Using a Bayesian Network Framework to Predict Permafrost Thaw in the Arctic
Abstract
As temperatures continue to rise across the Arctic, the vulnerability of permafrost to thaw is expected to increase. This change to the state of the cryosphere is closely coupled to other landscape-scale patterns and processes, such as hydrological and vegetation changes. However, there are important limitations in our ability to simulate permafrost thaw under projected climatic and ecological conditions. Most permafrost models do not represent vegetation or its ability to modify ground insulation, and Arctic observational data are sparse. Here, we outline a new modeling framework using a Bayesian Network (BN) to simulate permafrost thaw in the continuous permafrost region of the Arctic to address some of these limitations. A key advantage of the BN is its capacity to link variables using cause-effect relationships that are informed by observational data, model inputs, and expert assessments. Since processes are represented probabilistically, computational time and model complexity are reduced. The BN can also track the uncertainty associated with each cause-effect relationship. To our knowledge, the creation of an Arctic BN has only been attempted once (Webster and McLaughlin 2014). Our model development process involved: 1) identifying variables relevant to permafrost thaw via literature review and collaboration with experts; 2) pre- and re-validating the model via expert assessment in both prognostic and diagnostic manners; and 3) calibrating the model using observational data. Results from the first two steps show promise of this approach, as we were able to consistently link increased air and soil temperatures and low soil moisture conditions with increased permafrost thaw in prognostic mode. In diagnostic mode, we were also able to consistently link these variables with increased permafrost thaw, but a medium warming scenario was favored for all thaw extents. The third step is still underway and involves comparing the model to an existing local case study (Wilcox et al. 2019) and a circum-Arctic statistical model (Aalto et al. 2018). The final results from this study are expected to provide less uncertain predictions of permafrost thaw that can then be applied to carbon modeling studies, infrastructure hazard assessments, and policy decisions aimed at mitigation of and adaptation to permafrost thaw.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2020
- Bibcode:
- 2020AGUFMB080.0003B
- Keywords:
-
- 0414 Biogeochemical cycles;
- processes;
- and modeling;
- BIOGEOSCIENCES;
- 0428 Carbon cycling;
- BIOGEOSCIENCES;
- 0439 Ecosystems;
- structure and dynamics;
- BIOGEOSCIENCES;
- 0475 Permafrost;
- cryosphere;
- and high-latitude processes;
- BIOGEOSCIENCES