There are several ways to make a forecast that might be better than flipping a coin or asking a magic 8-ball.
- Persistence Forecasting. Forecast of no change in current weather conditions. Persistence forecasts are generally good only for short periods of a few hours and become less accurate as the time period lengthens. It is a pretty good bet that the forecast for 1 minute from now will be like conditions currently, but 5 days from now it might be less valid. During quiet weather patterns without frontal passages (i.e., the same air mass in place), a persistence forecast that tomorrow is going to be the same as today is usually quite accurate.
- Climatological Forecast. Forecast derived using the average value of weather elements for a particular time period. This forecast is based on the assumption that the weather will "average" for a given day. Remember the saying "climate is what you except; weather is what you get"? A climatological forecast turns this around a says that "weather is what you expect". If making a forecast of temperature for Spokane you might visit a climate page to look up the average high and low temperature for a month [http://www.wrcc.dri.edu/cgi-bin/cliMAIN.pl?wa7938].
- Analogue method. Forecast that uses weather conditions on past days with similar large scale patterns as a guide. The premise is that previous patterns that match well with forecasted patterns should yield similar weather conditions. For example, if from previous weather maps it is seen that an intense ridge at 500-hPa during the warm months over the west coast produces a surface, high-pressure center located over the Nevada-Utah regions and this pattern produced strong Santa Ana winds along the west coast of the United States, then when a forecaster sees a pattern developing which has a strong ridge at 500 hPa developing near the west coast, a forecast of strong, easterly winds at coastal stations will likely be accurate. Recognition of patterns can be done in an automated sense, or comes with experience in forecasting.
- Model Output Statistics. Forecast that uses statistical relationships between the element being forecast (e.g., high temperature) and values calculated by a numerical model. For example, the minimum temperature occurring at a location might be a function of day of year, cloud cover, 500-hPa heights, and wind speed. Bulletins containing the MOS forecasts are computer generated for various stations in the United States and are a tool used by forecasters in preparation of their local forecasts.