Search the RAMMB website
To evaluate the GPCE version of the MC probability model, a subset of the Atlantic cases from the 2008 season were re-run. These cases were run with the operational MC Model and the experimental version where the track error distributions are a function of the GPCE estimate.
For the GPCE evaluation, forecasts where the initial position was within 1,000 km from the U.S. coastline were chosen. The sample includes 167 cases, summarized below.
| ATCF No. | Storm Name | #Cases | First Case | Last Case |
| AL02 | Bertha | 5 | 08071512 | 08071612 |
| AL03 | Cristobal | 16 | 08071900 | 08072218 |
| AL04 | Dolly | 16 | 08072012 | 08072406 |
| AL05 | Edouard | 8 | 08080312 | 08080506 |
| AL06 | Fay | 28 | 08081618 | 08082312 |
| AL07 | Gustav | 21 | 08082718 | 08090118 |
| AL08 | Hanna | 24 | 08083118 | 08090618 |
| AL09 | Ike | 25 | 08090712 | 08091318 |
| AL11 | Kyle | 6 | 08092706 | 08092812 |
| AL13 | Marco | 7 | 08100600 | 08100712 |
| AL17 | Paloma | 11 | 08110700 | 08110912 |
The GPCE model will be evaluated by two methods. First, the 34, 50 and 64 kt wind probabilities for each case will be generated with the operational and GPCE models on a 0.2 deg. grid surrounding the forecast tracks. The 0-120hr cumulative probabilities for the operational and GPCE versions are shown below in the model results section. Difference plots will be added as well.
Second, the probabilities will be generated at the U.S., Caribbean and South and Central American breakpoints from the NHC web site. All points except the west coast of Mexico and Hawaii will be included. Hawaii and the Mexican west coast are excluded to reduce the number of verification points with very small or zero probabilities, the verification will be performed at the remaining breakpoints.
Click on links below to see the comparison of the operational and GPCE MC Model 0-120 hr cumulative probabilties for each case. For each run the GPCE values are listed at the top of the page, for each 12 hr. period. The track error tercile (L=low, M=medium, H=high, X=missing/unused) used at each time period is also shown. The basin-wide distributions were used if the GPCE value was missing.
Bertha (Operational: 100%) (GPCE: 100%) (Table: 100%)
Cristobal (Operational: 100%) (GPCE: 100%) (Table: 100%)
| 08071900 | 08071906 | 08071912 | 08071918 |
| 08072000 | 08072006 | 08072012 | 08072018 |
| 08072100 | 08072106 | 08072112 | 08072118 |
| 08072200 | 08072206 | 08072212 | 08072218 |
Dolly (Operational: 100%) (GPCE: 100%) (Table: 100%)
| 08072012 | 08072018 | ||
| 08072100 | 08072106 | 08072112 | 08072118 |
| 08072200 | 08072206 | 08072212 | 08072218 |
| 08072300 | 08072306 | 08072312 | 08072318 |
| 08072400 | 08072406 |
Edouard (Operational: 100%) (GPCE: 100%) (Table: 100%)
Fay (Operational: 100%) (GPCE: 100%) (Table: 100%)
Gustav (Operational: 100%) (GPCE: 100%) (Table: 100%)
| 08082718 | |||
| 08082800 | 08082806 | 08082812 | 08082818 |
| 08082900 | 08082906 | 08082912 | 08082918 |
| 08083000 | 08083006 | 08083012 | 08083018 |
| 08083100 | 08083106 | 08083112 | 08083118 |
| 08090100 | 08090106 | 08090112 | 08090118 |
Hanna (Operational: 100%) (GPCE: 100%) (Table: 100%)
Ike (Operational: 100%) (GPCE: 100%) (Table: 100%)
Kyle (Operational: 100%) (GPCE: 100%) (Table: 100%)
Marco (Operational: 100%) (GPCE: 100%) (Table: 100%)