High Resolution Modeling, Clouds, Precipitation and Climate Pier Siebesma, Ramon Mendez Gomez, Jerome Schalkwijk, Stephan de Roode Jisk Attema, Jessica Loreaux, Geert Lenderink, Harm Jonker 1. Precipitation and High Resolution Modeling 2. Extreme Precipitation 3. Ultra-High Resolution
Atmospheric Models at KNMI mm 10 m 100 m 1 km 10 km 100 km 1000 km 10000 km Cloud microphysics DNS turbulence Cumulus clouds Cumulonimbus clouds Mesoscale Convective systems Extratropical Cyclones Planetary waves Large Eddy Simulation (LES) Model (GALES, DALES) HARMONIE RACMO EC-Earth
1. Precipitation Ramon Mendez Gomez, Pier Siebesma, Jisk Attema, Stephan de Roode
Climatology Figure: 20-year moving average of mean, coastal and inland precipitation for summer in the Netherlands (Lenderink et al. 2008).
Climatology Trend in the growth of the inland-coastal precipitation difference. Hypothesis: Due to warmer sea surface temperatures Test in August 2006 case (record wet month with a record warm North Sea) Figure: 20-year moving average of mean, coastal and inland precipitation for summer in the Netherlands (Lenderink et al. 2008).
High Resolution Climate Simulations Geert Lenderink, Erik van Meijgaard and Frank Selten: Climate Dynamics (2008) Obs RACMO (6km) Using observed SST RACMO (6km) Using SSTclimatology Strong effect of SST on precip. Effect limited to the coastal areas (Calls for high resolution!!) Mind you: High resolution RACMO but still hydrostatic model!! How does HARMONIE perform?
RACMO(173mm) HARMONIE (224mm) RADAR (194 mm) Harmonie: picks up better the higher rain intensities underestimates overall precipitation amount
Precipitation as a function of distance to the coast HARMONIE: 280 HARMONIE Right position of the maximum Precipitation (mm/month) 230 180 130 HARM_HD RADAR GROUND OBS RACMO_HD_CTL HARM_CL RACMO obs Overestimation of the total amount Working on the why 80-150 -100-50 0 50 100 150 Distance to coastline (km) Overprediction of precip not uncommon for models operating in the Grey Zone (1~5 km)
2. Precipitation Extremes Jessica Loreaux, Geert Lenderink, Pier Siebesma, Stephan de Roode
observaties Neem de 90%, 99%, 99,9% percentielen van de meest extreme neerslag sommen: groepeer ze als functie van de (dauwpunts) temperatuur gebruik deze (dauwpunts) temperatuur als proxy voor hoe extreme neerslag veranderd in een opwarmend klimaat.
observaties Neem de 90%, 99%, 99,9% percentielen van de meest extreme neerslag sommen: groepeer ze als functie van de (dauwpunts) temperatuur gebruik deze (dauwpunts) temperatuur als proxy voor hoe extreme neerslag veranderd in een opwarmend klimaat. dagsom Precip intensity (mm/day) Td 7% toename per graad
observaties Neem de 90%, 99%, 99,9% percentielen van de meest extreme neerslag sommen: groepeer ze als functie van de (dauwpunts) temperatuur gebruik deze (dauwpunts) temperatuur als proxy voor hoe extreme neerslag veranderd in een opwarmend klimaat. dagsom 10 minuten som 7% toename per graad Frontaal 14% toename per graad!! Convectieve buien
Vervolgstappen: kunnen we deze toename verklaren? Kunnen we de geobserveerde schaling op deze manier aan thermodynamica & dynamica koppelen? Conceptueel model Large Eddy Simulatie model (temp laten toenemen, RH constant) Hoe relateert gevonden schaling in het huidig klimaat tot schaling door veranderend klimaat? Harmonie
3. Ultra-High Resolution Modelling Jerome Schalkwijk, Harm Jonker, Pier Siebesma
.History LES at KNMI 1992 : 40X40X40 grid points Dutch Atmospheric Large Eddy Simulations(DALES) State of the art supercomputer: 1024X1024X512 gridpoints
Does not come for free... Disruptive technology
GPU for general purpose Graphical Processing Unit video card Navier Stokes
Performance of GALES (GPU Atmospheric Large Eddy Simulations) Performance of DALES on Huygens (SARA) on a single node (32 processors) Wall clock time per time step per grid point 2 hour of simulation on 128^3 grid points in 3 minutes. Drawback: maximum on the amount of gridpoints ~ 512^3 at present
Applications
Climate Monitoring and Evaluation Tool for the Cabauw Site The KNMI testbed Neggers, Siebesma & Heus, accepted for BAMS 2012 φ φ φ φ φ" true state" = + + t t t τ LS LES, SCM
Climate Monitoring and Evaluation Tool for the Cabauw Site The KNMI testbed Neggers, Siebesma & Heus, accepted for BAMS 2012 GALES Gridpoints: 256X256X200 Resolution: Δx=Δy=100m Δz=40~100m Domain: 25X25X12km φ φ φ φ φ" true state" = + + t t t τ LS LES, SCM
Year of Gales; March 2011 March 2012; Evaluation Still early days. GALES obs Cloud fraction as seen by the Lidar versus simulated lidar at a single gridpoint in GALES
Year of Gales; March 2011 March 2012 Low clouds BL-clouds High clouds
Outlook Easy to implement at other locations (Schiphol) Relatively cheap (hardware costs ~5 Keuro) Many applications (local weather, impact weather on pollution, etc, etc) Performance depends on Quality of the model in which it is embedded in (HARMONIE, RACMO?) Initialisation (obs, obs & obs)
Multiple GPU s?
Nesting it into a regional model (RACMO) 25km 25km 100m How to couple the GPU s?
GALES as a NWP model Couple 16X16 GPU s of each 25 km^2 at 100 m resolution Allows for a 400X400 km domain First try: Boundary GPU s are nudged toward host model (RACMO) Fully interactive interaction between the ïnner GPU s First realistic LES on such a domain Supported by Bull, Paris : massive parallel GPU machine Hardware price ~ 250 Keuro
July 6, 2004
Temperature at 200 m clouds
Conclusions and Outlook GPU is a extremely suitable vehicle for fluid dynamics problems especially as a additional evaluation tool around supersites. Many applications Education, Evaluation, Monitoring, NWP, impact studies (what if scenario s on local scale), apps To be done: Mature Interactive radiation in CUDA Improved microphysics Albedo maps Vegetation maps Creation of structures (buildings etc, rather than rougness lengths) Fully compressible (to make the system more local) Comprehensive evaluation