A new approach to develop large-scale land-use models using publicly available data

Abstract

Developing a land-use model for large-scale cases is a topic that has received less attention in the literature, while it is crucial for transportation engineers and urban planners to analyze the effect of various policies in multi-jurisdiction metropolitan areas and to some extent on a statewide scale. While gravity-based models are too simplistic, microsimulation models require extensive data and massive computation. This paper presents a land-use model that can be applied to large-scale geographies using publicly available data and be able to forecast demographic and socioeconomic attributes with reasonable accuracy and acceptable computational time. The proposed model incorporates Putman’s Integrated Transportation–Land-Use Package (ITLUP) and Kockelman’s Gravity-based Land-Use Model (G-LUM) fundamentals with enhanced formulation of newly added variables and structural changes. Considering the nonconvex and nonlinear nature of the proposed model, we utilize an enhanced genetic algorithm for base year calibration. Further, we assess the accuracy of the model with backcasting validation. We utilize the state of Tennessee as the case study area and utilized all open-source data available to the model application. The model results show reasonably accurate estimates of households by size, employment by industry, and land utilization by condition. As applicable, the model outperforms G-LUM by accuracy ( R 2 and Percentage of Good Prediction ( PGP)) and error measures (Mean Absolute Percentage Error ( MAPE)). The proposed land-use model has the potential to be applied for medium to large-scale geographies with reasonable accuracy in predicting socioeconomic, demographic, and land condition estimates by using publicly available data.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 7 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.