Comparison of Feature Extraction to Test Dryness and Moisture Levels in Burned Restoration Areas Using Linear Discriminant Analysis

Abstract

The growth process after forest fires takes a long time because much land is often burned again, and some soils are less fertile. This research was conducted in Katingan, Central Kalimantan, as an area of interest. In this case, many human resources do not realize that they are trying to re-utilize the land that was burned by fires in that location. Land that has been burned should be rechecked a few months after the fire to see vegetation density. Using Landsat 8 OLI imagery to detect changes after the fire occurred, this research needs to compare feature extraction to test the level of dryness and humidity in the burnt restoration area. The feature extraction used is Normalized Difference Drought Index for drought detection and the Normalized Difference Moisture Index for detecting post-fire humidity. Furthermore, it will be tested using the Linear Discriminant Analysis algorithm to assess the classification of results from 1 June 2020 - 31 December 2020 in the area of interest carried out. The results of this research obtained for feature extraction NDDI has a lightness level with a range of 0.01 - 0.15 to 0.15 - 0.25, which indicates moderate to a severe drought that occurred in Katingan, Central Kalimantan, with a precision value of 98.33%, Recall 98.33% and Accuracy 98%. While NDMI has humidity in the range of 0.2 to - 0.4 to 0.4 to 0.6, which shows no growth to low growth with a precision value of 68.57%, Recall 68.57% and Accuracy 68% obtained.