Pengembangan Model Empiris untuk Estimasi Akumulasi Nitrogen Tanaman Padi Berbasis Citra Satelit Sentinel 2

Abstract

Plant Nitrogen Accumulation (PNA) merupakan jumlah total (akumulasi) N dalam tajuk tanaman per satuan luas tanah. PNA adalah sebuah indikator yang menunjukkan status N pada tanaman. Penelitian ini mengusulkan sebuah model empiris untuk menghitung PNA pada lahan sawah yang luas, dengan cepat dan dekat waktu nyata berdasarkan data Sentinel 2. Hasil analisis terhadap hubungan indeks vegetasi dengan data PNA sampel tanaman padi yang diambil dari plot-plot percobaan menunjukkan bahwa status N tanaman padi dapat dideteksi secara dekat waktu nyata menggunakan data Sentinel 2. Analisis statistik juga menunjukkan bahwa variasi temporal dan spasial indeks PNA tanaman padi pada perlakuan teknik tanam dan varietas yang berbeda dapat diestimasi berdasarkan indeks vegetasi dari Sentinel 2. Analisis hubungan data pengukuran PNA tanaman padi dengan delapan metode indeks vegetasi, membuktikan bahwa perhitungan PNA berbasis NDRE merupakan model optimum yang ditunjukkan oleh nilai R2 = 75.17%, RMSE = 22.5 kg/ha, dan akurasi = 80.80% terhadap data training. Validasi model perhitungan PNA berbasis NDRE memiliki performa yang baik dengan 65.08%, RMSE = 45.2 kg/ha, dan akurasi = 74.22%. Model PNA ini telah digunakan untuk memetakan distribusi spasial status N tanaman padi yang berbeda waktu tanaman pada lahan baku sawah Kecamatan Rawamerta, Kabupaten Karawang. Hasil penelitian telah berkontribusi terhadap penerapan prinsip-prinsip pertanian presisi, dengan memberikan landasan penting bagi pengembangan sistem rekomendasi pemupukan berbasis teknologi penginderaan jauh.

Plant Nitrogen Accumulation (PNA) is the total amount (accumulation) of N in the plant canopy per unit area of land. PNA is an indicator that shows the N status in plants. This study proposes a method for computing PNA over large areas of rice field, faster and in near real-time based on Sentinel 2 data. The research was proven that the N status of rice plants can be detected in real time using Sentinel 2 data, through empirical analysis of the relationship between the Sentinel 2-based vegetation indices and PNA of rice taken from the experimental plots. Statistical analysis was shown that the temporal and spatial variations of the PNA of rice on different planting techniques and varieties can be determined based on the vegetation indices. Empirical analysis of the PNA based on eight vegetation indices methods were proven that the NDRE-based PNA model was an optimum model, indicated by the value of R2 = 75.17%, RMSE = 22.5 kg/ha, and accuracy = 80.80% to the training data. The performance indicators of the PNA model were calculated with the results of R2 = 65.08%, RMSE = 45.2 kg/ha, and accuracy = 74.22%. The PNA model has been used to map the spatial distribution of the N status of rice plants at different planting stages over rice field area in Rawamerta District, Karawang Regency. The research results have contributed to the application of the principles of precision agriculture, through providing an important basis for the development of a fertilization recommendation system based on remote sensing technology.