Presentation is loading. Please wait.

Presentation is loading. Please wait.

Validation of the FSU/COAPS Climate Model Mary Beth Engelman [T. LaRow, D.W. Shin, S.Cocke, and M. Griffin] CDPW Meeting – Tallahassee, FL October 22,

Similar presentations


Presentation on theme: "Validation of the FSU/COAPS Climate Model Mary Beth Engelman [T. LaRow, D.W. Shin, S.Cocke, and M. Griffin] CDPW Meeting – Tallahassee, FL October 22,"— Presentation transcript:

1 Validation of the FSU/COAPS Climate Model Mary Beth Engelman [T. LaRow, D.W. Shin, S.Cocke, and M. Griffin] CDPW Meeting – Tallahassee, FL October 22, 2007

2 Outline: I.Introduction to work II.Background (Previous Studies) III.Data IV.Methodology V.Results VI.Summary/Conclusions VII.Future Work

3 Introduction: The study What I’m Doing… –Scientific Objective: Validating the FSU/COAPS Climate Model –Variables: »Surface Temperature (Maximum and Minimum) »Precipitation –Comparing: »Model forecast vs. Observations »Real-time runs (persisted SSTA) vs. Hindcast runs (observed weekly SSTA) –Scientific Questions: 1. How “well” does the FSU/COAPS climate model forecast surface temperature and precipitation over the southeast United States? 2. How do forecasts made from persisted SSTA compare to hindcasts made from weekly updated prescribed SSTA?

4 Introduction: Reasons for study And Why… –Benefit to society: Crop models can assist decision makers in minimizing loses and maximizing profits in the agricultural community (Hansen et al., 1997) –Benefit to the science: Evaluation of the need for a coupled ocean-atmosphere climate model Validation potentially leads to improved climate models/forecasts

5 Previous Studies: Persisted vs. prescribed SSTA –Why use persisted anomalies for climate forecasts? Currently, coupled models do not out perform predictions given by persisted SSTA at lead times less than 3 months (exception: ENSO events) –Why compare real-time runs to hindcasts runs? Overestimates skill, but evaluates model’s possible potential Determines how much skill is lost from using persisted SSTA –Performance of real-time vs. hindcasts runs? Similar values of skill, unless large SSTA errors (Goddard and Mason, 2002)

6 Previous Studies: SST and the southeast climate SOI and Southeast United States El Nino Signals (Ropelewski and Halpert, 1986) –Consistent signals: Above normal precipitation (October-March) Below normal temperature –Physical mechanisms causing precipitation and temperature signals (teleconnections vs. direct link) Tropical Pacific SSTA and ENSO Signals (Montroy, 1997) –Strongest signals: Precipitation in the Southeast US Positive: November - March Negative: July-August Teleconnections (Gershunov and Barnett, 1997) –Strong teleconnection between tropical Pacific SSTs and wintertime precipitation over the coastal southeast US (negatively correlated in the interior states)

7 Previous Studies: The model The FSU/COAPS Climate Model: Large Scale Features: –Capable of reproducing large scale features associated with ENSO (Cocke and LaRow, 2000) –Capable of reproducing teleconnections response (Cocke et al., 2007) Temperature and Precipitation: –Regional and global model are both in “reasonable” agreement with observations (Cocke and LaRow, 2000) –Alterations to the physics of the model resulted in a strong wet bias and cold surface temperature bias, upgrading the model (inclusion of the NCAR CLM2 as land parameterization) reduced biases (Shin et al., 2004) –Regional model rainfall events: Over predicts small rainfall events “Accurately” produces higher magnitude rainfall events (Cocke et al., 2007) –“Seasonal variations, anomalies, and changes in extreme daily events for maximum temperature are successfully forecasted” (Lim et al., 2007)

8 Previous Studies: Summary What to expect in my study (Hypothesis): – Model Forecast vs. Observations: Model forecasts will show a wet bias in precipitation and cold bias in maximum surface temperature (unless bias corrected) Forecasts should show “reasonable” skill (anomalies, seasonal variations, various skill scores) Forecasts will tend to over predict the number of small rainfall events – Real-Time Runs vs. Hindcast Runs (Persisted vs. Prescribed SSTA ) In general, hindcast runs will out perform real-time runs Real-time runs should be comparable to hindcasts runs when equatorial Pacific SSTA remain fairly constant throughout the duration of the forecast Loss of skill in the real-time runs will occur –When equatorial Pacific SSTA take a sharp change –Where areas are greatly influenced by SSTs (high teleconnection signal) Skill of real-time runs will decrease after 3 months

9 Data: Model Details of model (Spectral Model): –Resolution: GSM: T63 (1.875˚ x 1.875˚) NRSM: 20 x 20 km –Bias: Precipitation: wet Max Temperature: cold Real time run: –Persisted SSTA –Model Runs (5 ensembles): January 2005 and 2006 (3 month forecasts) Hindcast runs: –Prescribed SSTA (updated weekly) –Model Runs (5 ensembles): October 2004 and 2005 (6 month forecasts) –Model Climatology (20 ensembles): 1987-2005 FSU/COAPS Climate Model Model Details

10 Data: Observations COOP Observation Network: Daily/Monthly Temperature and Precipitation (January 1986-June 2007) Station to grid: –172 long term stations (1948- present) –For each grid point… Finds the closest stations (using the great circle distance) Assigns the station’s temperature and precipitation to that grid

