Forests influence climate change by storing C (carbon) that could alternatively be held in the atmosphere as CO2 (carbon dioxide)(1). However, forests are also influenced by climate change, as temperatures control tree physiology (2). If this feedback is negative and more C is stored in forests with warming air temperatures, forests mitigate climate change. If the feedback after a time lag is positive, global warming will continue despite a complete cessation of direct anthropogenic emissions. Knowledge concerning the impact of warming temperatures on forests is necessary for quantifying global warming caused by a given anthropogenic emission level.
Multiple methodologies are available to empirically study the impact of changing temperatures on forests. Perhaps the most straightforward conceptual approach is to simply document how growth or old-growth biomass has changed in recent decades and then assume the same trend to continue in the future. Canadian inventory data does not indicate a general change in trunk basal area growth (3). Similarly, height growth in Finland seems to have increased very little or not at all (4). In the Amazon rain forests, biomass growth increased in the 1990s but has since levelled off (5).
In addition to changing temperatures, changing CO2 concentration, N (nitrogen) deposition and other factors, such as changing precipitation, have influenced the long-term trends in tree growth. Disentangling their impacts is possible by focusing on year-to-year variability, e.g. based on annually repeated diameter measurements, tree rings or dendrometers. Diameter growth has been faster during warmer years in northern Finland (6). In contrary, diameters grew less in lowland Costa Rica during years with warmer diurnal minimum temperatures (7). In line with these local studies a global dataset reported positive correlation between temperature and tree ring widths in northern boreal but negative in other biomes (8). Even when focusing on year-to-year growth variation, the temperature signal may be masked by covarying meteorological variables, such as cloudiness, or by time lags e.g. if trees build reserves during good years instead of adding to their growth or if e.g. waterlogging kills deeper roots limiting growth over following years. These challenges can be avoided with long-term experiments. Warming all the tissues of large trees is practically impossible but warming of the soil in a temperate forest increased biomass growth (9), and whole-plant warming of potted tropical trees had mixed impacts on their growth (10).
Results on tree growth are valuable but insufficient to understanding the forest-warming-feedback for which stored C matters, as mortality is likely to change as well (11). For example, if mortality increases more than growth, a warming-triggered boost in growth will lead to educed biomass. Therefore, studies reporting changes in biomasses are especially valuable. Unfortunately, most forests are typically successional, even in remote areas in biomes influenced by large-scale disturbances, such as fires, and are far from an old-growth state in which biomass can be assumed to remain unaltered without global change impacts. However, a biomass change in remote regions or tropical rain forests lacking fires can be assumed to be naturally steady-state and any change can be attributed to global change. Forests in the Amazon basin (5), equatorial Africa (12), and Borneo (13) have all been shown to increase in biomass, but as with studies regarding growth, disentangling the impact of warming temperatures from other global change drivers is difficult. Furthermore, measuring large trees is surprisingly challenging (14), and even minor inaccuracies may significantly bias subtle trends (15). An additional challenge is that a seemingly stable old-growth forests may be actually recovering e.g. from anthropogenic disturbances that occurred centuries ago or from more recent natural disturbances such as drought (16).
The above-presented empirical approaches have provided valuable views from numerous perspectives, but overall understanding unfortunately remains blurred and even the direction of global forest biomass change caused by warming is questioned. Furthermore, all these empirical approaches neglect longer time lags in future biomass change, which are important in relation to precipitation (17). For example, an increasing temperature may weaken and kill the dominant tree species adapted to colder temperatures (18) before a species adapted to the new warmer climate is able to accumulate biomass above the initial level. If the new species were already present in the forest, the transitional phase could last for decades, but would take longer if tree migration is involved. These challenges suggest that modelling may significantly improve our understanding concerning the long-term impacts of global warming on forest biomass from that based on empirical research.
Models on global biomass patterns range in the levels that they are based on physiological processes determining biomass rather than the direct statistical estimation of biomass. Direct statistical modelling is not well suited for global future biomass modelling, as some future climates are currently found nowhere on Earth (19). Models including physiological processes vary greatly in their complexity. Earth system models, which attempt to quantify all major processes in the ecosphere, are extremely complex and therefore potentially very accurate at predicting future changes. Unfortunately, they are not good at predicting current AGB (old-growth above-ground biomass) variation (15), and their complexity makes open discussion on their structure and assumptions difficult. There is clearly a substantial need to model current global AGB and apply the same approach to predicting long-term future changes with a simple physiological model.
The normal way to understand biomass based on physiological processes is grounded on the production and decay of materials (20). Based on this thinking, exceptionally high biomass can result from large production or long residence time, i.e. turnover, of the material produced, or both. The focal “material” ranges from focusing only on woody biomass (21) to all ecosystem C (22). When applied to the AGB:
where NPP is above-ground net primary production and D is its decay, i.e. the reciprocal of turnover. NPP can be quantified with field measurements and understood based on physiological theories. However, even though D is intuitively reasonable, it is difficult to explain physiologically or make guesses about its climate-driven variation, unlike when focusing on dead tissue that decays predictably (23). In general, D increases with decreasing size, as small trees have a smaller proportion of biomass stored in long-term pools such as trunks. However, plants vary also in how often they replace leaves and in a given climate. Deciduous trees have both higher NPP and D than evergreens, possibly resulting in the same AGB. Because of these challenges, in practice, D is obtained statistically as the quotient of NPP and AGB.
Perhaps less intuitive, but better for numerous reasons, is focusing on the input and output of energy, i.e. chemical energy in both structural and non-structural material, instead of focusing on more permanent structures only. Maintaining living biomass consumes energy. As the input of energy in the ecosystem, GPP (gross primary productivity) does not vary much during the succession after canopy closure (24), and there must be a maximum biomass that the available energy can support and that can be estimated based on GPP and maintenance cost. In this energetic approach, Eq. 1 is modified to:
where b is a parameter and MCB is the maintenance cost per unit biomass, not only including autotrophic respiration but also heterotrophic respiration resulting from the turnover of tree parts and mortality (25). Parameter b was added, as autotrophic respiration does not increase linearly with increasing tree size (26), and was parametrised to 0.4 with a global data set (25, 27). In practice, similarly as with D in Eq. 1, MCB is best understood as the quotient of GPP and AGBb, and this understanding can then be used to model AGB. MCB in Eq. 2 can be understood better based on temperatures than D in Eq. 2 as it is composed in addition to the challenging heterotrophic respiration of autotrophic respiration with a well know temperature dependency (28) and data on GPP are readily available from eddy flux towers and can be modelled straightforwardly based on temperatures.
Water is crucial for trees, and many global patterns are well explained by annual precipitation (17), and its influence on AGB is certainly important (29) and probably increasing in importance (8). However, it is challenging to incorporate water into AGB modelling, as trees may obtain water from deep layers of the soil (30) and therefore even a dry climate may therefore be conducive to tree growth. Even reaching the water tables of forested regions of the world seems to be biomechanically easy compared to building massive trunks upwards that resist winds and gravity. Instead, much of the precipitation-impact on AGB could be caused by the covarying vapour-pressure deficit or even more indirectly via wildfires that increase with dryness (31), further complicating quantification of the mechanisms.
The energetic approach (Eq. 2) explained well the current global AGB variation in humid forests around the world (25, 27). As the approach is temperature driven, it is ideally suited to approximate long-term future changes in the AGB of humid forests. Our objective was to quantify changes in AGB caused by climate change in humid regions of the world based on this energetic approach and to determine the causes of these changes.