Specific leaf area (SLA; m² g C⁻¹) is related to leaf longevity, representing the leaf economics spectrum (Wright et al., 2004). Plants across growth forms, biomes and climates show a similar range of feasible leaf investment strategies along a gradient from “acquisitive” leaves with high SLA but low leaf longevities, to “conservative” leaves with low SLA but increased leaf longevity and persistence in harsh environments (Díaz et al., 2016; Wright et al., 2004). The generalized functional relationship used in the TREED model represents an intermediate form of the leaf economic trade-off used for needle-leaved and broadleaved plants with prescribed leaf longevities in the LPJ dynamic vegetation model (Schaphoff et al., 2018; Fig. 2): 1SLA=2×10-4⋅1DMc×102.25-log⁡all⋅12 with DM_c being an assumed average dry matter carbon content of 0.47 g C per g of dry matter and all the leaf longevity in years. The generalized relationship implies that modelled plants will have leaf longevity and SLA combinations in between two endmember strategies: short leaf longevity/high SLA strategies, representing broadleaved deciduous or evergreen plants, or long leaf longevity/low SLA representing evergreen, needle-leaved plants (Fig.  2). For a given leaf carbon pool, a high SLA will result in a high leaf area and light interception potential but is associated with high annual carbon investments to rebuild the leaf carbon pool due to the short leaf longevity. Whether such a strategy is favourable over long leaf longevities and reduced annual leaf carbon investments, as well as the exact duration of the leaf longevity, depend on local climatic conditions (precipitation, radiation and temperature) and is evaluated in the optimization procedure (see Sect. 2.3). By affecting the carbon assimilation potential as well as annual carbon turnover costs, the leaf investment trade-off further contributes to determine whether an evergreen or a deciduous phenology (i.e., no leaves during part of the year) is more favourable under temperate climatic conditions (see Sect. 2.2.3). The consideration of a predicted and continuous range of possible leaf longevities and SLA strategies represents an important difference to plant functional type-based vegetation models with prescribed leaf traits.

Following the allometric scaling relationships from the LPJ vegetation model (Schaphoff et al., 2018; Sitch et al., 2003), the size of a plant individual's average leaf carbon pool (Cleaf) and SLA determine the average leaf area of a plant (LA; m²): 2LA=SLA⋅Cleaf Following the pipe model (Shinozaki et al., 1964; Waring et al., 1982), an individual's leaf area and sapwood cross-section area (SA; m²) are assumed proportional (all pre-defined model parameters are listed in Table 1): 3LA=kla:sa⋅SA Leaf and fine root carbon allocation (Cfineroot; g C per individual) are related by a fine root to leaf carbon ratio (r:s; dimensionless): 4Cleaf=r:s⋅Cfineroot Plant height and diameter (D; m) are assumed to relate according to (Huang et al., 1992): 5H=kallom2⋅Dkallom3 Further, diameter and crown area (CA; m²) are assumed to scale following: 6CA=min⁡kallom1⋅Dkrp,CAmax with an assumed maximum crown area (CA_max) of 15 m² (Sitch et al., 2003). The CA represents the average surface area occupied by a single average plant in a model grid cell.

An average plant's individual leaf area index (LAI_individual; m² m⁻²) can be calculated as: 7LAIindividual=Cleaf⋅SLACA It is assumed that grid cells are fully covered by the local average plant that is characterised by the five key traits (Fig. 1a). Assuming no canopy overlap, the number of plant individuals per grid cell (Nindividuals) is thus Nindividuals=cell areaCA.Consequently, the grid cell's leaf area index (LAI), is equal to the leaf area index of the average individual (LAI = LAI_individual): 8LAI=LAindividual⋅Nindividualscell area=LAindividualCA=LAIindividual The light interception potential within a grid cell, i.e., the fraction of photosynthetically active radiation absorbed by leaves (fAPAR; dimensionless), is estimated using the Lambert-Beer law: 9fAPAR=1-e-k⋅LAI Thereby, we account for a tendency of an increased light extinction coefficient (k) for broadleaved, short-lived and high SLA leaves compared to needle-leaved, long-lived and low SLA leaves (Zhang et al., 2014), by linearly relating k in the range between 0.6 and 0.4 to the average leaf longevity, according to: 10k=0.6-0.04⋅all A plant's sapwood carbon pool (Csapwood; g C per individual) is derived from the plant's sapwood cross section area and height, assuming a cylindrical geometry and a constant wood density (WD; g C m⁻³): 11Csapwood=SA⋅H⋅WD The model considers a globally generalized plant physiology, and it is therefore assumed that for small plants that would not be considered trees, Csapwood describes an overall structural carbon pool that is proportional to a plant's height.

