Gradients removed in MODISA night SST

[schckngs]

Hello ocean colour folks,

I have been making 1 km SST layers by downloading night-time MODIS-Aqua L2 SST images, and binning / mapping into composites with SeaDAS.
However the non-uniformity flags result in valid pixels along SST fronts and gradients being removed. This removes quite a lot of data in my region (Pacific North America) especially during summer. Some example images are attached below.

I have been using the "best quality" option in l2bin (qual_max = 0) and experimenting with other quality levels doesn't make much of a difference. Is there a recommended method that I could try to reduce this effect? E.g.
- This appears most pronounced with night-time data, it might be worth just using day-time data instead?
- Perhaps it's worth switching to VIIRS for more recent years / merging it with the MODISA SST?
- Instead of downloading L2 I could process from L1A - are there SeaDAS settings recommended to reduce or mitigate this issue?

The code I'm using to bin and map is below.
Thanks very much,
-Andrea

Code: Select all

yr=$1
L2BIN_RES="1" 
L3MAP_RES="1km"
l2_suffix="L2.SST.nc"
north=56.5
south=46
east=-122
west=-139.5

months=$(seq 1 1 12)
days=$(seq 1 1 31)

for month in $months; do
  for day in $days; do 
	m=`echo $month | awk '{printf("%02d", $1)}'`; 
	d=`echo $day | awk '{printf("%02d", $1)}'`; 
	echo $m $d;
	if compgen -G "*${yr}${m}${d}*$l2_suffix" > /dev/null; then
		
		# l2bin
		ls *${yr}${m}${d}*$l2_suffix > tmp_l2bin.txt; 
		l2bin ifile=tmp_l2bin.txt ofile=AQUA_MODIS.${yr}${m}${d}.L3b_NSST.nc l3bprod=sst resolution=$L2BIN_RES night=1 qual_prod=qual_sst qual_max=0 suite=NSST; 

		# l3bin
		l3bin ifile=AQUA_MODIS.${yr}${m}${d}.L3b_NSST.nc ofile=AQUA_MODIS.${yr}${m}${d}.L3b_NSST.x.nc prod="sst" latnorth=$north latsouth=$south loneast=$east lonwest=$west;
		
		# L3map
		l3mapgen ifile=AQUA_MODIS.${yr}${m}${d}.L3b_NSST.x.nc ofile=AQUA_MODIS.${yr}${m}${d}.L3m_NSST.x.nc resolution=$L3MAP_RES product="sst" north=$north south=$south east=$east west=$west interp=area projection="+proj=aea +lat_0=45 +lon_0=-126 +lat_1=50 +lat_2=58.5 +x_0=1000000 +y_0=0 +ellps=GRS80 +towgs84=0,0,0,0,0,0,0 +units=m +no_defs +type=crs";
		
		# Make .png
		INFILE=AQUA_MODIS.${yr}${m}${d}.L3b_NSST.x.nc
		l3mapgen ifile=$INFILE ofile=${INFILE:0:-3}.png oformat=png resolution="1km" product="sst" apply_pal=yes scale_type=linear mask_land=yes north=$north south=$south east=$east west=$west interp=area datamin=-1.89 threshold=1.0 full_latlon=no datamax=30 projection="+proj=aea +lat_0=45 +lon_0=-126 +lat_1=50 +lat_2=58.5 +x_0=1000000 +y_0=0 +ellps=GRS80 +towgs84=0,0,0,0,0,0,0 +units=m +no_defs +type=crs"

	fi 
  done 
done

[OB SeaDAS - knowles]

The nonuniform areas are being determined as clouds similar to what happens along the coastline. The sst daytime scenes have the advantage of being able to mask out clouds based on light intensities in the near infrared. However the nighttime sst scenes have advantages where the skin surface temperature better matches the temperature just beneath the surface of the water.

If you want to help fill in these gradients lines of missing data, you could interpolate when you bin. To do this, add the parameter area_weighting=1 to your l2bin run.

If desired, to further fill in missing values, you could consider dropping the bin resolution to 2km. MODIS is only 1km resolution at nadir, so 2km is a reasonable value if pixels are at higher scan angles. In this case, the mapping resolution could be maintained at 1km to help keep the land mask image looking sharp.

Danny