Landcover mapping of tanjung jabung barat, jambi using Landsat - ALOS PALSAR data fusion and object-based hierarchical classification

Land cover mapping is fundamental activity for most land-resource-based studies such as land suitability analysis, carbon stock analysis, habitat analysis, forest resource management and many others. Recently, production of land cover map derived from satellite image classification is largely depended to Landsat systems. This is mostly due to the characteristics of Landsat's spatial and temporal coverage, resolution, data availability and cost-effectiveness. However, similar to most optical remotely sensed image that uses passive sensor, cloud cover became an obstacle in classifying the image to produce quality land cover map. Moreover, if the cloud covers large area or specifically most area that became the study areas, then the quality of land cover map produced would decrease significantly since its map contains many no data. Different from Landsat, ALOS PALSAR image as a product of active remote sensing sensor has ability to penetrate clouds, so that the image can be free from clouds. The characteristics of two different types of sensors have inspired land cover mapping in Tanjung Jabung Barat to combine Landsat with ALOS PALSAR in order to improve the quality of land cover map. The objective of this study is to improve land cover classification conducted in Tanjung Jabung Barat district, Jambi, Indonesia, by combining Landsat and ALOS PALSAR images. Types of land cover that will be derived from this study as follows: undisturbed forest, logged-over forest, logged-over swamp forest, undisturbed mangrove, logged-over mangrove, acacia (Acacia crassicarpa) plantation, oil palm (Elais guinensis) plantation, coffee (Coffea robusta) agroforest, betel-nut (Areca catechu L.) agroforest, rubber (Hevea brasiliensis) agroforest, and rubber monoculture. The classification scheme was based on 4 level of hierarchical classification where land cover classes were identified using combination of spectral and spatial information. In this study, Landsat and ALOS PALSAR images have been combined into one fusion image which consist of 6 bands of Landsat and dual-polarization of ALOS PALSAR. Spectral and texture information has been used in our classification rule sets using hierarchical object-based classification approach. The result shows that land cover map generated from fusion of Landsat and ALOS PALSAR has higher accuracy (82.26%) compare to the one generated from Landsat (79.83%) or ALOS PALSAR only (54.38%). We conclude that fusion of Landsat and ALOS PALSAR may provide an alternative solution to minimize the problem caused by lost area due to cloud as well as improve accuracy of land cover map produced.

---

• View source