Woody tissue represents the by far most important component of global above ground biomass (Brunner and Godbold, 2007). Therefore, a generalized power law derived from global canopy height (Lang et al., 2023) and above ground biomass density estimates (Huang et al., 2021; Santoro et al., 2021) is used to describe the relationship between the modelled volume of an average plant in a grid cell (H⋅CA) and total stem biomass (Cstem=Csapwood+Cheartwood): 12Cstem=ki⋅(CA⋅H)kp Coarse root carbon (Ccoarseroot; g C per individual) is assumed to be around a fourth of the stem biomass, as estimated from current above ground and below ground biomass data (Huang et al., 2021): 13Ccoarseroot=0.25⋅Cheartwood+Csapwood All allometric constants (kallom1,kallom2,kallom3,krp,kla:sa, ki, kp) were calibrated using present-day canopy height and above ground biomass data and are listed in Table 1. In the calibration procedure, we minimized the root mean square error (RMSE) between predicted above ground biomass density, derived using the present-day observed vegetation height and allometric Eqs. (5), (6) and (12), and the observed biomass density from Huang et al. (2021) and Santoro et al. (2021). The resulting height to biomass relationship is outlined in Sect. 5.3 and Fig. 6b.

The model's generalized plant physiology (Sect. 2.1) and carbon flux calculations (Sect. 2.2) allow to calculate a carbon balance for variable combinations of the traits Cleaf, H, all and phenology (deciduous or evergreen). Following optimality principles (Franklin et al., 2020; Harrison et al., 2021), we assume that competition, environmental filtering and natural selection will result in traits of an average individual at a location to evolve towards combinations that maximize fitness. We define fitness as the potential for height growth while maintaining a positive NCG. The choice of height as the fitness measure assumes that plants in a community are in continuous competition for light as a major limiting resource, resulting in increased height being associated with a competitive advantage. It is important to note that this approach represents a fitness measure at the community level and not a single-plant optimization, considering that carbon investments into structural tissues in the absence of competitors may not represent an optimal strategy. It can be assumed that height as fitness measure results in an approximation of an evolutionary stable strategy, resulting in a trait combination that is likely stable and dominant in a given climatic environment (Franklin et al., 2020; Valentine and Mäkelä, 2012). Height growth is a compound trait, and its optimization implies finding a leaf longevity, phenology and carbon allocation strategy that maximizes photosynthetic carbon gains over losses through growth and maintenance respiration, as well as tissue turnover carbon costs, all of which increase with increasing height. To predict the trait combination of Cleaf, H, all and phenology that maximize height subject to NCG≥0 under given climatic conditions, an Evolutionary Centers optimization algorithm from the Julia package Metaheuristics.jl (Mejía-de-Dios and Mezura-Montes, 2022) is employed, with the target function being: 49minimizefCleaf,H,all,phenology,climate=NCG-ω⋅H where ω(=10) is an arbitrary weighting parameter to ensure stable solutions. By minimizing NCG, the model will approximate a steady state vegetation at maximum height, i.e., the height at which carbon gains from photosynthesis are balanced by carbon costs from respiration and turnover. Whether height growth is maximized under a deciduous or evergreen phenology is only evaluated in locations with more than three months with average temperatures below 3 °C, which is considered a necessary condition for a deciduous phenology.

