Microsimulation Of Integrated Land Use And Transportation Based On Gis

Concerns about ineffective land use development, urban congestion, deteriorating air quality, and increasing traffic accidents have spurred interest in investigating and modeling the interactions between land use and transportation. Traditionally, land use models allocate land use based on individual traffic analysis zones (TAZ). They account for the amount of vacant land, zoning policy, land use policy (e.g., urban growth boundaries, land accessibility), population, and employment growth at county levels. However, traditional models do not consider agglomeration factors, the market equilibrium of land supply and demand, and are insensitive to policy changes. To address these issues, this study presents an integrated cellular automata (CA) and agents model that has been developed to simulate the characteristics of land use and transportation, their temporal and spatial dynamics, and the economic situation. Orange County, Florida is used as a case study.A bi-level integrated proto model was first developed to quantify the interactions between transportation and the spatial distribution of land use. To manage dynamic land use changes in both temporal and spatial dimensions, the upper-level model (land use) employs cellular automata (CA) to simulate the increasing land demand, while the bid-rent agent model represents household location choice. The cell-based land allocation strategy and residential location choice generated in the upper-level model are fed into the lower-level (transportation) model to reflect new transportation demands, travel costs, and transportation accessibility. Then, the travel cost and transportation accessibility that are produced in the lower-level model are fed back into the upper-level model. To optimize land use allocation, a combination of Genetic algorithm and Frank-Wolfe algorithm are used to minimize transportation system costs. The application of this model in a fictitious urban area showed that the optimal allocation significantly reduced the system cost of transportation, by 30.8-90.2%, demonstrating a potential improvement in transportation efficiency.In 2009, the Florida Department of Transportation (FDOT) Central Office surveyed FDOT and MPO modelers asking for input on the current state of modeling. The results show that modelers would like to see integrated land use and transportation models (e.g., the Florida Standard Urban Transportation Model Structure, FSUTMS). These can be used to help control air quality and promote sustainable development, as well as provide a close linkage between land use and transportation planning.At the grid cell level (50m×50m), the Multinomial Logit (MNL) relationship (explicit) and Artificial Neutral Network (ANN) ("black box") were used separately in CA to define the transition rules for land use changes in the temporal and spatial dimensions. Further model implementation requires both GIS and additional data, including digital elevation model files, soil data, land use/cover data, parcel data, zoning boundaries data, census data, and transportation network skim files. When compared to the MNL-CA model, the multi-layer ANN-CA model was found to better forecast transportation demand related land types (87.7% vs. 74.4%) and all land types (89.8% vs.76.7%) by using a confusion matrix method. Increasing the hidden layers from one to three improved the prediction accuracy but required more training time. Although ANN accurately simulated large sized lands, it had low effectiveness for small sized lands. For instance, the small number of industrial land cells (2010 cells) developed from land Type D in 1990 had extremely poor prediction accuracy (about 2.6%)The use of agents-based models (e.g., household, employment, and developer agent) in CA models is benefitable. CA analyzes the spatial suitability of land use change, while the agent models represent policy making and micro human decision making entities effects on land use dynamics. The integrated model predicted pretty well for transportation demand land using the MNL-CA-Agents (87.6%) and ANN-CA-Agents model (87.7%). Because the ANN models employ a "black box" technique, MNL-CA-Agent land use model provides a more obvious relationship between spatial variables and land use change, allowing for a better interaction with FSUTMS. For the MNL-CA-Agent land use model, most allocation errors fell within a range of [-50,50] for households (～52%) and employment (～37%). When the difference range is within 200 ([-200,200]), household and employment change was predicted at an accuracy for 75.7% and 69.9% of the zones, respectively.LandSim, a graphical user interface software for simulating land use change, is designed based on the MNL-CA-Agents land use model developed in this study. Integration of LandSim with FSUTMS follows a two-step procedure. First, LandSim generates socioeconomic inputs for FSUTMS. Second, the accessibility and travel time produced from FSUTMS are fed back into LandSim. By comparing the performance of transportation network and emission results, the integrated model framework efficiently evaluates urban land uses, transportation policy, and the complicated relationship between transportation and the environment. Four scenarios were examined for Orange County in 2000,2012, and 2025 business-as-usual, land use integration model, urban growth boundary options, and mixed land use. Results of the model indicate that the integrated model can effectively relieve transportation congestion and reduce transportation emission, which may help to achieve a sustainable urban development.