Penentuan Bobot Optimum dengan Pengganda Lagrange untuk Penggabungan Nilai Dugaan Ekstrim Curah Hujan

Abstract

The phenomenon of extreme rainfall could give negative impact such as a high risk of failure in agriculture and plantation productions. The estimation of extreme values using ensemble method to combine the estimates of modified Champernowne distribution and Generalized Pareto Distribution (GPD) is expected to anticipate the risk of production failure. The goals of this research are to determine optimum weight by Langrange multiplier in estimating the ensemble two extreme rainfall values based on modified Champernowne distribution and GPD, to compare the weighted results using Langrange multiplier by weighting iteratively and linear regression methods, and to forecast extreme rainfall values. The research used daily rainfalls data from January 1, 1985 to March 31, 2011 in Darmaga-Bogor Climatology Stations. The estimations using GPD tends to overestimate while the modified Champernowne distribution tends to underestimate. The ensemble method to combine both estimates using Langrange multiplier resulted a better estimate. The Langrange multipliers could be used as an alternative method to determine the optimum weight without trial and error process as in the iteration process. Forecasting one month ahead is better using the ensemble method.