[Paper Review] Fourier Analysis of an Expanded Gravity Model for Spatio-Temporal Interactions

Yanguang Chen, Fahui Wang|arXiv (Cornell University)|Jun 17, 2013

Urban Transport and Accessibility4 references3 citations

TL;DR

This paper proposes an expanded gravity model that incorporates time into spatial interaction by modeling city interactions as temporal processes using Fourier analysis and cross-correlation functions. By calibrating the model through spectral decomposition, it improves upon the conventional instantaneous gravity model, enabling dynamic spatio-temporal interaction analysis with demonstrated applicability in a case study.

ABSTRACT

Fourier analysis and cross-correlation function are successfully applied to improving the conventional gravity model of interaction between cities by introducing a time variable to the attraction measures (e.g., city sizes). The traditional model assumes spatial interaction as instantaneous, while the new model considers the interaction as a temporal process and measures it as an aggregation over a period of time. By doing so, the new model not only is more theoretically sound, but also enables us to integrate the analysis of temporal process into spatial interaction modeling. Based on cross-correlation function, the developed model is calibrated by Fourier analysis techniques, and the computation process is demonstrated in four steps. The paper uses a simple case study to illustrate the approach to modeling the interurban interaction, and highlight the relationship between the new model and the conventional gravity model.

Motivation & Objective

• To extend the conventional gravity model by incorporating time-dependent attraction measures such as city sizes.
• To model spatial interactions as temporal processes rather than instantaneous events.
• To develop a computationally feasible method for calibrating the expanded model using Fourier analysis and cross-correlation.
• To demonstrate the model's capability to capture dynamic interurban interactions through a case study.
• To establish a theoretical and methodological bridge between temporal dynamics and spatial interaction modeling.

Proposed method

• The model extends the traditional gravity model by introducing time-varying attraction terms, such as city population or economic output, over a time interval.
• Fourier analysis is applied to decompose the time series of interaction flows into spectral components for frequency-domain analysis.
• Cross-correlation functions are used to identify time lags between interaction flows and attraction measures, revealing temporal dependencies.
• The calibration process is structured in four steps: data preparation, cross-correlation computation, Fourier transformation, and parameter estimation.
• The model is validated using a case study of interurban interactions, comparing results with the conventional gravity model.
• The method enables the integration of temporal dynamics into spatial interaction modeling without requiring complex time-series modeling.

Experimental results

Research questions

• RQ1 How can time be systematically incorporated into the gravity model of spatial interaction to reflect temporal dynamics?
• RQ2 What is the role of time lags between city attractions and interaction flows in shaping spatial interaction patterns?
• RQ3 How can Fourier analysis be used to calibrate a spatio-temporal gravity model more effectively than traditional methods?
• RQ4 To what extent does the expanded model improve upon the conventional gravity model in capturing real-world interaction patterns?
• RQ5 What are the spectral characteristics of interurban interactions when modeled as time-varying processes?

Key findings

• The expanded gravity model successfully captures temporal dynamics in interurban interactions by modeling attraction measures as time series.
• Fourier analysis enables effective calibration of the model by identifying dominant frequency components in interaction flows.
• Cross-correlation analysis reveals significant time lags between city size changes and interaction flows, indicating delayed responses in spatial interactions.
• The model demonstrates improved theoretical consistency by treating interactions as cumulative over time rather than instantaneous.
• The case study confirms that the expanded model provides a more nuanced and realistic representation of urban interaction compared to the conventional model.
• The four-step calibration process is computationally feasible and provides a systematic framework for applying spectral methods to spatial interaction modeling.