The PR ensures that all lat/lon values are rounded to the same level of precision in output datasets.
This is in response to an issue between the mars and aws SEAS5 outputs. I'm still not sure exactly where this was introduced, but the output COGs from the mars data had lat/lon values that were very slightly (ie. 10e-13 order of magnitude) off from the AWS outputs (which had coords precise to 1 decimal point). This caused issues when trying to concatenate the datasets. This discrepancy is in spite of totally matching CRS and input extents.
See more exploration in the .md file included in this PR.
I'm not convinced it's a perfect solution, but rounding to 4 decimal places seems to fix the issue, and I don't think gets rid of any precision that we actually care about.
The PR ensures that all lat/lon values are rounded to the same level of precision in output datasets.
This is in response to an issue between the mars and aws SEAS5 outputs. I'm still not sure exactly where this was introduced, but the output COGs from the mars data had lat/lon values that were very slightly (ie. 10e-13 order of magnitude) off from the AWS outputs (which had coords precise to 1 decimal point). This caused issues when trying to concatenate the datasets. This discrepancy is in spite of totally matching CRS and input extents.
See more exploration in the
.mdfile included in this PR.I'm not convinced it's a perfect solution, but rounding to 4 decimal places seems to fix the issue, and I don't think gets rid of any precision that we actually care about.
Closes #14