Random Walk of the Penguins

Competitors built hundreds of algorithms to predict changes in Antarctic penguin populations from the most comprehensive counts available. These algorithms give researchers a greater understanding of penguin population dynamics, a leading indicator of climate change. #climate

$16,000 in prizes
jun 2017
603 joined

alt-text

Not only did the competition demonstrate the use of algorithms not commonly applied in ecological studies, but it has also provided us with a suite of models that can be incorporated into ensemble model forecasts for the Antarctic conservation community.

– Dr. Grant Humphries, seabird biologist and lead web developer for MAPPPD

Why

Penguins - in addition to being the best dressed animals in the Antarctic - are important indicators of the health of their fragile ecosystem. Monitoring and modeling their population dynamics over time supports long-term management of the species and provides important information about broader environmental changes.

The Solution

Using data from MAPPPD, the most comprehensive counts of Antarctic penguin populations, competitors built models to predict changes in four species of penguins based on data about their environment. The competition allowed a massive number of alternative prediction approaches to be explored, unbiased by researchers’ prior expectations for how the system should work.

The Results

The winning models outperformed the top model previously created by the team of domain experts, and halved the error from the initial benchmark. These models suggested less signal from large scale environmental factors that can be measured from satellites, compared with fine-scale factors such as precipitation.

In addition to fueling published research, these findings provide for useful integrations of predictions into marine resource management.


RESULTS ANNOUNCEMENT + MEET THE WINNERS

WINNING MODELS ON GITHUB