HELIOS: Household Employment and Land Impact Outcomes Simulator FLORIDA STATEWIDE IMPLEMENTATIONDevelopment & Application
Stephen Lawe RSG
Michael Doherty URS
MAY 2013
Presentation Outline1.Model Introduction
Model Goals & Process Model History
2.Model Development Data Used Estimation and its challenges Model structure
3. Model Evaluation Southwest Example
4. Model Use Model Interface Projects where HELIOS was used
Goals during development HELIOS
Consistent Land Use Forecast Initially used to refine the Turnpike ME process Integrate into the statewide and MPO models Support multiple geographic levels (parcels,
different TAZ boundaries, etc.) Sensitive to accessibility. Able to integrate into
any transportation model. Modest Runtimes:
The model runs the entire state of Florida in less than 5 minutes for a specified time period
Analysis of policy measures other than changes in accessibility run in less than 2 minutes
Model History
2005 2006 2007 2008 2009
Model applied to Southwest Florida Investment Grade Project. Land use forecasts were used in Florida’s Turnpike Planning out to 2045.
Model applied to Central Florida Investment Grade Project. Forecasts used to 2060
Implemented for Florida DOT at the Statewide Level.
Implemented for Florida’s Turnpike at the Statewide Level.
Setting the Stage
Peer Reviewed 2010 The Florida MPO Land Use Model Task Force
reviewed a broad range of tools and has suggested that HELIOS be made available across the state.
Application History 14 Managed Lane Studies Wekiva Parkway Suncoast 2 Suncoast 3 [ Tampa to Jacksonville Corridor ] I295 / I95 Managed Lanes Turnpike System Forecasts [ Traffic Trends Process
]
HELIOS Process
1.Determine control totals to be allocated
2.Apply Developments of Regional Impact (DRI) growth to known TAZs
3.Allocate remaining:A. Determine land
availability including converting some agriculture land
B. Apply a probability model to distribute remaining growth to vacant lands and underutilized developed areas
BEBR Forecast (Control Total)
DRIs in County
Allocate remaining
GIS Processvacant residential, non-
residential, and converted agricultural land
1
Probability Model Appliedwith iterative scaling
3
2
A B
Model Development
Data Inputs for Model
Estimation Parcel-level land use Urban Planning Boundaries (urban growth
constraints) Geoprocessing with GIS (proximities) Base Year Socioeconomic Data
Model Implementation Generalized land use Developments of Regional
Impact
DRIs
Data Cleaning – InfoUSA Employment
Land Use Data – Development History (parcel level)
Land Use Data – Lumpy by Year
1,132
849
0
200
400
600
800
1000
1200
1985 1990 1995 2000 2005
Vacant
Build Year
Land Use Data – Spatial Variability (parcel variability)
Land Use Data – Urban Growth Boundaries (legal constraints)
Model Structure – Two Stage Logit/Linear
1. Logit Model Estimates: Pi = Probability of Development
Pi = Pr(Yi = 1 | Xi) =ἀ + β1x1 + β2x2 +…+ βkxk e
ἀ + β1x1 + β2x2 +…+ βkxk 1 + e
Y(g) = β1x1 + β2x2 +…+ βkxk
Where (g) is a log link function
2. Linear Model Estimates: Y = Intensity of Development
Resulting outcomes scaled to control totals
Model Structure – Parameters of Model
Variable Source Effect on Development
Accessibility Travel Model Positive
Distance from Coast GIS Negative
Distance from Arterials & Interchanges GIS/Travel Model Negative
Density of Current Use Parcel Non-Linear, generally positive
Undeveloped Area Parcel Non-Linear, generally positive
Mix of Land-Uses Parcel Homogeneity is positive
Urban Growth Boundary GIS layer Less development outside UGB
Res. & Non-Res. Development History Parcel (dependent variable)
Model Evaluation in Lee-Collier
Planning Region
Model Structure – Calibration Results (Residential Growth 1980 - 2005)
Pearson’s CorrelationTAZ= .58ZIP = .81
Observed Growth Modeled Growth
Model Sensitivity Test - Accessibility
Significant bridge capacity added during 1990s
Large subsequent observed increase in development in Cape Coral
Modeled removal of new bridge capacity
25% decrease in HH growth in Cape Coral
40% decrease in Employment growth in Cape Coral
Legend% Difference HH
-0.53 - -0.40
-0.39 - -0.30
-0.29 - -0.25
-0.24 - -0.20
-0.19 - -0.15
-0.14 - -0.10
-0.09 - -0.05
-0.04 - 0.00
0.01 - 0.11
Model Use
Model Structure – Land Conversion Algorithm
According to US Census of Agriculture, FL farmland declined from 10.4 M acres in 2002 to 9.2 M acres in 2007
We assume this continues; each modeling period, a fraction of agricultural land becomes available for development
Agricultural Land
HELIOS - Final Observations
Inconsistent Inputs to HELIOS It is possible to give the model an “inconsistent set of inputs”. An
example would be growth control totals that exceed the available landHELIOS warns the user and then “softens” the constraint assumptions
to allow full allocation.This provides “what if” testing but also requires user
consideration
Recession Impacts: comparison of 2006 and 2012New Growth assumptionsRevised Occupancy assumptionsRevised Agricultural land conversion assumption
Policy ShiftsShifting DRI Designation
Model Interface
Windows Executable
Text file inputs and outputs
Ability to turn on/off key features (accessibility and distance calculations)
Questions