How do we forecast?

In modern times, we use sophisticated computer models to simulate the current and future (forecast) state of the atmosphere. This field is known as Numerical Weather Prediction (NWP), and meteorologists have been developing and improving NWP techniques for over 60 years now. Close cousins of these models are also used to simulate the climate, and determine how it will change as greenhouse gases build up in our atmosphere.

Many developed countries have NWP models, and forecasters with the National Weather Service (NWS) generally look at the forecasts generated by many models. This tells the forecaster whether there is any consensus for a certain forecast. As we all remember from 2012, multiple models forecast the northward movement of Hurricane Sandy, but there were substantial model differences in exactly where and when the storm would make landfall, if at all – the map below shows the wide range!

On that occasion, the European model appeared to produce the best forecast.

The main models used in the US are: the Global Forecast System (GFS) model and the North American Model (NAM), both of which forecast for the entire planet; and the Weather Research and Forecasting Model (WRF – pronounced “WORF”), which produces forecasts for limited regions such as the Bay Area. Forecasts produced by all of these are processed at SJSU and then posted online at our Weather Center page.

Right now – in the middle of our really long, dry period – all of us check model forecasts every day to see if there is any chance of rain. And thus far, as we all know – nothing. But – an interesting pattern has been noted! Every time we have checked – for at least as far back as Thanksgiving! – the forecast has been roughly the same: dry this week, rain returning next week or the week after. As you all know, this hasn’t worked out at all well! When models get the forecast right, we say that they have “verified”, but lately – they haven’t.

So today, 8 Jan 2014, the GFS model again has us dry all the way out to 18 Jan, except for maybe a light sprinkle over the upcoming weekend. After the 18th, the model is trying again to make it rain here! It’s forecasting rain on the 18th and again the 20th. It really IS interesting how the models are constantly forecasting that the big, bad ridge offshore (see my blog on 1/6/14 on our “Crazy Winter Weather” about ridges) is going to break up soon. It’s just as interesting as the fact that the ridge has been stuck out there, blocking storms from bringing our winter rain, for almost 2 months now!!

Here’s the forecast for rain on the 20th (PST). The yellow and green splashes show regions where the model forecasts that rain will have fallen within the previous 12 hours. As you can see, the model is forecasting a storm moving onshore in northern California, with rain spreading in over the Bay Area!

Will this forecast verify? Probably not! Meteorologists have established that the forecasts beyond about 10-14 days generally have little accuracy, although forecast accuracy has improved measurably in the last 60 years! However, we now understand that atmospheric behavior is a chaotic, and hence unpredictable beyond 10-14 days given our current state of knowledge of chaos. Incidentally, did you know that one of the founders of chaos theory was Dr. Ed Lorenz of MIT – a meteorologist!

So, we probably can’t trust the GFS forecast, but you can follow the progress of the forecast of rain around the 21st by checking back to this page daily, and seeing if the forecast of rain is still on track! Fingers crossed!!