Two types of grids are included, whose difference is clarified later:
Briefly, the Pathfinder algorithm differs from the MCSST algorithm used since 1981 to process AVHRR in that it includes a number of additional instrument corrections, recovers about twice as many SST values in the presence of clouds as the original MCSST, and uses a 5 month moving window to estimate the coefficients for the central month (the coefficients are used to linearly combine the radiances at the 5 AVHRR frequencies into an SST estimate). It also differs from the widely-used Reynolds SST grids which use as input the MCSST calculation and then combine them with in-situ data using a mapping algorithm to estimate weekly, 1 degree grids (see Reynolds SST; the Reynolds data can be obtained by ftp).
, longitude
, time t is a linear combination of observed
temperatures at points i as follows:
T(
,
,t) =
i
[T(
i,
i,ti)
* exp(-0.6931 * ((
-
i)2/
w
2 + (
-
i)2/ w
2
+ (t - ti)2/wt2))]
where w
= 1 deg, w
= 1 deg, wt = 2 days
The
is performed over all points i inside
the volume
|
-
i |
<= 2.5 deg, |
-
i
| <= 2.5 deg, | t - ti | <= 5 days
The first pentad is "centered" on January 3, 1990. The specific dates, summary statistics on each grid, and browse images of each, can be found in the Table of AVHRR Grids with Summary Statistics. A Java program to animate a time series of images between any two dates is also available.
For historical reasons, the output of the interpolation program is scaled to one byte integers. Therefore, the smallest temperature difference that can be seen from these grids is 0.15 .
|
These plots illustrate the change in data processing for 1990,1993-1999
versus 1991-1992 versus 2000-2001: the 1991-92 data, dubbed version
4.0, were found to have inadequate flagging code; the problem was
corrected during January 1998, and reprocessing started with data year
1994. Unfortunately, at the time this DVD folder was generated,
the 1991-1992 had not yet been reprocessed with the new flags.
A postscript version of this figure is also available. |
![]() |
| The figures corresponding to the 1.0-degree interpolated
grids are significantly different from the binned-grid statistics. The
standard deviations are higher because the data gaps are due to clouds
which obscure the same regions of the ocean much of the time, systematically
in certain higher and lower energy regions of the ocean. The mean temperature
values are lower because a greater number of interpolated lower temperatures
(higher latitudes) are included.
A postscript version of this figure is also available. |
![]() |
0.5 Degree 9-Year Mean
1.0 Degree Gaussian Interpolated 9-Year Mean
All data files contain the following variables:
See the SST 0.5° Variables and Attributes and SST 1.0° Variables and Attributes for more detail.
The filenames for the NetCDF files are of the form sst05d19900103.nc or sst10d19900103.nc, depending on whether they are 0.5-deg binned, or 1.0-deg interpolated; this is a WOCE Data Products Committee agreement. However, the filenames for the corresponding GIF files are sst05d0002.gif or sst10d0002.gif, where the date is given in cumulative days, with day 0000=1990-01-01, because the data animation software expects monotonically increasing numbers in the filename. The correspondence between both conventions is in the Table of SST statistics (0.5-deg) or the Table of SST statistics (1.0-deg)
gunzip will decompress files created by gzip. On Windows 95, 98, 2000
or NT use Winzip or pkzip to decompress .gz files.
On Mac systems Stuffit will decompress .gz files.
For more information on gzip, please refer to the gzip home page at
http://www.gzip.org