Pemetaan Sebaran Mangrove Menggunakan Google Earth Engine dengan Algoritma CART dan Random Forest di Kota Tegal, Jawa Tengah

Abstract

Kota Tegal merupakan kota dengan kepadatan penduduk yang tinggi di daerah pesisir utara Jawa yang masih memiliki ekosistem mangrove. Namun, kegiatan pemantauan keberadaan mangrove dan penelitian tentang sebaran mangrove di Kota Tegal tergolong minim. Penelitian ini bertujuan memetakan sebaran dan luasan mangrove di Kota Tegal menggunakan Google Earth Engine (GEE) dengan citra Sentinel-2A serta membandingkan dan mengevaluasi dua algoritma supervised machine learning yang diaplikasikan. Algoritma yang digunakan adalah Classification and Regression Tree (CART) dan Random Forest (RF). Kedua algoritma tersebut menghasilkan sebaran dan luasan mangrove berdasarkan kelas klasifikasi dan titik lapang yang telah dimasukkan. Sebaran mangrove di Kota Tegal banyak ditemukan di pinggiran tambak, dibandingkan pada daerah bibir pantai. Luas total mangrove di Kota Tegal pada 2023, yaitu 138,81 ha menggunakan algoritma CART dan 101,47 ha dengan algoritma RF. Berdasarkan hasil nilai mangrove accuracy (MA), overall accuracy (OA), dan kappa coefficient (KC) kedua jenis algoritma punya kinerja yang baik untuk memetakan mangrove di Kota Tegal. CART memiliki MA sebesar 78,89%, OA sebesar 84,56%, dan KC sebesar 0,72, sedangkan RF mempunyai MA sebesar 90,82%, OA sebesar 93,20%, dan KA sebesar 0,87.

Tegal is a city with a high population density in the northern coastal area of Java that still had a mangrove ecosystem. However, activities to monitor the presence of mangroves and research on their distribution in Tegal City were relatively minimal. This research aimed to map the distribution and extent of mangroves in Tegal City using Google Earth Engine (GEE) with Sentinel-2A imagery, and to compare and evaluate two applied supervised machine learning algorithms. The algorithms used were Classification and Regression Tree (CART) and Random Forest (RF). These two algorithms produced the distribution and extent of mangroves based on equations, classification classes, and field points that had been entered. The distribution of mangroves in Tegal City was mostly found on the edges of ponds rather than along the shoreline. The total area of mangroves in Tegal City in 2023 was 138.81 ha using the CART algorithm and 101.47 ha using the RF algorithm. Based on the results of the mangrove accuracy (MA), overall accuracy (OA), and kappa coefficient (KC) values, both types of algorithms demonstrated good performance with acceptable accuracy values. CART had an MA of 78.89%, OA of 84.56%, and KC of 0.72, while RF had an MA of 90.82%, OA of 93.20%, and KC of 0.87.