PECAS Recent and Upcoming Work June 2010 John E Abraham with the HBA and UCDavis PECAS Team
Part 1: Recent Work Socioeconomic/Floorspace Rent Transport costs Setup of production Activity Allocation module for Long Distance Commercial Vehicle base matrices Software improvements ·
Part 2: Upcoming Work Floorspace Synthesizer Disaggregate SD estimation Cluster analysis of technology options
Socioeconomic/Floorspace Population Synthesizer Employment Synthesizer Floorspace type allocation Floorspace quantity calculation Eric Lehmer, Shengyi Gao, John Abraham, Nicholas Linesch
Synthesizer Software “Combinatorial Optimization” Random changes to optimize objective function Add, remove or subtract PUMS households from TAZ’s “Simulated Annealing” Early on, accept changes that do not improve objective function with some probability Stops process from getting “stuck” at local optimum Based on annealing of crystals in metallurgy (Not found to be necessary with well chosen target weights)
Population Samples Data source: census 2000 PUMS Synthesizing all people in each TAZ Sampling: taking all the samples within the PUMA in which a TAZ nests and 20% random samples from the adjacent PUMAs Duplicate unique IDs based on their weights to get a full population Control variables in the sample: household size, housing type, income group, age, number of cars, occupation, student status, group quarter flag Automated with Python
Population Targets Data source: census SF tables Control variables: household size, housing type, income group, age, number of cars, occupation, student status, group quarter flag, number of rooms, number of bedrooms Targets: population by TAZ and each control variable Automated with Python
Population Synthesizer Programmed in Java Maximum iterations: 1,000,000 per PUMA Outputs: the same unique person IDs as in the PUMS by TAZ Goodness of fit: simulated volume/targets > 0.99 Synthetic population by TAZ: each TAZ has all the population (unique person and household IDs) with all the census variables and aggregated PECAS household types, industries, occupations, and synthetic floorspace
Processing of Population Synthesis Outputs Java code outputs a CSV file with 12208610 synthetic households. Each record has the TAZ and PUMS Household ID number (SN). This output was joined to a modified PUMS household table (hh) to get detailed information on each hh. The PUMS hh table was modified in the following ways: (1) ALL SENIOR HOUSEHOLDS: In order to get the population into the correct PECAS household types, households with all seniors had to be determined. This was done by examining the person records for the members of each household. Those households with all members > 65 yrs old were set to all senior. (2) HIGHRISE PUMA: In an attempt to capture the residential highrise spacetype we visually identified the PUMAS with existing highrise buildings and and this was then attached to the hh using the PUMA id. (3) URBAN PUMA: Visual representation of urban PUMAs. This was used as a criteria for the housing value calculations of the spacetype determination. (4) MILITARY HOUSEHOLD: Single person households where the person was active military living in group quarters were identified. This is so they can be put into gq spacetype.
Calculation of Residential SpaceType and SpaceQuantity • A SQL Server view was created based in part on work done by SANDAG. This view split the hh into different spacetypes based on several criteria: (1) PUMS BLDGSZ attribute. The building size value determined whether it was mobile home, sfd, multi-family, etc. (2) ACRES attribute. Determined whether it was rural, acreage, or SFD. (3) The Urban PUMA combined with hh VALUE and hh RENT attributes determined whether it was Econ or Lux spacetype. (4) The Highrise PUMA attribute allong with the hh income category helped us determine which large apartment buildings might be highrise buildings. (5) Military single person hh were put into GQ The quantity of space was also assigned using the PUMS attributes of ROOMS, BEDRMS,YRBUILT and square foot estimates derived from SANDAG.
Employment synthesizer
Employment Sample and Targets Data sources: census 2000 CTTP and PUMS; infoUSA Synthesizing all jobs in each TAZ by PECAS industry Sample: all persons with a job in PUMS with CTPP industry and occupation Targets: 1) CTTP employment by PECAS industry or occupation; no crosswalk; weight 1; 2) InfoUSA employment by TAZ, PECAS industry and occupation; weight 0.2 0.01
Employment Synthesizer Programmed in Java Maximum iterations: 1.5 billion Output: PUMS unique person IDs by TAZ Goodness of fit (GOF): simulated volume/targets at TAZ level; 1) for CTPP targets, GOF > 0.99; 2) for InfoUSA, GOF = 0.65; 3) for the pooled CTPP and InfoUSA targets, GOF = 0.82 Synthetic employment: all employees with PUMS variables within each TAZ Floorspace use: Floorspace use rate: surveyed average floorspace rate by PECAS industry A floorspace type is assigned based on existing land use type within each TAZ (if a space type exists, then assign it by allowable use; otherwise, randomly assign a space type)
Processing of Employment Synthesizer Outputs: Java program outputs a CSV with 14285953 synthetic employees. Each employee has a TAZ and PUMS person table id attached. The PUMS person table was augmented in the following ways: (1) PECAS Industry and Occupation reclassification of PUMS records to fit employees into correct PECAS activities. (2) Blue and White Collar labor split. (3) Agriculture and Construction activities randomly distributed to finer classification than PUMS using California Statistical Abstract proportions. For example, PUMS does not differentiate cattle production. We randomly assigned them to: CATTLE RANCHING AND FARMING production, DAIRY CATTLE AND MILK production, or NONCATTLE ANIMAL production.
