気象研究所研究報告 第 58 巻 73-83 頁 平成 19 年 8 月 Papers in Meteorology and Geophysics Vol. 58, 73-83, August 2007 doi:10.2467/mripapers.58.73 © 2007 by the Japan Meteorological Agency / Meteorological Research Institute Evaluation of Seasonal Cycles of Hydrological Processes in Japan Meteorological Agency Land Data Analysis by Tosiyuki Nakaegawa 1 , Takayuki Tokuhiro 2 , Akira Itoh 2 and Masahiro Hosaka 1 1. Climate Research Department, Meteorological Research Institute, Tsukuba, Japan 2. Japan Meteorological Agency, Tokyo, Japan (Received October 16, 2006; Accepted May 7, 2007; Published August 1, 2007) Abstract The present study evaluates the reproducibility of the climatological seasonal cycles of three hydrological variables, soil moisture, river discharge and terrestrial water storage in the Japan Meteorological Agency Land Data Analysis (LDA) by means of a system consisting of the LDA and a global river-routing model (LDAG). LDAG satisfactorily reproduces the seasonal cycle of soil moisture at most of the observation stations; however, negative correlations are found at stations in the mid latitudes of east Eurasia. The amplitudes of river discharge are well reproduced only in a few rivers. The relative errors of the amplitudes exceed 20% in many of the rivers. The correlation coefficients of the seasonal cycle are more than 0.5 in most of the 70 rivers investigated. The phase differences occur within one month in most of the rivers. The relative errors of the amplitude of terrestrial water storage are less than 10% in five of the top 10 largest river basins, and the correlations are high at the river basins located in areas such as the tropics and the Asian Monsoon region. The LDAG phases are generally earlier, similar to those of river discharge. Close examinations demonstrate that better reproducibility requires improvements in both the numerical model and the forcing data. Despite good reproducibility only in some river basins, the hydrological variables produced by LDAG may be useful for land surface initial conditions for weather forecasts and hydrological studies. 1. Introduction The impact of initial soil moisture on seasonal prediction has been investigated using the global climate model (e.g. Kanamitsu et al., 2003; the GLACE team, 2004; Nakaegawa and Kanamitsu, 2006), since soil moisture significantly influences seasonal predictability as a lower boundary condition (e.g. Shukla et al., 1998). However, in-situ and satellite observations have their own disadvantages, and no global soil moisture dataset is available. The in-situ data have been archived at the Global Soil Moisture Data Bank (GSMDB; Robock et al., 2000) but cover only limited areas and periods. This often leads to wide use of the reanalysis data, especially for initial conditions of weather forecasts (e.g. Kuma, 2000; Kanamitsu et al., 2002b). The Japan Meteorological Agency (JMA) routinely operates the Land Data Analysis (LDA; Tokuhiro, 2002) to produce the global land-surface dataset. The land-surface model is run in an offline mode forced by the outputs obtained from the global atmospheric analysis (GANAL) and the snow depth analysis (SDA). A one-month hindcast experiment of the JMA global model using the LDA outputs as land surface initial conditions alleviated negative biases of the ground temperature over Eurasia (Tokuhiro, 2002) and confirmed the significance of the initial land surface data. Reproducibility of the LDA was investigated with respect to soil moisture and snow depth in the developing stage when the European Center for Medium-Range Weather Forecast 15-year Reanalysis (ERA-15; Gibson et al., 1999) was used as forcing data (Tokuhiro, 2000, 2001), but reproducibility in the operational mode has not been fully investigated until now. Corresponding author: Tosiyuki Nakaegawa Meteorological Research Institute, 1-1 Nagamine, Tsukuba, Ibaraki 305-0052, Japan. E-mail: tnakaega@mri-jma.go.jp