Permafrost thaw is a key climate-related driving factor of change in the lowland discontinuous permafrost regions of the Taiga Plains; this thawing phenomena is transforming the spatial pattern and distribution of the dominant land cover types (permafrost plateaus, fens, and isolated bogs) and their hydrological properties. Many recent investigations have illustrated that regions of the Taiga Plains previously dominated by peat plateaus are transforming to those dominated by connected wetlands (fens) and isolated wetlands (bogs), and that this change has impacts upon streamflow. To address the effects of permafrost thaw on the transition of land covers in discontinuous permafrost regions, we developed a machine learning-based model which is able to estimate the evolution of the main hydrologically-important land cover types. The input data of the trained multinomial time series model (TSLCM) includes a set of spatio-temporal variables obtained by remote sensing: the estimated summertime land surface temperature anomaly (LST), the distance to land cover interfaces, time horizon (from 0 to 38 years), time-accumulated land surface temperature, and classified land cover maps from 1970-2008. The model can output historical or future spatial estimates of land cover distribution.
The devised TSLCM helps us to simulate historical land cover transitions, capture spatial patterns of change over time, and replicate the historical long-term evolution of land cover at the Scotty Creek Research Station (SCRS) in the Northwest Territories, and similar discontinuous permafrost landscapes. The model generates landcover maps which represent the spatial distribution of fen, bogs, and peat plateaus consistent with a default 50\%\ threshold applied on the predicted probability maps. We here use the TSLCM to simulate land cover change under multiple plausible futures scenarios by using a recent set of climate model projections.
The predicted time series land cover maps generated up to the year 2100 suggest that permafrost plateaus are rapidly transforming to fens and increase in the proportion of the landscape covered in fen is accelerating. In addition to examining uncertainty due to climate uncertainty, a probabilistic approach is used to sample the threshold value to generate a range of land cover realizations.
