Mapping land use classes can be challenging especially in highly-urbanized regions due to the diversity of materials and structures. We aimed to generate a land use classification for Metro Manila by combining spatial data derived from Sentinel-2 image, IFSAR DEM and DSM and segmented WorldView2 images and classified land use using Support Vector Machine.

Mapping land use classes can be challenging especially in highly-urbanized regions due to the diversity of materials and structures. We aimed to generate a land use classification for Metro Manila by combining spatial data derived from Sentinel-2 image, IFSAR DEM and DSM and segmented WorldView2 images and classified land use using Support Vector Machine. We were able to generate a land use classification with an overall accuracy of 81.6%. Our results show that the addition of informational layers such as height of the structure, dimensions, texture, distance and density improved the classification accuracy by 13.8% higher than when the RGB image classification. In addition, informal settlements can be classified more accurately (PA=85.86% and UA=79.07%). There was a great difficulty in accurately classifying industrial (PA=69.06% and UA=69.32%), commercial areas (PA=66.34% and UA=73.46%) and residential areas (PA=65.92%, UA=70.52%). These results can help in estimating the informal settlement population and exposure to various hazards in Metro Manila.

We also mapped the land use for 2011 using LiDAR DEM, DSM and orthophoto for Metro Manila, Pasig City and Barangay Batasan Hills in Quezon City to apply the methodology using another dataset. Using the same methodology, but without ground truthing activity, results appeared to have overestimated the informal settlements for Metro Manila and Quezon City. The best land use classification was for Barangay Batasan Hills. Even without ground truthing and relying on visual assessment, this exercise showed that adding more spatial information layers enable land use classification. In addition, this study showed that the collection of sufficient training and ground truthing sites are important in improving the land use classification accuracy.

RELATED STUDIES

Geospatial I 🔗 | Land Cover Change Analysis in Metro Manila and Marikina Watershed Philippines (2009-2018)

---

Geospatial II 🔗 | Assessment of Land Cover Changes to River Runoff and Scenario-based Flooding

---

Geospatial IV 🔗 | Using GIS to Visualize Risk in Metro Manila

---

Geospatial V 🔗 | Informal Livelihoods Survey along Commonwealth, Quezon City

RELATED SITES

Ateneo Innovation Center →

Climate and Disaster Resilience Laboratory →

SOSE Arise →

Department of Physics →

Department of Environmental Science →

DR. CELINE VICENTE
MS. FLORDELIZA DEL CASTILLO