications). The intention here, rather, is simply to generate a reasonably comprehensive list of issues, as a framework for organizing =d discussing the modeling state-of-the-artlpractice in Chapter 5. In addition, they provide the basis for the development of model evaluation criteria, discussed in Section 4.4. , c The identified design issues have been grouped in Table 4.l into five categories: physical system representation, representation of "active agents" in the system, representation of processes, "generic" issues (which cut across virtually- all physical system, active agent and process representation considerations)' and issues associated with the implementation of the model design within an actual computational environment. The first three categories deal with the substance of We system being modeled: the physical entities. Me behavioral entities, and the processes by which these physical and behavioral entities evolve over time. The last two categories are more methodological in nature. dealing with how the representation of these entities and processes is actually implemented within an operational modeling system. Each of these groups of issues are cliscussed in turn in the following sub-sections. 4.3.2 Physical System Representation Fundamental to model design are decisions concerning the representation of the physical elements of ~e system: time, land (space), buildings and transportation networks. These decisions fundamentally affect the precision and accuracy of the models its data and computational requirements, arid options for the representation of behavior within the physical urban system. 6 Note that these need not be the same thing! Theoretically crude approaches may well be very computationally intensive, with all other combinations oftheoretical elegance and computational requirements potentially existin, as well. - :8
TChP H-17 Final Report 1 1 Table 4.l Integrated Urban Mode} Design Issues Physical System Representation O Time O Space (land) O Building stock O Transportation networks O Services Representation of Decision Makers O Persons o Households ° Private firms O Public authorities Representation of Processes o Land development O Location choices o Job market O Demographics O Regional economics O Automobile holdings o Activity/trave} demand 0 Network performance "Generic Issues" O Level of aggregation/disaggregation 0 Endogenous versus exogenous treatment 0 Level of "process type" O Model specification Implementation Issues Dam requirements Computational requirements ,9
TCRP H-12 Final Report - Treatment of Time. All forecast~n~ models must predict how an urban system state in some base year is likely to evolve into the filture, typically up to some user-specified forecast "horizon year". Choices of mode! base and horizon years, arid the time increment or step used to move the system from the base to horizon year are fundamental design questions. Also fundamental is the treatment of ''dynamics" within the model. Many models assume that system equilibnum is achieved in each time step, and so are able to appeal to the mathematical conditions for equilibnu~r~ to solve for the system state at the end of each time step. Ordinary gravity models are a classic example of this approach. Other models do not assume equiiitnum. Rather, they explicitly simulate the evolution of the system state Tom one point in time to another as a function of various assumed processes. The question of system dynamics is further complicated by the fact that different processes at work within the urban system operate on different timeframes. Land development processes operate over time periods of decades or more; many househoid-leve! decisions are perhaps made on approximately a yearly basis (e.g., residential relocation decisions automobile transaction decisions, household structure evolution), marry activity/travel decisions change Tom week-to-~-eek arid from day-to-day, road network operating conditions (and, hence, energy consumption and tailpipe emissions) vary Tom m~nute-to-minute and second-to-second. Reconciling this wide combination Of "sIou-" (or long-run) and "fast" (or short-rur~) dynamics within an overall modeling system is challenging, to say the least. Treatment of Space. The spatial nature of urban systems represents one of the major sources of complexity in the analysis and modeling of these systems. Space enters both in terms of the locations of activities and in terms of the flows of people, goods, etc. between these activity locations. Design issues include zone system definitions, degree of use of/interface with GIS software arid the degree to which "micro" neighborhood design attributes are incorporated into the set of spatial attributes maintained within the model. Building Stock. While we often talk rather loosely about "land-use", most urban activities actually occur within buildings of one type or another, and the built environment, to a large extent, determines the nature of which activities occur where. The extent to which building stock (by amount type, etc.) is explicitly represented within the model represents an important design decision. and is found to vary considerably from one mode! to another. 7 Even with advanced GIS capabilities, it is likely that a zone system is required, at a minimum for data display purposes. Certainly all current and immediately forseeable integrated models depend on a spatial zone system. - 60
TCRP H-12 Final Report - Transportation Networks. Appropnate representation of both road ar~d trar~sit systems is clearly an essential component of any integrated urban model. Issues here include ma~nt~nin$ consistency in level of detail with Me zone system being used; appropriate representation of transit walk access/egress; and appropn ate representation of parking supply. 4.3.3 Representation of Active Agents "Active agents" are the decision-making units -- the people. households, fimns, etc. who actually cause We urban area to exist and to evolve over time. through their various activities. People buy and sell homes, participate in the labor market; travel to and from work. school, shopping, etc. every day, get married and (sometimes) have babies; age and (eventually) die; etc. Firms similarly face location-relocation decisions; go through ~ life-cycle process of birth, aging' perhaps with growth' perhaps eventually lldyingl,, make land development decisions; supply the goods and services which people buy; provide jobs for workers; etc. Implicitly or explicitly, integrated urban models must address how they are going to represent the two primary active agents within cities: people arid firms. People generally live within either family or non-family units generally referred to as households. For many important activities such as residential location choice and automobile holdings choice. the household is in most cases the natural decision-making unit, rather than the individual. Thus, the possibility exists that one might wish to explicitly represent both individual persons and households as ~nter-related but identifiably separable decision-making units within the model. Other active agents obviously exist within urban areas which have direct impacts on the transportation - land-use interaction, notably various government agencies, transportation service providers, etc. The extent to which such agents are explicitly incorporated within all integrated urban model is another design decision, although, in general such agents are usually assumed to act exogenously to the processes being explicitly modeled. 4.3.4 Representation of Processes Table 4.! lists Me primary processes which collectively define the transportation - iand-use interaction. Most of these processes have already been discussed in Section4.2, as Bell as to varying degrees in Chapter 3 and Appendix A. Additional points to note concerning process representation within integrated urban models include the following. 1. As has already been observed in Section 4.2, many of these processes are market- driven. These include land development~buildirlg suppler, residential and commercial real estate markets, labor markets, and travel markets. "Proper" representation of both demand and supply processes within each of these markets is - 61
ACRE H-12 Final Report essential to modeling such processes successfully. Implicit In this observation is that prices must be explicitly represented within the model Ed must be endogenously determined through the demand-supply interaction. Different models wall, of course, deal with these processes in venous ways, including implicitly through the combination of two or more processes within a single sub- model. For example, a simple Lowry-type model, In which the residential population is allocated over the urban area using a gravity mode! (given known employment locations) effectively combines residential land development, residential location choice and job 1ocahon choice into a single "net" or "reduced-form" model. One way or another, however, every integrated mode! must deal with each of the processes listed in Table 4. I. 3. 4. Network performance modeling has not be explicitly discussed to this point in the report (it has been implicitly subsumed within "activity/travel" to this noint1. Design r . ,. 1 ~ - ~ - ~= Issues here Include choice ot static versus dynamic route assignment procedures, deterministic versus stochastic procedures, equilibrium versus non-equilibrium assumptions. level of network detail. compatibility wi~/sunoort for emissions and . , , ~ ~ ., ~ ~ ~ ~ r · . .. ~ . ~ A, ·, , ~ energy use models, and degree ot Integration between road arid transit network representation arid processes. "Regions economics' are shown as one of the major "hovers" of the urban system · _ · . · ~ ~ In f lgure 4.1, with the Implication that they are exogenous to the modeling system. Given that the magnitude and nature of the economic activity which occurs within an urban area depends in no small way on regional economic factors some integrated urban models incorporate input-output models of the local and regional economies as a major component of He overall modeling system. Indeed, In the limit, the entire modeling system can be developed as a mode] of a spatially distributed economic system, with the consumption of Carol, buildings and travel as being but three out of many economic sectors being modeled. 4.3.~ Generic Design Issues Integral to the design of the representation of the physical system Me behavioral agents and the processes at work within the system are fundamental choices conceding aggregation level, boundaries between what is endogenous to the model arid what is not, and "process type". Each of these is briefly discussed below. L.eve} of aggregationIdisaggregation. Most currently operational integrated models are quite aggregate in both space and time, often using less than ~ 00 zones to represent an entire urban area and working in time steps of ~ or even 10 years. At the other extreme, many researchers are - 62