Calculation of Industry SpaceType and SpaceQuantity: Based on the synthetic employee’s PECAS activity type we can determine • which spacetypes are possible for that employee to occupy. Based on the synthetic existing land use layer we can determine which • spacetypes are available in the employee’s TAZ. Unable to determine the proportions that employee’s of a particular • activity consume spacetypes we decided to randomly assign employees to spacetypes that were available or if no spacetype was available we randomly assigned them to a possible spacetype and call that non- conforming space. Quantity was based on activity type with each employee consuming a • certain amount of floorspace. This was then summed up by TAZ and SpaceType to form FloorspaceI for Industry.
Rent Estimation
Rent Estimation Separate rent (use value of space) into two parts LUZ value based on accessibility and supply/demand relationship Modifiers based on site and structure specific attributes
Estimation Results (Residential Space) Rural Economic Residential* Rural Luxury Residential Single Family Detached Economic Single Family Detached Luxury Residential Residential Max Coefficient Function Max Coefficient Function Max Coefficient Function Max Coefficient Function Distance Distance Distance Distance Age -.176 Power -.176 Power -.007 Power .180 Power Coast 1.5 Shift. Expo 1.5 Shift. Expo 1.5 .012 Shift. Expo 1.5 .218 Shift. Expo Freeway 2 -.198 Shift. Expo 2 -.198 Shift. Expo 3 -.005 Shift. Expo 3 -.056 Shift. Expo Highway 2 -.050 Shift. Expo 2 -.050 Shift. Expo 2 .030 Shift. Expo 2 .022 Shift. Expo Park 1 .289 Shift. Expo 1 .289 Shift. Expo .25 .002 Shift. Expo .25 .003 Shift. Expo Rail 1 .379 Shift. Expo 1 .379 Shift. Expo .5 -.018 Shift. Expo .5 -.012 Shift. Expo Ramp .75 Shift. Expo .75 Shift. Expo .5 -.002 Shift. Expo .5 -.013 Shift. Expo School 1 -.262 Expo 1 -.262 Expo 1 .018 Expo 1 .035 Expo Water 2 .072 Shift. Expo 2 .072 Shift. Expo .5 .005 Shift. Expo .25 -.002 Shift. Expo body Transit 1.5 Shift. Expo 1.5 Shift. Expo 1 -.036 Shift. Expo 1 -.011 Shift. Expo Center
Estimation Results (Residential Space) Acreage Economic Residential Acreage Luxury Residential High-rise Economic Residential High-rise Luxury Residential* Max Coefficient Function Max Coefficient Function Max Coefficient Function Max Coefficient Function Distance Distance Distance Distance Age -.036 Power .081 Power -.264 Power -.264 Power Coast 5 .002 Shift. 1.5 .079 Shift. 1.5 1.804 Shift. 1.5 1.804 Shift. Expo Expo Expo Expo Freeway 1.5 .070 Shift. 1.5 -.014 Shift. 2 -.398 Shift. 2 -.398 Shift. Expo Expo Expo Expo Highway 2 .017 Shift. 2 .001 Shift. 1.5 .766 Shift. 1.5 .766 Shift. Expo Expo Expo Expo Park .25 .071 Shift. .25 -.013 Shift. .25 .056 Shift. .25 .056 Shift. Expo Expo Expo Expo Rail 1 -.037 Shift. 1 -.024 Shift. 1 -.255 Shift. 1 -.255 Shift. Expo Expo Expo Expo Ramp .5 -.042 Shift. .5 -.010 Shift. 1 .436 Shift. 1 .436 Shift. Expo Expo Expo Expo School 1 .029 Expo 1 .036 Expo 1 .125 Expo 1 .125 Expo Water .25 .079 Shift. .25 .033 Shift. .5 Shift. .5 Shift. body Expo Expo Expo Expo Transit 3 .090 Shift. 1.5 .041 Shift. 1.5 .033 Shift. 1.5 .033 Shift. Center Expo Expo Expo Expo
Recommend
More recommend
