Please use this identifier to cite or link to this item:
http://repository.ipb.ac.id/handle/123456789/168824| Title: | Penerapan Algoritma Random Forest Regression untuk Deteksi Kandungan Mangan Tanaman Kelapa Sawit Berbasis Citra Sentinel-2A |
| Other Titles: | Application of Random Forest Regression Algorithm for Manganese Content Detection in Oil Palm Plants Based on Sentinel 2A Imagery |
| Authors: | Seminar, Kudang Boro Sudradjat Zahra, Khairunnisa Az |
| Issue Date: | 2025 |
| Publisher: | IPB University |
| Abstract: | Kelapa sawit merupakan komoditas utama di Indonesia dengan luas areal
perkebunan yang terus berkembang. Namun, produktivitasnya masih tergolong
rendah. Salah satu faktor penting yang memengaruhi hasil panen kelapa sawit
adalah ketersediaan unsur hara, termasuk mangan (Mn) sehingga pemantauannya
sangat krusial untuk meningkatkan produktivitas kelapa sawit. Namun, metode
konvensional berbasis sampel dan penggunaan laboratorium untuk mengukur
kandungan mangan umumnya memerlukan waktu yang lama, biaya tinggi, dan
cakupan spasial yang terbatas. Penelitian ini bertujuan untuk memprediksi
kandungan mangan (Mn) pada perkebunan kelapa sawit di lahan gambut, mineral
dan data gabungan dari lahan gambut dan mineral menggunakan citra Sentinel-2A
dan algoritma Random Forest Regression. Penelitian ini mengintegrasikan data
citra Sentinel-2A dan indeks vegetasi sebagai variabel bebas, serta data sampel
daun kelapa sawit sebagai variabel terikat. Data citra yang diperoleh melalui tahap
pra-proses, yang meliputi resampling resolusi citra, cropping sesuai dengan peta
batas kebun, dan ekstraksi citra menggunakan point sampling tool. Data kemudian
dibagi menjadi 90% untuk pelatihan dan 10% untuk pengujian model. Hasil
penelitian menunjukkan bahwa model spesifik untuk lahan mineral menghasilkan
correctness yang baik sebesar 86,87%, sedangkan model spesifik pada lahan
gambut menghasilkan correctness sebesar 76,48% yang masih dapat dikategorikan
layak. Model gabungan yang diuji pada keseluruhan data memberikan correctness
sebesar 77,28%, serta masing-masing 85,27% dan 73,74% ketika diuji pada data
mineral dan gambut secara terpisah. Oil palm is a major agricultural commodity in Indonesia, with plantation areas continuing to expand. However, its productivity remains relatively low. One of the key factors affecting oil palm yield is the availability of nutrients, including manganese (Mn), making its monitoring crucial for improving productivity. Conventional sample-based and laboratory-based methods for measuring manganese content are generally time-consuming, costly, and spatially limited. This study aims to predict manganese (Mn) content in oil palm plantations located on peat soil, mineral soil, and a combination of both using Sentinel-2A imagery and the Random Forest Regression algorithm. The study integrates Sentinel-2A imagery and vegetation indices as independent variables, and manganese content from oil palm leaf samples as the dependent variable. The imagery data underwent preprocessing steps, including resolution resampling, cropping according to plantation boundary maps, and pixel value extraction using point sampling tools. The dataset was then divided into 90% for training and 10% for testing the model. The results show that the site-specific model for mineral soil produced a high correctness of 86,87%, while the model for peat soil achieved a correctness of 76,48%, which is still considered acceptable. The combined model tested on all data yielded a correctness of 77,28%, with 85,27% and 73,74% when tested separately on mineral and peat soil data, respectively. |
| URI: | http://repository.ipb.ac.id/handle/123456789/168824 |
| Appears in Collections: | UT - Agricultural and Biosystem Engineering |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| cover_F1401211048_a1a063ff020e48a88ae1ce52b9e1e40f.pdf | Cover | 460.05 kB | Adobe PDF | View/Open |
| fulltext_F1401211048_80d466ae1e3d426a855591b0d61a63cb.pdf Restricted Access | Fulltext | 1.86 MB | Adobe PDF | View/Open |
| lampiran_F1401211048_133176daeadc4192a83c35a836e86b3e.pdf Restricted Access | Lampiran | 1.39 MB | Adobe PDF | View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.