Environmental change, including climate change, increasing atmospheric carbon dioxide (CO2) concentrations, nitrogen deposition, and other factors, will continue to influence inter-annual variation and trends in tree growth, mortality, and recruitment rates, influencing the sustainability of forest management in Canada over the next century and beyond. Forest management decision making is increasingly vulnerable to error because current growth and yield models do not adequately account for these environmental changes. The related forest management information needs are diverse and wide ranging. Management interventions to improve the carbon balance of forest stands, to increase the supply of biomass or wood volume to meet society’s needs, or to enable a faster recovery of forest habitats following disturbances, would all benefit from better forecasts of future forest establishment, growth, and mortality at the landscape level across very large areas. However, model inter-comparison studies of process models often find disagreement on the magnitude and direction of growth responses to climate change, and they receive little operational use, despite a long history of scientific development. As a result, there is currently a disconnect between scientific and operational forest management approaches to improving growth models. We undertook a review of the current capacity for environmentally and climate sensitive growth and yield modelling in Canada, and consulted broadly with scientists and forest management agencies across the country. Here we present the results of this review, including some recommended actions and research activities to further the development and future implementation of these environmentally sensitive growth models.

Boreal forest composition and structure is a result of interactions between disturbance, climate, and belowground conditions. The Functionally Assembled Terrestrial Ecosystem Simulator (FATES) is a cohort-based dynamic vegetation model that operates within an Earth System Model framework as a component of the Community Terrestrial Systems Model (CTSM). It assembles ecosystems dynamically through plant interactions, degradation and loss by natural and disturbance based mortality. Within the boreal region, previous models demonstrate that consideration of vegetation structure and vegetation-climate-soil linkages is necessary for successful simulation. The University of Virginia Forest Model Enhanced (UVAFME) is an individual tree-based gap model that has been validated within the North American boreal zone. UVAFME was recently updated to relate fire ignition and spread directly to litter and fire weather conditions, allowing for further interactions between vegetation, soils, fire, and climate. As UVAFME performs well in the boreal region and operates at the inherent scale of trees, we can use it to facilitate calibration and testing with FATES. We compare the performance of UVAFME and FATES in boreal North America, directly comparing tree/cohort response to environmental drivers by feeding CTSM-simulated vegetation drivers (e.g. temperature, soil moisture, active layer depth) into UVAFME. These results advance our understanding of the importance of scale in forest models, and how we may use a model that explicitly considers individual tree dynamics to benchmark coarser-scale models.

Tree growth is affected by climate and future growth variability will likely depend on how climate will change. Modeling tree growth using monthly climate data can easily result in tens or hundreds of individual input variables. Fitting conventional statistical models with so many variables is difficult or even impossible, even when interactions between variables are omitted. In contrast, artificial neural networks and deep learning have made great progress in recent times and are suitable for such complex analyses; though they require large amounts of training data.

Thus, to predict radial growth of white spruce across its range in North America, we first compiled a large data base with more than 800,000 tree-ring width measurements from 399 sites, as well as 80 climate variables for each individual tree-ring (4 climate variables √ó 20 month). We then trained an artificial neural network (implemented in Python with TensorFlow) to predict tree-ring width from the climate data. We found that the network did partially learn to generalize the effect of climate, tree size and age on tree-ring width. By using the network to estimate tree-ring width over the past decades, we found that on average climate change had a positive effect on white spruce growth. Though, additional data like tree height, stand density and competition are needed to quantify forest-level productivity changes.

While artificial neural networks typically do not offer the deep mechanistic understanding that agent-based models or process models provide, our results indicate that they do allow potentially highly accurate predictions which might for example be used to quantify growth-rates under different climate change scenarios.

Tree ring-width series play an important role in many research fields. There is variability in the radial growth of trees, whether there is a specific-direction climatic relationship is worthy of research. At present, only a few studies have investigated the climate sensitivity of different directions within trees. In our study, we sampled four cores from different sampling directions each tree. chronologies and the growth-climate relationships were studied in Larix olgensis. In order to study whether there are direction-specific growth variability and its impact on the relationship between climate growth, Chronologies for different sampling directions were built. Signal strength differences and the climate-growth relationship between different sampling directions cores were compared. The chronologies and radial growth trends are basically the same between different sampling directions. The signal strength differences between different sampling chronologies were very small, through the detection of PCGA and climate-growth relationship, no differences in sampling directions were found in Larix olgensis. For Larix olgensis in the plain forest, there are no direction-specific effects on radial growth variability, and the impact of sampling direction on the analysis of the climate-growth relationship is negligible. There was no significant difference in the climatic-growth relationship of different sampling direction cores in Larix olgensis.

Considerable evidence from field provenance trials suggests that climate change could impact tree and stand growth and mortality. However, many operational-level growth models lack climate sensitivity. In this study, a transfer function for survival and a supporting individual tree mortality model were developed using long-term, range-wide, repeatedly measured, provenance data for lodgepole pine (Pinus contorta var. latifola Douglas) in British Columbia, Canada. The models were formulated to work with Tree and Stand Simulator (TASS), an individual tree growth and yield model. The transfer function predicts population-specific percent survival as a function of the climate distance seed is transferred and the climate at the seed origin (provenance). The shape of the transfer function is dynamic over time on account of its use of top height. Population response to climate transfer was found to depend on the provenance climate. Transfer to slightly warmer climates was associated with increases in survival rates for populations from the coldest climates, whereas populations from mild or warmer climates were associated with reduced survival when transferred to warmer climates. This suggests that local populations at cold sites could fare better under climate warming than local populations at warm sites. Modifiers derived from the survival transfer function are used to alter stand-level survival rates in TASS. Ultimately, the goal is to use modifiers for both survival and growth to explore seedlot deployment strategies within the context of an uncertain future climate.

Empirical individual-based models (IBM) have become widely used in forest management planning over the last decades. Given the anticipated climate change, it is essential that these models be climate-sensitive. Some efforts have been made to model diameter and height increment as well as tree mortality in function of climate variables. However, the impacts of climate on tree recruitment, which is an essential component of IBMs, have only been recently addressed.

In this study, we modelled tree recruitment for 33 species in the Province of Québec, Canada, using abiotic predictors related to climate and topography as well as biotic predictors that accounted for competition and the species presence. We quantified the relative contribution of these predictors to the model fit. It turned out that the degree-days were a major driver of recruitment occurrence but not so much of recruitment abundance. The species presence and stand basal area were the most important biotic predictors of recruitment occurrence and abundance. Under RCP 4.5, an increase in the recruitment of balsam fir is to be expected by 2050 in the boreal and transition forest zones.