Spatial Modelling and Prediction with the Spatio-Temporal Matrix: A Study on Predicting Future Settlement Growth

In the past decades, various Earth observation-based time series products have emerged, which enabled studies and analysis of global change processes. Besides their contribution to understanding processes, datasets hold enormous potential for predictive modeling thereby meet demands decision makers on future scenarios. order further exploit these data, a novel pixel-based approach has been introduced, is spatio-temporal matrix (STM). The integrates historical characteristics specific land cover at high temporal frequency in interpret spatial information neighborhood given target pixel. provided can be exploited with common models algorithms. this study, was utilized evaluated prediction urban/built-settlement growth. Random forest multi-layer perceptron were employed prediction. tests carried out training strategies based one-year ten-year span urban agglomerations Surat (India), Ho-Chi-Minh City (Vietnam), Abidjan (Ivory Coast). slope, use, exclusion, urban, transportation, hillshade (SLEUTH) model selected as baseline indicator performance evaluation. statistical results from receiver operating characteristic curve (ROC) demonstrate good ability STM facilitate settlement growth its transferability different cities, area under (AUC) values greater than 0.85. Compared SLEUTH, STM-based achieved higher AUC all test cases, while being independent additional restricted preferential development areas.