1. Precipitation characteristics in a matrix of atmospheric GCMs with increasing resolution
M.-E. Demory, P.L. Vidale, J. Donners,
M. Roberts, A. Clayton
With thanks to Christoph Frei (MeteoSwiss)

2. What is the role of resolution in climate research ?
Matrix of coupled models, with increasing resolution:
What is the impact of resolving eddies in the ocean ?
What is the impact of resolving weather ?
What are the crucial scales for proper coupling ?
What are the emerging processes ?
Vertical resolution is fixed: 38 atmospheric levels, 40 oceanic levels
Hundreds of years of simulations have been completed, including 25 years of AMIP2
∆x = 135 km
∆x = 90 km
∆x = 60 km
Flux coupler
Completed in 2007
1o - 1/3o ocean model
1/3o ocean model

3. General study & motivations: the hydrological cycle
Role of resolution in representing components of the hydrological cycle over land: precipitation, evaporation, runoff
Do high resolution models improve our understanding of the hydrological cycle?
Interactions with the global circulation and the weather systems (storms)?

4. Global mean precipitation
HadGAM (N96)
HiGAM (N144)
NUGAM (N216)
Trenberth (2007)
Global precipitation
1154 (+201)
1149 (+196)
1148 (+195)
953
Land precipitation
821 (+65)
(+55)
869 (+113)
756
Ocean precipitation
1320 (+286)
1291 (+257)
1266 (+232)
1034
Weighted average in kg/year/m2 (excess of precipitation)
Overestimation of precipitation (too much LH and net radiation)
Very little improvement with higher resolution models

5. Impact on precipitation over the Alps
HadGAM: N96 (~130 kms)
Do high resolution models represent precipitation distribution in a better way?
Is the precipitation frequency-intensity changing with resolution?
Is that due to spatial resolution (better resolved orography) or to better resolved weather systems?
HiGAM: N144 (~90 kms)
NUGAM: N216 (~60 kms)

6. DJF-MAM mean precipitation

7. JJA-SON mean precipitation

8. High-impact precipitation: QQ plot
HiGAM
Percentiles
Alpine dataset (C. Frei)
NUGAM
HadGAM

9. 95th percentile precipitation distribution (mm/day) in DJF
Alpine dataset
HadGAM
HiGAM
NUGAM

10. Summary (1/2)
Precipitation better represented in high resolution models
But mean precipitation biases remain the same in all resolution models, especially in summer
=> problem with the global circulation?
Same biases in regional models (MERCURE and PRUDENCE)
NUGAM shows better skills in representing high intensity events
Better location, especially along coastlines
=> impact on local river flows?

11. European river flows: Po
MAM
MAM
Outflow (1000*kg/s) in Po river
MAM

12. European river flows: Rhine & Danube
MAM
DJF
DJF
Rhine

13. Summary (2/2) Improvement of river flow seasonal cycle around the Alps
Due to more localized precipitation
Better snowmelt timing
Too dry in summer due to too much evaporation and too little rainfall
Snowmelt (mm/day)
HadGAM (MAM)
NUGAM (MAM)

14. Future work
High-impact precipitation better represented as the resolution increases
Is that due to orography? => Run a NUGAM test run with low resolution (N96) orography
Is that due to better resolved weather systems? => Storms analyses in Europe