Earth Wise

A look at our changing environment.

You are here: Home / Archives for machine learning

machine learning

The Sounds Of Coral Reefs | Earth Wise          

By Leave a Comment

Coral reefs around the world face multiple threats from climate change, pollution, and other impacts of human activity.  Reef conservation and restoration projects must be able to monitor the health of reefs and that is not such a simple matter.  Surveying reefs generally is labor-intensive and time consuming.  But in a new study, scientists at the University of Exeter in the UK have found a new way to do it.

The fish and other creatures living on coral reefs produce a vast range of sounds.  The meaning of these various sounds is for the most part unknown, but reefs nonetheless have distinctive sonic signatures.

The Exeter researchers decided to make use of machine learning technology.  They trained a computer algorithm using multiple recordings of both healthy and degraded coral reefs.  This essentially taught the computer to learn the difference between them.  A computer can pick up patterns that are undetectable to the human year.  This application of artificial intelligence can tell us faster and more accurately how a reef is doing.

The computer was then used to analyze a set of new recordings, and successfully identified reef health 92% of the time.  The team then was able to use this technique to track the progress of reef restoration efforts.

It is generally much cheaper and easier to deploy an underwater hydrophone on a reef and leave it there instead of having expert divers make repeated visits to a reef to survey its status.  Sound recorders and artificial intelligence could be used around the world to monitor the health of coral reefs and determine whether efforts to protect and restore them are working.

**********

Web Links

AI learns coral reef “song”

Photo, posted January 11, 2015, courtesy of Falco Ermert via Flickr.

Earth Wise is a production of WAMC Northeast Public Radio.

Deep Learning And Dirty Air | Earth Wise

By Leave a Comment

Poor air quality is a major global problem.  According to the World Health Organization, exposure to air pollution is linked to the premature deaths of an estimated seven million people every year.  In fact, 9 out of 10 people breathe air that contains more pollutants than what the WHO considers safe.  Air pollution is the fourth largest threat to human health, trailing only high blood pressure, dietary risks, and smoking.

But predicting pollution levels at a given place and time remains challenging.  According to a new study recently published in the journal Science of the Total Environment, scientists are turning to deep learning to improve air quality estimates. 

According to researchers, satellite observations and ground observations both measure air pollution, but both have major limitations.  For example, satellites may collect data at the same time and at the same location each day, but they miss how emissions may vary throughout the day.  Ground-based observations from weather stations do continuously collect data, but they only do so in a limited number of locations.    

As a result, scientists have turned to deep learning – a type of machine learning – to analyze the relationship between satellite and ground-based observations of nitrogen dioxide around Los Angeles.  Nitrogen dioxide is associated with emissions from traffic and power plants.  The researchers were able to rely on the learned relationship to take daily satellite observations and create hourly estimates of atmospheric nitrogen dioxide levels in approximately three mile grids.     

According to the research team, this study could be repeated for other greenhouse gases, and applied to different cities and regions – or even whole continents. 

**********

Web Links

Scientists turn to deep learning to improve air quality forecasts

Air Pollution

Photo, posted November 4, 2019, courtesy of Ninara via Flickr.

Earth Wise is a production of WAMC Northeast Public Radio.

Climate Change And Harsh Winters

By Leave a Comment

In recent years, there have been some unusually harsh winters in North America and Central Europe. This past January, the Midwestern US experienced extreme cold temperatures. We have all become familiar with the term ‘polar vortex’ and its role in sending cold air to middle latitudes and it is generally agreed that unusual behavior by the jet stream is the primary cause of the extreme winter weather.

For years, climate researchers around the world have been investigating the question as to whether the increasingly common wandering of the jet stream is a product of climate change or is a random phenomenon associated with natural variations in the climate system.

The jet stream is a powerful band of westerly winds over the middle latitudes that push major weather systems from west to east.  These days, the jet stream is increasingly faltering.  Instead of blowing along a straight course parallel to the equator, it sweeps across the Northern Hemisphere in massive waves, producing unusual intrusions of Arctic air into the middle latitudes.

