Time series of vegetation indices derived from satellite imagery are useful in measuring vegetation response to climate warming in remote northern regions. These indices show that productivity is generally declining in the boreal forest, but it is unclear which components of boreal vegetation are driving these trends. We aimed to compare trends in the normalized difference vegetation index (NDVI) to forest growth and demographic data taken from a 10 ha mapped plot located in a spruce-dominated boreal peatland. We used microcores to quantify recent growth trends and tree census data to characterize mortality and recruitment rates of the three dominant tree species. We then compared spatial patterns in growth and demography to patterns in Landsat-derived maximum NDVI trends (1984-2019) in 78 pixels that fell within the plot. We found that NDVI trends were predominantly positive (i.e., greening) in spite of the ongoing loss of black spruce (the dominant species; 80% of stems) from the plot. The magnitude of these trends correlated positively with black spruce growth trends, but was also governed to a large extent by tree mortality and recruitment. Greening trends were weaker (lower slope) in areas with high larch mortality, and high turnover of spruce and birch, but stronger (higher slope) in areas with high larch recruitment. Larch dominance is currently low (~11% of stems), but it is increasing in abundance as permafrost thaw progresses and will likely have a substantial influence on future NDVI trends. Our results emphasize that NDVI trends in boreal peatlands can be positive even when the forest as a whole is in decline, and that the magnitude of trends can be strongly influenced by the demographics of uncommon species.

Boreal peatlands are frequently underlain by permafrost, which is thawing rapidly. A common ecological response to thaw is the conversion of raised forested plateaus to treeless wetlands, but unexplained spatial variation in responses, combined with a lack of stand‐level data, make it difficult to predict future trajectories of boreal forest composition and structure. We sought to characterize patterns and identify drivers of forest structure, composition, mortality and recruitment in a boreal peatland experiencing permafrost thaw. To do this, we established a large (10 ha) permanent forest plot (completed in 2014), located in the Northwest Territories, Canada, that includes 40,584 mapped and measured trees. In 2018, we conducted a comprehensive mortality and recruitment recensus. We also measured frost table depth, soil moisture, soil humification and organic layer thickness within the plot between 2012 and 2018, and used habitat association tests to link these variables to forest characteristics and dynamics. Forest composition and structure varied markedly throughout the plot and were strongly governed by patterns in permafrost presence and organic layer thickness. Overall, there was a net loss of trees from the plot at a rate of 0.7% year−1. Mortality of black spruce, the dominant tree species, was more than double that of recruitment and was strongly associated with permafrost thaw. In contrast, recruitment of larch was over four times greater than mortality, and occurred primarily in low‐lying, permafrost‐free wetlands with mineral soil near the surface. Synthesis. The trends in tree demography and underlying drivers suggest that spruce‐dominated permafrost plateaus will be converted into larch‐dominated wetlands as permafrost thaw progresses in boreal peatlands, particularly in areas where mineral soil is near the surface. In the longer term, thaw could increase the hydrologic connectivity of the landscape, resulting in widespread drainage and re‐vegetation by spruce, but we did not find evidence that this is occurring yet. Given the increasing rates of permafrost thaw, and positive feedbacks between thaw and forest change, we predict that larch abundance will continue to increase in boreal peatlands over the coming decades, leading to shifts in ecosystem function, wildlife habitat, albedo and snow dynamics.