TCRP H-12 Final Report experimenting with "m~cros~rnulation" models. In which individual households, building, firms, etc. are the basic mode] braiding blocks.8 Choice of aggregation level wall have profound effects on data r ~ _ an, A_ _ _~-~ r - - requlrements, options tor models processes. computation requirements, etc. arid represents one of the primary, distinguishing decisions in any model design. We are most used to thinking of the aggregation issue in terms of spatial aggregation (i.e., use of zones instead of ~ndividu~ people as the unit analysis; size of zones used; etc.~. Aggregation decisions, however, are made wad respect to every entity (physical or behavioral) arid every process included in the model. Use of a five-year time step to represent a process which occurs on a yearly (or shorter) basis constitutes temporal aggregation. Not including potentially salient personal attributes (say, for example. education level or occupation type) in decision-making models represents aggregation over Attribute spaces. And so on. Endogenous versus exogenous factors. Any agent or process which is explicitly modeled within the model so that its attributes and/or behavior are determined within the model is said to be endogenous to the model. Conversely factors which affect system perfonnar~ce but whose values are simply provided to the model as inputs are called exogenous factors. A fundamental step in any model design involves "drawing the boundaries" around the model; that is, determining, what is to be included within the model versus what wit] be excluded. As with the aggregation discussed above, these decisions wall directly affect data and computing requirements. policy sensitivity, and process modeling options. Process type. Decisions must be made concerning how to mode! each endogenous process within Me model. While a near-continuum of options exists these can be broadly defined as falling into two cate~ones: "transition models" and 'choice models" Awakener ~ 9951. Transition models use simple . ~ , . ~eterm~rust~c or pro~ao~st~c rules tor determining changes In attnbutes. system states, etc. over time. Examples of transition models include most models for most demographic processes, such as deterministic population aging models (i.e., add ~ year to each person's age for each year being simulated) arid fertility models which express the probability of a women giving bird to a child as a simple function of her age, mantal stems. etc. Choice models, on the other hand, attempt to model explicitly the choice process underlying a particular decision or action (random utility choice models and computational process models are both obvious examples of this class of model). Residential location choice, employment location choice. auto ownership, arid activity/trave] decisions are all examples of processes which one might typically be modeled as choice processes within an integrated urban model. See, for example, Mackett [198:b] Spiekermann and Wegener [1993], Oskarnp [1997] or Miller and Salvini [1998] - 6:
TCRP H-12 Final Report While some processes may Obviously fall into one category or the other (aging is a pure transition process -- we have no choice in the matter whatsoever!), allocation of a given process to one type of modeling approach or Be other is at least partially dependent on the application context, available data and modeling methods, computational resources, etc. Household formation arid evolution for example, in "real life" certainly is Me result of complex interpersonal decision-making. In most integrated urban models, however, such processes (if endogenously modeled at all) are represented using relatively simply transition models. Mode! specification. This includes both He selection of mode! functional form (logit model, etc.) and the explanatory variables to be included within the model. This issue is so integral to all model- building that there is perhaps little that needs to be said with respect to it. except to point out the obvious facts Cat mode! specification determines theoretical soundness (and hence the fundamental credibility of the model), computational intensity, data requirements and policy sensitivity (if a particular policy-relevant variable is not included in the model, then the model obviously will not be able to respond to the given policy). 4.3.6 Implementation Issues All models require data, computational resources and technical support to be developed, implemented and maintained as an operational tool. Each of these issues is briefly discussed below-. Data requirements. Historical data are required for both mode! estimation/caTibration arid validation. Estimation usually refers to the statistical estimation of model parameters which cause the mode] to "best fit" (in a statistically well defined sense) to observed. historical data (e.g.. use of maximum likelihood estimation to estimate logit choice mode] parameters, or use of linear regression analysis to estimate trip generation mode] parameters). Calibration usually refers to post-est~rnation "parameter adjustments" which "force" the model to better replicate observed data (e.g., use of K-factors in gravity trip distribution models to force the mode] to reproduce observed screenTine or cordon counts). Given the complexity of most integrated urban models (typically involving many sub-models, each one of which possessing its own level of complexity, often exercised within a simulation framework), a considerable amount of calibration as opposed to estimation is usually required In order to get these models "working properly". This, in turn, implies the need for considerable experience and good professional judgement to be applied to the model development process. Once a model has been estimated/calibrated, it should be validated as a forecasting tool by performing "historical forecasts" between two or more points in time in the past, for which historical data are available. For example, a model may be calibrated using data from ~ 970 and ~ 980. Using ~980 as a base, it then may be used to "forecast" 1990 conditions. This ~990 "forecast" can then be compared with known data for 1990 in order to assess the ability of the model to predict beyond the time period covered by the calibration data - 64
TCRP H-17 final Report The foregoing discussion indicates that integrated urban models typically require a considerable amount of historical data from multiple time periods, in order to be calibrated and validated.9 The likely availability of historical data (what variables at what level of spatial detail for what years at what level of reliability, etc.) must be considered in the model design process, since there is no use in desi~,~ng a modeling system which can not possible be implemented due to data resections. Known, ir~unnountable data limitations will often drive the model design with respect to such important factors as time step, level of spatial aggregation, and choice of model specification. . . Once a model is operational, it requires a new type of data to be used as a forecasting tool: estimated values of the exogenous inputs to the model for the future year(s) being simulated by the model. These estimates may come from policy scenarios, professional judgement, other models, etc., but, one way or another, they must be provided by the analyst to the model so that it can be run. These input data can be quite extensive, difficult to generate, and, of course, subject to error. In general. a classic trade-off exists in model design between "specification error" (which is built into the model due to model simplifications, abstractiorls, etc. which cause the model to fail to perfectly capture real world behavior) and "forecast error" (error introduced during the forecasting process by inaccurate inputs). Figure 4.6 broadly illustrates this trade-off, which generally is assumed to imply some "optimal" level of model specification/detail for a given problem application. As with the model development data requirements, the forecast input data requirements must also be considered during the model design process, and, again, may well impose significant practical constraints on model design with respect to the temporal, spatial and/or behavioral representations which are feasible to achieve. Integrated urban models are well known to be extremely " data-hungry". At any point in time, data availability may well prove to be the single biggest cor~traint on model design and application. At least two more positive observations, however, with respect to data are the following. The H~t~t~ nv~il~hl~ to minnow integrated urban model. have improved 1 . ~ ,_ at ~ ~ in, me, _~ ~ ~r ,~ ~ ~ _ dramatically over the last twenty-five years, and can be expected to continue to improve as we move even more deeply into the "Information Age". A. The need for improved datasets to support Integrated modeling can prove to be a very positive stimulus for improving the overall planning database in urban areas. That is. while perhaps initially motivated by modeling requirements, once collected and assembled, databases (if properly Barraged) can take on a life of their own and can provide very useful support for a range of planning applications. Typically at least three time periods are required: two for model estimation I calibration, and a third for validation. - 65
