Beyond Standalone Geo-Embeddings: Weighted Multi-Model Ensemble Prediction for Tropical Land-Cover Mapping

Article Article scientifique Article Beyond Standalone Geo-Embeddings: Weighted Multi-Model Ensemble Prediction for Tropical Land-Cover Mapping eng 2026 2026 author Havinga Alessandra Marion 880253 Département des sciences végétales (BIVEG) author Giuliani Gregory 0000-0002-1825-8865 94488 Services communs de l'Institut des sciences de l'environnement (ISECO) author Nobel Sophie 957746 author Ranirison Patrick author Nusbaumer Louis Paul Gustave Alvin 529053 Département des sciences végétales (BIVEG) author Martin Pascal Jule André 531311 Département des sciences végétales (BIVEG) 2952 10.3390/rs18172952 1104-3792 https://www.mdpi.com/2072-4292/18/17/2952 Accurate land-cover mapping in tropical regions remains challenging because of high environmental heterogeneity, complex vegetation structure, and persistent cloud cover. Recent geospatial foundation models provide pre-computed Geo-embeddings that offer a new representation of Earth Observation data (EO), yet their potential for detailed tropical land-cover mapping and their performance relative to, and in combination with, conventional Sentinel-1 and Sentinel-2 satellite image time series remains largely unexplored. This study evaluated AlphaEarth Foundation Geo-embeddings for detailed vegetation mapping in northern Madagascar and investigated whether combining them with satellite image time series processed using the Satellite Image Time Series (SITS) R package could improve classification performance. Geo-embeddings were classified using a Random Forest model (GEO), while four supervised classifiers were trained on Sentinel-1 and Sentinel-2 time series. GEO achieved the highest standalone performance (OA = 0.74), outperforming all classifiers trained on the satellite image time series. Probability-level ensemble models were then used to assess whether both data representations could be beneficially combined. The best-performing ensemble, combining GEO (75%) with Temporal Convolutional Neural Network (TempCNN; 25%), increased overall accuracy from 0.74 to 0.82. These results suggest that Geo-embeddings provide an effective standalone representation for detailed tropical land-cover mapping and indicate that combining them with conventional satellite image time series may exploit complementary information. As geospatial foundation models continue to evolve, understanding how they can be integrated with established Earth observation workflows may support future operational land-cover mapping. Geo-embeddings Ensemble learning Land-cover mapping Tropical ecosystems Sentinel time series TempCNN Madagascar Dewey 333.7-333.9 Dewey 580 18 Institut des sciences de l'environnement 0DB 45 Département des sciences végétales 163 false 673 Envirospace Lab 938 ISE Pôle Sciences 970b71ee-b344-4440-bc52-c9af263975d0 Living Earth Lab 941 ISE Pôle Informatique environnementale 1057 TIGERS 859 Global Resource Information Database (GRID) Remote sensing 18 17 1b800bc9-a3f7-4385-927f-23c33f2b10ba Article (Published version) Public remotesensing-18-02952.pdf 6005375 application/pdf CC BY-4.0 0e87069ed810b0328c63fb631b058178fd1f6acf285a0ec9baa227ad97c11ef7 openAccess Centres et instituts / Institut des sciences de l'environnement Faculté des sciences / Section de biologie / Département des sciences végétales

Title	Published in	Access level	OA Policy	Year	Views	Downloads

Beyond Standalone Geo-Embeddings: Weighted Multi-Model Ensemble Prediction for Tropical Land-Cover Mapping	Remote sensing			2026	1	0

Archive ouverte UNIGE

Nobel, Sophie