A major task of our society is to manage forests in a way that their resources are preserved to meet future generation needs (Forest Europe et al., 2015). Current scenarios of climate change effects are making this task extremely challenging (Kirilenko and Sedjo, 2007). Climate shifts will impact forest vitality and affect goods and services forests provide, including carbon sequestration and climate change mitigation (IPCC, 2014). To guide sustainable forest management, forest researchers are asked to provide concrete answers about forest resilience in response to expected climatic trends, and extreme climatic events (Lindner et al., 2014). This is not an easy task, because responses of trees and forest ecosystems to environmental conditions are often non-linear and moreover vary on spatial and temporal scales (Smith, 2011; Anderegg et al., 2012; Reichstein et al., 2013). For instance, although drought is one of the most frequent and widespread climatic extremes affecting forests worldwide (e.g., Allen et al., 2010), the assessment of its impact on future forests is currently under intense debate. Mechanisms behind tree growth and mortality are complex (McDowell et al., 2008, 2011; Fatichi et al., 2014; Anderegg et al., 2015; Meir et al., 2015). Besides strength or frequency of external factors, such as extreme events, also the tree's ability to resist and recover is relevant, which, in turn, is largely determined by intrinsic factors such as the tree's life stage, life history, and genetic characteristics.

In this paper, we advocate for a tree-centered approach. By providing an improved mechanistic understanding of physiological and growth responses of trees growing under various conditions we can define the tree's capacity to respond to external stress factors. This concept can valuably contribute to the debate on how to shape future forests toward resilient forest ecosystems.

Current spatiotemporal simulations on future forest growth responses to changing climate conditions are performed with dynamic global vegetation models (DGVMs; Wullschleger et al., 2014). These models—usually generalizing tree species as plant functional types (PFTs)—provide valuable descriptions of the evolution of natural vegetation at a grid cell level under several climate scenarios. Such approaches are powerful in assessing growth responses related to the interaction between vegetation and atmosphere (including anthropogenic impact). However, although first steps toward representation of tree species, size classes, and forest structure in a DGVM were recently made (e.g., Naudts et al., 2016) they often lack to explain the variability between and within species, and often do not adequately explain growth responses (Fatichi et al., 2014) under varying site conditions, and to climatic extremes (Anderegg et al., 2015). These aspects are however extremely relevant to evaluate plasticity of tree individuals and tree species and the resilience of forests under changing climatic conditions, especially considering changing frequencies and intensities of climatic extremes (Reyer et al., 2013).

The tree-centered approach proposed here considers the individual tree as main source of information for understanding variability in growth responses. Comprehensive investigations using well-selected trees growing under different environmental conditions foster a better understanding of projected large-scale forest responses to changing climate. In comparison to generalizing approaches using PFTs, the tree-centered approach yields information with less spatial coverage but with the potential to convey more details on specific tree responses to a given climatic factor. This knowledge complements other approaches and can for instance support forest managers in tree species and/or provenance selection to better prepare specific forest stands to cope with expected challenges.

The incentive for the tree-centered approach is gaining a process-based understanding on tree responses to changing environmental conditions on temporal scales varying from short-term responses to climatic extreme events to long time periods matching the life cycles of tree populations. This can be achieved through an ensemble of observational studies on a selection of trees from different species and life histories, growing in diverse settings (forest types, species composition, successional stages, management regimes), and exposed to different climates and extreme climatic events. Establishing such a model framework requires the following elements:

The COST Action STReESS—a 4-year European framework initiative to promote networking among researchers of several plant-research disciplines to study tree responses to extreme events—is demonstrating the potential of such an integrated bottom-up approach. Starting from an enhanced understanding of the physiological processes behind wood formation, the STReESS Action established a modular process-based approach, which will eventually result in a model framework for explaining tree responses to climate extremes (Figure 1).

The COST Action STReESS contributed to the following main interlinked elements of the concept along the causal path: environmental trigger—structure—function—performance.

There is a fundamental difference between generalized PFT-based approaches (i.e., DGVMs) and tree-centered approaches. While PFT-based approaches perform spatially explicit “scenarios” of future global responses, the interdisciplinary process-driven tree-centered approach has potential to also provide practical support for local management decisions based on a solid understanding of tree functioning under specific site conditions. The COST Action STReESS has improved our understanding on the variability of responses to climate trends and extreme events. After 4 years of collaboration, the consortium has collected indications of the usefulness of such an integrated approach with continuous “methodological” development, and creativity. Through the integration of monitoring studies (e.g., time series of dendrometers, wood formation, and forest inventories), dendrochronological approaches, manipulation experiments (e.g., induced drought stress), and process-based models (e.g., at the cell, plant, or vegetation level) there is potential to collect valuable characterization to build process-based tree models accounting for variability between species, provenances, sites, and climatic events. This will contribute to unraveling a large set of yet unanswered questions related to processes of tree mortality (e.g., McDowell et al., 2011) or (mal)-adaptation (e.g., Martinez-Meier et al., 2008). Such a process-based approach will help to reduce uncertainty on tree performance under future environmental conditions.

Actual plans include the extension of the twittering-tree network for further development of a near real-time detection of tree processes and environmental impacts (Steppe et al., 2016) as well as the extension of global data networks and harmonization of protocols for high-resolution growth measurements (e.g., dendrometer, xylogenesis).

Despite still many implementation challenges ahead, we believe that the tree-centered approach offers an additional opportunity to assess forest management sustainability at the profit of the whole society.

All authors developed the concept and structure of the manuscript. USK and PF wrote the first draft of the manuscript. EMRR developed and designed the figure in cooperation with PF and USK. AB, EMRR, KS, JG, and PC authors accomplished and checked the first version and read and approved the submitted version.