TCRPH-12 Final Report - Fi ure 4e6 Mode} Specification Versus Forecast Input Error Forecast Input Errors Combined Error / ~ = Specification Error + Forecast Input Error ~ . . my' S , . ' - 66 Specification Error Model Complexity Spatial/Temporal Precision
TCRP H-12 Final Report _ A good historical database can support a wide range of histoncal analyses which result in improved understandings of processes and issues at work within the urban area; that is He database cart and should support various descriptive and diagnostic analyses as well as model-bu~ding and forecasting applications. Computational requirements. Integrated urban models by definition are computer-based. The size of the computer (CPU, memory, disk space, etc.) required to house the model, the time required to execute a single run of He mode} (troth obvious tradeoff between run time and computer size) and . ~ ^. · ~ . · ~ . ~ . . ~ ~ the software required to Implement aria support the model Me. the actual computer code within which the mode] is implemented, as well as the ancillary software -- operating system, GIS, DBMS, statistical analysis systems, etc.) are all of critical concern within the mode] design process. Historically, the computing power cost-e~ectively available to researchers arid Earners has imposed significant limitations on the scale and scope of integrated urban models. Past arid continuing avarices in computer technology, however' are fast removing these barriers. The amazing power of desk-top computers, the contimiing emergence of parallel processing, the explosion of software, etc. are all extending He boundanes of what is feasible. to the coins that computing Dower ner se is probably no longer the constraint on practical modeling systems. 1 ~7 1. Technical support requirements. The discussion to this point has focussed on the mode! design and development process. Implementation of a mode} within a given planning agency, and then the ongoing maintenance =d use ofthe mode] within this agency, requires significant technical support. In-house staff must be dedicated to the operation of the model. where this staff must have appropriate professional backgrounds and have been properly trained in the understanding and use of the model. An institutional, management-level commitment must exist within the planning agency to provide the time money and moral support required to get the mode] implemented and then to keep it operating effectively and efficiently. And adequate and on-going support must also be available from the model developers (who usually wili be external to the planning agency) with respect to trainings ~ouble-shoo~g and on-gong system maintenance and upgrading. While largely implementation and operations, rather tears design, oriented, the design implication of this issue is that an overly complex mode] design which is difficult to understand" operate and maintain, or which is not "robust" with respect to its ease of use within art operational planning environment Will not be an attractive or even practical mode! for application within such contexts. 4.4 Evaluation Criteria 4.4.! Introduction Table 4.2 lists the set of evaluation criteria which are used in the review of the integrated urban modeling state-of-the-ardpractice presented in Chapter 5. While these criteria obviously relate directly to the design issues discussed in the previous section, they also represent a different (and generally somewhat more abstract) "slice" through the problem. These criteria have been divided - 67
TCRP H-12 Final Report into three groups relating to: the credibility of the models for use within operational plying applications, the feasibility of implementing the models within operational contexts and the usability of these models once Hey are implemented. Each of these groups of criteria is discussed briefly in turn in the following sub-sections. 1 1 Table 4.2 Mode! Evaluation Criteria Credibility O Theoretical soundness O Policy sensitivity Precision (spatially, temporally) Validation o Representation of/sensitivi~ to transit Feasibility O Computational requirements O Data requirements O Technical support requirements O Cost leasability O Ease of input preparation O Mode} run the ° Output/presentation capabilities O Portability/transferability O FIexiblity/adaptability 4.4.2 Credibility' Criteria This set of criteria deal with the basic confidence one has in a given model. as well as the suitability of the mode] as a policy analysis tool. Each criterion is briefly discussed. - 68
TCRP H-12 Final Report Theoretical soundness. If a model is not theoretically soured then one can have little confidence in its predictive capabilities and sensitivities. Aspects of theoretical soundness include: the model captures key behavioral relationships and includes key actors, processes, etc.; the model is consistent with our current theoretical and empirical understanding of urban processes, the model is internally self-consistent; and it makes use of statistically and logically valid methods and procedures. Policy sensitivity. To be of practical use, the model must be capable of responding to the range of policy issues of interest. These can and should include: land-use policies (zoning, taxation, growth management policies. land servicing- etc.~: transportation policies (transits road TDM. etc.~; auto ~ -- - ~ ~ ~ ~ ~ .L ~ ~ ~ ownership related (vehicle technology options, taxation etc.), arid VarlOUS combinations thereof. Taking a different slice through the problem, the model should be able to analyze a wide range of infrastructure investment, operating, regulatory, and financial options, which target the demand and/or supply sides of the venous processes/markets at work within the urban Lea. Spatial and temporal precision. The model must be able to analyze the urban system and provide forecast outputs at sufficiently precise spatial and temporal scales to address policy questions in adequate detail. The defection of "sufficiently precise", of course, various from one application to another. In some cases, very "broad brush" results at a very gross spatial scale over one or more very large time steps maY well suffice. In others' much more spatial andior temporal detail will be · ~ ~ ~ 1 1 , 1 1 1 ~ 1 ~ 1 _ t 1 _ ~ _ It_ _ _ _ ~ ~ ~ ~ 1 1 ~ _ ` 1_ _ 1 _ _ _ _ ] _ ~ ~ ~ I required. ideally. the model Should be able to window in and Out to me level OI pleClSlOn reqUlreO for a given analysis. In practice, however. such flexibility is difficult to achieve and may imply a very high cost in terms of program complexity, computational and data resources. etc. Validation. Rather than speak of model accuracy (which is a very difficult Wing to determine in a- , ~ , ,, · r ~ ~ '' it'd . Validity is established in at least two ways. First a mode! has face validity if its results generally are in agreement with best professional judgement and win empirically observed past and current trends. Second, a model can be historically validated (as discussed In Section 4.3.6) in order to demonstrate its performance in replicating historical trends which lie outside the time period used in model calibration. practical teens) it 1S perhaps more useful to speak or model vallalty Representation o£/sensitivi~ to transit. Mile this criterion is effectively already covered by the "policy sensitivity-" criterion discussed above, it has been listed separately here in order to reflect the motivating objective ofthisproject: to assess the extent to which integrated urban models adequately capture the transit - land-use interaction. This includes "both directions" of the interaction: the extent to which transit services and facilities effect urban land-use, arid the extent to which urban form affects transit usage. - 69
