Estimasi Cadangan Karbon Hutan di KHDTK Wanagama I Menggunakan Citra Foto Udara dan Algoritma Random Forest

Eko Joko Nurcahyo, Bekti Larasati, S.Hut., M.Sc., M.Agr.Sc ; Ir. Budi Mulyana, S.Hut., M.Si. Ph.D, IPM

2026 | Skripsi | KEHUTANAN

Estimasi stok karbon hutan diperlukan untuk mendukung mitigasi perubahan iklim dan pengelolaan hutan berkelanjutan. Pendugaan karbon melalui inventarisasi lapangan dan persamaan alometrik memiliki ketelitian yang baik, tetapi memerlukan waktu, biaya, dan tenaga yang besar. Citra Unmanned Aerial Vehicle (UAV) RGB yang dikombinasikan dengan algoritma Random Forest dapat menjadi alternatif untuk mengestimasi stok karbon secara lebih efisien. Penelitian ini bertujuan mengestimasi stok karbon berdasarkan data lapangan, membangun dan mengevaluasi model Random Forest berbasis citra UAV RGB, serta memetakan sebaran stok karbon di KHDTK Wanagama I. Penelitian dilakukan di KHDTK Wanagama I, Kabupaten Gunungkidul, Daerah Istimewa Yogyakarta, pada November 2025 hingga Juni 2026. Sampel ditentukan secara stratified sampling pada 12 strata tegakan. Berdasarkan perhitungan koefisien variasi, diperoleh 63 plot sampel berbentuk lingkaran dengan luas 0,04 ha. Biomassa atas permukaan dihitung menggunakan persamaan alometrik Chave et al. (2014) berdasarkan diameter setinggi dada, tinggi pohon, dan berat jenis kayu. Biomassa dikonversi menjadi stok karbon menggunakan fraksi karbon 0,47. Variabel prediktor berupa indeks vegetasi Visible Atmospherically Resistant Index (VARI), Excess Green Index (ExG), dan Normalized Green Red Difference Index (NGRDI) diekstraksi dari orthomosaic UAV RGB beresolusi 0,368 m. Pemodelan dilakukan menggunakan Random Forest dengan validasi silang 20-fold. Model awal menggunakan 63 plot menghasilkan R² sebesar 0,517, RMSE 21,416 ton C/ha, dan MAE 17,634 ton C/ha. Setelah eliminasi tujuh outlier, model akhir menggunakan 56 plot menghasilkan R² sebesar 0,685, RMSE 13,736 ton C/ha, dan MAE 11,711 ton C/ha. Hasil ini menunjukkan bahwa citra UAV RGB dan Random Forest berpotensi digunakan untuk estimasi dan pemetaan stok karbon hutan pada skala tapak. 

Forest carbon stock estimation is essential for supporting climate change mitigation and sustainable forest management. Carbon stock estimation using field inventories and allometric equations provides reliable results; however, it requires considerable time, cost, and labor. The use of Unmanned Aerial Vehicle (UAV) RGB imagery combined with the Random Forest algorithm offers an alternative approach for estimating carbon stocks more efficiently. This study aimed to estimate carbon stocks based on field data, develop and evaluate a Random Forest model using UAV RGB imagery, and map the spatial distribution of carbon stocks in KHDTK Wanagama I. The study was conducted in KHDTK Wanagama I, Gunungkidul Regency, Special Region of Yogyakarta, from November 2025 to June 2026. Sampling was conducted using stratified sampling across 12 stand strata, resulting in 63 circular sample plots of 0.04 ha each. Aboveground biomass was estimated using the allometric equation developed by Chave et al. (2014), based on diameter at breast height, tree height, and wood density. Biomass was converted into carbon stock using a carbon fraction of 0.47. Predictor variables, including the Visible Atmospherically Resistant Index (VARI), Excess Green Index (ExG), and Normalized Green-Red Difference Index (NGRDI), were extracted from UAV RGB orthomosaic imagery with a spatial resolution of 0.368 m. Random Forest modeling was performed using 20-fold cross-validation. The initial model based on 63 plots produced an R² of 0.517, an RMSE of 21.416 tons C ha?¹, and an MAE of 17.634 tons C ha?¹. After removing seven outlier plots, the final model based on 56 plots produced an R² of 0.685, an RMSE of 13.736 tons C ha?¹, and an MAE of 11.711 tons C ha?¹. These results indicate that UAV RGB imagery and the Random Forest algorithm have potential for estimating and mapping forest carbon stocks at the site scale. 

Kata Kunci : stok karbon, UAV RGB, Random Forest, indeks vegetasi, KHDTK Wanagama I.