The TREED modules described under Sect. 2.1–2.3 can be used as a stand-alone model to obtain a prediction of vegetation that is in eco-evolutionary equilibrium with the environment, i.e., every grid cell is occupied by vegetation with a stable trait combination that maximizes height. Alternatively, the model can also be used to track vegetation structural and functional changes through time while adapting from a starting state and trait distribution to the predicted optimum state under new or changing climatic conditions. Two pathways of vegetation adaptation are represented: adaptation through evolution of traits and adaptation through dispersal and a redistribution of traits.

To evaluate the performance of the TREED model, we run one time step of the model using multi-year monthly average temperature, precipitation, downwelling shortwave radiation and cloud cover data from the years 1981–2010 from CHELSA (Karger et al., 2017) at 0.5° resolution in longitude and latitude. Derived carbon fluxes are evaluated using MODIS-derived GPP and NPP estimates, considering average fluxes from the years 2001–2010 (Kern, 2024). Vegetation height is compared to a combined satellite data and machine-learning based estimate for the year 2020 from Lang et al. (2023). Aboveground and belowground biomass carbon densities are compared to a machine-learning-based upscaling approach from field measurements and satellite data (Huang et al., 2021; Santoro et al., 2021). Evapotranspiration rates are compared to estimates from the Global Land Evaporation Amsterdam Model of the year 2010 (Miralles et al., 2011). Predicted latitudinal gradients of leaf longevity and leaf mass per area (LMA = 1/SLA) are compared to data-based trends derived from the Glopnet data set (Wang et al., 2023; Wright et al., 2004). To evaluate the TREED prediction of potential plant biodiversity, we compare the modelled diversity index to an ensemble prediction of global-scale plant species richness derived from observational records and extrapolation using machine learning and statistical methods by Cai et al. (2023).

The optimality-based prediction of vegetation functioning using the TREED model reasonably approximates the major distribution of photosynthetic carbon assimilation and evapotranspiration (Figs. 3, 4a–c). For NPP and GPP, similar spatial patterns in model biases are observed (Fig. 3a–f). A tendency for an overprediction of carbon assimilation is observed in the Afrotropics and Indomalayan tropical rain forest climates, as well as the subtropics of North and South America. A pronounced productivity underestimation is observed in low latitude regions associated with a pronounced seasonality in precipitation, including tropical savannah and arid steppe climates of Africa, South America or Southeast Asia. Biases in these regions are likely related to the limited temporal resolution of the model, using monthly average climate information. Thereby, the model may not capture seasonal climate extremes and higher frequency precipitation, temperature, and productivity variations in these regions. Biases are expected to further emerge from the simplified ecosystem structure in the model, with one community-representative plant modelled in each location. Thus, multi-layered vegetation structures of the tropics or the open woody-grassland savannah vegetation are represented in a simplified manner. However, globally, the errors tend to be well balanced with a RMSE of 408 g C m⁻² for GPP and 201 g C m⁻² for NPP. Further, the TREED steady-state prediction of the global NPP of 52 Pg C is in good correspondence with an estimated mean global NPP of 56.2 ± 14.3 (± standard deviation) from multiple data- and model-based NPP estimates compiled by Ito (2011). The estimated global GPP of 133 Pg C lays in the intermediate range of GPP estimates derived from remote sensing products by Stocker et al. (2019).

The model reproduces the global distribution of evapotranspiration, and thus reasonably approximates the carbon-water exchange during photosynthesis. However, there is a tendency for a slight but consistent underprediction of evapotranspiration rates (Fig. 4c), particularly in mid-latitude, temperate regions and the South American tropics (Fig. 3g–i). Given that carbon fluxes are well reproduced in the model, with more balanced biases, a possible explanation for this consistent underprediction is the lack of an explicit representation of non-photosynthesis related water fluxes (see Eq. 35), including bare soil evaporation or evaporation of water intercepted on vegetation surfaces.

