Labeling the Harvard Forest Wood Image Collection for AI: Towards an Improved Estimation of Land Carbon Sequestration and Better Climate Reconstructions
- HF-ID: HF2019-18
- Category: Research Files
- Creator: Rademacher, T., Seyed-Nasrollah, B., Basler, D.
- Associated Person: Mandra, T.
- Date: January 1, 2020 to March 31, 2021
- Location: Harvard Forest
- Media: Paper
- Contents:
- Microsoft AI for Earth Azure Grant #342.
- HU GMAS #7871085-01. Nickname: Tree-Ring Labeling. 1 File Folder. 8.5 inches x 11 inches.
- Keywords: carbon cycling, climate change, tree rings
- Abstract:
- Tree rings are an important proxy for climate reconstructions over thousands of years and invaluable resources to improve our understanding of wood formations, hence carbon sequestration in forests. We propose to digitize and label one of the most diverse collections of scientific wood samples in the world housed at The Harvard Forest and
- make the labelled data product available publicly. The data set will provide suitable training data necessary to automate crucial and time-consuming parts of the analysis process such as detection and measurement of tree rings, species recognition, or crossdating. Detecting tree rings remains a time-consuming job executed by specialists, which limits the sample size of previous studies. The Harvard Forest scientific wood collection is particularly well suited, due to their sufficient but manageable size and unrivaled diversity. The characteristics of the collection will make it a key resource to tackle the major challenges in tree and climate science. Expected breakthroughs in the
- visual processing of scientific wood samples are very likely to lead to improved climate reconstructions and a better understanding of carbon sequestration in wood.
- Archives Location: Middle Room, stack 16, drawer 4
- Access: Active