TCRP H-12 Final Report 4.4.3 Feasibility Criteria This set of criteria deal web Me technical arid financial feasibility of a modeling system. The three technical cnter~a have already been discussed at some length in Section 4.3.6. Cost has been added as an explicit criterion, given its obvious importance in any decision process. Costs include both implementation and on-going operating costs. In both cases, costs wait exist for the mode] software system itself, computer hardware and ancillary software, data collection and maintenance, and the technical support staff (both in-house and external) required to implement arid operate the modei. For large-scare integrated urban models, bow Implementation and operating cost components can be significant and can represent a major con-quaint or even banner to model adoption and usage. 4.4.4 Usability Criteria Ease of use represents another important dimension for evaluating any model. Each of the "usability" criteria listed in Table 4.2 are briefly discussed below. Input data preparation. Even conventional four-stage travel demand models typically have very onerous. time-consuming, and often error-prone input data preparation requirements, which can significantly limit the number of alternatives investigated, the extent of sensitivity testing, etc. relative to what one would ideally like from a planning perspective. Integrated urban models can easily possess the same or even greater problems in this regard. Model run time. Mode! run time is a related, although generally not quite so significant issue. Ideally one would like to set up a rare, go have a cup of coffee, and come back to get the run results. It can be argued, however, that within fairly broad limits, mode] run time is not overly critical in most applications: if it takes an overnight run to generate results which you might Men spend a week or more analyzing, the actual run time is fairly unimportant. Output/presentation capabilities. Much more important is the question of result output and - r - - 1 ~ presentation capab~t~es. One runs a mode] in order to obtain useful outputs. The nature of these outputs and the ease who which they can be stored' accessed, marupulated and presented are critical to the overall utility ofthe modeling system. Indeed, no matter how substantively credible the results might be, if they can not be readily used to provide real insight into the problem being addressed, and if they do not easily trar~sTate into formats arid messages which cart be understood by non- technical decision-makers, then they are of little practical use. Models are ultimately decision- support tools and their effectiveness in this role depends in no small way on their output/presentation capabilities. Given the power of ar~cilIary software readily available today (spreadsheets, database management systems, graphical presentation software, GIS's, etc.), much of this capability may reside outside of the modeling system per se. In such cases, the key concern is the ease of interface - 70
TCRP H-12 Final Report between the mode} and such software packages. In other cases. some or much of the posts analysis and display- capability may reside within the Dackace itself thereby minimizin_ the need to move data trom place to places as well as' perhaps" providing efficient. customized analysis and · . - report-generat~on capabilities. Regardless of how the post-rur~ alysis and display tasks are performed, however, the more fiandamer~tal concern is Mat the right information is computed and stored within the model for later retrieval and analysis. This. of course, returns one to fundamental questions of model design. In this case, however, the design is driven from the "bottom up" in terms of beginning by asking what information is needed from We model, rather than from the "top down" (as has implicitly been the case in much of the discussion to this point) in terms of what information the mode! can provide easily (given available theory, data, etc.~. Portabilit~r/transferabilib. The greater the extent to which a model can be transfered from one application (urban area) to another, the more credible and useful it is, for at least three reasons. First, transferability implies generality, which, in turn, implies that the model must be theoretically sound since it seems to hold across a variety of applications. Second, development and implementation costs should be reduced, given that they are spread across several users. And third. shared models generate a community of users who can share expenences. help each other with common problems and collectively contribute to mode] improvements over time. FIexibility/adaptability. Models are not static entities. Rather, they (should) grow' and evolve as data, theory, computational capabilities, experience, financial constraints. etc. change over time. Similarly, application contexts are constantly evolving as new issues arise, new- alternatives are suggested decision-makers and their interests change, etc. Thus, models should ideally- be flexible and adaptable so that they can evolve and change over time in response both to new- challenges and new opportunities. 4.5 Specification of An "Ideal" Integrated Urban Modeling System 4.5.! Introduction The purpose of this section is to farther elaborate the specification of the ideal modeling system first sketched in Section 4.2 so as to provide a more detailed benchmark against which current models can be compared (Chapter 5), as well as to define a desired end state towards which model research and development activities can be targeted (Chapter 6). Section 4.5.2 begins this process by stating an explicit set of axioms or fundamental assumptions which guide our model design process. Section 4.5.3 then works through this model design in detail, using the set of design issues introduced Table 4.2. Finally, Section 4.5.4 provides a brief summary of the key elements in the model design. - 71
TCRP H-12 Final Report 4.5.2 Mode} Axioms In travel demand modeling it is fairly safe to say that a reasonable consensus exists concerning the "paradigm" within which this activity is undertaken in practice. Elements of this paradigm include: . . adoption of the four-stage structure as a means for dealing way the complex multi- dimensional nature of travel; use (implicitly if not explicitly) of entropy maximization concepts to build trip distnbunon models (i.e.- gravity models); · use of random utility models to mode] mode choice (e.g., nested iogit); and · use of user equilibnum methods to perform route assignment. All of these elements are contained within (if somewhat loosely at times) an overall behavioral framework drawn Tom neo-cIassical microeconomics. This paradigm has evolved and been institutionalized over more than a forty-year period. It represents the embodiment of the collective experience of a large number of individuals and agencies worId-wide who have engaged In operational travel demand modeling over this very long time period. A very strong case can be made that a "paradigm shifts' is underway to an explicit "activity-based" view of travel behavior which will likely result in quite different mode} structures tPas, 19904. The key point to rote, however. is that this involves replacing one well defined paradigm win another one which is rapidly gaining similar widespread support and consensus.~° It is also fair to say however, that no similar paradigm or consensus exists with respect to larld-use modeling. Far fewer people have attempted to build land-use models aIld so we have far less empirical experience with them, especially In operational settings. There tend to be a few "key players" In the field, each of whom tends to have strong and open somewhat divergent views of the process and how best to model it. Without art agreed-upon conceptual structure, however, it is obviously difficult either to evaluate alternative approaches or to design a sensible research arid development strategy (the two main objectives of this project). Figure 4. ~ represents our first step in building a paradigm for at least our own purposes within this project. At the highly abstract level at which it is posed, this figure is probably not very if Shone differences exist amon, many researchers concerning the details about how the paradigm should be implemented (at the risk of considerably over-simplifying the debate, this often pits economics-based ';utili~ maximizers" versus sociology-based "mle-based modelers"). Nevertheless, considerable consensus exists concerning, the overall nature of and need for the activity-based approach. - 72
TCRP H-12 Final Report . controversial. It also however, requires considerable more detain In order to become operational. As a second step, Table 4.3 presents a set of axioms which we wall use in our specification of art rtideal" model In Me next section. Some of these axioms involve assumptions concerning real- worId behavior, while some express basic strategies for modeling this behavior. Ale label them axioms in that they are hypotheses which are largely untestable but if accepted as "true" they can form the basis for an internally consistent modeling system. As with Figure 4. ~ we believe that most of these axioms are relatively non-controversial. One possible exception is the assumption of no system equilibnum (Axiom 61. While we believe it to be sound behaviorally. Axiom 6 certainly is at odds with the fairly strong equilibrium assumptions which are made in most currently operational models. How-ever, the application of Axiom 6 -- and other axioms -- to this discussion wiiT become evident below (see. for example, Section 4.5.3.2 regarding Axiom 6~. 4.5.3 Mode! Design 4.~.3.! Introduction Table 4. ~ in Section 4.3 lists a set of ~nte:~,rated urban mode] design issues. Each of the issues listed in the table was briefly discussed in Section 4.3 at a very general level. The focus of that discussion was on motivating the importance of each issue within the model design process. rather than on deciding how to best address each issue within a particular mode} implementation. In this section we attempt ~ more detailed discussion of these issues in teens of how an "ideal" model would address each issue. In so doing we cumulatively sketch out the specification for what such an ideal modeling system would look like. The discussion focuses in the following sub-sections on issues related to the representation of the three key components of the urban system (Axiom 21: the phi sical ss stem, decision-makers within this system, and their decision processes, respectively. The "generic issues" listed in Table 4. ~ are discussed throughout. as required. The primary purpose of this discussion is the specification of an ideal modeling system towards which any given practical system would presumably stnve. Therefore, consideration of practical constraints associated with implementation of this ideal system (data computing, etc.) are deferred until Chapter 6, which deals with them as part of its discussion of research. development and implementation issues and options. In particular, Axiom 1 1 ("the ideal model should be conceptualized at a very fine level of representation") is maintained throughout this discussion while recognizing that any practical implementation of this ideal model may well involve "backing off'' from this extreme level of disaggregation due to data limitations, computational constraints, etc. - ?^ 1 ~
