月降水量に適した情報の与え方と2変数最大エントロピー分布

概要

論文の詳細を見る
Methods for obtaining optimum information for a 2-variable maximum entropy distribution (2MED) were examined to estimate the hydrologic variable for the design of a water-supply system between two basins. Information consisting of statistical moments, and combined information of statistical moments and the expectations of an exponential function were tested with a set of rainfall data in Nagano, Nagoya and Tokyo. The goodness-of-fit (Log-likelihood,Akaike's information criterion and histogram) and stability of the return period were used as evaluation criteria in selecting suitable information. Though the 2M(x², exp(-4x / Mx), y², exp(-4y / My), xy) distribution for combined information ranked first in terms of goodness-of-fit, it did not provide good stability of the return period because it exhibits a sensitive reaction with regards to removed data. The 2M(x, x², y, y², xy) distribution for statistical moments only showed excellent results for both evaluation criteria. Therefore, it was judged from these results that the 2M(x, x², y, y², xy) distribution was the optimum distribution for the data set used in this investigation.
信州大学工学部の論文
1991-09-30