RIRW Verification
![Website header](/eval/hwrf_rirw/images/header_rirw.jpg)
RIRW Verification: 2014 Model Forecasts
RI Contingency table: 30 kt in 24-h
OBS | |||
---|---|---|---|
Yes | No | ||
FCST | Yes | 128 | 253 |
No | 1623 | 37654 |
• The distributions of the difference of change in a-deck values in the last 24-h and change in b-deck values in the last 24-h (ABDEL) are examined and for HWRF-14 forecasts. The hits (green boxes) are below the zero line, indicating that if the forecasts predicts a RI event, it under-estimates the magnitude of the intensity change.
![](../images/H214_rirw_boxplot.png)
Click for larger image.
• Similar boxplots of CTCX and OFCL show that both also under-estimates RI magnitude, even for hits. After 48-h, there are no OFCL RI forecasts.
![](../images/CTCX_rirw_boxplot.png)
![](../images/OFCL_rirw_boxplot.png)
Click for larger image.
• The 2014 HWRF model has difficulty predicting RI when the initial intensity is hurricane strength at the initial time.
• The 2014 HWRF model has higher POD (15.3%) and lower FAR (58.7%) for RW cases.
2014 HWRF RW Contingency table: H215 30 kt in 24-h
OBS | Yes | No | |
---|---|---|---|
FCST | Yes | 324 | 461 |
No | 1793 | 37080 |
• The ABDEL boxplots for 2014 HWRF and CTCX models show that when the models detects RW, they predict the average intensity change. The OFCL tends to underestimate the reduction in intensity by 5 to 10 kt for hits.
![](../images/H214_rw_boxplot.png)
![](../images/CTCX_rw_boxplot.png)
![](../images/OFCL_rw_boxplot.png)
Click for larger image.