|
|
Annual Mean Absolute Error Charts
Overview
In cooperation with the Climate Prediction Center (CPC),
the HPC has verification data for its medium-range minimum
and maximum temperature forecasts for days 3, 4, and 5.
The first full year for which mean absolute error (MAE) statistics
are available is 1972. Verification for 1996 or 1997 is not available.
The HPC began forecasting for days 6 and 7 in 1998.
The primary guidance used by the HPC through 1995 was the Klein-Lewis forecast
that provided departure from normal (DN) values.
There were several other forecast strategies that incorporated various
weightings of the Klein-Lewis DN forecast (KL) and a 10-day running bias of the KL (E10).
These forecast strategies included:
| 1. KL - E10 |
| 2. KL/2 |
| 3. (KL-E10)/2 |
| 4. KL - (E10/2) |
| 5. 3*(KL-E10)/4 |
1983 was the first full year that Linear Regression (LR) forecasts were available.
These utilized equations different from the Klein-Lewis forecasts, and included
a 60-day running bias correction. MOS forecasts were available starting in the
early to mid 1990's, with verification statistics starting from 1998.
Description of "HPC -vs- Guidance" Charts
The red line shows the MAE of the HPC forecasts of
max or min temperature for day 3, 4, or 5. HPC MAE data is available 1972-1995 and
1998 to present.
The green line indicates the MAE of the forecast
against which HPC was offially compared. During 1972-1995 the HPC was scored against
the Klein-Lewis forecasts. By 1998 MOS became the standard against which HPC
forecasts were compared, and is currently the only guidance available.
The blue line depicts the MAE of the BEST performing
strategy for that year. This plot terminates after 1995. The HPC has full-year MAE data for
the following strategies (as described above) and times:
| KL | 1972-1995 |
| #1 | 1972-1982, 1984-1995 |
| #2 | 1972-1982, 1984-1995 |
| #3 | 1973-1982, 1984-1995 |
| LR | 1983-1995 |
|
|
www.weather.gov 
Rainfall