TCRP H-12 Final Report Table 4.3 Integrated Urban Systems ~ Modeling Axioms 1. 2. 4. 5. In referring to the urban system, Me focus Is on Hose elements that influence and/or interact with the transportation system. Notwithstanding, the model should be extensible as appropriate. The urban system consists of physical elements, actors and processes. representation of this urban system must contain all three of these. The modeling The transportation system is inherently multi-modal and involves the flows of bow people and goods. Markets represent the basic organizing principle for most interactions of interest within the urban area, providing price and time signals to producers (suppliers) and consumers (demanders) of housing, transportation services, etc. Flows of people, goods, ~nforrnation and money through time and space arise as a derived demand from market interactions that are distributed in time and space. 6. Urban areas are open, dissipative systems subject to external forces. As such, they never achieve a state of equilibrium. 7. The future is path-dependent. In order to generate a forecast year end state, the model must explicitly evolve the system state over time. 8. The mode} must address both shorthand (activity/travel) and lon~,-run (land development, transportation infrastructure, etc.) processes. There must be feedback / interaction between both processes. 9. Some factors and processes are clearly exogenous to the urban system per se. Others may be treated as exogenous as a modeling strategy. 10. Some activities within the urban area are 'basic' in the sense that they arise in response to external demand. 11. The ideal model should be conceptualized at a very fine level of representation (i.e., analytical units) so as to maximize "behavioral fidelity" in the representation of actors and processes, recognizing that any practical implementation probably will occur at higher levels of aggregation. - 74
TCRP H-l ~ Final Report 4.~.3.2 Physical System Representation Time. As stated In Axiom 6, it is unlikely that urban areas are ever In a state of formal equ~libnum. Reasons for this assertion include: . - 3. Urban areas are "open" systems which interact with the woAd around them. People, money, etc.. flow into and out of urban areas in time-vary~ng, non-balancing ways (i.e.. inflows rarely if ever balance outflows). External "forces" such as taxation policy, interest rates. immigration policy, gasoline prices. etc., act upon urban areas over time. Under such conditions, urban areas are never "left alone" long enough to achieve equilibrium. Endoger~ous. dynamic demographic and economic processes provide additional forces which drive the evolution of the system state over time. For an urban system to achieve equilibrium at the end of each time step it must be the case that all processes must be "fast" relative to the time step so that all information can flow hom agent to agent and so that all markets can "settle" into an equilibrium state. This is an extremely strong assumption, for at least three reasons: In the real world of imperfect information. time lags must inevitably exist between action arid reaction: today's decisions must be based on yesterday's information arid it Will only be tomorrow when we wall find out if today's decision vitas correct or not (at which time we may decide to make some further adjustment). Many land-use (and perhaps also some location choice) processes are very long term in nature. Thus, at any particular point in time many actions may be "in progress" and so, by definition, not in art equilibrium state. Considerable inertia exists with respect to many decision processes, especially major ones such as land-use and location choice. This is likely due to a number of factors including: imperfect information (an actor may not know that there is a better alternative available), risk avoidance (one may not act until one is very certain that the new alternative is superior to the status quo) and unobserved transaction costs (it is judged not to be "worth" taking action until a significant gain can be achieved). The result tends to be "lumpy" decision processes characterized by typically longer periods of inactivity in the given market, punctuated by typically short periods of active participation in the market, where this activity might be triggered by a number of different factors tMiller and Sargeant, 19874. This type of - 73
TChP H-12 Final Report . . decision process is not directly consistent with the marginal compensatory models of decision-making generally assumed in equilibrium models (i.e., models in which increment changes in attributes result in incremental changes in behavior). If a mode! is to be constructed which forecasts fixture equilibrium end states. it must be assumed that the pointers) in time Mom which the observed data are drawn to calibrate the model also represent equilibrium points. Since we inevitably observe the urban system at arbitrary points in time. as well as generate forecasts for similarly arbitrary points in time, it logically follows that we are assuming that the urban area is always in an equilibrium state. This further accentuates the problems discussed above. Thus, while many processes may be "equilibrium-chasing'" it is unlikely that they achieve equilibrium at any given point in time. As a result, integrated urban modeling systems should use models of decision processes and market interactions which do not depend on equilibrium assumptions. Exceptions to this broad statement can occur for sub-processes which are demonstrably "fast" relative to the mode! time step. For example, it may well be perfectly reasonable to argue that the modeiina of mowing peak-period flows on the road network for a "typical" day within a one-year time step can be handled within an equilibnum framework (given current land-use, etc.~. This is because the adjustment process associated with these daily peak-penod flows is very fast relative to the land-use and location choice processes which are the determinants of the activity patterns generating the ori~in-dest~nation flow patterns being assigned. ~ ~ is. Axiom 7 states that urban systems are path dependent. This simply means that tomorrow's system state depends rather heavily on We state ofthe system today; and that we move into the fixture by making decisions today, given today's understanding of what the "best" course of action is. If a building is currently standing on a parcel of land, it will be there tomorrow, unless we explicitly decide today to demolish it. If this building is currently occupied by a given tenant. no-one else will be able to occupy it in the fixture until such time as the current tenant vacates the building, thereby making it available for someone else's use. For a path-dependent system, the only way to forecast a future year end state is to "know'' the current system state arid then explicitly trace out how this system state is likely to change over time. The combination of Axioms 6 (no equ~libnum) arid 7 (path dependency) implies the need for a simulation framework for integrated urban models. That is, the mode] assumes the existence of art initial system state which it then explicitly evolves over time -- typically using a fixed time step to incrementally update the system state over time -- until the future end state is reached. The model, therefore consists of: - 76
TORT H-17 Filial Report . the set of rules or procedures which determine how Me system state changes in each time step as a function of both its current state arid any exogenous forces/inputs which are currently acting on the system, arid Me data management system (or "accounting procedures") required to keep track of the typically complex representation of the system state which is stored in the computer. Simulation is essentially Me only method available for modeling dynamic, path dependent complex systems. All current dynamic or quasi-dynamic urban modeling systems use some form of simulation structure for their Implementation. It is generally assumed that the initial system state is obtained Tom observed base year data. As urban simulation models become more disaggregate, however, it is not always the case that sufficiently detailed observed data are available to define fuller the initial system state for the model. In such cases, the detailed input data required must be synthesized in a statistically appropriate manner from more as; regale available data. Base data synthesis procedures are currently receiving considerable attention by researchers [Beckrnarl~ et al. ~995; Miller, ~996] and are fast approaching a practical, operational capability. Given the long time-frame for lar~-use and location choice processes, integrated urban models should be feasibly applied to forecast time periods of at least 20 years, and perhaps up to as much as :0 years, into the future (recognizing the Increasing uncertainty associated with moving further and further into the future). From a modeling point of view, this largely has implications for the computational and data storage requirements of the model. as well as the reliability of exogenous inputs required by the model over the entire span of the forecast penod. Most content operational models are in principle flexible with respect to the time step used to move the system state into the future. However, in practice, five year time steps tend to be the norm. This choice is largely driven by a combination of data availability for model calibration and computational considerations. A five year time step, however, introduces considerable "temporal aggregation" into the model, with the associated potential for sigrnficant aggregation bias. In particular. many real-estate and other market processes tend to exhibit ~ 8-24 month "boom-bust" cycles, while many household-level decisions (demographic and economic) tend to occur on approximately a yearly basis. Given these observations, it is highly recommended that a one-year time step be adopted as a reasonable compromise between accuracy in the representation of temporal processes (which are, of course, actually occurring continuously throughout time) and computational requirements. Axiom ~ identifies the existence of multiple time-Eames within the urban system. Some processes are very short-run in nature: in the limit. for example, roadway operating conditions are - 77
