We utilize the ICON version 2.6.3 with upper-atmosphere extension as distributed by the German weather service (DWD). The ICON model is a collaborative project of DWD and the Max Planck Institute for Meteorology, striving at providing a unified modeling system to seamlessly allow simulations from climatological time scales to large-eddy simulations as well as for global numerical weather prediction (Zangl et al., 2015). In addition to the upper-atmosphere physics package implemented in UA-ICON, the dynamical core is extended from the shallow to deep atmosphere dynamics (Borchert et al., 2019). In our setup, the UA-ICON model is set up with the horizontal resolution of R2B4, which corresponds to a grid mesh of approximately 160 km with 120 levels up to a height of approximately 147 km. The time step of the simulation is 360s. The data output interval is set to 6 h, which is essential in the calculations of the Eliassen-Palm (EP) flux divergence. The mid-monthly sea surface temperature (SST) and sea ice concentration (SIC) values produced by the Program for Climate Model Diagnosis and Intercomparison (PCMDI) for the Atmosphere Model Intercomparison Project (AMIP, Taylor et al. (2000)) served as lower boundary condition data. Interactive chemistry was not used in the current simulation. Instead, the concentrations of CO2, CH4, N2O, CFC-11, and CFC-12 were taken from the historical greenhouse gas volume mixing ratios for CMIP6 (Meinshausen et al., 2017). The atmospheric ozone concentrations were prescribed based on the input4MIPs project (https://esgf-node.llnl.gov/search/input4mips). Here we conduct 30-year long time-slice experiments with the UA-ICON model by employing repeated annual cycles of SST, SIC, and greenhouse gases of the year 1985. This year is appointed as both El-Nino southern oscillation and Pacific decadal oscillation were in their neutral phase and no major volcano eruption has occurred, hence conditions in this year can serve as a useful proxy for the multi-year mean conditions and an estimate of their internal variability. The control run is carried out where both the sub-grid scale orography (SSO) scheme and non-orographic gravity waves scheme are used. In UA-ICON, the entire SSO drag is treated after Lott and Miller (1997), and the non-orographic GW drag parameterization is based on Warner and McIntyre (1996) and Scinocca (2003). It is worthwhile to mention that in addition to the orographic GWD, the SSO scheme used in UA-ICON also contains the effect of low-level blocking and wake drag. As the simulations are the time-slice experiments, only the months , days and hours have their true/usual meanings and years in the name of output files do not have their true/usual meaning. In other words, all the outputs files with different years (1986, 1987, ..., 2015) are identical to year 1985 as identical boundary conditions as year 1985 are used to simulate them. For example, 0001 and 0002 represent the first and second years of simulations, respectively and 0030 is the last (30) year.

We utilize the ICON version 2.6.3 with upper-atmosphere extension as distributed by the German weather service (DWD). The ICON model is a collaborative project of DWD and the Max Planck Institute for Meteorology, striving at providing a unified modeling system to seamlessly allow simulations from climatological time scales to large-eddy simulations as well as for global numerical weather prediction (Zangl et al., 2015). In addition to the upper-atmosphere physics package implemented in UA-ICON, the dynamical core is extended from the shallow to deep atmosphere dynamics (Borchert et al., 2019). In our setup, the UA-ICON model is set up with the horizontal resolution of R2B4, which corresponds to a grid mesh of approximately 160 km with 120 levels up to a height of approximately 147 km. The time step of the simulation is 360s. The data output interval is set to 6 h, which is essential in the calculations of the Eliassen-Palm (EP) flux divergence. The mid-monthly sea surface temperature (SST) and sea ice concentration (SIC) values produced by the Program for Climate Model Diagnosis and Intercomparison (PCMDI) for the Atmosphere Model Intercomparison Project (AMIP, Taylor et al. (2000)) served as lower boundary condition data. Interactive chemistry was not used in the current simulation. Instead, the concentrations of CO2, CH4, N2O, CFC-11, and CFC-12 were taken from the historical greenhouse gas volume mixing ratios for CMIP6 (Meinshausen et al., 2017). The atmospheric ozone concentrations were prescribed based on the input4MIPs project (https://esgf-node.llnl.gov/search/input4mips). Here we conduct 30-year long time-slice experiments with the UA-ICON model by employing repeated annual cycles of SST, SIC, and greenhouse gases of the year 1985. This year is appointed as both El-Nino southern oscillation and Pacific decadal oscillation were in their neutral phase and no major volcano eruption has occurred, hence conditions in this year can serve as a useful proxy for the multi-year mean conditions and an estimate of their internal variability. The non-orographic gravity waves scheme is not used in this simulation. In UA-ICON, the entire SSO drag is treated after Lott and Miller (1997), and the non-orographic GW drag parameterization is based on Warner and McIntyre (1996) and Scinocca (2003). It is worthwhile to mention that in addition to the orographic GWD, the SSO scheme used in UA-ICON also contains the effect of low-level blocking and wake drag. As the simulations are the time-slice experiments, only the months, days and hours have their true/usual meanings and years in the name of output files do not have their true/usual meaning. In other words, all the outputs files with different years (1986, 1987, ..., 2015) are identical to year 1985 as identical boundary conditions as year 1985 are used to simulate them. For example, 0001 and 0002 represent the first and second years of simulations, respectively and 0030 is the last (30) year.

