Debiasing | Notion

A debiasing option is available through the API for forecasts and historical reanalysis data to localize our forecast to the sub-grid level. The debiasing feature uses quality-controlled weather station data to correct the gridded forecasts if a station is available within the grid.

<aside> 💡 Note: Bias correction when using forecast time series only applies to the values and values-by-quantile fields of temperature, wind speed and precipitation variables, and when frequency is weekly or monthly. Bias correction when downscaling applies to the temperature (average, min, max), wind speed, and precipitation variables.

</aside>

Sample Request

Below is a sample API requests to make it easy to copy, paste, and modify.

Downscaling to a single point with debiasing:

https://api.salientpredictions.com/v2/downscale?lat=37.42&lon=-1.59&date=2022-01-01&variables=temp&members=11&debias=True

Sample Output

Screenshot 2024-02-13 at 5.35.27 PM.png

For each variable , three variables will be returned in the output when debias=False. A fourth variable will be added when the variable is temp, wspd or precip and debias=True, which will be “Variable name_bias_correction”.

Variable	Description
variable name	Bias corrected variable prediction
variable name_anom	Deviation from 30-year climatology. This value is not impacted by debiasing
variable name_clim	30-year climatology. This value is not impacted by debiasing
variable name_bias_correction	Bias correction factor. This is an additional variable when bias correction is enabled (debias = True). The bias correction factor is calculated on a monthly basis based on the difference between ERA5 historical averages and the nearest observation station.

Therefore, using temp, daily mean temperature:

temp = temp_clim + temp_anom + temp_bias_correction

If you want to reconstruct the original temperature before bias correction:

temp = temp_unbiased - temp_bias_correction

<aside> 💡 Note: The temperature bias correction factor is additive. Wind, sun, and precipitation bias correction factors are multipliers.

</aside>

Observation Datasets

The debiasing system currently uses two global observation datasets. The system will select the closest observation station for debiasing.

Global Historical Climatology Network Daily (GHCNd) - Daily climate summaries from >100,000 stations in 180 countries and territories. NCEI provides numerous daily variables, including maximum and minimum temperature, total daily precipitation, snowfall, and snow depth. About half the stations only report precipitation. Both record length and period of record vary by station and cover intervals ranging from less than a year to more than 175 years.
Integrated Surface Database (ISD) - A global database comprising hourly surface observations at >20,000 stations compiled from >100 sources. ISD includes numerous parameters such as wind speed and direction, wind gust, temperature, dew point, cloud data, sea level pressure, altimeter setting, station pressure, present weather, visibility, precipitation amounts for various periods, snow depth, and various other elements as observed by each station.