The forecast error is computed each month by comparing daily rainfall amounts to that which had been forecast by both the morning forecast and the afternoon update. The average daily error (RED for the MORNING FORECAST and LIGHT BLUE for the AFTERNOON) is then compared to various objective forecast schemes.
There are 3 basic objective schemes:
ZERO-CAST (GREEN) is a forecast of no rain every day. This scheme is terrible during wet season months when it usually rains somewhere in the District every day but it is a good false alarm check during the dry season
PERSISTENCE (DARK BLUE) is a forecast that the same amount of rain will fall today as did yesterday. Contrary to popular belief, this is usually not a good forecast and the scheme serves as a good check against trying to issue the same forecast everyday.
CLIMATOLOGY (ORANGE) is a forecast of what historically has happened on a particular date in the past. This scheme ignores all conditions currently going on in the atmosphere but it serves as a check against trying to always forecast the seasonal norms.
There are also 2 derived objective schemes:
CLIPER (PURPLE) is the average between Climatology and Persistence. This scheme combines seasonal norms with data from events which are currently occurring and it generally works well in tropical environments. However, CliPer will never forecast a sudden change.
BEST-CAST (BLACK) is a scheme that selects the best forecast of either Zero-Cast, Persistence, or Climatology on a daily basis. This scheme is relatively hard to beat since it usually has the lowest error for the easy (or "no brainer") forecasts and it never forecasts a false alarm of a heavy rainfall event. The only opportunities to solidly beat Best-Cast is to correctly forecast sudden increases in daily rainfall amounts.
It is the unofficial goal of the meteorology section to have an average daily forecast error which is lower than the Best-Cast on a monthly basis. Obviously, the best way to have low forecast error is to get the forecast exactly right every day. Barring that development, the Best-Cast provides a good benchmark for comparison.