RAINFALL FORECAST USING KALMAN FILTER METHOD FOR SETTING CROPPING PATTERNS : Abstract

• Main Author: Estiningtyas <br> NIM. 224 02 002, Woro
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/6920
• Institution: Institut Teknologi Bandung

</i><b>Abstract: <i></b><p align="justify"> Deterministic model applied to forecast rainfall in tropical region in which its determinants are quite complicated, dynamic and random is unmanageable. Therefore, it needs renewable statistical model in real time. Crops water balance is computed by using local rainfall pattern, but by increasing the intensity and frequency of climate anomaly the computed water balance needs to be renewed more frequently through cropping pattern setting based on forecast aspects. Many approach models have been applied to forecast climate using statistical model, deterministic model or its between independent and dependent variables empirically. It is more practical to analyze the indications but it needs validation anytime and anywhere. Kalman filter combines physical and statistical model approaches to stochastic model renewable anytime for objective of on line forecasting. Validates model correlating rainfall and SST gives correlation coefficient value of more than 75%. It implies that predicting model using Kalman Filter is feasible to forecast montly rainfall. Correlation model between rainfall and SST mostly applied is ARMAX (46%). The shortest time series data period employed to run the model is between 7 to 10 years. Results of cropping pattern scenarios based on predicted rainfall show there are periods with harvest losses more than 20% especially in the locations with unequel annual rainfall distribution. Thus, it is not recommended to plant seasonal crops.</p>