Pengembangan Model Estimasi Kadar Lengas Tanah Berbasis Citra Satelit Sentinel-1 dan 2 dengan Pendekatan Empiris dan Machine Learning untuk Mendukung Pertanian Lahan Kering di Kawasan Karst: Studi Kasus Kawasan Karst Kendeng dan Semanu
AFINAFGHANI DUTA PRATAMA, Hanggar Ganara Mawandha, ST., M.Eng., Ph.D. ; Dr. Prieskarinda Lestari, S.T., IPM.
2024 | Skripsi | TEKNIK PERTANIAN
Soil moisture content is a reflection of the water balance conditions in the upper layers of the soil surface. This condition can provide information related to cropping patterns and plant health. Karst area is one type of land that requires more attention for cultivation because it is not able to store surface water for a long period of time. Kendeng and Semanu Karst Areas are two karst areas that have potential as dryland agricultural land and water reserves for the surrounding community. Observation of soil moisture content in both karst areas cannot be done directly due to varying topographic conditions. Therefore, this research aimed to develop a model for estimating soil moisture content based on remote sensing techniques by utilizing Sentinel-1 and 2 imagery and combining them with DEM.
The development of the estimation model is based on three methods, the DuboisTopp combination empirical model, the Stepwise Multi Linear Regression (SMLR) mathematical model and the Artificial Neural Network (ANN) machine learning model. Soil moisture samples were taken in Kendeng and Semanu Karst Areas on October 15 and 23, 2023 (dry season) and March 14 and 26 and April 7 and 23, 2022 (wet season) based on slope and land cover levels. Model accuracy was tested with coefficient of determination (R2 and Adj R2), Pearson correlation coefficient (r), and Mean Square Error (MSE) and Nash Sutcliffe Efficiency (NSE) error tests.
The results showed that the ANN model was able to provide a very strong accuracy test (r = 0,689; R2 = 0,467; Adjusted R2 = 0,475; MSE = 0,183, dan NSE = -60,26) it the best model developed. In addition, the three models have been able to visualize the spatial distribution with SMLR as the model that is able to show the heterogeneity of soil moisture distribution based on elevation variables.
Kata Kunci : kadar lengas tanah, kawasan karst, machine learning, model estimasi, penginderaan jauh