Climate Action: Making data widely available

In addition to conventional and traditional knowledge, there are many hidden resources on environmental change. These include private photographs, landscape paintings, limited circulation expedition reports, ships’ captains’ log books and harbor master records of harbor ice etc. Often, these resources extend the records on environmental change in the ArcticDefinitions of the Arctic vary according to environmental, geographical, political, cultural and scientific perspectives. Some scientists define the Arctic as areas having a high latitude, long winters, short, cool summers,... More back in time before satellites and sometimes before research stations were established.

We aim to:

• Discover examples of these old resources in text and/or image format and use breaking science technology to demonstrate some aspects of environmental change over long periods combining this data with conventional data.
• Increase awareness among station managers of ground breaking methods such as machine learning, what they can achieve and the prerequisites for using them.
• Produce a best practice scheme for the use of this methodology at ArcticDefinitions of the Arctic vary according to environmental, geographical, political, cultural and scientific perspectives. Some scientists define the Arctic as areas having a high latitude, long winters, short, cool summers,... More research stations.

This Work Package (WP 6) is led by AFRY who will utilize the latest technologies within machine learning (ML) and artificial intelligence (AI). The work package will benefit from and use the wide range of expertise from engineers within AFRY and also include contributions from master thesis projects:

• Computer Vision for Camera Trap Footage: Comparing classification with object detection, Github: here
• Automated Digitization and Summarization of Analog Archives: Comparing summaries made by GPT-3 and a human, Github: here
• Searching and Recommending Texts Related to Climate Change, Github: here
• Image Augmentation to Create Lower Quality Images for Training a YOLOv4 Object Detection Model, Github: here
• Deep Learning for Iceberg Detection in Satellite Images, Github: here

One of the initial tasks for the work package will be to work closely with station managers (WP2), Virtual Access (WP3) and scientists to retrieve historical archived data (e.g. derived from photos, paintings, maps and reports) and to identify and classify it as well as finding solutions to transfer and store relevant information needed to be processed. If necessary, the process may also use data retrieved from various modern technologies such as satellite and aerial images or other sensor data.

Below, the presentation from the mini-workshop on Artificial Intelligence that was held in June 2020.