Determination of stratification boundary for forest and non forest multitemporal classification to support REDD+ in Sumatera Island

Abstract

Multitemporal classification for forest and non-forest is a method to classify forest and non-forest by considering the missingdatacaused by cloudcover that frequently occur in tropical area such asIndonesia using multi temporal satellite data. Multitemporal classification is done after annual classification of forest and non-forest. To gain an optimum result of forest and non-forest classification, it is needed a stratification zone that describes the difference of vegetation conditiondue to different of vegetation type, soil type, climate and land use/cover associations. Thisstratification zonewill be useful toindicatethe different of biomass volume which relating to carbon content for supportingthe REDD+ project. This paperwasstudied aboutdetermination of stratification boundary to perform multitemporal classification in Sumatera Island using Landsat imagery in 25meter resolution and Quick Bird imagery in 0.6meter. Rough stratification was made by considering land use/cover, DEM and landform, using visual interpretation of moderate spatial resolution of satellite data. High spatial resolution data was also provided in some areas to increase the accuracy level of stratification zone. The stratification boundary was evaluated using forest classification indices, and it was redetermined to obtain the final stratification zone for the focused area. The indices was generated by Canonical Variate Analysis (CVA) method, which was depend on training samples of forest and nonforest in each previous stratification zone. The amount of indices used in each zone were two or three indices depends on the separability for the forest and non-forest classification. The suitable indices used in each zone described forest as 100, non-forest as 0, and uncertain forestbetween 50-99. The result shows that there were 20 stratification zones in Sumatera that spread out in coastal, mountain, flat area, and group of small islands. The stratification zone will improve the accuracy of forest and non-forest classification result and their change based on multitemporal classification.