検索対象:     
報告書番号:
※ 半角英数字
 年 ~ 
 年

福島における放射性物質分布調査,8; 陸域動態モデルの高度化と福島への適用

Investigation on distribution of radioactive substances in Fukushima, 8; Improvement of the MERCURY model and its application to the Fukushima

佐久間 一幸   ; 町田 昌彦  ; 山田 進  ; 操上 広志  

Sakuma, Kazuyuki; Machida, Masahiko; Yamada, Susumu; Kurikami, Hiroshi

河川を経由し海洋へと流出する放射性物質量をより精度良く推定するために、放射性核種流出推定モデルMERCURYにGISを活用したパラメータ設定方法及びパラメータ最適化手法を組み込んだ。対象河川流域内の標高(100m間隔),傾斜度(10$$^{circ}$$間隔),土地利用,土壌,表層地質の面積比率をGISで算出し、MERCURY内5つのパラメータを目的関数,流域毎の面積比率を説明変数として重回帰式を作成した。前田川,熊川,請戸川,阿武隈川で作成した重回帰式と高瀬川で取得した面積比率からパラメータを決定し計算を実施した。自動キャリブレーション機能としてニュートン法,PSO法,SCE-UA法およびベイズ最適化法の4つの最適化手法を実装し、上記5河川に適用した。高瀬川を対象とした検証の結果、半年間程度の河川流出量の解析において相対二乗誤差RSEが0.44程度となり、重回帰式に供した河川数は4河川と少ないが良好な結果となった。一方、自動キャリブレーション機能については、RSEの幅がニュートン法(0.29-1.5),PSO法(0.28-0.56),SCE-UA法(0.18-0.39),ベイズ最適化法(0.29-0.42)となった。パラメータ空間は多峰性があり、ニュートン法では十分に最適化できないものの、PSO法,SCE-UA法およびベイズ最適化法で十分な精度を確認できた。

In order to estimate the amount of radioactive materials discharged into the ocean via rivers more accurately, a GIS-based parameter setting method and parameter optimization method were incorporated into the radionuclide discharge estimation model MERCURY. The elevation (at 100 m intervals), slope (at 10-degree intervals), land use, soil, and surface geology area ratios in the target river basins were calculated using GIS, and multiple regression equations were developed using the five parameters in MERCURY as objective functions and the area ratios for each basin as explanatory variables. The parameters were determined from the multiple regression equations created for the Maeda, Kuma, Ukedo, and Abukuma Rivers and the area ratio obtained for the Takase River, and calculations were conducted. Four optimization methods, the Newton, the PSO, the SCE-UA, and the Bayesian optimization, were implemented as automatic calibration functions and applied to the above five rivers. The results of the verification for the Takase River showed that the relative squared error RSE was about 0.44. Although the number of rivers used in the multiple regression equation was only four, the results were good. On the other hand, for the automatic calibration function, the RSE ranged from 0.29-1.5 for the Newton method (0.29-0.5), 0.28-0.56 for the PSO method, 0.18-0.39 for the SCE-UA method, and 0.29-0.42 for the Bayesian optimization method.

Access

:

- Accesses

InCites™

:

Altmetrics

:

[CLARIVATE ANALYTICS], [WEB OF SCIENCE], [HIGHLY CITED PAPER & CUP LOGO] and [HOT PAPER & FIRE LOGO] are trademarks of Clarivate Analytics, and/or its affiliated company or companies, and used herein by permission and/or license.