The TREED model allows a reasonable approximation of the first-order distribution of vegetation height and biomass (Figs. 5, 4d–f). For vegetation height, the model generally results in an overly smooth height distribution and tends to overestimate vegetation height compared to the data of the year 2020. The height bias (Fig. 5a–c) needs to be understood in context of the model's underlying structure and bias in NPP predictions. In the model's optimization-based trait allocation (Eq. 49), a surplus NPP will result in a prediction of increased height and biomass growth, until respiratory and tissue turnover carbon investments exceed the surplus carbon from photosynthesis. Through this procedure, the model reproduces the first-order relationship between vegetation height and NPP derived from observational records (Fig. 6a). However, for regions with a positive bias in productivity, such as subtropical South America, the model will overpredict vegetation height. For regions with a small bias or an underprediction of NPP, an overprediction of vegetation height may be caused by an underestimation of carbon turnover processes that reduce the carbon availability for height growth. For example, some environments exhibiting pronounced height overpredictions such as tropical savannahs or the continental interiors of North America may be associated with frequent environmental disturbances, including fires, heat or cold conditions. Such disturbances reduce the longevity of vegetation structures and increase the annual average carbon investments into tissue turnover. A reduced availability of carbon for height growth may also represent carbon investments into trait trade-offs not currently represented, for example a spatially variable partitioning between above-ground and below-ground tissues to optimize water acquisition in seasonally dry environments, or investments for biotic interactions, including defence and competition mechanisms. Finally, in the height comparison it should be noted that the model does not consider present-day human land-use, nor herbivory dynamics. In areas of agricultural production (e.g., cropland and pastures), land use is expected to more strongly affect vegetation structures, such as height and biomass, compared to carbon fluxes like NPP and GPP.

There is a strong relationship between vegetation height and biomass carbon storage, which is reproduced due to the calibration of the allometric constants in the model (Fig. 6b), and which drives the modelled spatial distribution of aboveground and belowground biomass (Fig. 5d–i). Due to the strong biomass to height relationship, mismatches between modelled and observed belowground- and aboveground biomass primarily occur in regions where there is a mismatch in the productivity and height predictions as described above.

The TREED prediction of deciduous vegetation phenology agrees with the present-day distribution of temperate deciduous broad-leaved forests, with highest frequencies of such growth forms in East Asia, Europe and eastern North America (Fig. 7). Together with the tropics, these regions are associated with predicted leaf longevities of around a year or less. Persistent leaves with longevities of longer than two years are primarily observed in arid or cold regions with a limited growing season length. In line with the carbon economics spectrum (Wright et al., 2004), this tendency indicates a more conservative leaf strategy with lower annual leaf building costs and light acquisition potential but longer persistence. The model captures the major latitudinal patterns of leaf longevity and leaf mass per area (LMA = 1/SLA) on the globe, with increasing LMA and leaf longevities for evergreen plants, and a decreasing latitudinal trend observed for deciduous plants (Kikuzawa et al., 2013; Reich et al., 2014; Wang et al., 2023). Latitudinal variations are driven by the availability of light, colder temperatures towards the poles and therefore, decreasing growing season lengths. Deciduous and evergreen plants follow contrasting strategies to adjust to these environmental changes (Wang et al., 2023), which are captured by the model and its optimality-based trait prediction. For deciduous plants, shorter growing season are associated with reduced leaf longevities. To compensate a reduction in the carbon sequestration potential due to the shortened growing season length, deciduous plants require more acquisitive leaf structures and a reduced LMA. Evergreen plants, on the other hand, compensate the shorter growing seasons, reduced light, and colder temperatures by reducing annual carbon investments into leaves, increased LMA, and longer persistence.

