Weather Data for SE Asia

We first looked at rainfall data from the Global Daily Summary CDROM from the National Climate Data Center. We then set this aside to look at more contemporary time frames where better sources of rain data were available.


This shows our study basins, the 2 buffer used for selecting weather stations, and the 5 used for selecting radiosonde stations. The rectangle indicates the size of the data grids being passed on the the VIC model.

For January ,1997 through September, 2000, we downloaded the daily records from the National Climatic Data Center 1254 stations fall within 2 degrees of our study area. The number of stations with data on any given day is:

(This graph is actually the count of valid temperature stations; the number of wind and rainfall stations varies slightly.)
The initial phase of our processing is done in geographic coordinates. This gives a bias to the weights of stations to the east or west of a point, but the effect is small at these latitudes. Subsequent calculations involving areas all use corrected values. Max temperature, min temperature, wind, and precip, were extracted from the NCDC ASCII files and placed in 1368 daily files. These files were converted to ARC/INFO "coverage" format. For the five missing days, grids were created by temporal interpolation.

Quarter-degree grids of wind speed were created for each day by using an inverse-distance-weighted algorithm built into ARC/INFO. The algorithm averages all stations within 0.18 degrees of the grid center, but extending the search until at least 4 stations are found.

Precipitation grids were made in a similar manner, but with less confidence. We judged the NCDC GPCP (Gloabal Precipitation Climatology Project) one-degree rainfall to be superior to any product we could devise, so we created our grids only for comparison.

TRMM year 2000 data

We needed a different data source for rain in 2000, so Sarah downloaded TRMM data from NASA Data file names are: 3B42.990101.5.HDF etc. The data is column-major, south-north, west-east. We uncompress, skip to the data, clip out our (86,-12) to (140,36) rectangle, and rotate in one operation, e.g.
zcat 3B42.990101.5.HDF.Z | cutout 320 54 48 112 85414 4 90 > 990101
OR
foreach month (01 02 03 04 05 06 07 08 09 10 11 12)
  foreach fil (3B42.??${month}*)
    zcat $fil | cutout 320 54 48 112 85414 4 90 > `echo $fil | cut -d '.' -f2`
  end
end
Comparing 990101 to the NCDC 1dd data (after multiplying by 25.4), the mean difference is 0.22 mm, with a with abs(diff) of 3.33 mm, while all rainfall patterns are maintained.

Compare GPCP and TRMM rainfall (as we resampled to .25 degrees) for January 1, 1999. The no-data gaps appear in about half to TRMM data, and are filled by simple extrapolation. Greenish lines indicates basins being modeled. The GPCP file shows sharper peaks, but the mean values vary by only 1.4%


Here is the wind interpolated from station data for that day.


This is minimum temperature, adjusted to sea level and interpolated. Weather stations with wind data are shown.


and this is the same data, adjusted back to surface elevation.

Here are graphs of the datasets.



These graphs represent unmasked rectangles. The data points were selected near our study basins, so there are some severe extrapolations in areas which will be masked out in the VIC model.

DOWNLOAD NCDC GPCP rain data 1/97-11/99, including 1/10/98 fix
DOWNLOAD replacement NCDC GPCP rain data for Jan 10, 1998(see note below)
DOWNLOAD clipped TRMM data
DOWNLOAD wind data interpolated from stations.
DOWNLOAD minimum surface temperature interpolated from stations. (based on constant lapse rate)
DOWNLOAD maximum surface temperature interpolated from stations. (based on constant lapse rate)
DOWNLOAD new NCDC GPCP rain data 11/99 - 12/2000 A comparison of this November 1999 with the previous one shows that NCDC made minor updates to their files. The file names in this dataset differ from the old ones in having 4-digit years.

DOWNLOAD new TRMM rain data 10/2000 - 03/2001

DOWNLOAD newer TRMM rain data 4/2001 - July 4,2001 DOWNLOAD still newer TRMM rain data July 1,2001 - November 30,2001

Note: There were only 43 weather stations for January 10, 1998. Therefore, temporal interpolations of temperature data were used. The NCDC GPCP rainfall record for that date looks plausible enough (It is, afterall, a synthesis of satellite and ground data.), but it happens to be be the only daily rainfall record for which the maximum is less than 25.6 mm. This caused problems with the export routine, which wrote an 8-bit file instead of a 16-bit file. Two different fixed files are contained in gpcp_jan10.zip. The file 980110.bil has a cell with rain=25.6mm tweaked into the northeast corner, forcing our ARC/INFO software to write it as a 16-bit file. An alternative version (980110b.bil) is the average of Jan09 and Jan11.

Questions: Harvey Greenberg