Atmospheric researchers at the Alfred Wegener Institute in Germany have now developed a climate model that can accurately predict the frequently observed winding course of the jet stream.  The breakthrough combines their global climate model with a new machine learning algorithm on ozone chemistry.  Using their new combined model, they can now show that the jet stream’s wavelike course in winter and subsequent extreme weather outbreaks are the direct result of climate change.  The changes in the jet stream are to a great extent caused by the decline in Arctic sea ice, according to the results of the investigations.  The results are not surprising but there is now a detailed model to support the hypothesis.

**********

Web Links

A warming Arctic produces weather extremes in our latitudes

Photo, posted January 11, 2011, courtesy of Carl Wycoff via Flickr.

Earth Wise is a production of WAMC Northeast Public Radio.

Fish And Ships

By Leave a Comment

Solutions to overfishing of certain tuna and shark populations have been hindered by some significant unknowns:  where the fishing is happening, and where the fish are.  But researchers from Stanford University have recently shed some light on this mystery. 

According to a paper recently published in the journal Science Advances, the researchers have developed a map of shark, tuna, and ship movements that could help ocean managers identify regions of the sea where vulnerable species may be at risk. 

The researchers’ work builds on a 2018 study in which four years of satellite vessel movement data was used to develop a machine learning algorithm that mapped the footprint of 70,000 different fishing vessels around the world.  In their paper, the researchers zeroed in on the activities of 900 vessels from 12 countries in the northeast Pacific Ocean to figure out to what degree fishing fleets, sharks, and tunas overlapped. 

The researchers combined the vessel data with the ocean habitat preferences of sharks and tunas obtained from a decade-long tracking program called Tagging of Pacific Predators (or TOPP).  According to the IUCN, most of the 876 individuals tagged in TOPP belong to species that are either threatened or near-threatened.

By synthesizing this data, researchers were able to map where sharks and tunas would have the highest overlap with commercial fishing vessels.  Increasing the transparency of where fish meets fleets will allow ocean managers to determine where international protections may be needed. 

The United Nations is currently developing the world’s first legally binding treaty to protect international waters.  The Stanford University researchers hope their findings can help with this treaty’s formulation. 

**********

Web Links

Tunas, sharks and ships at sea

Photo, posted June 20, 2011, courtesy of Mike Baird via Flickr.

Earth Wise is a production of WAMC Northeast Public Radio.

A Breakthrough In Animal Identification

By Leave a Comment

Researchers from the University of Wyoming have developed a computer model that can identify wild animals in camera-trap photographs with remarkable accuracy and efficiency.

This breakthrough in artificial intelligence (AI), detailed in a paper recently published in the scientific journal Methods in Ecology and Evolution, represents a significant advancement in the study and conservation of wildlife. According to the paper’s authors, “the ability to rapidly identify millions of images from camera traps can fundamentally change the way ecologists design and implement wildlife studies.”

This study builds on previous research from the university in which a computer model analyzed 3.2 million images captured by camera traps in Africa.  The A-I technique called deep learning categorized animal images at a 96.6% accuracy rate.  This was the same accuracy rate as teams of human volunteers achieved, but the computer model worked at a much more rapid pace. 

In the latest study, UW researchers trained a deep neural network on a powerful computer cluster to classify wildlife species using 3.37 million camera-trap images of 27 different animal species.  The model was tested on nearly 375,000 images at a rate of about 2,000 images per minute. It achieved a 97.6% accuracy rate, which is likely the highest accuracy to date in using machine learning for wildlife image classification. 

Artificial intelligence has been used in environmental science in other ways as well. For example, AI has been used to increase agricultural yields in farm fields and to help predict extreme weather. 

Maybe artificial intelligence can prove to be a game changer for the environment.   


**********

Web Links

Researchers Successfully Train Computers to Identify Animals in Photos

Photo, posted January 8, 2012, courtesy of the U.S. Fish and Wildlife Service via Flickr. 

Earth Wise is a production of WAMC Northeast Public Radio.

Big Data + Technology = Improved Global Health

By

Scientists are calling for the creation of a global early warning system for infectious diseases. Such a system would use computer models to tap into environmental, epidemiological, and molecular data – gathering the intelligence needed to forecast where disease risk is high and what actions could prevent outbreaks or contain epidemics.

[Read more…] about Big Data + Technology = Improved Global Health

WAMC Northeast Public Radio

WAMC/Northeast Public Radio is a regional public radio network serving parts of seven northeastern states (more...)