Disentangling Responses of boreal ecosystems to climate change

Evidence for large-scale boreal ecosystem change

Some of the fastest warming rates on earth occur in the northern hemisphere high-latitudes. As a result, we expect large changes in ecosystem functioning. Satellite observations reveal indeed that Northern hemisphere boreal forests are ‘greening’. Based on atmospheric CO2 records, we also know that Northern hemisphere land regions are  strong carbon sinks. Furthermore  the amplitude of mid and high latitude atmospheric CO2 records is increasing in the Northern hemisphere mid and high latitudes.  The amplitude increase reflects a substantial increase in seasonal vegetation carbon uptake and release. It is, however, not clear, whether this increase is caused by changes of boreal forest functioning alone. Other biomes and processes may also play a role. It is also not clear whether the increased amplitude parallels a net carbon sink. It could instead be caused by carbon neutral forest demographic changes, specifically disturbance regimes. Fire levels are unprecedented over recent years.

Possible nature of changes

The most likely overall driver for these observed changes is the strong temperature increase. Support for this mechanism comes from an analysis of the atmospheric CO2 drawdown in spring observed at Barrow, a high latitude observation site. This record correlates with spring temperature anomalies. Intriguingly this correlation weakens over the past decades. This suggests a recent change in vegetation functioning. On the other hand the correlation between between spring vegetation greenness and the atmospheric CO2 record at Barrow remains stable over time. Vegetation greenness is measured from space and is an indicator of vegetation productivity. Thus altogether the numerous pathways and feedbacks through which temperature, and associated changes in fires, droughts, insect outbreaks, and permafrost melting, affect boreal vegetation functioning needs better understanding.

Project aims

The aim of this PhD project is to gain deeper insights in the possible drivers behind observed changes in boreal forest functioning. Itwill focus on the following three drivers:

  1. Expansion of boreal forest cover. Temperature increases should cause a shift of the tree line, but very few studies have demonstrated this empirically and there are very few estimates of the rate of forest expansion in this region. How much forest expansion contributes to the net global carbon sink is also poorly known.
  2. Changes in forest disturbance regimes. Fire, droughts and insect attacks are important disturbances affecting boreal vegetation functioning, and are all likely increasing with global change. More frequent fires may lead to shifts towards more deciduous trees, while increases in any disturbance types will lead to decreases in mean forest age. Productivity of regrowing young forests is much higher than older forests, and shifts in forest age thus affect carbon uptake and loss.
  3. Changes in primary productivity or respiration of existing vegetation. Increases in season length and mean growing season temperature will increase both photosynthetic productivity and respiration, nonetheless we currently know neither the magnitude of these increases nor the balance between productivity and respiration.



The successful applicant will assess the contributions of the described drivers to boreal vegetation functioning. He/she will  complement remote sensing data  with on-the-ground forest plot data and tree ring data. He/she will furthermore use Landsat 5-9 datasets to quantify expansion of forests northwards (driver 1) connecting this to Land Surface Temperature trends. The project will aim to build a machine learning model (such as a gradient boosted tree or GAM) to predict forest expansion rate from pixel-wise temperature, DEM and other suitable remote sensing derived inputs. As part of the investigation into driver 2 the project will investigate existing algorithms Landsat-based Detection of Trends in Disturbance and Recovery (LandTrendr), and Breaks For Additive Season and Trend Monitor (BFAST), both of which have been developed for forest disturbance detection from Landsat data.

The project will also use highly detailed information on forest age, composition, biomass and productivity. The data are from over 30 thousand sites distributed across Quebec. It will use historical images from the same region to ground-truth the remote sensing data and train AI classification models. To study effects of temperature on productivity and respiration (driver 3) the student will combine various datasets. Specifically he/she will assess and compare the length of growing season, derived from optical Landsat 5-9 datasets, with tree ring data from Quebec and the global tree ring database (ITRDB). He/she will also use these data to assess long-term changes of tree responses to climate variation.