Low cloud forecasting challenge
San Francisco International Airport (SFO) has two closely spaced parallel runways. Both runways can only be used at the same time when the approach is completely clear of any cloud. Maximum flow rates change from 35 aircraft landing per hour when there is cloud, to up to 54 per hour once the clear conditions are reached. Therefore an accurate forecast of cloud clearance can be of considerable benefit for planning for both Air Traffic Control and the airlines coming into SFO.
Low cloud over the ocean off the San Francisco coast comes in over the airport on an almost daily basis through the summer months. Forecasting the exact clearance time of this cloud over the runway and approach zone poses a challenge as conditions can vary considerably day-by-day and most weather models struggle to capture some of the finer detail in this area.
Aviation Weather Testbed
This August the Met Office participated in the Aviation Weather Testbed experiment, which runs at NOAA’s Aviation Weather Center in the USA and is where new tools and processes in aviation meteorology are assessed. For this the Met Office aviation and defence applications teams ran a high-resolution weather model over a limited area around SFO, producing forecast data with a 330 metre horizontal resolution. This data was analysed against other models and observations over a two-week period to see how this high-resolution modelling capability may add value for these types of local forecasting challenges.
Modelling weather at a 330 metre resolution
Here is a case study from Wednesday 16th August 2017, showing the added value the 330m model can provide. This example shows this model nicely capturing the structure and timing of the cloud clearance. Below is the cloud cover forecast valid at 15 UTC, compared against satellite data showing the actual cloud cover at this time. The model was run at 00 UTC, showing that it can have skill at this scale even 15 hours ahead, which would prove useful for some planning timescales.
Snapshot comparing the cloud forecast valid at 15 UTC with the satellite observations for the same time. The box indicates the region over which the high-resolution model was run. The yellow marker shows the location of SFO.
The more detailed terrain and improved representation of clouds in this model can help to better capture the cloud features and give a more realistic looking trend in the clearance, compared to other weather models. This case represented the closest match between model and observation over the two week period but demonstrates the promise of this exciting new modelling capability. Further research and analysis will continue in this area, to keep working towards providing more useful forecasting guidance for both our meteorologists and customers.
Claire Bartholomew, David Bentley and Piers Buchanan