著者
蒲地 政文
出版者
社団法人 日本流体力学会
雑誌
日本流体力学会誌「ながれ」 (ISSN:02863154)
巻号頁・発行日
vol.13, no.6, pp.440-451, 1994-12-31 (Released:2011-03-07)
参考文献数
45
被引用文献数
3

Data assimilation has been recently considered as a key and essential component of understanding of oceanic phenomena, development of parameterization process in numerical ocean modeling, mapping out of observation strategies and forecasting of ocean state. Variational adjoint method is introduced as one of the data assimilation method in this introductory review. The basic concept of the variational adjoint method is to combine observations via variational method (estimation theory or Lagrange's multiplier method) with a dynamical model. The method is applied to a parameter estimation in a simple model of an unsteady Ekman flow in physical oceanography. Recent developments of the method are also introduced briefly.
著者
石崎 廣 五十嵐 弘道 荒井 頼子 蒲地 政文 石川 洋一 齊藤 誠一
出版者
Faculty of Fisheries Sciences, Hokkaido University
雑誌
Memoirs of the Faculty of Fisheries Sciences, Hokkaido University (ISSN:24353361)
巻号頁・発行日
vol.60, pp.1-31, 2021-12

When a histogram-based method for front detection was applied to the sea surface temperature (SST) and chlorophyll-a (CHL) data by ‘Himawari’ and ‘Shikisai’ (GCOM-C) satellite, the obtained frontal patterns were scale-selective, corresponding to the window scales (W). On this basis, the optimum initial smoothing condition as the data preprocessing was searched for, that maximizes the frontal edge point detection rate to the given W. As the smoothing filter, the median filter (MF) and the Gaussian filter (GF) were used solely or co-used. As the result, it was found that the frontal edge point detection rate was maximized when the original data were smoothed until the scale of about a half of W, that is, when the disturbances with the scales less than about a half of W were removed, for the low-latitude SST data of ‘Himawari’ with the co-use of MF and GF. Namely, the optimum smoothing scale (D) that maximizes the frontal edge point detection rate is roughly proportional to W. The mean values (Rm) of the ratio of D to W (R=D/W), averaged over the practical range of W for various data and regions, fall in a range 0.3 < Rm < 0.5.