Deep learning application able to predict El Niño events up to 18 months in advance

Posted: September 20, 2019 by oldbrew in climate, ENSO, methodology, predictions, research


In 24 out of 34 cases anyway, which is said to be better than existing methods.

A trio of researchers from Chonnam National University, Nanjing University of Information Science and Technology and the Chinese Academy of Sciences has found that a deep learning convolutional neural network was able to accurately predict El Niño events up to 18 months in advance, reports Phys.org.

In their paper published in the journal Nature, Yoo-Geun Ham, Jeong-Hwan Kim and Jing-Jia Luo, describe their deep learning application, how it was trained and how well it worked in predicting El Niño events.

El Niño-Southern Oscillation events are periods during which water warms above normal temperatures in tropical parts of the Pacific. When that warm water moves east, it leads to more rainfall and other weather events, such as hurricanes, in the Americas, and less rain in Australia and Indonesia.

Current models can accurately predict such events using data from water temperature gauges spread across the globe up to a year in advance. Scientists would like to be able to predict such events even sooner, however, because they can have a big impact on areas where the weather changes.

Knowing when a drought is coming in Indonesia, for example, could help officials prepare food stores to feed people suddenly unable to grow their food for a period of time.

In this new effort, the researchers took a different approach to predicting El Niño events using a deep-learning neural network rather than conventional weather forecasting models.

The researchers report that they trained their system using data collected from weather stations over the years 1871 to 1973. Data from such sources included a variety of weather and environmental measurements such as sea temperatures and average ocean heat content.

The researchers also trained it on 300 El Niño events that occurred between the years 1961 to 2005. Once the system had been taught to recognize the conditions that lead up to El Niño events, they tested it using data from 1984 to 2017.

They report that their system was more accurate than current weather models, correctly identifying 24 out of 34 events, compared to only 20 of the same events identified by conventional modeling. The system was also able to do so 18 months in advance.

The researchers also report that their system was able to recognize other events that are believed to lead to El Niño events, such as an Indian Ocean dipole.

Source here.

Comments
  1. erl happ says:

    The models are all over the place as per usual, as indicated by the use of the word ‘plume’. This is what the latest summary at https://www.cpc.ncep.noaa.gov/products/analysis_monitoring/enso_advisory/ensodisc.pdf has to say.

    The majority of models in the IRI/CPC plume (Fig. 6) continue to favor ENSO-neutral (Niño-3.4
    index between -0.5C and +0.5C) through the Northern Hemisphere spring. Interestingly, the statistical model averages favor Niño-3.4 values above the El Niño threshold (+0.5C) during the fall and winter,while the dynamical model average indicates values near +0.2C. Forecasters are leaning toward the dynamical model average, which is also supported by the current tendency of the ocean toward cooler conditions.

  2. Stephen Richards says:

    colour me sceptic

  3. Bob Koss says:

    They included 22 years of training data(1984-2005) in their comparison to the models. Only 12 years were independent.

    How many not found by the models came from the training period?

    Were all the ones found by the models found by their spiffy new neural net.

  4. ivan says:

    The big question is about the data they used for training. Since we know that almost all weather data sets have been ‘massaged’ by the UN Church of Climatology to give the results they expect and with computers if you put garbage in you get garbage out.

    The only conclusion is that what they are seeing is ‘massaged’ garbage that has been tweaked to give some results – GIGO.

  5. oldbrew says:

    The researchers also trained it on 300 El Niño events that occurred between the years 1961 to 2005.

    300? — maybe a typo for 30? Depends what they mean by ‘El Niño events’.

  6. tom0mason says:

    Does anyone have access to the paper which this write-up is based.
    It’s Yoo-Geun Ham et al. Deep learning for multi-year ENSO forecasts, Nature (2019). DOI: 10.1038/s41586-019-1559-7

  7. tom0mason says:

    I find it interesting that the UN-IPCC, and their modelling colleges, currently seem to think they can project (predict in layman’s terms) the future of the climate for many decades or even centuries ahead. Surely for that to be so the Pacific Ocean and the El Nino/La Nina cycles must be either very predictable (if it is climatically significant), or are not significant, therefore these convolutional neural network (CNN) computations are just art for art sake, or perhaps help with making better short term weather forecasts.
    Alternatively maybe the UN-IPCC and climate modellers just don’t really understand, maybe these climate models are just an excuse to keep the money rolling in.

    Ho-hum, all is unfair in taxation, superstitions and climate change.

  8. Graeme No.3 says:

    The current system is to predict coming El Niños regularly (at least once a year). Occasionally Nature obliges, and they chalk up a win (and forget the failures).

  9. Kip Hansen says:

    Well,they got 70% of events they had trained their model to predict — pf course, they trained the model on the years 1961 to 2005 and then tested the model on 1984 to 2017. quite an overlap.

    This is called weather prediction….what weathermen do.

  10. I do not think SOI data which is available for over 150 years has been fiddled. Look at this https://data.longpaddock.qld.gov.au/static/products/pdf/WetDryDroughtPoster.pdf It has a graph of SOI and IPO . It will be noted that the periods have regularity in length as well as occurrence. There is considerable evidence that both SOI and IPO are affected by planetary alignments. I have noted that daily SOI patterns are related to tides at Darwin (there is little tide change at Tahiti). There is a 28 day cycle due to the moon. Roger and others have looked at pattern recognition. There is long term data such as SOI available to look at patterns..No model will be successful if it includes the CO2 scam..

  11. oldbrew says:

    Landscheidt: SOLAR FORCING OF EL NIÑO AND LA NIÑA (2000)

    Correlation between ENSO events and sunspot cycle

    http://www.mitosyfraudes.org/Calen/NinoLand.html

    Lots of theory there.

  12. Gamecock says:

    ‘a deep learning convolutional neural network’

    I remember the good ol’ days when we just wrote software.

    Maybe I would have been paid more if I convinced the bosses that I had a a deep learning convolutional neural network.

  13. phil salmon says:

    So what’s coming next?

  14. oldbrew says:

    What happens next in terms of ENSO may depend on when the current solar minimum ends and sunspot activity takes off again. One or two signs of life this month…
    https://www.solen.info/solar/cycle25_spots.html

  15. oldbrew says:

    Atlantic hurricane season is 2 weeks past its peak date now.


    – – –
    SEPTEMBER 24, 2019
    NASA-NOAA satellite finds wind shear taking a toll on tropical storm Jerry
    by NASA’s Goddard Space Flight Center

    https://phys.org/news/2019-09-nasa-noaa-satellite-toll-tropical-storm.html

    But there’s no El Niño…

    In general, wind shear, the change in wind direction and/or speed with height, is stronger in the Atlantic Basin in El Niño hurricane seasons. Wind shear tends to reduce the number of named storms and hurricanes. In addition, there is often more sinking air in portions of the Atlantic Basin in El Niño years. Sinking air is hostile to tropical cyclone development.

    https://weather.com/storms/hurricane/news/2019-07-12-el-nino-weakening-enso-neutral-busier-atlantic-hurricane-season

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s