Biodiversity Forecasting Challenge

Predict future species abundance based on historical data

About this challenge ▸ About this challenge ▾

This is an experiment to understand how well machine learning models compare to human intuition. It is also a great tool to introduce interested people to the perks of predicting insect population trends. As you will see, it is not an easy task, even for the longer time series. This is because insects show large year-to-year fluctuations. Knowledge about the species behaviour and environmental variables can help make predictions, but some of the time series are hard to predict even for well-trained models. Thanks for playing with us and trying to beat us at predicting biodiversity trends!

Species information: Descriptions are generated with the support of artificial intelligence and reviewed by ecologists involved in research.

This is an initiative led by EBD-CSIC and UPM. The modeling framework used is developed by UPM and uses Reservoir Computing. The data comes from the BIOTIME database, a public repository of biodiversity time series.

EBD-CSIC UPM

Loading case study...

Instructions

  1. Look at the following time-series graph showing a historycal trend.
  2. Use the context green buttons to learn additional information (optional).
  3. Use the sidebar to predict the future value on the graph.
  4. Submit your prediction to see how you performed.

Time Series:

Time Series

Predict the next year Predicting the trend in 5 years

Context:

Location

Species Information

Temperature Data

Precipitation Data

Hover over the blue point and drag it up/down to make your prediction

Make Your Prediction

Drag the blue prediction point on the chart up or down to adjust your forecast

Current Prediction:

Results

Prediction Error
Your Prediction
Machine Learning Prediction
Other Users Average
Real Value