TCRP H-12 Final Report changing on a second by second basis; while at the other extreme, major land-use changes play out over time periods of several decades. Interfacing short-run activity/travel processes arid longer-mn land-use and location choice processes within a fillly integrated modeling structure is a very difficult task on bow practical and theoretical levels. Even In an "ideal" model, many practical compromises are likely to be required wad respect to this issue. These compromises are driven both by computational limitations and limitations in our theoretical understanding on how short-run experiences contribute to long-run adaptation. Given our current expectations concernung both of these constraints, it is reasonable to expect models to be developed with the following features: As discussed above, land-use. location choice (by households and finals), household auto choice and demographic processes are modeled on a one-year time step basis. A "typical weekday" of ac~vity/~avel behavior is modeled, perhaps yearly, perhaps on a longer penod (e.g., every five years), depending on need =d computational capabilities. Ideally this "day in the life" is modeled on a 24-hour basis; at a minimum, peak and off-peak conditions are suitably modeled so that ener~y/emissions calculations, etc.. cart be credibly estimated. Ideally, activitr/uave] for a "typical weekend day" is also modeled. Feedback from the activity/trave! model to the Tonger-term components is lagged from one or more previous time periods. For example, travel utilities from the previous the period may be used to define accessibilities which enter into location choice decisions in the current time period. Land. The basic unit of land in an ideal mode! is the individual lot. Attributes of each lot include: . . . x-y coordinates defining the Tot centroid and boundary lot area, permitted uses, densities, etc.; physical pre-existing conditions or constraints (e.g., if previously an industrial site with contaminated soil, it may be precluded from having residential development), current usage, accessibilities to work, school etc. by both transit and road (i.e., how the lot is connected to the transportation network); market potential (i.e., suitability for development of various types); current value; and other attributes as required and available. Once a land database is established at this level of detail, it can be aggregated up to any zon system for analysis and display purposes. Similarly, attributes of adjacent lots (current usage, etc.) - 78
TIP H-12 Final Report Can be identified, and their impact on a given lot's market value, etc. accounted for. Current Geographic Information Systems (GIS) provide the ability to operate at this extreme level of spatial detail (and even beyond. As with time, representation of space is fundamental to the conceptual integnty and operational capabilities of the model structure. A major difficulty with current models is their extreme spatial aggregation. This is narticularlv troublesome from the point of view oftransit oolicY ~. . .. . _ ~ _ analysis, since the large zones typically used in current models simply cannot properly capture transit access/egress effects (which are so fundamental to the determination of transit ridership," the development impacts of rail infrastructure (subway station location are lost within the large area of the zone), or the impacts of neighborhood micro-design alternatives on transit and non-motorized travel. All of these issues are, in principle, addressable once one moves to the individual lot as the basic level of analysis. Building Stock. In addition to land, the build environment must be explicitly represented in an ideal model, since people and firms occupy buildings, not land per se. Households and firms can only move into vacant buildings/floor space. The amount of built space on any unit of land represents the current upset limit on hove- many people/hov~- much activity can occur at that point in ...... . .. .. , . ~ space. without an explicit representation of building stock, it is impossible to constrain properly location choice decisions. Built space can be measured either In terms of the number of 'units' (typically the most useful measure when dealing with residential housing) or Moor space' (typically the more useful measure when dealing with commercial activities). These measures represent the commonly used variables to which Ravel activities cart be related. In addition to housing and "normal" commercial buildings, major, "special generators" such as stadia airports, universities, etc., should be explicitly represented in teens of the annbutes to which travel activities can be related (number of seats, enrolment, etch. On each lot there either is or is not a building with a given amount of floor space. Other attributes of bu~dings/floor space may include: It is recognized that very few, if any, urban areas are currently completely "there ' in terms of having a 1 006/0 .'clean" and operational database such as that envisioned here, and the time and effort involved in developing comprehensive GIS databases should not be underestimated. Nonetheless, the state-of-the- practice is moving towards the required level of spatial detail in a number of urban areas. 12 We refer here to the ways in which a "piece" of led Is connected to the transit network, including the need to ensure that a particular piece of land is connected unambiguously to transit stops or stations (which relates to the level of spatial a;,are:,ation or disa~gre~ation), access times between the piece of land and the transit network, etc. - 79
TCRP H-17 Final Report function (retail, office, residential, etc.~;~3 size/density; tenure (ownlrent); puce; age, state of repair/quality; occupancy status (occupiedlvacarlt); and other attributes as required and available. Transportation Networks. By this we mean the representation of the supply of urban transportation facilities (infrastructure) and services. These may be (and generally are) supplied by public agencies or private firms, both of which offer services to the public (e.g., public transit agencies and for-hire trucking firms), or by "own-account" individuals and firms who "supply themselves" (drivers of personal cars, own-account trucking by firms etc.~. In an ideal mode} all relevant infrastructure would be represented, including: . roads (all major roads in Me functional hierarchy -- freeways, arterials, etc. -- down to art "appropriate" level in the hierarchy); transit facilities (rat! lines, dedicated busways, etc.~; sidewalks, bicycle paths, etc.; parking; arid relevant components of control systems, as required by the network perfonnance models used. In addition, all relevant modes would be included (Axiom 31: . · auto (SOY, HOV; driver versus passenger); · transit (bus routes, commuter rail routes, etc.; accounting also for "modal combinations" in which the auto is used to access transit services); paratraIlsit (taxis, jitneys, shuttles, etc.), . . walk, bicycle; truck, goods movement; and rail, goods movement. A~butes of the network, on a link-by-link basis include: is Note that multiple uses often exist within a single building,, so we are really talking about the distribution of functions which are existing or feasible within the building. - 80 -
TCEP H-12 Final Report physical attributes (number of sanest speed limit. lane capacity, etc.); connections (i.e. intersection confi;~,urations' delay, control type etc.); service attributes (frequency, reliability, etc.), travel time; travel cost; comfort, safety, personal security, flow; operating cost (i.e., cost to Me service supplier as opposed to the "travel cost" -- the cost to user of the service), emissions generates ; energy consumed; and other attributes or outputs (e.g., accidents, side friction due to parking, etc.) as required and available. Central to the design of the transportation network representation is the nature ofthe network period ce and route assignment mode] to be used. The current, dominant trend in this regard is to the development of dynamic, microsimulation network models (TRANSIMS, DYNASMART, INTEGRATION, PARAMICS etc.~4 which require very detailed representation of the (road) network arid which generate time, cost and flow information for each link dynamically over time. Such models are generally viewed as being required to generate Improved estimates of emissions and energy consumption. as well as to represent system behavior under re~-t~me control and other ITS related options. However. the extent to which this level of detail is required in art integrated urban model, or (if it is not required) how such detailed operational models impact arid He impacted by integrated models. are questions for fimber research and testing. Services. The development of tared depends on He provision of many services in addition to transportation. These include: . . . . sewers water elecmcity/~as communications (telecommunications. fiber optic networks. etc.) heatin;,/cooling proximity,- to emergency services (fire stations, hospitals. etc.) ]4 These are all road assignment software packages. However, disaggregate transit assignment packages also exist. These transit assignment procedures are comparable to the road procedures in terms of level of detail of representation of the transit network and In assigning point-to-point Hips over the network. They do not, however, have the same dynamic representation of these flows through time, with both being more traditional "static ' procedures. See, for example, Chapleau [19863. - 81
