/precip_bias_correction

This repo contains the script for bias correcting precipitation data.

Primary LanguagePython

General steps:

  1. We use ERA5Land, downloaded from the Ulysses project. The following procedure will be done at 30 min resolution
  • Containing bias, usually too wet.
  • Containing drizzle.
  • But covering until the year 2023 (until now).

  1. To remove the bias: We use W5E5, but this covers 1979-2019 only. For every month, we will calculate the monthly climatology of precipitation: Pclim_W5E5 and Pclim_era5land Then the corrected era5land: P_corrected_era5land = (Pclim_W5E5 / Pclim_era5land) x P_era5land

  2. For the drizzle correction: We use W5E5, but this covers 1979-2019 only. For every pixel, we will identify the minimum precip above zero. Let’s assume this as Pmin_W5E5

Pmin_W5E5 on velocity: /scratch/sutan101/forcing_for_beda/w5e5/timmin_precipitation_without_zero_w5e5_1979-2019.nc

Then, to remove the drizzle we will assume if P_corrected_era5land < Pmin_W5E5, P_corrected_era5land = 0.0

  1. Remove the bias again. The above step will introduce ‘bias‘. Therefore we have to do additional bias correction. Then, we implement the extra bias correction. extra_corrected era5land = (Pclim_W5E5 / Pclim_corrected_era5land) x P_corrected_era5land

  2. As era5land covers only LAND, we will cover all missing values (e.g. at coastal regions) with the nearest values.