Welcome to the Earthdata User Forum! Here, subject matter experts from several NASA Distributed Active Archive Centers (DAAC) can discuss general questions, research needs and data applications. Users can query how to access, view and interpret the data.

by st10873 » Mon Apr 01, 2019 10:14 am America/New_York

I have downloaded two SeaWiFS level-3 .nc-file sets for chlorophyll a for a specified lat/lon boundary within the period 1998-2010; one is for 8-day periods (8D), and the other for monthly periods (MO). I read and process the files in Matlab, and would like to calculate the monthly averages (all years combined) for some given bins.

I thought the MO data would work best, but the chlor_a variable is denoted as sum and sum_square. However, I can not find the appropriate way to estimate the mean chlor_a based on the sum. What do I divide by - days in the month? I can't find any weight or N variable in the file. Likewise for the 8D files. Are these chlor_a variables also sums, or are they averages? I have always assumed the 8D was already averaged over 8 days - although I've never used them before.

To get the mean values you need to divide each "sum" value by the corresponding "weights" value. You will find those two data sets in two different places in the level-3 file. h5dump -A S19980011998031.L3b_MO_CHL.nc |grep -A 8 BinList|head -8 DATASET "BinList" { DATATYPE H5T_COMPOUND { H5T_STD_U32LE "bin_num"; H5T_STD_I16LE "nobs"; H5T_STD_I16LE "nscenes"; H5T_IEEE_F32LE "weights"; H5T_IEEE_F32LE "time_rec"; }

h5dump -A S19980011998031.L3b_MO_CHL.nc |grep -A 4 'DATASET "chlor_a"' DATASET "chlor_a" { DATATYPE H5T_COMPOUND { H5T_IEEE_F32LE "sum"; H5T_IEEE_F32LE "sum_squared"; }

Regards, Norman

P.S. Do not attempt to use the "time_rec" data for anything. That field is present for historical reasons, but it has never contained the information that it was originally intended to contain.

by st10873 » Mon Apr 01, 2019 11:21 am America/New_York

Thank you for the fast reply. Does "weights" represent a weighted observation count? I.e. can I divide the sum_square by the same "weights" to get the variance?

To get the variance, you need to divide the "sum_squared" value by the "weights" value and then subtract the square of the mean from that.

Norman

P.S. Note that rounding errors can sometimes cause the variance to go slightly negative when you subtract the mean squared. In those cases, I just set the variance to zero.

by gnwiii » Mon Apr 01, 2019 7:01 pm America/New_York

You may find Guide to the Creation and Use of Ocean-Colour, Level-3, Binned Data Products. IOCCG Report Number 4, 2004 helpful. The OBPG binned data use an extended version of the Naïve Algorithm for single pass variances. As Norman remarked, this method has some numerical limitations. In those cases it may be useful to define new level-2 products (rescaled, centered values). With the NetCDF4-CF format it is relatively simple to do this with any "matrix" language that supports NetCDF4-CF. You will need to add an entry for your new product to the product.xml file. Working with level-2 data does, however, require more compute time and storage than using OBPG binned data prodcuts.

Volume 32 of the SeaWiFS Prelaunch Technical Memorandum series, where the binning scheme we use was first described, does indicate that data should be log-transformed before they are binned. Even before the SeaWiFS mission got underway, however, the log transformations and maximum likelihood estimator were found to be undesirable and were abandoned. So, our sum and sum_squared values are straight up sums with no log transformation.

by reghe1984 » Mon Apr 08, 2019 4:59 am America/New_York

Hi all,

I am a PhD student at Stockholm University working on eastern equatorial Pacific Ocean paleoceanography. I am about to finish to write my PhD thesis and I would like to have an image of the eastern equatorial Pacific Ocean in the cover page. The image should be as the ones in the "Image gallery" folder in this web page showing phytoplankton blooms but of the area comprised between ±20° latitude from the equator and between 70°W and 160°W of longitude.

My thesis would greatly benefit from such an image. I would adequately acknowledge Ocean Color and NASA for this. Thank you Very much.