TCRP H-12 Final Report . . . As currently envisioned, even an "ideal" mode] would probably have very limited representation of these services. At a minimum a simple "serviced/not serviced" flax may well prove adequate. The architecture of the mode] software, however. should be extensible to include a broader arid more detailed representation of services as time. opportunity and need rams. 4.5.3.3 Representation of Decision-Makers There are four fundamental decision-makers in urban areas: persons, households, private ferns, and public authorities (government and other public agencies). All four tripes of decis~on- makers must be explicitly represented in the model. While there is obviously a strong "mapping" back and forth between persons and the households within which they live, both representations are required within the model. Individuals are born, age arid eventually die. They go to school work at jobs etc. Households. on the other hand, are the appropriate decision-making units for residential location decisions automobile transaction decisions, and for providing Me framework for understar~din~ the activity/travel patterns of individual tr~p-makers. Private florins provide the majority ofthe "economic energy" feeling urban processes. They occupy land arid buildings, they demand and supply goods and services they employ workers. Trio tripes of few which are of particular Interest within the model are developers (whose business it is to develop land and construct buildings) and transportation firms (whose business it is to provide transportation services to themselves and/or others). Much of the behavior of public authorities lies outside the domain ofthe mode} per se in that it represents the political and bureaucratic processes of public policy debate and decision-making, the outcomes of which become exogenous inputs to the model. In addition however, public authorities are typically major employers arid consumers (and providers) of land, floor space, transportation and other goods and services within urban areas and so have endo~enous roles to play within the mode] as well. Thus, it is import within the conceptualization of the model structure to recognize the role of public authorities within urban areas, both as "exogenous drivers" of the urban system, as well as endo~enous participants in the urban community's daily life. Table 4.4 presents a representative list of possible attributes of persons. households, firms and public authorities which might be included in an ideal model. In this table "utility" simply is short-hand for whatever measures of person and household "satisfaction", "well-being", etc., are being used within the mode] (discussed farther below). - 82
TCRP H-12 Final Report Table 4.4 Decision-Maker Attributes ~ : ~Actors i: Aft: ~ | ~ ::: : Attributes ~l Persons ·Quality of life: - utility being achieved - travel {consistent with actrri~-based travels · Work . · Over - place (location) utility - other consumption / activity - total utility': ~ Gavel, place. other utilities · Age · Sex · Driver's licence · Mobility restrictions · Income Awake plus other; also taxes) · Household role · Employment (related attributes;: - potential occupations - current employment status: - not in labor force - unemployed - employed: -part time/fulltime - occupation - job location - u age (and taxes) · Education related attributes - highest level of schooling attained - current educational status: - not a student - part time / fills time - level - location - 8:
TCRP H-12 Final Report Table 4.4 Decision-Maker Attributes i: Actors -I ~ : :- i: ~ : ~ Attributes : ~ ~ ~ :: ~ ~ ~: --it ~ : .. . Households · Dwelling unit - location - attributes - property taxes · Tenure · Household utilities = ~ person utilities ·Vehicle availability - number -type · Life cycle point of Be household (e.~., defined by number of children etc.; will influence place / travel / other utilities) Private Establishment · Industry type fi.e., a f rm, or a unit of · Finances thatfirm, that has a - gross revenues unique location. (NB: - gross costs (occupancy costs; including taxes, carriers and developers especially property taxes) fit within this framework, · Inputs but are Two categories - capital that have special roles / - number of jobs by occupation type impacts in this model. - other roods and services (commodities, etc.) Although these may be - space (location) publicly-owned, they tend · Outputs (goods, services and infrastructure) lo behave lilts private · Own-account shipping (yes / no): establishments.)y - If yes: - mode - fleet size Public Authorities fit e., · Industry type according to unique · Finances functional location) - revenues - costs · Inputs - capital - number jobs by occupation type - other goods and services Commodities etc.) - space (location) · Outputs (goods, services and infrastructure) · Own-account shipping (yes / no): - If yes: - mode - fleet size - 84
TCRP H-17 Final Report - In the market framework to be descnbed fur~er below, each of these four actors generally has several roles; sometimes as a producer or supplier of goods and services, and sometimes as a consumer or dem=der of these goods and services. For example, persons supply labor to firms, who "consume" this labor. Retail fimns produce retail goods which are consumed by shoppers. Retailers, in turret, are demanders of these same goods from wholesalers. And so on. Table 4.5 illustrates some of the key production/consumption roles played by the four actor types with respect to key sectors within the mode] (transportation, floor space and other goods and services). 4.~.3.4 Representation of Decision Processes Market Processes. As outlined in Axioms 4 and i, a fundamental org~z~ng pnnciple for urban areas is the market within which goods. services. =d money are exchanged between producers and consumers. Prices for these goods and services are detennined through the market interaction; these prices, In turn, determine the level arid nature of the exchanges Cat occur within the market between supplier and demander. Without explicit representation of the demand and supply processes at work within urban markets-- and of the price signals which mediate between demand and supply -- it simply is not possible to mode! adequately the outcomes from these markets. Table 4.6 lists the key markets at work within urban systems along with the demanders and suppliers in each case. As is clear Tom this table, market processes determine virtually every element of interest within the land-use - transportation system, Including land development. location choice processes, travel, and automobile ownership. As has already been discussed, individual actors (persons, firms, etc.) participate in these markets in complex ways, sometimes as producers and sometimes as consumers. Development of an "ideal" model involves the explicit specification of demand and supply processes for each of the markets listed in Table 4.6. While beyond the scope or capabilities of this current study, this specification involves Get ring for each process the actorks) involved, the specific decisions which they must make, the attributes of the actors which are salient to their decision- making process, the attributes of the alternatives and the decision context which affect the decision, and the "market clearing" mechanism by which demanders and suppliers interacts ultimately resulting in the determination of prices and the exchange of goods and services. Central to the development of these models is the issue of actor motivations, which is discussed briefly next. - 8:
- . . 1: =,,: : i, ~ ~ ! I - := ~ 3~^ ~1 - z~ r ~ ~ 1 :~ ~ _ ~4'== 1 00 1 ~4 ~ . .- _ _ A, _ _ O ,£4- ._ ~ A,,= _ O A_ ~ C ._ ~ it, O ~ ~ _ _ ~ _ A, _ _ _ : ~ ~ ?;~ hi. _ o- o ~ rat ~ _ O ._ ~ _ _ ~ ; rat ~ SAC ~ I_ O~ ~ ~) C ~ I. ~, q~ ·_ _ ~ ~ _ HE ~
TCRP H-12 Final Report Table 4.6 Markets in the Urban System .... -I . - ~ - . I. ~ ~ . . . . ; . .. . - ~ . .. .. -:~: --Market :::- ~ :::: --I :- -:~::Demanders : :~ : -:: Suppliers Housing Market Household Public authorities Developers Household Floor Space Market ~Firms Developers Public authorities Firms Goods and Services ~ Persons Firms Market (includes Households Public authorities education) Firms Public authorities Job Market Firms Persons Public authorities Personal Transportation Persons Persons Market Households Firms Public authorities _ Goods Movement Market Persons Persons Firms Households Public authorities Firms (couriers etc.) _ Inhas~ucture Market Persons Public authorities Households Firms Firms (couriers, etc.) Public authorities _ Auto (Vehicle) Market* Households Firms CAB: exogenous) Firms . Public authorities * Interest is primarily in the personal vehicles. Motivational Frameworks. In order to model actor decision processes in either a supply or a demand context' one must have some understanding of the motivations driving the actor ~ question. For persons and households, we believe it is reasonable to assume that this motivation is the maximization of the "joint utility" of household members, subject to income, time and capability constraints. At this level of the conceptualization, "utility" is used in a very loose sense of "well- being" and does Ilot necessarily imply any particular modeling paradigm (e.g., random utility theory). - 87
