Climate Modeling
Then: Fallible Forecasting
Now: State-of-the-Art Predictions
The suburban town of Cranford, New Jersey, inundated with severe flooding following heavy rains.
As climate change increasingly impacts human health and infrastructure, the importance of accurate climate models has never been greater. These models provide essential data to policymakers, enabling them to make informed decisions that can save lives and reduce property damage. Tripti Bhattacharya, the Thonis Family Professor of Paleoclimate Dynamics, discusses how technological advancements over the past 25 years have improved climate modeling and her predictions for the future of this field.
How have advancements in technology over the past 25 years enhanced scientists' ability to study ancient atmospheric conditions and in what ways have these advancements improved climate modeling and enhanced researchers' ability to predict future climate trends?
Tripti Bhattacharya (TB): Humans have always been interested in the past, and throughout human history, fossils of plants and animals have given clues about ancient landscapes and climates. In the past 25 years, advancements in instrumentation means we have better quantitative constraints on variables like temperature, rainfall, ecosystem structure and fire.
Simultaneously, computers have gotten more efficient, allowing us to run longer simulations of more complex climate models at higher resolution. Together, these advances allow us to use the past as a test bed for seeing how well our state-of-the-art climate models capture climate states that are unlike the present. When models reproduce the past, we have more confidence in their ability to simulate the future.
What are your predictions for the future of climate models over the next 25 years, and why will this be significant?
(TB): The next set of innovations in climate modeling will involve improving our ability to provide regional-scale climate predictions that are useful for adapting to our changing climate. We have tools like ‘variable resolution’ climate models that allow us to embed high resolution grid cells over particular regions, while keeping a course resolution globally. This reduces overall computational cost and allows us to capture both global and local influences on a region’s climate evolution. Moreover, machine learning will continue to be extremely useful as a way to test the way we represent different processes in our model code.
Overall, the goal of the future will be to provide reliable climate model simulations that serve human needs. Our models are overwhelmingly reliable at a global scale and agree about global trends in temperature and rainfall. However, the details of specific regional patterns of climate will benefit from innovations in modeling.
