Presentation is loading. Please wait.

Presentation is loading. Please wait.

Simulating daily precipitation variability.

Similar presentations


Presentation on theme: "Simulating daily precipitation variability."— Presentation transcript:

1 Simulating daily precipitation variability.
Is there any benefit from increased resolution ? Colin Jones1, Samuel Girard1 and Katja Winger1,2 1CRCMD/UQAM 2Environment Canada

2 There is a common held view that the number of wet days
will decrease in future climate conditions, while the number or intensity of extreme precipitation events will increase. These contentions arise mainly from precipitation changes simulated by Coupled GCMs and RCMs for the future climate. Although observations suggest this trend can be somewhat seen for the recent past decades. An obvious question is whether a given Climate Model can simulate present-day observed (daily) precipitation variability (Number of wet-days, No. & intensity of extreme wet-days Does increased resolution improve a models ability to simulate high time frequency precipitation variability ?

3 Run a Limited-Area version of GEM over North America forced by ‘’perfect’’ analysed boundary conditions and ask how well it simulates observed daily mean precipitation variability. Compare 0.45 simulations for the period to NCEP gauge based 0.25ºdaily precipitation accumulations for USA

4 Higher resolution may improve simulated precipitation variability
Improved topographic forcing ? Better resolution of convection-dynamical interaction/organisation ? 2.Analyse 3 GEM-LAM runs at 0.45º, 0.33º & 0.15º for the period 1993 and 1994 (all runs identical with Kain-Fritsch convection scheme)

5

6 GEM, Observed & Corrected-Observed
Mean Annual Cycle ( ) GEM, Observed & Corrected-Observed NW CNTR SE Corrections for wind effects & evaporation losses due to Ungersbock et al NW winter season accurate NW Summer season dry bias

7 GEM, Observed & Corrected-Observed
Mean Annual Cycle ( ) GEM, Observed & Corrected-Observed NW CNTR SE CNTR winter ppt accurate CNTR LARGE summer season dry bias: 50% underestimate of mean JAS precipitation

8 GEM, Observed & Corrected-Observed
Mean Annual Cycle ( ) GEM, Observed & Corrected-Observed NW CNTR SE SE winter season dry bias SE summer season accurate

9 How is the monthly/seasonal mean precipitation distributed in terms of the underlying daily mean intensities ? Can GEM-LAM simulate this distribution ?

10 X10 Fraction of total number of days in a given season for 1964-1998
mean that experience precipitation in a given daily accumulation bin e.g. SE USA Fraction of events per intensity band DJF mean Observations and GEM-LAM 0.45º X10 mm/day

11 Fraction of total precipitation over a given region coming from each
daily intensity bin, average for SE USA DJF fraction of total precipitation from a given intensity band Observations and GEM-LAM 0.45º mm/day

12 Absolute amount of precipitation (in mm/day) averaged over a given region coming from each daily intensity bin, mean for e.g. SE USA absolute amount of precipitation from a given intensity band Observations and GEM-LAM 0.45º mm/day

13 Cumulative intensity curves of total precipitation as a function of
intensity class for a given region averaged over e.g. SE USA DJF : Observations and GEM-LAM 0.45º

14 Cumulative intensity curves of total precipitation as a function of
intensity class for a given region averaged over e.g. SE USA DJF : Observations and GEM-LAM 0.45º

15 NW JJA (0.45º 1964-1998 mean) Fraction of events per intensity band:
X10 NW JJA (0.45º mean) Fraction of events per intensity band: Too many dry days. Too few wet days.

16 X10 NW JJA (0.45º mean) Fraction of total precipitation from each intensity band: Distribution quite accurate

17 X10 NW JJA ( mean) Absolute amount of precipitation from each intensity band: Significant underestimate for all bands due to underestimate of occurrence.

18 Central USA (CNTR) JJA 0.45º 1964-1998 mean
Same occurrence underestimate across all classes Underestimate of precipitation in all intensity bands in CNTR X10 mm/day mm/day

19 Central USA (CNTR) JJA 0.45º 1964-1998 mean Do things improve at
Same occurrence underestimate across all classes Underestimate of precipitation in all intensity bands in CNTR X10 mm/day Problems in triggering convection and maintenance of intense and organised convection are prime candidates causing these errors Do things improve at higher resolution? mm/day

20 NW JJA Some modest improvement with increasing resolution in
the NW (mountainous) region. Improved topographic forcing ? 11% 25% The number of intense days (>20mm/day) increases too much with increasing resolution

21 Central USA (CNTR) JJA Some very modest improvement
at 0.33º resolution, largely Lost again at 0.15º resolution. Convection in CNTR region is not topographically forced. Even at higher resolution all precipitation classes are underestimated in CNTR JJA except very intense classes which are overestimated.

22 DJF cumulative precipitation distributions by resolution (93-94)
NW CNTR SE

23 JJA cumulative precipitation distributions by resolution (93-94)
NW CNTR SE

24 JJA cumulative precipitation distributions by resolution (93-94)
NW CNTR SE

25 Summary Winter season precipitation variability is fairly well simulated by GEM at 0.45º. This remains true at higher resolutions. Summer precipitation is less accurate with an underestimate of occurrence across all intensity classes, especially higher intensity days Resolution helps a little in regions of topographic forcing: NW Resolution does not help in regions where convection is not topographically forced (Central North America). A more fundamental (convection) parameterisation problem. Very intense precipitation events (≥50 mm/day) become too frequent as resolution is increased.

26 Summary Winter season precipitation variability is fairly well simulated by GEM at 0.45º. This remains true at higher resolutions. Summer precipitation is less accurate with an underestimate of occurrence across all intensity classes, especially higher intensity days Resolution helps a little in regions of topographic forcing: NW Resolution does not help in regions where convection is not topographically forced (Central North America). A more fundamental (convection) parameterisation problem. Very intense precipitation events (≥50 mm/day) become too frequent as resolution is increased. Can we simulate summer daily precipitation variability at resolutions that require convective parameterisations ??? What can we do to make progress ??

27 Look at other models in a
variety of summer season precipitation regimes. Is the problem universal? Can we find pointers in other models how to improve summer season convection in GEM ? This is why we perform model intercomparisons e.g. ICTS-CEOP

28 Run multi-nested models, with the interior model convection-resolving.
compare convection-parameterising and convection-resolving domains to try to get pointers for improving the lower resolution simulations

29 NW JJA mean Some modest improvement with increasing resolution in the NW (mountainous) region. Improved topographic forcing ? mm/day But the number of intense days (>20mm/day) increases too much with increasing resolution

30 Central USA (CNTR) JJA 1993-1994 mean
Some very modest improvement at 0.33º, which is lost at 0.15º Convection in CNTR region is not topographically forced. mm/day Even at higher resolution all precipitation classes are underestimated in CNTR JJA except very intense classes which are overestimated.


Download ppt "Simulating daily precipitation variability."

Similar presentations


Ads by Google