We utilize the ICON version 2.6.3 with upper-atmosphere extension as distributed by the German weather service (DWD). The ICON model is a collaborative project of DWD and the Max Planck Institute for Meteorology, striving at providing a unified modeling system to seamlessly allow simulations from climatological time scales to large-eddy simulations as well as for global numerical weather prediction (Zangl et al., 2015). In addition to the upper-atmosphere physics package implemented in UA-ICON, the dynamical core is extended from the shallow to deep atmosphere dynamics (Borchert et al., 2019). In our setup, the UA-ICON model is set up with the horizontal resolution of R2B4, which corresponds to a grid mesh of approximately 160 km with 120 levels up to a height of approximately 147 km. The time step of the simulation is 360s. The data output interval is set to 6 h, which is essential in the calculations of the Eliassen-Palm (EP) flux divergence. The mid-monthly sea surface temperature (SST) and sea ice concentration (SIC) values produced by the Program for Climate Model Diagnosis and Intercomparison (PCMDI) for the Atmosphere Model Intercomparison Project (AMIP, Taylor et al. (2000)) served as lower boundary condition data. Interactive chemistry was not used in the current simulation. Instead, the concentrations of CO2, CH4, N2O, CFC-11, and CFC-12 were taken from the historical greenhouse gas volume mixing ratios for CMIP6 (Meinshausen et al., 2017). The atmospheric ozone concentrations were prescribed based on the input4MIPs project (https://esgf-node.llnl.gov/search/input4mips). Here we conduct 30-year long time-slice experiments with the UA-ICON model by employing repeated annual cycles of SST, SIC, and greenhouse gases of the year 1985. This year is appointed as both El-Nino southern oscillation and Pacific decadal oscillation were in their neutral phase and no major volcano eruption has occurred, hence conditions in this year can serve as a useful proxy for the multi-year mean conditions and an estimate of their internal variability. The non-orographic gravity waves scheme is not used in this simulation. In UA-ICON, the entire SSO drag is treated after Lott and Miller (1997), and the non-orographic GW drag parameterization is based on Warner and McIntyre (1996) and Scinocca (2003). It is worthwhile to mention that in addition to the orographic GWD, the SSO scheme used in UA-ICON also contains the effect of low-level blocking and wake drag. As the simulations are the time-slice experiments, only the months , days and hours have their true/usual meanings and years in the name of output files do not have their true/usual meaning. In other words, all the outputs files with different years (1986, 1987, ..., 2015) are identical to year 1985 as identical boundary conditions as year 1985 are used to simulate them. For example, 0001 and 0002 represent the first and second years of simulations, respectively and 0030 is the last (30) year.

