Thanks to the Chinese.

Western science, along with western civilization, appears to be in terminal decline.   Perhaps as a consequence the west was mostly ignored at the recent United Nations talkfest in Rio.  With the BRIC countries – Brazil, Russian, India and China – running the show there were no all night sessions, no rock-stars flying in at the last minute to put a deal together, and most importantly no grand announcements of more money to stop climate change.    

I am feeling particularly indebted to the Chinese at the moment.  Not only for saving us from Rio+20, but also for publishing my most recent contribution to science.   Amongst more than two dozen Chinese names, John Abbot and Jennifer Marohasy, can be found listed in the table of contents of the most recent issue of Advances in Atmospheric Sciences, a journal sponsored by the Chinese Academy of Sciences.  

It wasn’t the first place we sent the journal article, but it should have been. We got excellent feedback through their peer-review process and more. It's entitled ‘The Application of Artificial Intelligence to Rainfall Forecasting in Queensland’ and is a consequence of us deciding to have a go at rainfall forecasting after the terrible Queensland floods.  

Remember the flooding. 

Remember three-quarters of the state of Queensland was declared a disaster zone following torrential rains during the summer of 2010-2011.  That was soon after Tim Flannery convinced the state government to invest in desalination on the basis it would never rain again. Anyway, official weather and climate forecasts failed to predict the event with the Australian Bureau of Meteorology reliant on output from a general circulation model that also predicted more drought.  

In Asia, the use of artificial intelligence for short-, medium-, and long-term forecasting is an active area of research within the meteorology and mathematics and computing communities. Using the climate indices recognised as most relevant to the Australian continent, e.g. El Niño/La Niña Southern Oscillation, and a neural networks design used to forecast rainfall in Iran, John Abbot and I have come up with a prototype design that gives a more skilled medium-term forecast than the Australian Bureau of Meteorology’s general circulation model. (General circulation models are also used to predict global warming.) 

We are not claiming to have a particularly skilled rainfall forecast.  But we do consider it relevant that after working for less than 8 months on a standard laptop with off-the-shelf neural network software (Synapse from Pelatrion, Sweden), we have a more skilled forecast than the Bureau's general circulation model that had been in development for more than ten years and runs on the most expensive super computer in the southern hemisphere. We consider the neural network technology, based on artificial intelligence, to be a relatively inexpensive and complementary approach to the use of general circulation models and we are working on a second prototype.

If you would like a pdf copy of our paper in the latest issue of Advances in Atmospheric Sciences, send me an email (jennifermarohasy@jennifermarohasy.com) and I will send you a copy. 

Explained in more detail here-

http://jennifermarohasy.com/2012/05/using-artificial-intelligence-t...

and here-

http://jennifermarohasy.com/2012/05/forecasting-rainfall-neural-net...