Presentation is loading. Please wait.

Presentation is loading. Please wait.

ETA and GFS Validation (Southeastern Mexico) John M. Dickens RSIS/CPC.

Similar presentations


Presentation on theme: "ETA and GFS Validation (Southeastern Mexico) John M. Dickens RSIS/CPC."— Presentation transcript:

1 ETA and GFS Validation (Southeastern Mexico) John M. Dickens RSIS/CPC

2 Outline Models Models Data Data Methods Methods Results Results Conclusions Conclusions

3 Purpose To determine whether it is appropriate to replace the 20 km ETA model with the 50 km GFS model for forecasting precipitation amounts in Central America and the Caribbean. To determine whether it is appropriate to replace the 20 km ETA model with the 50 km GFS model for forecasting precipitation amounts in Central America and the Caribbean.

4 Models ETA 20 ETA 20 20 km 20 km CPC Domain CPC Domain GFS 50 GFS 50 Global Global 50 km 50 km Show model domain

5 Data Central America GTS Stations (75): The data from these stations are quality controlled downstairs and in the CPC, making them more reliable than their NOAA/USGS counterparts. Due to the sparse data these stations will not be used. NOAA/USGS Stations (101): These stations were set up as part of the Hurricane Mitch Program. No money was set aside for maintenance of these stations so their reliability is in question. Also, the exact units of the measurements have never been determined. Therefore, these stations will not be used. Mexico Co-Op Stations (~500 out of 1000): The Mexico stations will be used because they are fairly dense in the southern half of the country and therefore will be a decent representation of the Central American region.

6 Data Only used stations reporting precipitation amounts greater than zero Only used stations reporting precipitation amounts greater than zero Time period: May 1 to August 31, 2004 Time period: May 1 to August 31, 2004

7 August 2 Mexico Data: The Mexico data runs from 12Z to 12Z. August 1 Forecast August 2 Forecast * For the August 1 st forecasts of 24 hrs and 48 hrs the Mexico data for August 1 st and 2 nd are required. August 3 Forecast August 1 Forecast Times F00 F06 F12 F18 F24 F30 F36 F42 F48 F54 F60 F66 F72 F78 The forecasts represent the precipitation for the past 6 hours. F06 represents rainfall from F00 to F06 F12 represents rainfall from F06 to F12 ……. August 1 Mexico Data: The Mexico data runs from 12Z to 12Z. Data

8 Methods Match Co-op stations to model grid points in a point-to-grid comparison Match Co-op stations to model grid points in a point-to-grid comparison = model grid points = Co-op stations

9 Methods Regular Regular 24 hr / 48 hr 24 hr / 48 hr forecasts forecasts No Coast No Coast 24 hr / 48 hr 24 hr / 48 hrforecasts No Mountains No Mountains 24 hr / 48 hr 24 hr / 48 hrforecasts

10 Data Points

11 Methods Regular vs No Coast vs No Mountains Regular vs No Coast vs No Mountains 24 hr and 48 forecasts 24 hr and 48 forecasts 24 hr vs 48 hr forecasts 24 hr vs 48 hr forecasts ETA and GFS ETA and GFS Overall Overall

12 24 hr Mean Difference

13 24 hr Standard Deviation of the Mean Difference

14 48 hr Mean Difference

15 48 hr Standard Deviation of the Mean Difference

16 Categorical Summary OverallRegular No Coast No Mountains Mean Difference (lowest) 24 hr GFS (3.27 mm) GFS (4.37 mm) ETA (1.67 mm) 48 hr GFS (2.19 mm) GFS (3.19 mm) ETA (3.04 mm) Standard Deviation of Mean Difference (lowest) 24 hr ETA (4.43 mm) ETA (5.13 mm) TIE (.90 mm) 48 hr ETA (3.39 mm) ETA (4.28 mm) TIE (.10 mm)

17 24 hr vs 48 hr Forecasts 24 hr ETA48 hr ETA 24 hr GFS48 hr GFS

18 Conclusions Mean differences increase with increasingly large events Mean differences increase with increasingly large events Standard deviations of mean difference increase with increasingly large events Standard deviations of mean difference increase with increasingly large events

19 Conclusions Regular Regular Overall GFS smaller mean difference Overall GFS smaller mean difference Overall ETA smaller standard deviation of the mean difference Overall ETA smaller standard deviation of the mean difference No Coast No Coast Overall GFS smaller mean difference Overall GFS smaller mean difference Overall ETA smaller standard deviation of mean difference Overall ETA smaller standard deviation of mean difference No Mountains No Mountains Overall ETA smaller mean difference Overall ETA smaller mean difference Overall ETA and GFS have identical standard deviations of mean difference Overall ETA and GFS have identical standard deviations of mean difference

20 Conclusions ETA 48 hr forecasts more accurate than the ETA 24 hr forecasts ETA 48 hr forecasts more accurate than the ETA 24 hr forecasts Known initialization problems….additional 24 hrs for spinup Known initialization problems….additional 24 hrs for spinup ETA forecasts precipitation in the region better than the GFS ETA forecasts precipitation in the region better than the GFS Mean differences: both < ±6 mm Mean differences: both < ±6 mm Standard Deviations: both > 15 mm, ETA ~ 5 mm lower Standard Deviations: both > 15 mm, ETA ~ 5 mm lower Correlations: ETA.25-.30, GFS near zero Correlations: ETA.25-.30, GFS near zero

21 Thank You


Download ppt "ETA and GFS Validation (Southeastern Mexico) John M. Dickens RSIS/CPC."

Similar presentations


Ads by Google