Persistent Identifier
|
doi:10.5683/SP3/TIRAXJ |
Publication Date
|
2024-07-18 |
Title
| Peat depth and carbon storage of the Hudson Bay Lowlands, Canada |
Author
| Li, Yiyao (McMaster University) - ORCID: 0009-0008-2271-7631
Han, Daorui (McMaster University) - ORCID: 0000-0002-1871-7056
Rogers, Cheryl (Toronto Metropolitan University) - ORCID: 0000-0003-2792-1128
Finkelstein, Sarah (University of Toronto) - ORCID: 0000-0002-8239-399X
Hararuk, Oleksandra (Northern Forestry Centre, Canadian Forest Service, Natural Resources Canada) - ORCID: 0000-0001-5694-6813
Waddington, James (McMaster University) - ORCID: 0000-0002-0317-7894
Barreto, Carlos (Ontario Ministry of Natural Resources and Forestry) - ORCID: 0000-0003-2859-021X
Mclaughlin, James (Ontario Ministry of Natural Resources and Forestry)
Snider, James (World Wildlife Fund Canada) - ORCID: 0000-0003-4500-8121
Gonsamo, Alemu (McMaster University) - ORCID: 0000-0002-2461-618X |
Point of Contact
|
Use email button above to contact.
Gonsamo, Alemu (McMaster University) |
Description
| This dataset contains ground-measured peat depth and maps with the spatial distribution of peat depth, carbon stock and uncertainty in the Hudson Bay Lowlands, Ontario, Canada. The ground-measured peat depth data was collected by a Russian-type peat corer from 32 sites in 3 groups spanning from 51oN to 55oN in July and September 2022. The version 1 maps were produced in the Remote Sensing Lab, McMaster University, on March 2024. To generate the peat depth map, we used 495 peat depth records from Ontario Ministry of Natural Resources and Forestry data archive, topographic information, long-term satellite observations of land surface temperature, greenness, and polarization signatures in Synthetic-Aperture Radar (SAR) as well as machine learning models. Data was trained using multi machine learning algorithms. Based on the Root Mean Squared Error (RMSE) derived from 10-fold cross-validation, four models were selected for further prediction, including Gradient Boosting Machine (GBM), Deep Learning, Distributed Random Forest (DRF) and Extreme Gradient Boosting (XGBoost). For the final peat depth map, a second-level “meta-learner’ called stacked regression was applied to find an optimal combination of the 4 base models. Here we used generalized linear model (GLM) during the stacking process to map peat depth for the entire HBL. The uncertainty was estimated as ± one standard deviation around the mean estimates of all base models. The carbon stock map was estimated based on empirical relationship between the estimated peat depth and C stock. (2024-06-24) |
Subject
| Earth and Environmental Sciences |
Keyword
| peat depth, soil carbon stock, Hudson Bay Lowlands |
Related Publication
| Li, Y., Han, D., Rogers, C.A., Finkelstein, S.A., Hararuk, O., Waddington, J.M., Barreto, C., McLaughlin, J.W., Snider, J., & Gonsamo, A. (2024). Peat depth and carbon storage of the Hudson Bay Lowlands, Canada. ESS Open Archive. June 13, 2024. DOI: 10.22541/essoar.171828456.66424229/v1 |
Depositor
| Gonsamo, Alemu |
Deposit Date
| 2024-06-24 |
Date of Collection
| Start Date: 2022-07 ; End Date: 2022-09 |