Have A Question For Erik?

Sep 14, 2015 By Deepa Gopal
Deepa Gopal's picture

We hope you enjoy the series of articles by Erik. The first article is up this week, and you will see other articles coming up over the next two weeks.

It is a great opportunity for you to learn about the wonderful work that atmospheric scientists do, and what led them to follow their passion.

Perhaps you have a fascination for our natural world and may want to work in this area someday. Or you have questions on climate change and how scientists track what is causing the changing weather patterns. Well, here is your chance to ask Erik!

If you have any questions for Erik, be sure to ask him in the comment box below.

To add comments, log in or join
7 comments
4
Average: 4 (3 votes)

Comments

Pandas's picture
Pandas May 23, 2016 - 7:29am
How do fossil fuels affect our earth, and what can we do to stop them? How and why do fossil fuels release greenhouse gases?
4Bs's picture
4Bs February 6, 2016 - 10:03am

Do you know a lot about submarines?

vinays's picture
vinays November 10, 2015 - 3:29pm

yeah its right because if people ask questions it is so good

Arjun's picture
Arjun November 6, 2015 - 8:09pm

Are you using machine learning in your climate prediction models? Do you think machine learning could play an important role in climate modeling in the future?

larsoner's picture
larsoner November 10, 2015 - 12:22pm

Hi Arjun,
Good question. Machine learning is essentially computer algorithms or codes that identify patterns in data and can make predictions based on these. For instance, if you had an algorithm that predicted what you were going to eat next, it would track your eating habits and might correctly predict you would eat jelly with peanut butter, because you have done that in the past. Similarly, machine learning can be used to recognize patterns in weather and climate that may help improve predictions and forecasts. Machine learning is starting to be used to make climate predictions, however it is still not the base of the models. This is partly due to the nature of climate models which like to use physical equations to make predictions. For example, a machine learning code could recognize that after a hurricane the surface ocean temperature changes and then include that effect in a forecast. This could improve your forecast but it does not explain WHY the ocean temperature changed. Most scientists are very interested in understanding the cause of that temperature change, and thus use models that include equations based on known physics to try and understand and predict that temperature change.

I think machine learning definitely has a role to play in the future of climate modeling. However, I think it will always be complimentary to physical based models. Machine learning has a lot of potential to identify climate patterns that physical based climate models do not do a great job of predicting. That would point to climate processes that require further study.

Hope that answers your question. Take care,
Erik

diamond's picture
diamond October 6, 2015 - 2:16pm
I enjoyed learning about how weather and climate change is predicted. But I have sometimes wondered... How much do climatologists rely on past data (ice cores, etc) to model climate change and the factors that influence the earth's long term changes? How are the elements of the climate change models determined (since there are so many anthropogenic factors affecting the climate)? Do models constantly change with more available information? How difficult is it to predict future climate change, especially since things we do now affect the future and we may not the full impact of all our actions - example greener cars or ocean pollution?
larsoner's picture
larsoner October 15, 2015 - 12:42am

Hi Diamond,

Good questions. Climatologists use data from past climates to validate their models and put climate change in perspective. For instance, the earth was likely a few degrees warmer about 3 million years ago. This warm period also had higher concentrations of CO2, consistent with our understanding of greenhouse gasses and climate.

Climate models are constantly changing and being updated as computers become more powerful, new data becomes available, and algorithms become more sophisticated. Climate modelers try to add every element they can identify as important to the climate into climate models. State of the art climate models calculate properties about the atmosphere, land, ocean, and sea ice. They include biological carbon cycles as well. These models try to calculate the effects from all sorts of different anthropogenic sources of climate change including greenhouse gasses like CO2, methane, and CFCs. They include changes to Earth's albedo (surface reflectance) do to land use changes. They also include aerosol emissions. State of the art climate models include a tremendous amount of physics and chemistry.

When it comes to predicting climate change, the human element is the hardest part. We think we have a pretty good understanding of the physics that controls climate change. There are some things that we know we don't understand that well, such as how aerosols (small particles) affect cloud formation. Since we're aware that they affect clouds, at least we can try to estimate some uncertainty due to this unknown. However, the human aspect is much harder to predict. Global emissions of greenhouse gasses and aerosols (smog) are governed by consumer decisions, government policies, global economics, etc. These are harder to predict. For this reason scientists often run different simulations of the future with different concentrations of global emissions. In this way we can predict a range of possible climates depending on our actions. This range of possible climates also indicates how much we need to cut emissions if we want to stabilize the climate at a certain temperature.

Hope this answers your questions.
Erik