The climatological dataset was produced using the Weather and Research Forecasting (WRF) model, version 4.2.2, configured with two nested domains at 10 km (D1) and 2 km (D2) horizontal grid spacing. It covers most of the South Island of New Zealand and is centered over Brewster Glacier in the Southern Alps. The model was forced every three hours by ERA5 reanalysis data at its outer lateral boundaries. The dataset spans the period of 1 January 2005 to 31 December 2020, providing daily output in the outer domain (D1) and 3-hourly output in the innermost domain (D2). The data provided here are a selection of daily averages from the inner WRF domain (D2; 2-km grid spacing). They are distributed among three different file types containing 4-dimensional, 3-dimensional and time-invariant output variables, respectively. For the 4-dimensional fields, perturbation and base-state atmospheric pressure (WRF variables P and PB) and geopotential (PH and PHB) were combined to produce full model fields (PRES and GEOPT). Perturbation potential temperature (T) was converted to total potential temperature (THETA). Wind vectors (U,V, and W) were converted to mass points and rotated to earth coordinates. ------- Acknowledgements: The modeling and related research was supported by the German Research Foundation (DFG) grant no. 453305163. The authors gratefully acknowledge the scientific support and HPC resources provided by the Erlangen National High Performance Computing Center (NHR@FAU) of the Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU) under the NHR project b128dc / ATMOS ("Numerical atmospheric modeling for the attribution of climate change and for model improvement"). NHR funding is provided by federal and Bavarian state authorities. NHR@FAU hardware is partially funded by the German Research Foundation (DFG) – 440719683.

ModE-Sim (short for Modern Era Simulations) is a medium-size ensemble of model simulations using the ECHAM6 atmosphere general circulation model (model version 6.3.5p2, doi:10.17617/2.1810480). Its setup is based on the PMIP4 experiments, but uses a forced AGCM rather than a fully coupled model. ModE-Sim was originally designed to form the a-priori state for a climate reconstruction (Modern Era Reanalysis, ModE-RA, to be found as separate experiment within this WDC project) that uses an offline data assimilation technique to combine the output of ModE-Sim with historical climate information. However, beyond its original purpose ModE-Sim on its own can be used as a tool to study climate variability, providing a high number of posible climate states that are physically plausible under the given forcings and boundary conditions. This might include, e.g. the separation of internal variability from the response to externally forced signals, understanding of teleconnection patterns, or the study of extreme events. The ensemble uses observed/reconstructed forcings and boundary conditions, while accounting in uncertainties in these. For 1420 to 1850 we provide a 60 member ensemble grouped in three subsets. The subset 1420-3 provided in this dataset group has 20 members and uses PMIP4 radiative forcings. As ocean boundary condition 20 different realizations of SST reconstructions were used and for sea ice a climatology was computed from the years 1850-1900 from HadISST2 sea ice.

ModE-Sim (short for Modern Era Simulations) is a medium-size ensemble of model simulations using the ECHAM6 atmosphere general circulation model (model version 6.3.5p2, doi:10.17617/2.1810480). Its setup is based on the PMIP4 experiments, but uses a forced AGCM rather than a fully coupled model. ModE-Sim was originally designed to form the a-priori state for a climate reconstruction (Modern Era Reanalysis, ModE-RA, to be found as separate experiment within this WDC project) that uses an offline data assimilation technique to combine the output of ModE-Sim with historical climate information. However, beyond its original purpose ModE-Sim on its own can be used as a tool to study climate variability, providing a high number of posible climate states that are physically plausible under the given forcings and boundary conditions. This might include, e.g. the separation of internal variability from the response to externally forced signals, understanding of teleconnection patterns, or the study of extreme events. The ensemble uses observed/reconstructed forcings and boundary conditions, while accounting in uncertainties in these. For 1420 to 1850 we provide a 60 member ensemble grouped in three subsets. The subset 1420-1 provided in this dataset group has 20 members and uses PMIP4 radiative forcings. As ocean boundary condition 20 different realizations of SST reconstructions were used and for sea ice analogues were picked from the HadISST2 sea ice, based on the reconstructed SST fields.

This dataset contains the results of nine simulations performed for the validation of the snow cover and precipitation scheme used in the microscale, obstacle-resolving model MITRAS v3.0 (Salim et al., 2013; Schluenzen et al., 2018), v3.1 (Ferner et al. 2023), and v3.3 (Samsel et al. 2025, in review). The model domain extends 240 m x 210 m horizontally and includes orography, slanted roofs, obstacle corners and different surface cover classes. The simulations were performed using different model versions, initial temperatures, precipitation and processing modes. The simulations cover 62 minutes model time, starting at 7:30 am model time, with a temporal resolution of 5 seconds or 5 minutes. This dataset contains a selection of output variables, control variables are not included. The file names of the data sets are composed as follows. {temperature}_{intensity}_{model version}_{comment}_{case ID}.nc There are 3 intensities for temperature: low, medium, high Associated values are: T_low = 272 K, T_medium = 280 K, T_high = 288 K Three versions are considered: v3p0 = MITRAS v3.0 (initial version), v3p1 = MITRAS v3.1 (inclusion of warm rain scheme), v3p3 = MITRAS v3.3 (inclusion of winter parameterisation scheme) Special cases are 'noparallel', where the simulation was performed with the parallelisation mode switched off, and 'noprecip', where no precipitation is initialised. Example: T_high_v3p1_Hwr.nc

