Synoptic Control of MesoscalePrecipitating Systems in thePacific NorthwestKyle Swanson and Paul J. RoebberAtmospheric Sciences GroupDepartment of Mathematical SciencesUniversity of Wisconsin at Milwaukee
To what extent does an adequate representation of flowfeatures on the synoptic scale allow for skillful inferenceof mesoscale precipitation, in the setting of landfallingsystems on the U.S. west coast?• Study design separates initial and boundary condition effects fora range of synoptic systems:- Developed cyclone- Pineapple Express- Cut-offs- Frontal-waves• We seek to understand:- transfer of “information” from synoptic to mesoscale- dynamics of error growth in landfalling systems- ability of mesoscale models to “fill in” details absent fromlarger scale models (and the initial data - emergence)
Frontal WaveHigh error (RMS = 64.3 m)
MM5 ConfigurationThe “front-half” hemispheric domain, with 120 km grid spacing, is indicated by theouter boundaries of the map. The large and smaller inner domains, with 15 km gridspacing, are indicated by the bold rectangles. The hatched region indicates the targetzone in which precipitation verifications are conducted for landfalling cases.•One-Way•KF Cumulus•Blackadar PBL•Mixed Phase
Null hypothesis:Uncertainty in the large scale analysis has no impact on theskill of precipitation simulations in the target zone of the PacificNorthwest(uncertainty = initial conditions from the large-scale analysis or theflow of information through the mesoscale grid boundaries)No rejection1) large upscale error growth swamps predictability signal,regardless of analysis uncertainty2) mesoscale predictability is not strongly connected to thesynoptic-scale in the studied conditions
Separated using twin experiments with NCEP andECMWF reanalyses:1) inner domain will act to strongly amplify initialdifferences between reanalyses2) measures of the differences in inner domain will reveallittle growth during the simulationsRejectionImportant connection does exist between the quality of thesynoptic information and predictability at the mesoscale forthe studied conditions
Secondary hypothesis:Upstream buffer size has no impact on the skill ofprecipitation simulations in the target zone of thePacific Northwest.No rejection1) Large upscale error growth swamps predictability signal,regardless of quality of boundary information2) Mesoscale predictability is not strongly connected to thesynoptic-scale in the studied conditionsSeparated using twin experiments with NCEP andECMWF reanalyses.
Rejection1) LBC error exerts a significant influence on mesoscalepredictability.2) Boundary information strongly constrains evolution of themesoscale, such that small analysis uncertainty and smallbuffer produces superior results (consistent with higherquality information)Resolved via analysis of simulation data(see preliminary results).
4-66 February 1996 (Pineapple)500 hPa root mean square error (m) in the inner domains relative to the 120km outer domain (constrained to the NCEP reanalysis). Hovmoeller-typediagram showing the error as a function of integration time for sectionsnormal to the upper-level flow. The error for both domains is shown for thecommon area of the small inner domain.
Table 1:Precipitation skill scores (Kuiper Skill Score, KSS, and Equitable Threat Score, ETS)and model simulation fidelity for the 25 mm threshold for the period 1200 UTC 4-6February 1996Domain KSS ETS POD FAR BIAS BpRMSE(mm)Smallinner0.35 0.21 0.75 0.26 1.02 95.4 26.7Largeinner0.23 0.14 0.73 0.31 1.06 88.5 32.3Hemiouter0.16 0.09 0.58 0.32 0.84 55.2 40.8
Experiment:Two factor factorial design with repeated measure on onefactor (upstream buffer), of the following form:Analysis Uncertainty Small buffer Large bufferSmall dS1, dS2, … dSN dS1, dS2, … dSNLarge dL1, dL2, … dLM dL1, dL2, … dLMdSN = N landfalling cases with small analysis uncertaintydLM = M landfalling cases with large analysis uncertainty
The power of the statistic is computed, and wemake a test such that there is a low likelihoodof incorrect rejection.How is analysis uncertainty defined?Use NCEP and EMCWF reanalyses to construct a 10-year climatology. Where do the events fit in thesenorms? Also, an opportunity to explore this climatology(topic of today’s talk).
High LARGEtemporal errorvariabilityLow SMALLtemporal errorvariability
Analysis Difference (Power)Inner DomainWhat color is this??
EOF – Analysis DifferencesI.~6%~10 largescale patternsII.~3%~20%variance(rest whitein time)
Data VoidI.30 m trend!II.
Nino3 Regressed on 500 mbNOAA-CIRES/Climate Diagnostics Center•Same order as trend in I!
Composite Analysis Differences500 mb Vg (10%strongest – 10%weakest) # DAYS•Jet•Phasing orbaroclinic?•Targeted obs?