PENGARUH DISTRIBUSI SPASIAL SAMPEL PEMODELAN TERHADAP AKURASI ESTIMASI LEAF AREA INDEX (LAI) MANGROVE

Abstract

Leaf Area Index (LAI) has an important role in defining the health of mangrove forest. Remote sensing images able to estimate mangrove LAI, especially through semi-empirical approach. This approach needs appropriate selection of sample location and value distribution for both modelling and accuracy assessment purposes. However, both aspects are often neglected when selecting the sample for modelling. This research aims to explor and analyze the LAI field sample collected to answer (1) if the spatial and (2) value distribution of modelling samples affect the accuracy of mangrove LAI estimation. The method used was by developing regression models between Soil-Adjusted Vegetation Index (SAVI) pixel values derived from ALOS AVNIR-2 image (10m) and field LAI measurement using LICOR LAI-2200. The modelling samples were selected randomly and purposively through three simulations based on spatial distribution and value range of the samples. The accuracy of the estimation was assessed using 1:1 relationship plots and Standard Error of Estimate (SEE). The research results show that the accuracy of LAI estimation is dependent to the spatial distribution and the range value of the modelling samples. High estimation accuracy achieved when the sample location for modelling is evenly distributed and covers the range of the field sample values.