Pengembangan Sistem Pendugaan Unsur Kalium Tanah Gambut Perkebunan Kelapa Sawit Berbasis Citra Satelit Sentinel-1

Abstract

Palm oil play an important role in the Indonesian economy because it is the largest Indonesian export commodities. Sufficient soil nutrient will result in maximum production and better oil yield quality. One of the nutrients that is very influential on productivity is potassium. Soil samples and location coordinates taken from peat soil oil palm plantations were then adjusted to the Sentinel-1 satellite image. Sentinel-1 image obtained processed through preprocessing steps. Sentinel-1's image preprocessing steps are apply file orbit, thermal noise removal, border noise removal, calibration, speckle filtering, terrain correction, and convert to dB. The satellite image parameters used are, sigma naught, gamma naught, beta naught, local angle of incidence, projected local incidence angle, elevation, and incidence angle from ellipsoid. These parameters used as model’s independent variables. The actual soil nutrition from the lab results used as model’s dependent variable. The data was divided into two groups of training data with 90% data and test data with 10% data. Prediction model is made using Neural Network Regressor (NN). All models that were built were assessed for the quality of the model using MAPE. NN modeling is built using hyperparameter tuning so that the model quality is better. The NN model has a MAPE of 25,85%, meaning that the correctness is 74,15%.