Monthly averages of SO2

Hello,

I’m working with Sentinel-5P data for the first time. Averaging Level 3 products at 0.05 degree resolution processed with HARP (qa > 0.5) for July 2021 I get below map.

It’s very noisy and I’m unsure about the high SO2 values which seem to be linked to sea ice?
Is this generally what to expect or am I missing out on some more filtering / postprocessing which is typically done.

The image is also based on > 300 individual Level 2 files so I’m surprised about the orbit artefacts still in there.

I’m also temporally averaging the images in Python after using HARP. If I would do this using HARP as well, would the averaging take into account the uncertainty of the individual orbits or have any other advantages over my current approach?