A Berkeley Scientist Explains Why It’s So Hard To Predict The Weather

Tech Industry

Why aren’t we more accurate at predicting the weather? originally appeared on Quora: the place to gain and share knowledge, empowering people to learn from others and better understand the world.

Answer by Dr. Andrew Schwartz, Lead Scientist at UC Berkeley Central Sierra Snow Laboratory, on Quora:

In a phrase, quantum computing. The weather models that have been developed using physical equations actually represent the atmosphere exceedingly well and they can do a terrific job of predicting the weather that actually occurs. But, there’s a trade-off with these models that most people aren’t aware of – they can either predict highly-accurate weather on small scales (think exact thunderstorm position) for a short period or less accurate weather on large scales and longer times (weather patterns for the next two weeks). The reason is that they are incredibly computationally expensive.

Weather models work by simulating specific points over the Earth’s surface at many different elevations and they need to do it for every time that we’d like to forecast. They need to do billions of calculations every day and they use some of the largest supercomputers in the world to do it. In fact, it takes more computing power to run a weather/climate model than it does to model the entire development of the universe to this point in time, and weather models run several times each day!


Ultimately, the reason why we have short-range and long-range models comes down to our limits of computing power. For instance, if you want to simulate the weather of a specific area but have a limit of 20 calculations, you can do 4 points in space for 5 times (further into the future) or 10 points in space for 2 times (more detailed). The same is true for weather models. We have a limit in our current computing power even with the biggest of supercomputers that won’t be resolved until quantum computing is available. As such, you can either simulate a small area with a lot of data points for a short time that will give you exact positions of weather features or you can spread out those data points and have a forecast over a longer period.

Now, computing power isn’t the only limitation to our current weather prediction abilities but it is the biggest issue that we face right now. If you’re interested in looking at the different strengths of different model types, have a look at the High Resolution Rapid Refresh (HRRR) model that is often used for severe weather forecasting because it’s very detailed for the near future and then look at the Global Forecast System (GFS) model that is less detailed over a longer time.

HRRR Pivotal Weather

GFS Pivotal Weather

This question originally appeared on Quora – the place to gain and share knowledge, empowering people to learn from others and better understand the world.


Leave a Reply

Your email address will not be published.