DOE/LANL Jurisdiction Fire Danger Rating:
  1. LANL Home
  2. Media
  3. Newsletters
  4. STE Highlights
August 26, 2025

Los Alamos scientists win American Statistical Society award for computer model analysis method

This approach offers a new way to examine how two models behave together

Feature statistics
Clustering of 14 computer models based on co-active subspace analysis illustrates the relative similarity or dissimilarity of their input-response behavior. Credit to: Technometrics. Reproduction of this image may only be used for government purposes.

Los Alamos researchers Scott Vander Wiel, Kellin Rumsey, Zachary Hardy and Cory Ahrens (now of Pacific Northwest National Laboratory) demonstrated a method to help make better use of computer models to simulate physical systems. Their work won the 2025 Statistics in Physical and Engineering Sciences Award.

Given by the American Statistical Association, the award recognizes individuals or teams for their innovative use of statistics to solve high-impact problems.

Why this matters: Sometimes scientists use multiple computer models at the same time, each based on different physical assumptions or with their own attributes. For example, in experimental design, several different materials may be under consideration and need evaluation. In this demonstration, the Los Alamos team guided scientific decisions about a set of simulated high-explosive experiments based on information about the alignment between models.

What they did: Lab statisticians used sophisticated machine learning techniques to estimate (or learn) the co-active subspaces of different computer models. The team developed a method that compares how computer models will react to changes in their inputs in relation to one another — specifically, whether the models respond to inputs in similar or different ways. 

How it works: Using this co-active subspace method, scientists can: 

·      Identify which set of physical experiments will provide the most information about a physical system. 

·      Validate a new physical system by comparing it with trusted benchmarks, highlighting both alignment and areas that may need further investigation.

·      Guide the design of more efficient simulations, saving time and reducing demands on high-performance computing resources.

Funding: The Engineering and Technology Maturation Program and Weapons Program Support – Design for Manufacturing at Los Alamos National Laboratory.

The paper: Co-Active Subspace Methods for the Joint Analysis of Adjacent Computer Models

LA-UR-25-28707

Share

Stay up to date
Subscribe to Stay Informed of Recent Science, Technology and Engineering Highlights from LANL
Subscribe Now

More STE Highlights Stories

STE Highlights Home
Thumbnail Plus

New Plutonium Science Laboratory serves critical national security missions

New center prepares for fiscal year 2026 operations

Thumbnail Uranium

Unlocking uranium’s hidden influence in the behavior of plutonium alloys

Work advances understanding of nuclear materials performance

Thumbnail Mpa

Why stretched graphene conducts heat on metal surfaces

Study aids efforts for advanced heat management in next-generation coatings

Thumbnail Payload

Two ions with similar masses won’t sneak past this new space instrument

Advancing technology for space weather forecasting

Thumbnail Braido

Braido honored for accelerator control system breakthroughs

Given award at international conference

Thumbnail Soot

Coated black carbon finding makes Earth system models much more accurate

Although water repellent by nature, soot can uptake water as it ages