calculation, the pre-processing includes weighting by daylight hours.

Note on Easy Aerosol daylight hours error in UKCP runs

Elizabeth Kendon and Simon Tucker

13^th July 2020 Description of issue

An error has been found in the code used to generate Easy Aerosol ancillaries used in the UK Climate Projections (UKCP) Regional (12km) and Local (2.2km) projections. The error

resulted in daylight hours weighting, which is used to calculate shortwave aerosol radiative effects, being set at the value for September for all months, instead of varying month-by- month. This led to too much shortwave aerosol impact in summer and not enough in winter over Europe (Fig 1) in the 12km regional climate model (RCM) and Local 2.2km model.

The error,

was in the pre-processing code that converts diagnostics output from the 60km global climate model (GCM) to Easy Aerosol input files. In Easy Aerosol (Stevens et al 2017), aerosol optical properties (absorption, extinction and asymmetry) and cloud droplet number concentration (CDNC) are prescribed from the driving GCM (using time-varying monthly mean, full 3D spatial fields). For the short-wave

calculation, the pre-processing includes weighting by daylight hours.

Fig 1. Impact of fixing daylight hours error on short-wave aerosol properties. Shown is the ratio of the fixed value to the original value, for the six shortwave waveband groups (pseudo levels 1-6) used in the SOCRATES radiation code, for model level ten extinction, averaged over the EURO-CORDEX domain.

The daylight hours error was discovered on looking at clear-sky short-wave radiation differences between a low-resolution (60km) run of the RCM and the GCM, where differences are dominated by Easy Aerosol. In particular there were large differences in clear-sky surface downward shortwave radiation that were hard to explain, and pointed to a difference in the short-wave aerosol impact between the RCM and GCM. Since the use of Easy Aerosol should result in the aerosol forcing in the

GCM being approximately replicated in the RCM, this was not expected and indicated an underlying error.

In order to assess the impact of this error, simulations of the low-resolution 60km (LOWRES-RCM) and the 12km RCM have been performed for one ensemble member (specifically the standard member with no parameter perturbation applied) for 10 years with this error fixed. These test simulations are compared with the original LOWRES and UKCP RCM simulations in order to assess the impact of the error both for the present-day (1981-1990) and future changes (2061-2070 minus 1981-1990, under RCP8.5), focussing on temperature and precipitation.

Overall, we find that the impact of the error is small and generally less than the spread across the twelve-member UKCP Regional 12km ensemble. The impact is greatest over central/eastern Europe, with much smaller impacts over the UK where the impact is judged to be not significant.

Results from 12km RCM simulation

We now assess the impact of fixing the daylight hours error in the 12km RCM used for the UKCP Regional projections, both for the present-day simulation and future changes.

For present-day:

• Fixing the error leads to warmer temperatures by up to 0.5K over the East of the domain, but has little impact on precipitation, in both summer and winter (Fig 2 & 3). Differences on fixing the error are not significant (at the 95% level) compared to year-to-year natural variability across almost all of Europe (except for temperature over Eastern Europe).

Differences over the UK are small and discussed in more detail below.

• The impact of the error is smaller than the spread across the UKCP RCM ensemble for temperature across the entire domain (Fig 4). For precipitation there are some large local impacts, which are locally larger than the ensemble spread, however these differences are not significant compared to year-to-year variability and are largely just noise (reflecting the fact that we only have 10 years of data).

Fig 2. Impact of fixing the daylight hours error on present-day surface temperature and precipitation in the 12km RCM, in summer. Results are shown for (top) surface temperature and (bottom)

precipitation, for (left) differences between new run with error fixed and original RCM, (centre) differences between the new RCM and the GCM, and (right) differences between the original RCM and the GCM. 10 years of model data are used for 1981-1990. Stippling shows where local

differences are significantly different from zero at the 95% level, using a one-sample t-test.

