Prediction vs. Intervention - Weather and Climate

1. Let's start with the weather, 
then go from there to climate

Observation:

. . . My lab was involved in an interesting weather research project

. . . Later, let's discuss if the results apply to climate...

Predicting the Weather

Remember "Spoil Sports of the Prediction Game"?

. . . That's why weather forecasters have a bad reputation

. . . . . . It's not their fault!

. . . . . . They're doing the best they can

. . . They often do okay short term

. . . . . . We'll never predict the weather on this date next year

. . . . . . Delphi method: how far in advance do you think we'll get?

2. What about weather control?

. . . Keeping the streets open costs a lot!

. . . Like here in the US in, say...

. . . . . . Minneapolis (straight north of Little Rock)

. . . . . . Buffalo (lots of Great Lakes-related snow)

. . . . . . etc.

. . . Little Rock -

. . . . . . Snow equipment costs a bundle, hardly ever needed

. . . . . . Cheaper just to shut down the city

. . . . . . . . . That's expensive too (but at least you get an extra vacation day!)

Moscow has

(a) a lot of snow, and

(b) budget problems

Credit: Igor Tabakov, Moscow Times, 10/15/09, http://www.themoscowtimes.com/business/article/luzhkov-takes-snow-fight-to-the-sky/387446.html

The Russian Air Force sprayed clouds with liquid nitrogen, silver (iodide?), and cement
. . . in 2009
. . . see news article link under image above
. . . no word on whether it worked

3. Interestingly, you can buy weather modification equipment

. . . http://www.nawcinc.com/photos.html

. . . How well it works is another issue...


4. What's easier, predicting or controlling weather?

. . . What would you guess?







5. About research from my lab

. . . Experiments done by former grad student Dongping Xu

. . . . . . as-yet unpublished research

We asked:

What predictions are more reliable:

Forecasting the weather?

-or-

Forecasting the effects of weather modifications?


How we answered

1. Predict the weather using MM5 on given initial weather pattern

2. Add a bug to MM5, run it again

. . . what is a bug?

Source: http://en.wikipedia.org/wiki/File:H96566k.jpg, in http://en.wikipedia.org/wiki/Software_bug

3. Do this for many different bugs

4. Calculate the average amount of change caused by a bug

. . . that is, change to some quantitative weather value

Why care about how MM5 acts when bugs are added?


6. Now do another experiment

(This is a little more complicated)

1. Run MM5 to predict the weather

1a. Change an initial condition a tiny bit and run it again

1b. Determine the change in the prediction

1c. Calculate sensitivity s:

s  = change in output / change in input

   = change in predicted value / change in initial condition

   = Δ output / Δ input
(Delta is a Greek letter that is a triangle)
"Δ"  means "change"

2. Add a bug to MM5, run it again

3. Do this for many different bugs

4. Calculate the average amount of change (in sensitivity, this time) caused by a bug

So:

First part:

How resistant is weather prediction to software bugs?

Second part:

How resistant is weather intervention prediction to bugs?

Conclusion:

7. Weather prediction is considerably more bug-resistant than weather intervention prediction

Or

"We can't predict the effects
of weather modifications very well,
unless we get rid of all bugs,
and we already know we can't do that."

This leads to the big remaining question:

. . . Only a few people care about weather modification

. . . A lot of people care about changing global warming

. . . But can we tell how much change is actually needed?