The diversity calculations in TREED capture a large degree of the observed global plant species diversity distribution (Fig. 8), with an underprediction of the species richness in hotspot regions. The diversity potential (DP) in TREED is a compound product of three indices: a functional richness index (FDI), a landscape heterogeneity index (LHI) and a landscape fragmentation index (LFI) (see Eq. 63). The functional diversity index (Fig. 8a) strongly resembles the global pattern of vegetation height and carbon assimilation, indicating that it is primarily a function of the resource availability. In the model, locations with high levels of radiation and water, together with suitable temperatures for photosynthesis, result in a larger range of growth forms and trait combinations that can be sustained and potentially co-exist. The two indices of landscape complexity (LHI and LFI) capture regions of high topographic complexity, including the Andes, Himalayas, the east African mountains and the Alps. Topographic complexity in these regions results in a high variability of climatic conditions and thus vegetation habitats. Further, in topographically complex regions the arrangement of vegetation habitats shows a higher degree of fragmentation, which may be associated with dispersal barriers that promote allopatric speciation processes (i.e., speciation by isolation). Combining functional diversity as an indicator of the local diversity potential and the two landscape metrices effectively captures major biodiversity gradients on Earth, with a spearman ranked correlation between the modelled DP and observed species richness of 0.81. A major underestimation of the diversity potential is observed in South-East Asia, likely driven by an underestimation of the productivity potential in the region (Fig. 3a–f) and thus a reduced functional diversity potential. Similarly, an overprediction of diversity is observed in the central African tropics, likely related to an overprediction of the productivity and functional diversity potential in the region.

Plant carbon and water exchange, as well as biomass carbon storage, represent major forcings in Earth's carbon and climate system. The TREED model can serve as a computationally efficient, first-order approximation of the large-scale patterns of carbon assimilation and sequestration at present and under different planetary climate states in the past and future. Thereby, the model is computationally efficient and uses comparatively little input data of monthly average temperature, radiation, precipitation and cloud cover. This represents an advantage for the main purpose of the model of studying paleoclimate and vegetation dynamics across large timescales, and for conducting sensitivity studies. However, the coarse time stepping likely introduces biases in regions of high climate variability. Global trends in carbon and water fluxes (i.e., GPP, NPP, AET) are predicted with more confidence than vegetation biomass carbon sequestration, indicating that the implemented optimization procedure is better suited to capture the carbon assimilation potential but does not resolve the full variability of how the assimilated carbon is invested into growth and biomass. A discrepancy in the performance of models to predict productivity compared to biomass storage is a feature observed across many vegetation models and is partly explained by the difficulty to resolve carbon turnover times of biomass (Pugh et al., 2020). In the TREED model, the largest biomass carbon pools (heartwood, sapwood, coarse root) are associated with fixed turnover times in the range observed and modelled for the present day (Pugh et al., 2020), while functional relationships determine fine root and leave carbon turnover. In natural systems, turnover of large structural carbon pools is primarily a result of vegetation mortality, driven by a combination of biotic (e.g., competition, herbivory, diseases) and abiotic (e.g., fires, droughts, windthrow) factors (Franklin et al., 1987). A large range of mechanistic formulations exist to approximate these processes, but are associated with considerable uncertainty and differences among models (Pugh et al., 2020). Following other vegetation models (Schaphoff et al., 2018; Sitch et al., 2003), the TREED model further simplifies growth respiration and reproduction carbon costs as fixed fractions of GPP and NPP, respectively. The current implementation of the vegetation carbon turnover and mass balance represents a simple first approximation, with the model being flexible to further develop and test increased complexity formulations of carbon turnover that may depend on environmental conditions or functional trait trade-offs.