TCRP H-12 Final Report - Rather it simply represents the very basic assumption that we derive utilit~- fi:om (i.e.. we need) . .. . .. ~ . . .. . . . . . . . . . . . .. ~ ~ . . .. . sheller arid security tarlc so enter me nousm, market in order to obtain shelter). loon. clothing, etc. (and so enter the retail market to obtain these goods), and so on. ~ order to obtain these things within a market economy we need money, which we usually obtain by entering Me labor market and selling our labor skills and knowledge for wages. Further, we need access to opportunities (jobs, education, shopping, etc.) and so enter travel markets in order to gain physical access to the jobs, schools, goods, services' etc. which we need and want. Thus. persons arid their associated households enter various markets as demanders (housing, goods, services) or suppliers (labor. resale housing): first, to achieve their basic needs, then to maximize their personal well-be~ng or utility. Firms similarly can be assumed to be motivated by the desire to survive and prosper' where "prospering" might be measured in a number of different ways (net revenue, reman on investment, growth rate, etc.~. Regardless of the measurers) used, in order to survive, the fig must obtain revenue from the production and sale of goods and/or services In response to either an existing demand or anticipated (latent) demand; thus the final becomes active as a supplier in the market for .. . . ~ . . .. . .. ~ . . . . . . these goods or services. in order to Produce these goods the firm requires various inputs ( arid hence is a demander of): labor, space, capital. and other physical Inputs. l he process of producing and selling the goods/services also generates flows of goods tofDom the point of production. as well as flows of business travel, thereby making the firm a demander (and sometime supplier) of transportation services. Finally, it is assumed that the motivation of public authorities is to maximize social welfare, subject to societal budget, time and cacabili~v constraints. recoan~z~n~ that auantifvin~ this concept . .. it, .. . .. d. - 4 ~ ~ , _ ~ may be very o~cu~t In practice. In order to maximize (or at least to improve) social welfare, public authorities often directly enter markets as suppliers of services and/or infrastructure, open in situations in which private forms will not enter due to a perceived lack of profits. In addition, public authorities seek to guide markets towards socially beneficial states through a variety of policies, such as: provision of subsidies. taxation, income redis~bution~5 and regulation. In order to accomplish these tasks, public authorities (just like private firms) enter markets to demand labor, space, goods and services, and transportation. Demographics. While implicitly shown as an exogenous "driver" of the urban system in Figure 4. I, demographic processes are. in fact, an integral part of Me urban system and its internal dynamics. Births, deaths aging, household formation/evolution/dissolution are fundamental processes determining the characteristics of the population arid thereby the demand for housing, education, jobs, goods and services, etc. Much of the explanatory power of disaggregate models of human i5 This is somewhat redundant, since income redistribution is generally achieved through some combination of subsidy and taxation. Subsidies and taxation, however, can be used for other purposes than income redistribution (hence the distinction, for the purposes of this work). - 88
TCRP H-19 Final Report decision-making comes from being able to specify the atinbutes of the individuals involved, and hence to be able to say something with reasonable confidence about their tastes and preferences, etc. If such disaggregate decision models are to be employed effectively, then the overall modeling system must be able to supply these decision models with the required decision-maker attributes. As a result, a central and significant component of an ideal integrated modeling system is a strong, endogenous, dynamic model of person and household demographics. This includes both the capability to synthesize (if need be) the attributes of individuals and households in the base (initial) system state, and to "update" or "evolve" these attributes over time within the overall simulation run. Regional Economics. This relates to Me overall regional economic system within which individual firms operate and compete, arid which determines the overall flow- of goods and services both internally within the urban area and into/out of the urban area as import/export flows between the urban area and the "outside world." Some regional economic factors/processes are exogenous to the urban area and so to the model. such as interest rates, inflation, national immigration policy, and perhaps total production levels by industrial sector. Other aspects are endogenous. In current models, this component, if present, is usually handled through some form of input/output model, and this is the most likely approach for models to maintain for some time to come. 4.5.4 Summary Table 4.7 provides a brief summary of the attributes of the ideal integrated urban model discussed above. These attributes are grouped according to the three main categories defined in Table 4.1: physical system, decision makers and processes. Land development, location choice processes, and job-worker linkages are all modeled as economic markets with explicit supply and demand functions and procedures for price determination and "market clearing" (i.e., the allocation of supply to demand). The model is envisioned to be dynamic, disaggregate and behaviorally sound. As such, it will be sensitive to a wide range of land-use and transportation policies and able to trace the direct arid indirect impacts of any of these policies through time and space. No attempt has been made to specify; detailed formulations of individual sub-models within the overall modeling system. Many options typically exist here, and much research is required in order to translate this very general model into operational practice. Similarly, no attempt is made here to address the data and computational requirements of such a model, except to note that such a modeling system is almost certainly not beyond our current and emerging capabilities [Miller and Salvini, 1998]. - 89
TCRP H-12 Final Report Table 4.7 SuTr~nary of Ideal Integrated Mode! Attributes PHYSICAL SYSTEM Torte: Dynamic evolution of the system state In one-year time steps. System state generally not in equilibrium. Interactions between long-run and short-run processes are ';properly" accounted for. Land: The basic unit of land is the individual lot. Building Stock: Building stock is explicitly represented. Each lot has a certain amount of floor space, characterized by type, price, etc. Transportation Networks: Full, multimodal representation of the transportation system used to move both people and goods. Sufficient spatial and temporal detail to properly mode! flows, network performance, emissions, etc. Ideally, a twency-four hour network mode} to be used. Services: Sufficient representation of other services for the purpose of modeling land development decisions. DECISION MAKERS Persons and Households: Both persons and households are explicitly maintained (with appropriate "mappings" between the two entities) in sufficient detail to mode! the various processes of interest. Firms: Explicitly represented. Firms at least as important as households in the overall system: they occupy land / floor space; they employ workers; and they buy / sell goods and services from / to themselves and households. Firms are modeled in sufficient detail to capture adequately their behavior within these various roles. Public Authorities: Represented within the mode! to the extent they generate purely endogenous effects (employers of workers, demander / supplier of services; etc.~. Will remain represented largely by exogenous inputs to the model. - 90
TChP H-12 Final Report Table 4.7 Summary of Ideal Integrated Mode! Attributes, cont'd PROCESSES Markets: Land development, residential housing, commercial floor space and labor all function within economic markets which possess demand and supply components and price signals which mediate between demand and supply. These economic markets must be explicitly modeled if them behavior over the is to be captured properly. Demographics: Demographic processes should be modeled endogenously so as to ensure that the distribution of population attributes (personal and household) are representative at each point of time being modeled and are sufficiently detailed to support the behavioral decision models being used. Regional Economics: Essential components of urban production / consumption processes should be modeled endogenously. The mode} should also be sensitive to macro exogenous factors such as interest rates, national migration policies, etc. Activity / Travel: The Gavel demand component of He integrated mode] should be activity- based and sufficiently disaggre~ated so as to properly capture trip-makers' responses to a full range of transportation policies, including ITS and TDM. Automobile Holdings: Household auto holdings (number of vehicles, by type) should be endogenously determined within the model. 91
TCRP H-12 Final Report - 92
