Settlement Growth Prediction exploiting EO-based Time Series with the Spatio-Temporal Matrix Approach: a Case Study for the City of Hue, Vietnam

Satellite-based Earth observation (EO) time series data possess enormous potential for analyzing the past and forecasting future trends of urban/settlement development. While historic settlement extent maps with high spatial resolution can be generated from EO data, detailed local information such as intra-urban recreation spaces or restricted areas for specific land use types are hard to acquire. In order to overcome this data gap from which many modelling approaches suffer, the Spatio-Temporal Matrix (STM) was developed. The STM provides spatial and temporal characteristics of a target pixels’ neighborhood to be used for predicting the future urban/settlement growth with a machine learning approach. In this study, a multi-layer perceptron (MLP) was employed to utilize the STM for the settlement growth prediction of the City of Hue, Vietnam. The SLEUTH model was used as a benchmark for the performance evaluation. The results show that the STM-based model achieved a high accuracy in settlement growth modelling. Compared to the SLEUTH model, the STM approach simulated less growth in restricted areas without having to rely on external datasets.