- Provide high resolution regional climate data - Spatial and temporal downscaling - Bias adjustment of climate data - Weather generator - Uncertainty Analysis
## Provide high resolution regional climate data
### Regional climate data - Europe - Europe wide gridded and station observations - CORDEX-EUR11 ensemble of regional climate projections for Europe - Horizontal resolution: ~12.5km - Temporal resolution: daily (gridded observations, projections) to 1 hour (station observations, projections) - Time frame: 1971-2100 - Variables: precipitation, temperature, sea level pressure, global radiation and wind speed
Precipitation RCP85 (2070-2099 vs. 1971-2000, JJA)
# Spatial and temporal downscaling - **Coarse spatial resolution of climate data** **global: 100 to 250km** **regional: 10 to 50km** - **Coarse temporal resolution** **global: 3hr to 1d** **regional: 1hr to 1d**
#### Dynamical regional downscaling - Physically consistent simulation of 3D atmosphere - PIK capability: Horizontal resolution: $\\approx$5km Temporal resolution: $\approx$ 10min values - Computational demanding $o($1d$)$ to simulate 1 year

COSMO CLM model
ICON-CLM model
### Statistical regional downscaling - Statistically consistent simulation of 2D atmospheric fields - Spatial and temporal resolution depend on available observations - various Methods available
### Temporal downscaling - Dynamical downscaling using RCMs - Statistical downscaling using ML
# Bias Adjustment of climate data
## Example of Model Bias
JJA Temperatures, Bangladesh

# Weather Generator
Temporal consistency
Spatial correlation
Variable dependency
- Generates 'statistical' consistent evolution of weather paterns - Use cases: 1. Empirical statistical downscaling 2. Parametric Bootstrapping
### Parametric Bootstrapping - Artificially increase sample size based on statistical relationship in data - Typically: 50 years → 1000 years - Robust estimate and extrapolation of rare events
# Uncertainty Analysis
#### Temperature rcp85 ##### Domain: Alps 2070-2099 vs. 1971-2000 CORDEX-EUR11 projection
#### Generalized extreme value distribution --- Annual maximum precipitation 2071-2100 vs. 1981-2010 ---