In the current TREED version, the considered carbon economic trade-offs and main axes of trait variation include carbon investment into height growth, the leaf carbon pool, leaf longevity and phenology (e.g., Figs. 6 and 7). The underlying allocation dynamics and trade-offs of height growth and leaf carbon economics are well established in both present day vegetation systems (Falster and Westoby, 2003; Valentine and Mäkelä, 2012; Wang et al., 2023; Wright et al., 2004), as well as in the geologic past (McElwain et al., 2024; Soh et al., 2017). With these trade-offs, the model resolves the two main axes of variation in vegetation growth forms observed in plants today (Díaz et al., 2016). However, while capturing mean trends in the form of a single, community-representative average plant per location, the model does not resolve the full range of trait variability and complexity we observe in natural vegetation systems. An increase in the modelled trait variability or the range of independent axes of trait variation would require the formulation of additional trait trade-offs that capture the cost and benefits of growth strategies in given abiotic and biotic environmental conditions. Thereby, the trait trade-off should not be specific to a species or location, but generally applicable across plant species, space and time. The increasing availability of plant functional trait data from various environments at present (Bruelheide et al., 2018; Díaz et al., 2016; Kattge et al., 2020), as well as the geologic past (McElwain et al., 2024), is essential to establish such general trait relationships. The capacity of the TREED model to resolve major patterns in present-day vegetation structure and functioning based on a small number of functional trade-offs illustrates the potential of trait-based models in leading to more generalizable and self-consistent representations of vegetation systems, and may allow a reduction of uncertainty in model-predicted carbon and climate dynamics associated with vegetation parametrizations under environmental conditions of the future (Berzaghi et al., 2020; Fisher et al., 2014) or the deep past (Matthaeus et al., 2023).

Focusing on basic principles of trait functional relationships, a continuous trait space and no pre-parametrized vegetation classes, the TREED model is intended to provide a tool to understand how Earth's climate, topography and vegetation have co-evolved through time. While the fundamental trade-offs of height growth and the leaf economics spectrum that drive the model are expected to also have played out in the distant geologic past (Butrim et al., 2024; Falster and Westoby, 2003; McElwain et al., 2024; Soh et al., 2017), two limitations for the application to the geologic past should be noted. Even though applying a generalized physiology across all plants, the TREED model still takes a present day-derived perspective on traits and whole-plant architecture in the form of the applied allometric and functional relationships. Further, the model assumes a constant theoretically possible trait space for the main varying traits and their combinations. The model does not resolve how the availability of genetic variance may have limited the range of plausible trait combinations through time. The major advantage of TREED with regards to these uncertainties, however, is its capacity to track combined carbon fluxes as well as associated vegetation structures. The model results can thus be compared to combined geochemical records informing about vegetation productivity (Bowen, 2013; Denis et al., 2021), as well as trait structures derived from paleobotanical records (McElwain et al., 2024; Rogger et al., 2025), to assess the validity of currently implemented and future additions of functional trade-offs and allometric relationships.

To study co-evolutionary transitions between the physical environment and vegetation systems, TREED includes eco-evolutionary adaptation processes not currently considered in standard vegetation models, including the capacity of vegetation to adapt to environmental changes through trait evolution and dispersal dynamics (Berzaghi et al., 2020). A limited capacity of vegetation systems to respond to abrupt climatic changes and a loss of vegetation-mediated climate regulation are considered to have strongly shaped global carbon cycle dynamics in the geologic past, and particularly, during episodes of abrupt climatic change (Hull, 2015; Payne et al., 2004; Retallack et al., 1999; Rogger et al., 2024; Xu et al., 2022, 2025). An evaluation of the capacity of vegetation to respond to environmental change is of particular importance considering that the speed of current climate warming may have no precedent in the last 400 million years (Foster et al., 2017). However, our understanding of the plant-climate adaptation process through acclimatization and evolutionary adaptation, the speed, and its limits, remain an active area of research (Bennett et al., 2021; Jezkova and Wiens, 2016; Liu et al., 2020; Schneider et al., 2025; Stemkovski et al., 2025). By combining trait-enabled model predictions, geochemical and palaeobotanical records from Earth's past, the TREED model was developed to help constrain and test hypotheses about the vegetation response to past climatic changes, the speed and capacity of adaptation, and how adaptation may depend on the starting climate state, atmospheric CO₂, or the rate and spatial distribution of climate warming. Obtaining an understanding of these dynamics through geologic time will help to better predict the vegetation responses that may emerge under different scenarios of climate change over the next centuries.

