Main Article Content
Prediction of season onset is important for many sectors, particularly on agricultural practices, as its usage for reducing climate risk and planning activities. Current knowledge on season onset prediction mainly focused on large area, which remains research challenge for local level. This research developed model prediction of season onset for Malang Regency, East Java based on global climate data. The research specifically aimed to: (i) determine the onset date of rainy and dry season, (ii) generate equation for onset date prediction using principal component regression (PCR) approach, and (iii) evaluate the model performance. We depend on statistical model based on a combine of domain time and principal component analysis (PCA) for atmospheric variables, namely sea level pressure, outgoing longwave radiation, and zonal wind. We used the Tropical Rainfall Measuring Mission (TRMM) data for model evaluation, especially for determination of onset date. Based on cumulative anomalies rainfall, the onset date for dry season occurred in the early May, whereas for rainy season it was in early November. The results showed that regression models of the principal components had a good skill to predict onset date for both seasons. It has been confirmed by a low error and a high correlation. Visually, the dynamic of onset dates from model was mostly identical to the observation. The predictive model for rainy season had higher performance compared to the model for dry season. This finding was confirmed by insignificant difference resulted from the independent t-test between model and observed onset dates. The best model for dry season was conducted by domain time of February, whereas for rainy season was domain time of August. This research can be used to complement previous studies regarding season onset prediction in Indonesia.
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.