Naked Science Forum

Life Sciences => The Environment => Topic started by: thedoc on 31/01/2012 16:16:22

Title: How can we better predict thunderstorms and floods?
Post by: thedoc on 31/01/2012 16:16:22

2012 began with gale force winds, localised flooding and travel disruptions in the UK.  But not everyone is sorry to see the stormy weather. Here's Professor Robin Hogan telling us how he tracks, and predicts, thunderstorms.....
Read a transcript of the interview by clicking here (http://www.thenakedscientists.com/HTML/content/interviews/interview/1937/)

or  

Title: Re: How can we better predict thunderstorms and floods?
Post by: CliffordK on 01/02/2012 20:54:03

Here in the USA, we have fairly accurate 10-day weather forecasts across much of the country.

Hurricanes are monitored as they cross the Atlantic. 

Perhaps tornadoes are a bit more tricky to predict too far in advance, although one can predict heavy storms, and perhaps damaging hail a few days in advance, with actual tornado risks assessed a few hours in advance. 

Floods have a few somewhat predictable components.

On the West Coast, they are often a combination of snowpack and elevation of the snowpack, and heavy warm rains.  Of course, it is still complicated managing dams, and how much water to hold, or  dump and when.

East Coast floods may be caused by hurricanes and tropical storms, and would be related to single events dumping a lot of water.

Midwest floods would be related to diffuse rain patterns across a enormous flood basins, and generally slow moving rivers.

Anyway, I would imagine a lot of the same factors apply in the UK, and Europe.  However, the country is much smaller, and thus heavily dependent on weather out at sea, or over other countries.

A 1.5 km grid sounds quite good.  Is it in fact too fine of a grid?

I'm thinking of my home state, Oregon.  If I was making a grid, I would make large blocks, perhaps 20x20 miles, or 400 sq miles.  But, try to make the grid based on large topographic features.  I suppose it might be easiest to create a fine resolution grid, perhaps 1 sq mile, then lump it together so one might get approximately 400 sq mile lumps around features such as small towns, or along the valley floor, mountain ranges, or along roads through mountain passes.

Certainly things like flooding is based on events across the entire flood basin, so one needs to define the paths of the rivers and tributaries.

I wonder if they apply technology such as neural networks to the weather grids?