26 Feb. 2003
Kooiti Masuda (Frontier Research System for Global Change, Yokohama, Japan)
This text is almost verbatim translation of my abstract (in Japanese) for the 2003 Spring Meeting of the Meteorological Society of Japan (MSJ). This draft contains figures (included as HTTP links) and a reference list which do not appear in the abstract to be printed by MSJ.
We evaluate how useful the GAME Reanalysis data sets, which cover the period from April to October 1998, for discussions of the hydrological cycle over continents.
The major 70 river basins of the world, which was identified by Oki et al. (1995), is re-evaluated on a 1-degree latitude-longitude grid based on the gridded drainage network of TRIP (Oki and Sud, 1998).. They are drainage areas of stations where river discharge is observed. The stations are selected so as to represent as large area as possible of the river basin. The area of each basin is more than 100000 km2.
From GAME Reanalysis (see http://gain-hub.mri-jma.go.jp/GAME_reanal.html), two versions (Version 1.5 and Version 1.1) of "2-dimensional physical monitor" (2dmon) data sets are used. The 2dmon contains 12-18 hour forecasts in Version 1.5 and 12-24 hour forecasts in Version 1.1. The values are available in 1.25 deg. latitude-longitude grids at 6-hour time interval. The values of precipitation, evaporation, vertically integrated water vapor content (so-called precipitable water) and vertically integrated horizontal transport vectors of water vapor are extracted from the data set. Each term of the equation of mass balance of water vapor balance in the atmosphere is evaluated in monthly basis, interpolated to the 1-degree grid and summed over each basin.
For comparison, terms of water balance is similarly calculated from NCEP/NCAR Reanalysis (Kalnay et al., 1996) and NCEP Reanalysis 2 (Kanamitsu et al., 2002). In both of NCEP's products, precipitation and evaporation are 0-6 hour forecasts and horizontal transport vectors of water vapor are computed by Masuda from sigma-level analyses. They are available in the 192 x 94 Gaussian grid (about 1.9 degree interval) and interpolated to 1-degree grid here.
Also, following observation-based data sets of precipitation are used.
For Indochina Peninsula, also a data set tentatively called "GAME-gauge" is produced. It is based on daily precipitation (*1) from operational observations of Southeast Asian countries available from two sources: one is GAME-T CD-ROM (http://hydro.iis.u-tokyo.ac.jp/GAME-T/GAIN-T/) and another is Lower Mekong Hydrologic Yearbook 1998 published by the Mekong River Commission (http://www.mrcmekong.org). Monthly values at the stations are calculated and interpolated (*2) to 1-degree grids for this study.
Precipitation of GAME Reanalysis (both 1.5 and 1.1) matches observation-based data sets in Lena, Yenisey, Ob and Mackenzie. In these regional averages, difference between observation-based data sets are negligible in the time period of GAME Reanalysis (though there is problem of correction of rain gauges in winter).
On the other hand, GAME Reanalysis is not so good as NCEP's in the consistency of atmospheric water balance (i.e. the smallness of the residual in the budget equation when all terms are evaluated with data). It is suspected that both precipitation and evaporation are too large in NCEP's two versions, and only evaporation is too large in GAME's two versions.
Precipitation of GAME Reanalysis (both 1.5 and 1.1) matches observation-based data sets in many river basins in Southeast and East Asia, including Chao Phraya and Huai He which were areas of enhanced observations for GAME. NCEP's products have larger difference from observation-based products. Especially, precipitation in Indochina Peninsula according to NCEP Reanalysis 2 is exceedingly large.
In some basins in the tropics such as Amazon and Ganges, precipitation according to GAME Reanalysis is too large and NCEP gives better values (if we rely on observation-based data sets such as GPCP).
Distribution of precipitation in Indochina Peninsula in the rainy season (northern summer, actually May to August 1998) is compared. All data sets show maxima along coasts that face oceans to the west (upstream of the summer monsoon). According to GAME-Gauge (right panel), there is also minima in inland basins of Myanmar and Thailand and a maximum along the highlands of Laos. GAME Reanalysis (middle panel) shows smoother distribution with weaker contrast within inland area.
Distribution is also smooth according to GPCP (left panel) and GPCC (figure omitted) products. It is presumably because reports from Myanmar, Laos and Cambodia rarely reach German Weather Service (where GPCC is located) by the Global Telecommunication System and not incorporated in the production of these products.
See an example of counts of rain gauges in July 1998 that entered to GPCC Monitoring Product (left panel) and GAME-gauge product (right panel).GPCC also makes "Full Data Product", not yet open to public, which may incorporate reports from these countries.
Precipitation of GAME Reanalysis matches observations in many regions when averaged at space scale larger than about 500 km and time scale of one month. On the other hand, it does not depict spatial distribution of scales smaller than that.
http://EarthInteractions.org/].