ERA5-Land Precipitation: Negative Values After Deaccumulation

Hi everyone,

I’m working with the ERA5-Land Hourly Precipitation dataset, and I’ve encountered some issues while trying to derive hourly rainfall rates from the accumulated fields.

After performing deaccumulation (i.e., subtracting consecutive time steps to obtain hourly values), I noticed:

  • Negative precipitation values in certain hours.
  • Very small values (close to zero) across large areas in some time steps.

From my understanding, ERA5 precipitation is accumulated from the start of the forecast, and this deaccumulation approach can occasionally lead to such artifacts. I wanted to confirm:

  1. Is this a known issue when working with accumulated ERA5-Land precipitation?
  2. What are the best practices for handling these negative or very small values?
  3. Should I apply a minimum threshold (e.g., 0 or 0.1 mm/h) when using this data for analysis or modeling?

Any guidance or documentation references would be greatly appreciated.

Thanks in advance!

Sandeep

You’re seeing these artifacts because of GRIB packing errors. A good explanation can be found here.