Merged chlorophyll-a time series for the Drake Passage area

Mati Kahru, mkahru@ucsd.edu

 

·        Merged time series of chlorophyll-a data from full-resolution Aqua-MODIS, SeaWiFS and Terra-MODIS data are created for the Drake Passage and Scotia Sea area. The Level-2 data are being used. Both SeaWiFS MLAC (Merged Local Area Coverage) and GAC (Global Area Coverage) data as well as MODIS Level-2 are downloaded from the NASA Ocean Color website http://oceancolor.gsfc.nasa.gov/ . The sub-setting limits for the area were: -55-65 S; 70-45 W.

 

·        Two Mercator projections generated with the Terascan (SeaSpace Corp) utility master were chosen as standard maps for respectively, Regional and Local view of the area. The Regional view has the following parameters: center_lat=59S, center_lon=55W, num_lines=1100, num_samples=1280, pixel_width=1.0, pixel_height=1.0.  The Local view has the following parameters: center_lat=61S, center_lon=57.5W, num_lines=704, num_samples=824, pixel_width=1.0, pixel_height=1.0.  The sample images below are in the larger, Regional view.

·        Mapped composites were generated for both sensors for each calendar day using all available passes with a WAM utility wam_l2_map. Level-2 flags ATMFAIL, LAND, HIGLINT, HILT, HISATZEN, STRAYLIGHT, CLDICE, HISOLZEN, HITAU, LOWLW, CHLFAIL, NAVWARN, CHLWARN, DARKPIXEL, SEAICE, NAVFAIL were used to eliminate low-quality pixels. Other flags (BADANC, COASTZ, NEGLW, COCCOLITH, TURBIDW, ABSAER, TRICHO, MAXAERITER, MODGLINT, ATMWARN, FILTER, SSTWARN, SSTFAIL) were ignored. It was found that the pixels marked with the latter flags were statistically not different from neighboring pixels while their elimination would have decreased the amount of usable data. In addition to eliminating the flagged pixels, the cloud image determined with the flag CLDICE was dilated (expanded) to eliminate contaminated pixels near cloud edges. Cloud edges are often associated with erroneously high chl-a values. The resulting mapped daily composite was saved as HDF with the standard chlorophyll log-scaling. In addition, a 2-times reduced and annotated JPEG quick-look was saved as well. The filename pattern (e.g. S2004001_chl_a_mapped.hdf and S2004001_chl_a_mapped.jpg for SeaWiFS) shows the year, the Julian day of the year and the variable. For Aqua the first letter of the filename is “A”. Additional information is stored as HDF attributes.

·        When applied to MODIS-Aqua data the utility wam_l2_map produces both chl and SST daily mapped composites using all available passes during the calendar day (using UTC time). Whereas ocean color data is available only during light period of the day, infrared SST data is also available during night. It was found that SST data derived from processed MODIS-Aqua Level-2 datasets had considerable differences between different passes, due to natural diurnal variability and/or unresolved observation effects (e.g. viewing angle). As a result, composited daily SST images look patchy and unnatural. I will try to refine the SST compositing process, e.g. by eliminating passes with large viewing angles or imposing other restrictions.

·        For the 2002-2004 time period both MODIS-Aqua and SeaWiFS were available. Daily datasets from both sensors were composited into merged images with a WAM utility wam_composite_2sensors. The same naming convention was used, except the first letter of the filenames was replaced with “C”. The following is a sample image from January 28, 2004. A complicating issue is the different resolutions of SeaWiFS data. SeaWiFS MLAC (1 km) data was not always available during the time of SeaWiFS regular operations (1997-2005) and therefore GAC (4-km) data had to be used. Starting from the end of Decmber, 2005 only GAC SeaWiFS data is available (with a delay). When merging Aqua 1-km data with SeaWiFS 4-km mapped data, Aqua data were given precedence.

 

·        As daily images are largely covered by clouds, multi-day composites have to be used. As the first step, 5-day composites were created using either the single sensor daily composites (SeaWiFS MLAC or GAC or Aqua-MODIS) or the SeaWiFS-Aqua combined daily composites. The 5-day composites were created with the WAM utility wam_composite_5day. This created 2 types of HDF files: composited average concentration images and count images (the number of valid pixels used for averaging each pixel). The filename pattern (e.g. C2004001_C2004005_chl_a_comp.hdf and 2004001_S2004005_chl_a_count.hdf) shows the start year and day and the end year and day of compositing. For each average image a 2-times reduced JPEG quick-look was created as well. The following example shows a 5-day average for January 26-30, 2004. Note the increased coverage compared to the previous daily image.

 

·        As even the 5-day composites are partly blank due to clouds, 15-day, 30-day and 60-day compositings were performed with a WAM utility wam_composite_2x. This process averages adjacent 5-day composites, e.g. for days 1-5, 6-10 and 11-15. The filename pattern is similar to the 5-day composites (e.g. C2004001_S2004015_chl_a_comp.hdf) and shows the year, start and end days of compositing. The reduced JPEG image is also saved. The following example shows a 15-day average corresponding to the previous daily and 5-day examples.

 

·        An interactive web-based utility using ASP.NET can be used to show any of the two images (5-day, 15-day, 30-day) simultaneously and can be accessed at http://spg.ucsd.edu/Show2SetsOfImages/Show2SetsofImages.aspx. First you need to pick the category of image (“Pick 1st image of:”) and then the individual image in that category. Then you can do the same for the 2nd image. The images shown in the browser are compressed. You can download the full-resolution JPEG images by right-clicking on the image and choosing “Save Picture As:”.

·        The 1-km daily, 5-day and 15-day images are all available as HDF or JPEG. The HDF files have the numerical values for each pixel and can be used to generate statistics for any selected area. The HDF files can be read with any HDF-aware software (e.g. Matlab, IDL, and WIM). When read with WIM the geo-location and scaling are automatically retrieved. For geo-location with other software the latitude and longitude arrays corresponding to each pixel must be created.