Analysis of microclimate in a greenhouse in predicting flowering time and physiological maturity of tomato plants by using heat unit and artificial neural network method

Abstract

The objective of the research was to analyze the microclimate in a greenhouse in order to predict flowering time and physiological maturity of tomato by using heat unit and artificial neural network method. The research was conducted at Indonesian Agroclimate and Hydrology Research Institute (IAHRI), Cimanggu, Bogor during the period of August – December 2010. Determining heat unit was done by using temperature daily average data, and artificial neural network (ANN) by using Matlab software. Measured data were divided into two parts: one part was for training data, and the other part was for testing. The performance of ANN model was described by the value of correlation coefficience (R). The validation process that were ANN performance test on sample data never used before in the training was done by calculating the RMSE (Root Mean Square Error), Standard Error of Prediction (SEP) and Coefficient Variation (CV). The result indicated that the heat unit during the growth of the plants was 1661 oC day while the average temperature inside the greenhouse during the research was 27.1 oC, the average humidity was 74.2 %, and solar radiation intensity was 9.3 MJ/m2/day. The R values based on the prediction of flowering time was 0.51, with value of RMSE, SEP and CV were 4.88, 26.43 and 69%. The R values based on the physiological maturity was 0.63 while RMSE, SEP and CV were 2.1, 4.63 and 9 %, respectively. The result of mesurement in the field indicated that the average flowering time in the greenhouse was 34 dap (days after planting), and based on ANN simulation model flowering time was 31 dap. The result of measurement indicated that of physiological maturity was 49 daf (day after flowering), and based on ANN simulation model was 48 daf.

Penelitian ini bertujuan untuk menganalisis iklim mikro di dalam rumah tanaman untuk memprediksi waktu pembungaan dan masak fisiologis tanaman tomat dengan menggunakan metode heat unit dan artificial neural network (ANN). Penelitian dilaksanakan pada bulan Agustus sampai Desember 2010 di rumah tanaman Balai Penelitian Agroklimatologi dan Hidrologi, Cimanggu, Bogor. Penentuan heat unit dilakukan dengan menggunakan data rata-rata suhu udara harian di dalam rumah tanaman. Sedang analisis data untuk pemodelan ANN dilakukan dengan menggunakan perangkat lunak (software) Matlab. Dalam pemodelan ANN, data pengukuran di lapangan dipilah menjadi dua bagian; satu bagian digunakan untuk data pelatihan (training) dan satu bagian lainnya digunakan untuk data pengujian (testing). Model yang diperoleh dari data pelatihan digunakan untuk data pengujian. Untuk mengevaluasi performa model ANN atau kinerja jaringan ditentukan dari nilai koefisien korelasi (R) yang diperoleh dari data pelatihan (training), sedang untuk pengujian (testing) dihitung dari nilai RMSE (root mean square error), Standard Error of Prediction (SEP) dan Coefficient of Variation (CV) antara nilai hasil prediksi berdasarkan pemodelan ANN dan nilai pengukuran di lapangan (observasi).