Pemetaan Tekstur Tanah Digital Menggunakan Metode Random Forest Regression Kriging (RFRK) di Sebagian Desa Krasak Kabupaten Magelang

Anita Rahmawati, Dr. Eng. Guruh Samodra, S.Si., M.Sc.

2026 | Skripsi | GEOGRAFI DAN ILMU LINGKUNGAN

Tekstur tanah merupakan sifat fisik tanah yang berperan penting dalam infiltrasi, suplai hara, erosi tanah, dan aerasi. Informasi distribusi spasial tekstur tanah diperlukan untuk mendukung perencanaan dan pengelolaan lahan yang berkelanjutan. Penelitian ini bertujuan untuk (1) memetakan pola distribusi spasial tekstur tanah di sebagian Desa Krasak, Kabupaten Magelang menggunakan metode Random Forest Regression Kriging (RFRK) serta (2) mengetahui kinerja model tersebut. Metode RFRK merupakan pendekatan hibrida yang menggabungkan metode RF dengan Ordinary Kriging untuk memperhitungkan autokorelasi spasial dari residual RF. 

Hasil prediksi menggunakan metode RFRK menunjukkan bahwa wilayah penelitian didominasi oleh fraksi lempung. Pada lapisan topsoil, fraksi pasir memiliki rentang nilai individual 2,8–48,8%, fraksi debu 10,9–74,1%, dan fraksi lempung 20,1–82,4%. Pada lapisan subsoil, fraksi pasir memiliki rentang 2,3–46,1%, fraksi debu 12,5–75,6%, dan fraksi lempung 19,4–85%. Fraksi lempung memiliki proporsi lebih tinggi terutama pada lereng atas di bagian barat hingga barat daya, sedangkan fraksi debu tersebar pada lereng tengah hingga lembah di bagian timur laut dan fraksi pasir relatif lebih tinggi pada area lembah. Pola distribusi ini dipengaruhi oleh kovariat lingkungan, terutama elevasi, MRRTF, MRVBF, dan Slope. 

Evaluasi kinerja model pada lapisan topsoil menunjukkan nilai RMSE sebesar 5,79–9,55%, MAE sebesar 4,15–7,88%, dan R² sebesar 0,32–0,43. Pada lapisan subsoil, nilai RMSE berkisar 6,66–9,59%, MAE sebesar 5,13–8,28%, dan R² sebesar 0,24–0,45%. Model RFRK memberikan kinerja yang lebih baik pada lapisan topsoil untuk fraksi pasir dan debu, sedangkan fraksi lempung menunjukkan kinerja yang hampir sama antara lapisan topsoil (R² = 0,42) dan subsoil (R² = 0,45).

Soil texture is a physical property of soil that plays an important role in infiltration, nutrient supply, soil erosion, and aeration. Information on the spatial distribution of soil texture is essential to support sustainable land use planning and management. This study aims to (1) map the spatial distribution pattern of soil texture in part of Krasak Village, Magelang Regency using the Random Forest Regression Kriging (RFRK) method, and (2) evaluate the performance of the model. The RFRK method is a hybrid approach that combines Random Forest (RF) with Ordinary Kriging to account for spatial autocorrelation in RF residuals.

The prediction results using the RFRK method indicate that the study area is dominated by the clay fraction. In the topsoil layer, the sand fraction had an individual range of 2.8–48.8%, the silt fraction ranged from 10.9–74.1%, and the clay fraction ranged from 20.1–82.4%. In the subsoil layer, the sand fraction ranged from 2.3–46.1%, the silt fraction from 12.5–75.6%, and the clay fraction from 19.4–85%. The clay fraction had higher proportions mainly on the upper slopes in the western to southwestern parts of the study area, while the silt fraction was distributed from the middle slopes to the valleys in the northeastern part, and the sand fraction had relatively higher proportions in valley areas. This distribution pattern was influenced by environmental covariates, particularly elevation, MRRTF, MRVBF, and slope.

Model performance evaluation showed that the topsoil layer had RMSE values of 5.79–9.55%, MAE values of 4.15–7.88%, and R² values of 0.32–0.43. In the subsoil layer, RMSE ranged from 6.66–9.59%, MAE from 5.13–8.28%, and R² from 0.24–0.45. The RFRK model performed better for the sand and silt fractions in the topsoil layer, while the clay fraction showed similar performance between the topsoil (R² = 0.42) and subsoil (R² = 0.45) layers.

Kata Kunci : Random Forest Regression Kriging, Pemetaan Tanah Digital, Machine Learning, Tekstur Tanah, Transformasi ILR