Satellite Remote Sensing - Data Information
The quality of satellite images used for the generation of SST (Sea Surface Temperature) composites can be affected by the presence of undetected clouds or ice, or by an occasional malfunction of a satellite sensor. Temperature values computed over pixels exhibiting those imperfections then become incorrect.
Various levels of filtering are applied to images in order to remove bad pixel values, giving rise to various versions of the composites.
SST product types available on SLGO
The “Data type” drop-down menu holds four entries related to SST composites.
“Mean Temperature” represents the union of L0 and L1 products; the L1 version (more reliable) is always offered when both L0 and L1 versions exist for a given product. Unfortunately, an L1 version does not become available until at least a month after the underlying images have been acquired; therefore the L0 version is used when being the only available version. In fact, the L0 version is returned, so that users are not misled into thinking that no products are available for the most recent days.
Unfiltered Temperature (L0)
This level corresponds to the absence of any filtering other than the basic one allowed by the TeraScan system (www.seaspace.com). Composites at this level are computed from SST maps generated just a few hours after the satellite passes become available. They can exhibit wrong temperature values at various spots.
Temperature with cloud and ice filters (L1)
At this level, composites are produced using images subjected to a series of three distinct filters.
a. Due to limitations of the TeraScan SST computing procedure, temperatures outside the range [-1.7, 27.9] are excluded.
b. The second filter is aimed at excluding cold values originating from cloud contamination of an image. It is derived mostly from empirical findings. Bad values for an image of a given day ‘d’ are identified by comparing each pixel value with the maximum temperature value observed at this same pixel during the two seven day periods bracketing that day, e.g. days j-7 to j-1 on the left, and days j+1 to j+7 on the right at that day. A pixel value is excluded if it is colder than 5 degrees and if it is inferior by more than 4 degrees to the above defined maximum.
c. The third filter tries to eliminate wrong temperature values due to ice contamination inside an image. Each image is subjected to an ice-mask computed from daily ice coverage data published by the American NSIDC (National Snow and Ice Data Center, http://nsidc.org). A temperature value computed at a given pixel is excluded if it is less than 14 degrees and if the ice coverage shown for this pixel in the NSIDC map is more than 1%.
Temperature with climatological filter (L3)
Composites of this level are produced using images which are first filtered similarly to the L1 products. A fourth climatological filter is then applied.
a. Using L1 daily composites in year range [1985,2009], a 25 year climatology of the SST temperatures was first computed for each calendar day of the year. Those daily climatologies then are used to compute climatologies for the 7-days periods starting at each day of year (Jan 01 to Jan 07, Jan 02 to Jan 08, etc.). The resulting products are designated as L1 climatologies.
b. A new set of daily composites, designated as L2 composites, was produced by first subjecting the images to the three filtering steps defined above for the L1 composites, followed by a fourth filter based on the count, mean, and standard deviation computed for each pixel inside the 7-days L1 climatologies. For each pixel at a given day, a confidence interval is defined, centered on the mean for that day in the 7-day climatology, and whose width is a multiple of the standard deviation. The multiple is a function of the N count of values measured at that pixel, as given by the following table :
c. The climatology filter excludes a pixel inside an image if its value falls outside the confidence interval assigned to the corresponding pixel inside the 7-day climatology centered on the image day of year. A pixel is also excluded if its associated count is less than 3.
d. Using this new set of L2 daily composites, a refined version of the 25 year climatologies (both 1-day and 7-days) was computed. The resulting products are designated as L2 climatologies.
e. Level L3 composites are produced by subjecting the images to the L1 filtering procedure, followed by a climatological filter based on the 7-days L2 climatologies. The climatology filter is applied similarly to the one previously described for L2 daily composites.
L2 daily composites were used just as intermediate products in the process of producing more accurate climatologies. They are not made available.
a. L0 composites are normally computed with a 1 day delay. The most recent composites (those of the current month plus the last week of the preceding month) are all of level L0.
b. Constraints related to availability of the products needed in the filtering steps explain why the production of L1 and L3 composites suffers some delays, generally 4 to 5 weeks. Filtered composites are usually computed on the 8th of each month ‘m’, over the period covered by the last week of month m-2 and the first 3 weeks of month m-1.
WMS (Web Map Service)
SST composites are made available through a WMS service, so that a user equipped with a GIS software supporting WMS connections can easily access the temperature maps over the internet. In fact, the SLGO portal directly relies on this service for browsing through the SST image catalog when used in basic display mode.
The WMS service can be reached at the following basic URL that can be used when the GIS software requires the identification of the WMS service:
The above URL gives access to products of the “Mean Temperature” category available for the most recent day. Those will always be of L0 type (see “Availability” above).
Two custom parameters can be appended to the URL to select the category and the time period for which to return products. When used, a ‘?’ character must prefix the first appended parameter. When both are used, they must be separated by a ‘&’ character. No space should appear in the appended string.
coll=cat, where ‘cat’ identifies the product category. It must be one of
msst select “Mean Temperature” category; this is the default
msst-L0 select “L0 category”
msst-L1 select “L1 category”
msst-L3 select “L3 category”
period=per, where ‘per’ identifies the wanted time period. It can be entered as a single value, or a series of comma-separated values. A date value can identify a single day, or a time interval composed of a starting and ending date separated by a ‘/’ character.Both value formats can be intermixed. A date value is entered in the ‘yyyy-mm-dd’ format. For example:
2011-03-01 March 01 2011
2011-03-01,2011-10-10 March 01 2011 and March 10 2011
2011-03-01/2011-03-07 From March 01 2011 to March 07 2011
2010-03-01/2010-03-07,2011-03-01/2011-03-07 First 7 days of March 2010 and March 2011
By default, the products of the most recent day for which they are available are selected.
For example, using the following URL when identifying the WMS service inside the GIS tool will result in the tool asking the service to return layers pertaining to level L1 SST composites available for the first 5 days of July 2006 :
NOTE. The above URL is NOT a valid WMS command. The GIS WMS tool will take care of appending the parameters needed to turn it into a valid WMS command.
Connectivity to the WMS service can be checked by sending the service a ‘getcapabilities’ command similar to the one sent by a WMS client to query the service about what it has to offer. Enter the following URL in an internet browser :
The service should reply with an XML document formatted as per the OGC WMS specification.