Analisis Kinerja Deteksi Objek (Spesies Mangrove) Pada Citra Satelit Multiresolusi Menggunakan Linear Spectral Unmixing

Fultriasantri, Indah (2025) Analisis Kinerja Deteksi Objek (Spesies Mangrove) Pada Citra Satelit Multiresolusi Menggunakan Linear Spectral Unmixing. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6016231018_Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (38MB) | Request a copy

Abstract

Kawasan konservasi mangrove Pamurbaya di Surabaya Timur memiliki peran penting dalam perlindungan wilayah pesisir, namun rentan mengalami degradasi akibat aktivitas manusia dan perubahan penggunaan lahan. Peta sebaran spesies sangat penting untuk memahami fungsi ekologi seperti penyerapan karbon, toleransi terhadap salinitas, dan stabilitas ekosistem. Penelitian ini memanfaatkan data penginderaan jauh multiresolusi dari citra satelit WorldView-2 untuk memetakan mangrove hingga tingkat spesies secara rinci. Metode Random Forest digunakan untuk membedakan antara area mangrove dan non-mangrove, sedangkan teknik Linear Spectral Unmixing (LSU) digunakan untuk memetakan sebaran spesies mangrove secara lebih detail. Analisis lanjutan dilakukan untuk mengetahui pada resolusi berapa metode LSU bekerja secara optimal. Citra satelit dianalisis pada resolusi 0,5 meter dan kemudian dilakukan downsampling ke resolusi 5 meter, 10 meter, 20 meter, 30 meter, dan 50 meter. Hasil penelitian menunjukkan bahwa LSU mampu membedakan spesies mangrove berdasarkan endmember-nya dan bekerja secara optimal pada resolusi menengah (10–30 meter), dengan akurasi keseluruhan meningkat dari 70% (resolusi 10 m) menjadi 75% (resolusi 30 m), serta nilai Kappa meningkat dari 53,7 menjadi 60,41. Resolusi tinggi (0,5–10 m) memberikan detail spasial yang lebih tinggi, namun lebih sesuai untuk memetakan spesies yang tersebar secara kecil dan tidak merata. Sementara itu, resolusi rendah (20–50 m) cenderung menyebabkan estimasi berlebih (overestimation) atau penggabungan (agregasi) antar spesies.
=======================================================================================================================================
The Pamurbaya mangrove conservation area in East Surabaya is crucial for coastal protection, but it is vulnerable to degradation due to human activities and land-use changes. Species distribution maps are essential for understanding ecological functions, such as carbon sequestration, salinity tolerance, and ecosystem stability. This study utilizes multiresolution remote sensing data from WorldView-2 satellite imagery to map mangrove and detailed species-level. Random Forest is utilized to differentiate mangrove and non-mangrove, while Linear Spectral Unmixing allows for detailed mangrove species distribution. Further analysis was carried out to determine at what resolution the LSU works optimally. The imagery were served in 0.5 meter resolution and downsampled to 5 meter, 10 meter, 20 meter, 30 meter, and 50 meter resolutions. This study obtained that LSU were able to differentiate mangroves according to its endmember and working optimally at medium resolution (10–30 m), with overall accuracy increasing from 70% (10 m) to 75% (30 m) and Kappa value increasing from 53.7 to 60.41. High resolution (0.5–10 m) provides more detailed mapping but is optimal for species with small and scattered distributions. Meanwhile, low resolution (20–50 m) tends to cause overestimation or aggregation of species.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Citra WorldView-2, Mangrove, Random Forest (RF), Linear Spectral Unmixing (LSU). WorldView-2 Imagery, Mangroves, Random Forest (RF), Linear Spectral Unmixing (LSU)
Subjects:	Q Science Q Science > QH Biology > QH91.8.S64 Species diversity
Divisions:	Faculty of Civil, Planning, and Geo Engineering (CIVPLAN) > Geomatics Engineering > 29101-(S2) Master Thesis
Depositing User:	Indah Fultriasantri
Date Deposited:	29 Jul 2025 09:17
Last Modified:	29 Jul 2025 09:17
URI:	http://repository.its.ac.id/id/eprint/122907

Actions (login required)

View Item