11 Methodology In General: –Looking for the ability to forecast: Correct anomaly Extreme events –Heavy Precipitation/Drought –Freezing Temperatures Today’s Talk (Temperature): ability to forecast: –1. Average monthly temperature anomaly –2. Max temperature distributions (bias corrected)* –3. Freeze events (bias corrected)* *bias corrected: where:

12 Comparing JFM 05 vs JFM 06 SSTA 2005 2006 2005: SSTA switch from warm to cool 2006: SSTA remain fairly constant (figures provided by: http://www.pmel.noaa.gov/tao/jsdisplay/ )

13 Results: Ability to forecast anomalies Real-TimeHindcastReal-TimeHindcast Jan. Feb. Mar. 2005 2006 Error in Monthly Temperature Anomalies (˚C): Initial time:Jan. 05Oct. 05Jan. 06Oct. 06 Error in Temperature (ε = f ` - o`) 2006 real-time runs resulted in less error than the 2005 real-time runs 2005 and 2006 hindcasts runs showed approximately the same magnitude of error In 2006 the real-time runs and hindcast runs resulted in “similar” forecast error |ε| = 4.28 |ε| = 2.50 |ε| = 0.77|ε| = 3.14 |ε| = 4.86 |ε| = 2.72 |ε| = 0.87 |ε| = 1.96|ε| = 2.79 |ε| = 0.95 |ε| = 1.40|ε| = 2.65

14 Results: Ability to forecast distributions Daily Maximum Temperatures (January-March) Observed distributions more accurately captured by the models in 2006 Hindcasts vs. real-time runs differ more in 2005 than in 2006 In both years, model over estimates tails of the distribution

15 Results: Ability to forecast extreme events Deep Freeze Events (January – March) Model tends to over predicts number of freeze events Real-time and hindcasts runs are more similar in 2006 Observed Real-Time Hindcast 2005 20062005 2006 T min < -2˚CT min < -4˚C

16 Summary/Conclusions: –Model forecasts show a cold bias in surface temperature –The model tends to over predict “extreme” events –Real-time runs resembled hindcasts runs when equatorial Pacific SSTA remain fairly constant throughout the duration of the forecast (as in 2006) –When SSTA changed during the time period of the forecast (as in 2005), hindcasts runs outperform real-time runs even at lead times of 3 months

17 Future Work: –Evaluate Real-Time October runs: Currently comparing Januray real-time runs (0 month lead time) to October hindcast runs (3 month lead time) –Examine other verification methods/skill scores: Anomaly correlations Taylor diagrams Relative Operating Characteristics (ROC) Heidke Skill Score (HSS) –Further investigate SSTA in the tropical Pacific and their effects on model forecast errors

18 Questions?

19 References: Cocke, S., and LaRow, T.E., 2000: Seasonal predictions using a regional spectral model embedded within a coupled ocean-atmosphere model. Mon. Wea. Rev., 128, 689-708. Goddard, L, S.J. Mason, S.E. Zebiak, C.F. Ropelewski, R. Basher, and M.A. Cane. 2001. Current approaches to seasonal-to-interannual climate predictions. International Journal of Climatology. 21: 1111-1152. DOI: 10.1002/joc.636 Goddard L, Mason SJ, 2002. Sensitivity of seasonal climate forecasts to persisted SST anomalies. Climate Dynamics 19: 619–631 Gershunov, A., 1998: ENSO influence on intraseasonal extreme rainfall and temperature frequencies in the contiguous United States:Implications for long-range predictability. J. Climate, 11, 3192–3203.. Jolliffe, I.T., and D.B. Stephenson, 2003. Forecast Verification. Wiley, 240 pp. Lim, Y.K., Shin, D.W., Cocke, S., LaRow, T.E., Schoof, J.T., O’Brien, J.J, and Chassignet, E., 2007: Dynamically and statistically downscaled seasonal forecasts of maximum surface air temperature over the southeast United States. Montroy, D., 1997: Linear relation of central and eastern North American precipitation to tropical Pacific sea surface temperature anomalies. J. Climate, 10, 541–558.. Murphy, A.H., 1993. What is a good forecast? An essay on the nature of goodness in weather forecasting. Weather and Forecasting, 8, 281-293. Ropelewski, C. F., and M. S. Halpert, 1986: North American precipitation and temperature patterns associated with the El Niño/Southern Oscillation (ENSO). Mon. Wea. Rev., 114, 2352–2362. Shin, D.W., S. Cocke, T. E. LaRow, and J. J. O’Brien (2005), Seasonal surface air temperature and precipitation in the FSU climate model coupled to the CLM2, J. Clim., 18, 3217-3228. Wilks, D.S., 1995. Statistical Methods in the Atmospheric Sciences. An Introduction. Sans Diego: Academic Press. Wilks, D.S. and R.L. Wilby. 1999. The weather generation game: A review of stochastic weather models. Progress in Physical Geography 23:329-357.


Download ppt "Validation of the FSU/COAPS Climate Model Mary Beth Engelman [T. LaRow, D.W. Shin, S.Cocke, and M. Griffin] CDPW Meeting – Tallahassee, FL October 22,"

Similar presentations


Ads by Google