Electronic Theses and Dissertation

RINGKASAN

Nitrogen, phospor dan kalium merupakan unsur hara penting yang diperlukan
tanaman, sehingga penting untuk mengetahui kandungannya di dalam tanah sebelum praktik
pertanian. Untuk mengetahui kadar NPK biasanya digunakan metode analisis kimia di
laboratorium, namun metode itu memerlukan biaya yang mahal, waktu yang lama dan
kesulitan akses oleh petani. Kini teknologi seperti sensor soil NPK telah menjadi alternatif
dalam mendeteksi unsur hara NPK secara cepat, namun sebelum diaplikasikan secara luas
oleh petani sensor perlu diuji untuk memastikan bahwa nilai yang diperoleh sensor
mendekati nilai sebenarnya. Pengujian dilakukan dengan membandingkan nilai sensor
dengan laboratorium sebagai data acuan. Penelitian ini bertujuan untuk menganalisis
hubungan antara data hasil sensor dengan hasil laboratorium dan menganalisis tingkat
akurasi sensor dalam mendeteksi unsur NPK.
Penelitian dilakukan di tiga kecamatan di Kabupaten Aceh Besar yaitu Kuta Baro,
Blang Bintang, dan Darussalam pada bulan Agustus 2025. Setiap kecamatan diambil 4
sampel tanah (2 sawah irigasi dan 2 sawah tadah hujan) secara acak sehingga total terdapat
12 sampel. Selanjutnya, sensor soil NPK ditancapkan pada titik sampel untuk memperoleh
data sensor di lapangan dan sampel tanah diambil menggunakan bor tanah pada kedalaman
0-20 cm. Sampel tanah dari setiap lokasi dianalisis di Laboratoriun Penelitian Tanah dan
Tanaman, Fakultas Pertanian, Universitas Syiah Kuala. Selanjutnya data dianalisis untuk
melihat hubungan keduanya dengan parameter nilai R² dan r serta tingkat akurasi sensor
berdasarkan parameter nilai RMSE dan MAPE.
Hasil penelitian menunjukkan bahwa dari 12 data berpasangan antara hasil sensor
dan hasil laboratorium diperoleh rata-rata nilai laboratorium unsur N= 716,67mg/kg, P=
441,67mg/kg, dan K= 1.733,33mg/kg. Rata-rata nilai sensor yang diperoleh untuk N=
348,42mg/kg, P= 215,39mg/kg, dan K= 331,47mg/kg. Analisis data laboratorium dan sensor
menggunakan regresi linier sederhana menunjukkan keterkaitan hubungan linier yang kuat
dan bernilai positif atau searah. Pada unsur N diperoleh persamaan kalibrasi y = 4,3584x –
801,89 dengan R²= 0,8722 (sangat kuat), unsur P diperoleh persamaan y = 2,641x – 127,17
dengan R²= 0,7069 (kuat) dan unsur K diperoelh persamaan y = 8,6959x – 1.149 dengan R²=
0,5949 (cukup). Nilai koefisien korelasi (r) unsur N= 0,9339 (sangat kuat), P= 0,8408 (sangat
kuat) dan K= 0,7713 (kuat). Tingkat akurasi sensor diperoleh dengan nilai RMSE unsur N=
413,45mg/kg, P= 239,36mg/kg dan K= 1.956,97mg/kg. Nilai MAPE pada unsur N= 48,71%
(cukup), P= 50,13% (rendah), dan K= 70,03% (rendah). Hal ini menunjukkan bahwa sensor
belum cukup akurat digunakan untuk memprediksi kadar NPK total dalam tanah jika tidak
dilakukan kalibrasi. Namun, setelah dilakukan kalibrasi menggunakan persamaan garis
regresi untuk mengkoreksi data sensor agar nilai prediksi lebih mendekati nilai aktual,
diperoleh tingkat akurasi sensor meningkat dengan nilai MAPE unsur N= 8,63% (tinggi),
P= 10,22% (baik), dan K= 42,45% (cukup). Hal ini menunjukkan bahwa kalibrasi pada data
sensor perlu dilakukan untuk mengkoreksi hasil prediksi sehingga hasilnya lebih mendekati
nilai aktual sesuai dengan kondisi di lapangan.

Abstract Nitrogen, phosphorus, and potassium are essential nutrients required by plants, making it important to determine their content in the soil before agricultural practices. Typically, chemical analysis methods in laboratories are used to determine NPK levels; however, these methods are costly, time-consuming, and difficult for farmers to access. Nowadays, technology such as soil NPK sensors has become an alternative for rapidly detecting NPK nutrients, but before being widely applied by farmers, the sensor needs to be tested to ensure that the values obtained are close to the true values. Testing is done by comparing the sensor values with laboratory results as reference data. This research aims to analyze the relationship between the sensor data and laboratory results and to analyze the accuracy level of the sensor in detecting NPK elements. The research was conducted in three sub-districts in Aceh Besar Regency, namely Kuta Baro, Blang Bintang, and Darussalam, in August 2025. From each sub-district, 4 soil samples (2 from irrigated rice fields and 2 from rainfed rice fields) were taken randomly, resulting in a total of 12 samples. Subsequently, the soil NPK sensor was inserted at the sample points to obtain sensor data in the field, and soil samples were taken using a soil auger at a depth of 0-20 cm. Soil samples from each location were analyzed at the Soil and Plant Research Laboratory, Faculty of Agriculture, Universitas Syiah Kuala. The data were then analyzed to examine the relationship between them using the parameters R² and r, as well as the sensor's accuracy level based on RMSE and MAPE values. The results showed that from 12 paired data points between sensor results and laboratory results, the average laboratory values for element N= 716.67 mg/kg, P= 441.67 mg/kg, and K= 1,733.33 mg/kg were obtained. The average sensor values obtained were N= 348.42 mg/kg, P= 215.39 mg/kg, and K= 331.47 mg/kg. Analysis of laboratory and sensor data using simple linear regression showed a strong, positive linear relationship. For element N, the calibration equation y = 4.3584x – 801.89 with R²= 0.8722 (very strong) was obtained; for element P, the equation y = 2.641x – 127.17 with R²= 0.7069 (strong) was obtained; and for element K, the equation y = 8.6959x – 1,149 with R²= 0.5949 (moderate) was obtained. The correlation coefficient (r) values were N= 0.9339 (very strong), P= 0.8408 (very strong), and K= 0.7713 (strong). The sensor's accuracy level, obtained from the RMSE values, was N= 413.45 mg/kg, P= 239.36 mg/kg, and K= 1,956.97 mg/kg. The MAPE values for element N= 48.71% (moderate), P= 50.13% (low), and K= 70.03% (low). This indicates that the sensor is not yet accurate enough to predict the total NPK content in the soil without calibration. However, after calibration was performed using the regression line equation to correct the sensor data so that the predicted values are closer to the actual values, the sensor's accuracy level increased, with MAPE values of N= 8.63% (high), P= 10.22% (good), and K= 42.45% (moderate). This shows that calibrating the sensor data is necessary to correct the prediction results so that they are closer to the actual values according to field conditions.