How to apply e.g., '-1' in coda.fetch on S5P L2 products

CODA
1

I was trying out coda.fetch on a S5P L2 file, following the documentation at CODA Python. Unfortunately I don’t fully grasp the documentation, especially the optional use of the integers and a bit the terminology.

Would “datasetname” correspond to a netCDF variable name ?

I got the following commands to work:

s5p_rec = coda.fetch(path2file) # fetch entire file

s5p_rec_product = coda.fetch(path2file, "PRODUCT") # fetch everything below group PRODUCT

s5p_rec_cf = coda.fetch(path2file, "PRODUCT/cloud_fraction") # fetch 1 variable. Output is a numpy array.

s5p_rec_cf0 = coda.fetch(path2file, "PRODUCT/cloud_fraction[0]") # get 1 value of that variable

However, I don’t got the examples with the following syntax to work

coda.fetch(pf, "datasetname", 0)

coda.fetch(pf, "datasetname", -1, "dsr_time")

E.g., the following

s5p_rec_cf0 = coda.fetch(path2file, "PRODUCT/cloud_fraction", 0)

coda.fetch(pf, "PRODUCT/cloud_fraction", -1, 'scanline', 0) # Inspired by example here: Problem with HARP/Python within a loop - #3 by corisanucci

both return an error ValueError: number of specified indices does not match the dimensionality of the array

Part of the problem might be that I don’t know how the terms ‘dataset’, ‘field’, ‘dimension’ translate to the S5P L2 file.

If you could give working examples that would help my understanding of coda!

2

To know how to order the fields and array indices it is often best to inspect the file first with a codadump list command. This will show you that the cloud_fraction is stored using e.g.
/PRODUCT/cloud_fraction[1,4172,450].
Whenever there is a multi-dimensional array you will also have to provide a multidimensional index. So you can say coda.fetch(pf, "PRODUCT/cloud_fraction", [0,0,-1]) to fetch to first scanline.

3

Thanks! This is much more clear now.

codadump list is a commandline command. Does there exist a python interface for it that would output to e.g., a python tuple or list?

4

There is no function in the coda python interface that provides such a list, but you would be able to create such a function yourself by creating a coda.Cursor, inspect the type using the coda API functions, and recursively navigate through the product. This would be a nice coding exercise if you want to get more familiar with the CODA api.

5

Trying to.

With coda.get_fieldnames(cursor) I can get all nodes below the cursor, but how do I differentiate between variables and groups ?

6

By checking whether the field itself is a record or an array. Groups are records and datasets are arrays.