[Paper Review] Testing the ability of Multivariate Hybrid Spatial Network Analysis to predict the effect of a major urban redevelopment on pedestrian flows

Crispin Cooper, Ian F. Harvey+2 more|ORCA Online Research @Cardiff (Cardiff University)|Mar 28, 2018

Urban Transport and Accessibility12 references3 citations

TL;DR

This study introduces a Multivariate Hybrid Spatial Network Analysis (MHSpNA) model to predict changes in pedestrian flows due to urban redevelopment. By integrating multiple theoretical flows—such as those from stations, parking, and retail areas—into a calibrated spatial network, the model successfully forecasts pedestrian vitality in Cardiff’s town centre, validating its predictive power for 2010 and 2011 using 2007 baseline data.

ABSTRACT

Predicting how changes to the urban environment will affect town centre vitality, mediated as pedestrian flows, is important for environmental, social and economic sustainability. This study is a longitudinal investigation of before and after urban environmental change in a town centre and its association with vitality. The case study baseline is Cardiff town centre in 2007, prior and after major changes instigated by re-configuring Cardiff public and quasi-public street layout due to implementation of the St David's Phase 2 retail development. We present a Multivariate Hybrid Spatial Network Analysis (MHSpNA) model, which bridges the gap between existing Spatial Network Analysis models and four stage modelling techniques. Multiple theoretical flows are computed based on retail floor area (everywhere to shops, shop to shop, stations to shops and parking to shops). The calibration process determines a suitable balance of these to best match observed pedestrian flows, using generalized cross-validation to prevent overfit. Validation shows that the 2007 model successfully predicts vitality as pedestrian flows measured in 2010 and 2011. This is the first time, to our knowledge, that a vitality-pedestrian flow model has been evaluated for its ability to forecast town centre vitality changes over time.

Motivation & Objective

• To assess whether a multivariate hybrid spatial network model can predict changes in pedestrian flows following major urban redevelopment.
• To evaluate the model’s forecasting capability for town centre vitality over time, using empirical pedestrian flow data.
• To calibrate and validate a spatial network model that integrates multiple flow types (e.g., retail to retail, parking to shops) with real-world observations.
• To bridge the gap between traditional spatial network analysis and four-stage transport modeling in urban planning contexts.
• To demonstrate the model’s transferability and robustness in forecasting urban vitality under environmental change.

Proposed method

• Develops a Multivariate Hybrid Spatial Network Analysis (MHSpNA) model combining theoretical flows from multiple urban elements (e.g., retail, stations, parking).
• Computes four distinct theoretical flow types: everywhere to shops, shop to shop, stations to shops, and parking to shops.
• Applies generalized cross-validation during calibration to optimize the weighted combination of theoretical flows to match observed pedestrian counts.
• Uses 2007 pre-redevelopment data as baseline to train the model, then tests its predictive accuracy on 2010 and 2011 pedestrian flow measurements.
• Employs spatial network analysis techniques to model path-finding and accessibility across a modified street network after St David’s Phase 2 redevelopment.
• Validates model performance using observed pedestrian counts, treating the 2010 and 2011 data as out-of-sample forecasts.

Experimental results

Research questions

• RQ1 Can the MHSpNA model accurately predict changes in pedestrian flows following a major urban redevelopment?
• RQ2 How well does the calibrated model forecast actual pedestrian flows in post-redevelopment years (2010 and 2011) using only pre-redevelopment data?
• RQ3 What is the optimal combination of theoretical flows (e.g., from stations, parking, retail) that best replicates observed pedestrian movement?
• RQ4 To what extent does the MHSpNA model outperform traditional spatial network or four-stage modeling approaches in forecasting urban vitality?
• RQ5 Is the model robust enough to predict vitality changes over time in a real-world urban environment with complex infrastructure changes?

Key findings

• The MHSpNA model successfully predicted pedestrian flows in Cardiff town centre for 2010 and 2011 using only 2007 baseline data.
• The calibrated model achieved a balance of theoretical flows that closely matched observed pedestrian counts, with generalized cross-validation preventing overfitting.
• The model demonstrated predictive validity for urban vitality changes, marking the first known evaluation of such a model for forecasting over time.
• The integration of multiple flow types—especially from stations and parking—proved critical in improving prediction accuracy.
• The model’s ability to forecast post-redevelopment pedestrian flows confirms its potential for use in urban planning and sustainability assessments.
• The validation results indicate that MHSpNA effectively captures the impact of spatial reconfiguration on pedestrian movement patterns.