Electronic Theses and Dissertation

Prediksi Erosi Menggunakan Jaringan Syaraf Tiruan Algoritma Backpropagation Di Daerah Tangkapan Air (DTA) Waduk Keuliling Aceh Besar

ABSTRAK
Waduk Keuliling di Aceh Besar berperan penting dalam mendukung irigasi pertanian, namun mengalami ancaman serius akibat sedimentasi yang dipicu oleh erosi di daerah tangkapan air (DTA). Penelitian ini bertujuan untuk memprediksi erosi tanah secara akurat melalui integrasi antara metode Universal Soil Loss Equation (USLE) dan Jaringan Syaraf Tiruan (JST) algoritma Backpropagation. Data yang digunakan mencakup curah hujan tahunan 2015–2023, serta data spasial seperti penggunaan lahan, jenis tanah, kemiringan lereng, dan jaringan sungai. Perhitungan faktor-faktor erosi dilakukan menggunakan persamaan USLE, kemudian diproses dalam model JST dengan arsitektur 8-10-1 neuron menggunakan MATLAB R2013a. Hasil penelitian menunjukkan bahwa model JST menghasilkan tingkat akurasi tinggi dengan nilai Mean Squared Error (MSE) sebesar 0,00107 dan koefisien determinasi (R²) sebesar 0,99261. Prediksi erosi tahun 2024 menunjukkan pola musiman yang konsisten dengan data historis, memperkuat validitas model. Model ini memiliki implikasi praktis dalam perencanaan konservasi tanah dan pengelolaan sumber daya air secara berkelanjutan di kawasan DTA Waduk Keuliling.
Kata Kunci: Erosi tanah, USLE, Jaringan Syaraf Tiruan, Backpropagation, Waduk Keuliling, konservasi
tanah

Erosion Prediction Using Artificial Neural Network with Backpropagation Algorithm in the Watershed Area of Keuliling Reservoir, Aceh Besar ABSTRACT Keuliling Reservoir in Aceh Besar plays a vital role in supporting agricultural irrigation but faces severe threats from sedimentation caused by soil erosion in its catchment area. This study aims to accurately predict soil erosion by integrating the Universal Soil Loss Equation (USLE) method with an Artificial Neural Network (ANN) using the Backpropagation algorithm. The dataset includes annual rainfall from 2015 to 2023 and spatial data such as land use, soil types, slope gradients, and river networks. Erosion factors were calculated using the USLE equation and processed in an ANN model with an 8-10-1 neuron architecture implemented in MATLAB R2013a. Results indicate that the ANN model achieved high accuracy with a Mean Squared Error (MSE) of 0.00107 and a coefficient of determination (R²) of 0.99261. The 2024 erosion prediction revealed seasonal patterns consistent with historical data, strengthening the model’s validity. This integrated model offers practical implications for targeted soil conservation planning and sustainable water resource management in the Keuliling Reservoir catchment area. Keywords: Soil erosion, USLE, Artificial Neural Network, Backpropagation, Keuliling Reservoir, soil conservation.