3D geospatial mapping of Arctic permafrost carbon
Soil organic carbon (SOC) stored in the northern circumpolar permafrost region, so-called Arctic "permafrost carbon", has the potential to substantially amplify global warming. Although the northern circumpolar permafrost region makes up only 15% of the global soil area, it contains up to half of the global SOC pool and twice as much carbon as currently is in the atmosphere. At the same time, the Arctic is warming two to four times faster compared to the global average, causing the permafrost to thaw. There is a risk that substantial amounts of permafrost carbon are released into the atmosphere as greenhouse gases during this process. This makes permafrost carbon a potentially strong climate feedback that could further amplify global warming. However, considerable uncertainties persist in accurately mapping Arctic permafrost on a circumpolar scale, as previous approaches lack in capturing the high spatial and vertical variability of soil properties in the Arctic environment. Currently, only a few studies attempted to map permafrost carbon on a circumpolar scale, and they reported large uncertainties, particularly increasing with soil depth. While previous studies on modelling permafrost carbon focused on mapping its spatial heterogeneity, they lacked in capturing the complex vertical distribution of SOC. Furthermore, they often rely on discrete models to estimate the spatial variation. In this work, we applied digital soil mapping (DSM) to perform large-scale spatial predictions of SOC density across the Arctic at a high spatial resolution (30m). We investigated how a 3D DSM model compares to a conventional 2.5D DSM model, both in terms of predictions over depth and prediction accuracy. For that, we used a random forest machine learning algorithm to predict SOC density based on spatial covariates derived from a digital elevation model (DEM). This was applied for three large sub-regions in Alaska, Russia and Canada. We harmonized a Circum-Arctic Soil Permafrost Region (CASPeR) database from published datasets on soil profile observations, which we used as a reference for training and validation of the DSM models. Furthermore, we developed a modular Python package that enables large-scale predictions of soil properties, and facilitates easy integration and scalable computation of new covariate datasets. Results indicate that 3D DSM does not significantly improve predictions of SOC density compared to 2.5D DSM models. However, 3D DSM has distinct advantages (e.g., higher vertical resolution and preserved sample size), and we anticipate that integrating more spatial covariates (i.e., climate and land cover) could significantly improve the prediction accuracy. Our Python software provides an important foundation for scalable and advanced circumpolar mapping of permafrost carbon. Accurate depth-aggregated maps of permafrost carbon (SOC stocks) are crucial for validating and initializing Earth system models (ESMs) to reduce uncertainties in future climate projections.
Date
16.10.2024
Time
13:15–14:45 h
Place
- Bundesstr. 53, room 022/023
- Seminar Room 022/023, Ground Floor, Bundesstrasse 53, 20146 Hamburg, Hamburg