WxChallenge Model Output Page Tutorial Brian Tang University at Albany - SUNY 9/25/12 http://www.atmos.albany.edu/facstaff/tang/forecast/
Clicking on square brings up graphic for the specified variable (e.g. temperature) from the corresponding model run (e.g. 12Z NAM)
Two available models: GFS and NAM. Both are run at 00, 06, 12, and 18Z. Model output generally delayed 2-4 hours after the model initialization time Hourly output for the NAM, 3-hourly output for the GFS Verify the initialization time of the guidance on the title of each graphic! Sometimes data files are unavailable and graphics may be old.
Pressure (mb) Time-height cross sections display vertical profiles of an atmospheric field over the WxChallenge city out to 60 hours Temperature (C) = lower troposphere only Model initialization Time (Day/Hour UTC) Forecast at 60 hrs
Pressure (mb) Wind barbs are for the horizontal (u,v) wind. Shading gives the wind speed. Wind (knots) e.g. SE wind at 12 knots Time (Day/Hour UTC)
Pressure (mb) Top of boundary layer estimated on some figures Potential Temperature Wind TOP OF BOUNDARY LAYER Time (Day/Hour UTC) Time (Day/Hour UTC) Possible uses: Estimate surface temperature by extending dry adiabat down from top of BL Assess winds within the mixed layer
Temperature and Dewpoint (F) Time series display near-surface or convective variables 2 meter temperature and dewpoint Forecast Period Time (Day/Hour UTC)
Temperature and Dewpoint (F) 2 meter temperature and dewpoint Raw 2 meter temperature Raw 2 meter dewpoint Time (Day/Hour UTC)
Temperature and Dewpoint (F) Comparing deviations between MOS and raw model temperatures is important when there is behavior atypical of the diurnal cycle 2 meter temperature and dewpoint MOS temperature MOS dewpoint Time (Day/Hour UTC)
Temperature and Dewpoint (F) 2 meter temperature and dewpoint --- MOS high temperature --- MOS low temperature Time (Day/Hour UTC) Odd-time lows and highs automatically considered
Temperature and Dewpoint (F) Assess short term errors in the model forecasts 2 meter temperature and dewpoint + Observed temperature + Observed dewpoint Time (Day/Hour UTC)
Pressure (mb) Wind Speed (knots) 10 meter wind Boundary layer maximum wind Boundary layer average wind 10 meter wind Observed wind MOS wind Stochastic wind distribution Wind (knots) Time (Day/Hour UTC) Mean of distribution Boundary layer average and/or the stochastic wind will usually be superior. Be careful if there are complex terrain features nearby!
Precipitation (in) Precipitation Precipitation type Accumulated precipitation Watch out for misdiagnosed precipitation type during the cold season Time (Day/Hour UTC) Precipitation in the preceding 3 hours (GFS) or 1 hour (NAM) Accum. precipitation during forecast period Examine the precipitation time series in conjunction with time height cross sections
Convective Time Series Convective Available Potential Energy (CAPE) - Measure of integrated positive buoyancy of a surface parcel. CAPE allows for convection. Convective Inhibition (CIN) Measure of integrated negative buoyancy of a surface parcel. CIN inhibits convection. Lifted Index (LI) Difference in temperature between a parcel lifted to 500 mb and the environment at 500 mb. Negative LIs support convection. K-Index Measure of atmospheric instability for air-mass type thunderstorms. Values greater than 30 indicate a good chance for thunderstorms. Precipitable Water Total water vapor in a column of air. Values >1.5 indicate the potential for heavy rainfall. Convection is a major forecast challenge! A favorable convective index is not sufficient for convection. These convective time series are best used to supplement your forecast if you think convection will occur.
2 meter temperature Newest model Oldest model Look for trends and outliers Stochastic wind Oldest model Newest model
Probability GEFS High Temperature Probability Expected absolute error Expected Absolute Error (F) Ensemble member Control member Deterministic model Temperature (F) Useful for assessing uncertainty and not necessarily better than deterministic guidance High spread = more uncertainty = more reason to take a risk
Temperature error (F) High temperature errors Low temperature errors Meteorological history
Temperature error (F) NAM MOS GFS MOS Mean absolute errors 9/20 9/21 9/22 9/23 9/24 9/25 Observed high temperature on 9/21 Forecasted high temperature from the 9/20 12Z GFS MOS. Error of -1 F. Forecasted high temperature from the 9/20 12Z NAM MOS. Error of -4 F.
Temperature error (F) NAM MOS GFS MOS Mean absolute errors 9/20 9/21 9/22 9/23 9/24 9/25 Observed low temperature on 9/21 & Forecasted low temperature from the 9/20 12Z NAM MOS. No error. Forecasted low temperature from the 9/20 12Z GFS MOS. Error of -1 F.
Similar verification figures for the raw 2 meter low and high temperature for the NAM and GFS Which of the guidance products is doing best? Use mean absolute error. Are any of the guidance products biased (e.g. is the GFS MOS consistently too warm or too cold with the high temperature forecast)? Use mean error.
Temperature Meteorological history Dewpoint 9/14 9/15 9/16 9/17 9/18 9/19 9/20 9/21 Bounds of colored bars demarcate low and high temperature for day 9/19 High temperature 9/19 Low temperature
Sky Condition and Precipitation Clear Few/ Scattered Broken Overcast/ Obscured Light Precip. Moderate/ Heavy Precip. CLR FEW SCT BKN OVC BKN FEW SCT CLR RA RA+ RA- RA- Wind barbs (every 4 hours) Color key: <5 knots 5-20 knots >20 knots
Temperature error (F) Meteorological history can be used with low and high temperature errors to deduce conditional errors Given a scenario, what are the errors? e.g Given clear, calm mornings, what are the low temperature errors? Given northerly winds Given a rising dewpoint overnight Given a cold front passage Given cloudy, rainy conditions during the day
Wind error (F) Wind products are verified in the same manner as temperature Observed maximum 2-min. wind and forecasts from previous day s 12Z model Mean absolute errors show which wind guidance products are performing best Meteorological history can again be used to assess conditional errors
Example of NAM vs. GFS matchup for 10 meter wind and stochastic wind Mean absolute errors for raw 10 meter wind Mean absolute errors for stochastic wind In this particular case, both NAM and GFS raw 10 meter wind are biased low with GFS being a bit worse Stochastic wind mean absolute errors are small and neither model has advantage
High and low temperature climatology for the WxChallenge city One standard deviation from mean Mean Two standard deviations from mean
Wind and precipitation climatology for WxChallenge city Precip. climatology only considers days that have measurable precipitation Percentage of days precipitation > X Percentage of days precipitation X Median precipitation Mean precipitation Percentage of TOTAL days when there is measurable precipitation
Checking climatology is useful when Establishing guidelines on the range of weather to expect Guidance is calling for extreme temperatures, wind, or precipitation Data sources http://www.meteo.psu.edu/bufkit/ http://www.meteor.iastate.edu/~ckarsten/bufkit/data/ http://weather.rap.ucar.edu/ http://www.weather.gov/ http://ncdc.noaa.gov/ Acknowledgements Jonathan Moskaitis (ensemble graphics) Cegeon Chan (climatology graphics)