Fig 3. Impact of fixing the daylight hours error on present-day surface temperature and precipitation in the 12km RCM, in winter. Results are shown for (top) surface temperature and (bottom)

precipitation, for (left) differences between new run with error fixed and original RCM, (centre) differences between the new RCM and the GCM, and (right) differences between the original RCM and the GCM. 10 years of model data are used for 1981-1990. Stippling shows where local

differences are significantly different from zero at the 95% level, using a one-sample t-test.

Fig 4. Ratio of impact of fixing daylight hours error to UKCP RCM ensemble standard deviation, for present day. Results are shown for (top) winter and (bottom) summer, for (left) surface temperature and (right) precipitation.

For future changes:

• Fixing the error leads to smaller future increases in temperature, particularly over central/eastern Europe in winter (Fig 5). Impacts of fixing the error on future changes in precipitation are noisy and not significantly different from zero (at the 95% level) when compared to year-to-year natural variability. Differences over the UK are presented in more detail below.

• The impact of the error is smaller than the spread across the UKCP RCM ensemble for future changes in temperature across the entire domain, with impacts about 50-75% of the

ensemble standard deviation over central/eastern Europe (Fig 6). For precipitation changes there are some large local impacts, which are locally larger than the ensemble spread, however these differences as noted above are not significant compared to year-to-year variability and are largely just noise (reflecting the fact that we only have 10 years of data).

Ratio, temperature Ratio, precipitation

Fig 5. Impact of fixing the daylight hours error on future changes in surface temperature and precipitation in the 12km RCM. Shown are differences between future changes in the new run with the error fixed and those in the original UKCP RCM, for (top) winter and (bottom) summer, for (left) surface temperature and (right) precipitation. Future changes correspond to the difference between the future (2061-2070) and present-day (1981-1990) periods. Stippling shows where local differences are significantly different from zero at the 95% level, using a one-sample t-test.

Temperature Precipitation

Fig 6. Ratio of impact of fixing daylight hours error to UKCP RCM ensemble standard deviation, for future changes. Results are shown for (top) winter and (bottom) summer, for (left) surface

temperature and (right) precipitation.

Implications for UKCP Regional and Local projections

Easy Aerosol is used to approximate aerosol forcing in the UKCP Local 2.2km and Regional 12km models. It aims to replicate the aerosol forcing in the driving GCM, however it is only an

approximation of the real aerosol forcing and ignores any cloud-aerosol interactions (Voigt et al 2014). Therefore, there is significant uncertainty in the aerosol forcing in the UKCP Regional and Local projections through the use of Easy Aerosol, independent of any daylight hours error.

The above results indicate that the impact of the daylight hours error in the Easy Aerosol pre- processing code is small and judged not significant compared to the spread across the UKCP RCM ensemble members. On this basis, the daylight hours error does not suggest the need to rerun the UKCP RCM simulations or to retract any of the RCM results, although users should be aware that future increases in temperature over central/Eastern Europe may be overestimated (by up to 1K in winter and 0.5K in summer) in the UKCP Regional projections. Over the UK impacts are much smaller

Ratio, temperature change Ratio, precipitation change

(with maximum local impacts of 0.2-0.3K) and no special consideration is needed by users looking at the RCM results over the UK.

References

Stevens, B., S. Fiedler, S. Kinne, K. Peters, S. Rast, J. Müsse, S. J. Smith, and T. Mauritsen (2017) MACv2-SP: a parameterization of anthropogenic aerosol optical properties and an associated Twomey effect for use in CMIP6, Geosci. Model Dev., 10, 433-452, doi: 10.5194/gmd-10-433-2017 Voigt, A., B. Stevens, S. Bony, O. Boucher (2014) Easy Aerosol - a modelling framework to study robustness and sources of uncertainties in aerosol-induced changes of the large-scale atmospheric circulation. WCRP report (https://www.wcrp-

climate.org/images/grand_challenges/clouds/documents/easyaerosol_projectdescription_expprotoc ol.pdf )