Article

A New, Safer Way to Monitor Bear Reproduction

Computer models trained on bear collar data can accurately predict the presence or absence of cubs.


About this article

This article was first published online on February 27, 2026, as part of the Picturing the Unseen series.


A graph with many gray lines and one black line depicting modeled relationship between the accuracy of identifying cub presence and how many days it's been since an adult female bear left the den.
Computer models trained on real-world tracking data helped researchers find out how accurately they could tell if a female bear had cubs, based on her movements after she left the den. Gray lines show 100 trials. Black line shows median accuracy. See Fig. 3, Anderson and others, for details.

Anderson and others. 2025. Identifying presence or absence of grizzly and polar bear cubs from the movements of adult females with machine learning. Movement Ecology 13: 48.

They’re super cute in pictures, but you don’t want to get too close to bear cubs, especially since Mom is usually nearby. Yet some scientists have to do just that. That’s because information on bear reproduction is key to understanding their population trends. For some bears, like grizzlies and polar bears, this means monitoring how many cubs are present. Observing polar bears directly is especially dangerous, because they remain active in winter. They den for long periods only when pregnant. To collect reproductive data, scientists often watch polar bears long after they’ve left the den, a method that has its own shortcomings.

In a July 2025 study in Movement Ecology, researchers gathered location and movement data from tracking collars placed on female grizzly and polar bears. The data were collected over time periods of different lengths after den departure. The scientists used some of the data to train computer models called support vector machines (SVMs). They tested the accuracy of the SVMs by asking them to predict results for the remaining data. The models were up to 93 percent accurate in determining if a bear had cubs.

The scientists conclude that measuring bear reproduction remotely this way would improve estimates of reproductive success. When combined with other data, like environmental conditions, we can learn what helps or harms young bears. The method also worked well with older, less detailed data, enabling researchers to compare historic to modern information. Similar models could be used to study animals other than bears.


Republishing this article

This article is available for unedited republication, free of charge. We ask that you use the following credit: Originally published as “A New, Safer Way to Monitor Bear Reproduction on nps.gov, February 27, 2026.

Important: Some images may be copyrighted. Permission from the copyright holder is needed to reuse these. Check image credits for details.

Please let us know if you republish this article or have any questions.


Cite this article

National Park Service. 2026. "A New, Safer Way to Monitor Bear Reproduction." First published online on February 27, 2026. https://www.nps.gov/articles/000/psv40n1rr_a-new-safer-way-to-monitor-bear-reproduction.htm


Part of a series of articles titled Picturing the Unseen.

Last updated: February 28, 2026