METEOPROOF
Daily Forecast Verification
Operational Runs
--
💡
How to read this chart?
Show more
▼
This chart shows how the forecast for a specific target date changes as the date gets closer, from D-13 (13 days before) to D-0 (the target day).
Each step on the X-axis represents a new forecast run available in Meteoproof. For example, D-9_12z means the forecast from the 12:00 UTC model run, made 9 days before the target date.
As new runs become available, the forecast can move up or down as the models update their calculations with new information. This shows how the forecast develops over time and how it gradually settles as the target date approaches.
The cyan dotted line (Model Median) shows the median forecast across all available models for each run. This gives you a simple view of where the models are generally pointing, while reducing the influence of individual outliers.
Model runs are usually available in Meteoproof with a short delay after the official model run time. Times shown on the X-axis are in UTC (Z).
In short: the further right you go, the closer the forecast is to the target date. The final result can then be compared with what actually happened.
Multimodel Accuracy & Bias Verification
Target Date: -- | Location: -- (Max)
Reference: --. Calculated RMSE and Bias values (AVG D1-6 = average for D-1 through D-6):
| LEAD | ||||||
|---|---|---|---|---|---|---|
| AVG D1-6 |
||||||
| Bias D1-6 | ||||||
| D-1 | ||||||
| D-2 | ||||||
| D-3 | ||||||
| D-4 | ||||||
| D-5 | ||||||
| D-6 | ||||||
| D-7 | ||||||
| D-8 | ||||||
| D-9 | ||||||
| D-10 | ||||||
| D-11 | ||||||
| D-12 | ||||||
| D-13 |
💡
What is Accuracy & Bias Verification?
Show more
▼
RMSE (Root Mean Square Error)
RMSE shows how far a model's forecast was from the reference value. This is usually the official weather station observation. If that observation is not available yet, the model median is used instead.
RMSE is expressed in the same unit as the forecast, such as °C or km/h.
RMSE gives more weight to large errors than to small ones. This is useful because one big miss can matter more than several small ones. Two models can have the same average error, but the one with the larger outlier will have a higher RMSE.
Example: Suppose a model misses the reference by +1°C, -1°C, +1°C and +3°C across four runs for the same lead day. The average error is 1.0°C, but the RMSE is about 1.73°C. The larger 3°C miss pushes the RMSE up more than the smaller errors.
Lower is better. Green indicates better accuracy. The columns are ranked from best to worst, left to right, based on AVG D1-6, which is the average RMSE across D1 to D6 and is shown in the top row.
Bias D1-6
Bias answers a different question: does a model tend to be too warm or too cold? It is simply the average of forecast − observed across D1 to D6.
- ▲ Red = the model tends to be too warm
- ▼ Blue = the model tends to be too cold
- ● = no clear tendency
The separate blue/red colour scale is intentional. RMSE tells you how large the errors are, while bias tells you which direction they tend to go.
A model can therefore have a good RMSE and still have a noticeable warm or cold bias. The two metrics tell you different things.
Run-to-Run Stability
Target Date: -- | Location: -- (--)
How much a model's forecast changes between consecutive runs on each lead day (lower = more consistent).
| LEAD | ||||||
|---|---|---|---|---|---|---|
| AVG D1-6 |
||||||
| D-1 | ||||||
| D-2 | ||||||
| D-3 | ||||||
| D-4 | ||||||
| D-5 | ||||||
| D-6 | ||||||
| D-7 | ||||||
| D-8 | ||||||
| D-9 | ||||||
| D-10 | ||||||
| D-11 | ||||||
| D-12 | ||||||
| D-13 |
💡
What is Run-to-Run Stability?
Show more
▼
This table measures consistency, not accuracy.
It looks at how much a model changes its forecast for the same target date from one run to the next. A stable model tends to make similar forecasts as new runs come in. A less stable model can keep changing its mind.
This is useful because a model can be accurate on average, but still move up and down quite a lot between runs. This table shows you how steady its forecasts were.
Example: A model forecasts 20°C, then 21°C, then 19°C for the same target date. The changes between runs are +1°C and -2°C. The resulting stability score is 1.58°C.
Only runs that actually produced a forecast are compared. Missing or skipped runs are simply ignored, so they never create an artificial jump in the results.
Lower is better. Green indicates a more stable model. The columns are ranked from best to worst, left to right, based on AVG D1-6, which is the average stability score across D1 to D6 and is shown in the top row.
A low score means the model tends to stick with its forecast. A high score means it changes its mind more often as new runs arrive.