This experiment contains sensitivity test results (Ferner et al. 2023) of 11 simulations with the microscale, obstacle-resolving model MITRAS v 3.1 (Salim et al., 2018; Schluenzen et al., 2018) for a domain of 1.6 x 1.8 km² around Hamburg City Hall in Hamburg. The domain contains various street configurations, open spaces, water surfaces, orography and building heights. The simulations were performed with different initial wind speeds, rain amounts, wind directions, and domain configurations. The simulations cover 1:40 hours, starting at 7:30 LST (LST refers to Local Solar Time), with a temporal resolution of 10, 1 or 5 minutes. This experiment contains a selection of output variables, control variables are not included. The file names of the data sets are composed as follows. {precipitation}_{intensity}_{windspeed}_ {intensity}_{winddirection}_{value}_{case ID}.nc There are 3 intensities: low, medium, high Associated values are these: pr_low = 0.5 mm, pr_medium =0.9 mm, pr_high = 1.7 mm (after 10 minutes) ff_low = 2 m/s, ff_medium = 4 m/s, ff_high = 4 m/s Example: pr_medium_ff_low_dd_270_ML27.nc

ModE-Sim (short for Modern Era Simulations) is a medium-size ensemble of model simulations using the ECHAM6 atmosphere general circulation model (model version 6.3.5p2, doi:10.17617/2.1810480). Its setup is based on the PMIP4 experiments, but uses a forced AGCM rather than a fully coupled model. ModE-Sim was originally designed to form the a-priori state for a climate reconstruction (Modern Era Reanalysis, ModE-RA, to be found as separate experiment within this WDC project) that uses an offline data assimilation technique to combine the output of ModE-Sim with historical climate information. However, beyond its original purpose ModE-Sim on its own can be used as a tool to study climate variability, providing a high number of posible climate states that are physically plausible under the given forcings and boundary conditions. This might include, e.g. the separation of internal variability from the response to externally forced signals, understanding of teleconnection patterns, or the study of extreme events. The ensemble uses observed/reconstructed forcings and boundary conditions, while accounting in uncertainties in these. For 1420 to 1850 we provide a 60 member ensemble grouped in three subsets. The subset 1420-2 provided in this dataset group has 20 members and uses a 20-member ensemble of perturbed volcanic forcings from the easy volcanic aerosol (EVA) model to account for uncertainties in the strength and the timing of volcanic eruptions. As ocean boundary condition 20 different realizations of SST reconstructions were used and for sea ice analogues were picked from the HadISST2 sea ice, based on the reconstructed SST fields.

ModE-Sim (short for Modern Era Simulations) is a medium-size ensemble of model simulations using the ECHAM6 atmosphere general circulation model (model version 6.3.5p2, doi:10.17617/2.1810480). Its setup is based on the PMIP4 experiments, but uses a forced AGCM rather than a fully coupled model. ModE-Sim was originally designed to form the a-priori state for a climate reconstruction (Modern Era Reanalysis, ModE-RA, to be found as separate experiment within this WDC project) that uses an offline data assimilation technique to combine the output of ModE-Sim with historical climate information. However, beyond its original purpose ModE-Sim on its own can be used as a tool to study climate variability, providing a high number of posible climate states that are physically plausible under the given forcings and boundary conditions. This might include, e.g. the separation of internal variability from the response to externally forced signals, understanding of teleconnection patterns, or the study of extreme events. The ensemble uses observed/reconstructed forcings and boundary conditions, while accounting in uncertainties in these. For 1850 to 2009 ModE-Sim offers 36 members grouped in two subsets, all using PMIP4 radiative forcings. The subset 1850-2 provided in this dataset group has 16 members and uses linear combinations of HadISST2 realizations as SST and HadISST sea ice as ocean boundary conditions.