Bias Corrections

Bias corrections can be applied to ensure the temperature and precipitation associated with the future climate data is roughly consistent with the reference climate data.

Delta change

The temperature bias corrections follow Huss and Hock (2015) and use an additive factor to ensure the mean monthly temperature and interannual monthly variability are consistent between the reference and future climate data.

There are two options for the precipitation bias correction:

Option 2 follows Huss and Hock (2015) and uses a multiplicative factor to ensure the monthly mean precipitation are consistent between the reference and future climate data.

Option 1 modifies the approach of Huss and Hock (2015), since in dry regions, the adjusted precipitation was found to be unrealistically high when using multiplicative values. This option uses the interannual monthly variability and quality controls the bias corrected precipitation by replacing any values that exceed the monthly maximum with the monthly average adjusted for potentially wetter years in the future using the normalized annual variation. Note that Marzeion et al. (2020) suggests that Huss and Hock (2015) have updated their approach to also adjust the precipitation bias correction to account for monthly interannual variability as well.

Quantile delta mapping

The temperature and precipitation bias corrections follow the approach below:

Option 3 follows Lader et al. (2017) and uses a quantile delta mapping approach to capture both the future mean climate and future climatic extremes. This option determines the relative change between the future and historical simulated climate data at each percentile in the distribution and overlays the difference onto the reference dataset. For example, the 70th-percentile precipitation from a simulated future distribution is divided by the 70th-percentile precipitation from the simulated historic distribution to calculate the modeled ratio of change. This ratio is then multiplied by the 70th-percentile precipitation from the reference dataset to calculate the bias-corrected future value. This bias correction is applied in 20-year increments (e.g., 2020-2040, 2040-2060) to better represent the changing climate over time.

Note

These bias corrections significantly impact projections for all glacier evolution models. Current work is thus investigating different bias corrections and how they affect projections. We hope to add additional bias correction options in the near future.