We show that the generalized plant physiology and continuous trait space of the TREED model cannot only be used to predict vegetation functioning and structure, but also to estimate locations of a high plant diversity potential. Thereby, we combine measures of local scale diversity (i.e., functional diversity potential) and landscape-scale biodiversity metrics (i.e., landscape heterogeneity and fragmentation of vegetation habitats). The employed metrics are aimed to resolve predominant drivers of plant species richness, including the spatial distribution of productivity-limiting resources such as energy and water, and the distribution of environmental heterogeneity, promoting species diversity though niche differentiation and allopatric speciation (Antonelli et al., 2018; Cai et al., 2023; Kreft and Jetz, 2007; Stein et al., 2014). Our results support that the consideration of these biodiversity drivers captures a large degree of the broad-scale plant species richness distribution on Earth today. As such, in the context of paleoecological research, the model represents a tool to evaluate how topographic and climatic changes throughout Earth's history may have influenced the global distribution of the plant biodiversity either through alteration of the functional diversity potential or landscape dynamics. In the biodiversity estimation, the TREED model takes a purely environment-driven approach by estimating a diversity index based on the current timestep's abiotic environmental conditions. Individual species and their life histories are not resolved. As the biological processes and species life histories that ultimately lead to the realized global biodiversity distribution of a period are not resolved (e.g., speciation, movement, extinction), the predicted biodiversity indices should be considered indicators of an abiotic diversity potential rather than realized diversity.

The TREED vegetation model balances a mechanistic, trait-based prediction of vegetation functioning and biodiversity with computational efficiency, allowing for modelling experiments and sensitivity tests across large time scales and under various climate states without pre-defining vegetation types. Key limitations that should be considered in the application of the model include the coarse monthly time stepping, which may not capture non-linear variation in climate and productivity at high temporal resolution. Further, the model does not capture complex ecosystem and carbon turnover dynamics, due to its ecosystem representation by a single, community-representative average plant and simple biomass turnover rules. For the biodiversity estimation, the main limitation of the model arises from its purely abiotic diversity perspective, predicting biodiversity given a distribution of topography and climate, but not resolving species' life histories, biotic interactions, or genetically constrained innovation through time. Regarding eco-evolutionary, non-steady state simulations, the model can track vegetation structures and functioning under different adaptation speeds from no evolutionary adaptation to immediate response, allowing model–data comparisons with paleobotanical records. However, the rates of trait evolution are simplified to a user-defined rate, and there is currently no mechanistic description of evolution, driven by interactions of genetics, traits and the environment. The same applies to the implemented climate niche dynamics, which can be derived based on present-day or past species or biome distributions, while not resolving the underlying physiological mechanisms responsible for observed limitations in climatic tolerances.

Future developments of the TREED model will focus on the consideration and testing of additional functional traits and environment-dependent trait trade-offs. In combination with paleobotanical and present-day trait data, this may help to derive the necessary environment-dependent rules of trait evolution, and to resolve environmental limitations and growth forms not currently represented by the model. Targets for development particularly include functional trade-offs concerning carbon investments into hydraulic traits (e.g., maximum stomatal conductance or xylem conductivity), which are important both in determining the response of vegetation systems to environmental conditions as well as in their impact on the coupled global carbon and water cycle (McElwain et al., 2024). Related to water acquisition, the competition for pre-emptying water resources from competitors may be considered in future model versions, complementing the current light competition scheme. In drought-prone environments, a competitive advantage in securing water may favour the relative allocation of carbon into belowground biomass at the expense of height and aboveground biomass growth (Cabal et al., 2020), which is not currently resolved in the model that assumes a constant root to shoot carbon ratio. Further, alternative models concerning carbon turnover costs may be employed. Possible schemes include environment-dependent vegetation mortality, such as increased carbon turnover in heat, drought or wind-prone environments (Pugh et al., 2020), or susceptibility to biotic stresses depending on environment conditions and functional trait characteristics, e.g., increased herbivory pressures in high temperature environments and for low LMA leaf traits (Currano et al., 2008, 2010). Finally, it is aimed to include recently developed, optimality-based and parameter-scarce models of the photosynthetic carbon assimilation and evapotranspiration into TREED (Stocker et al., 2020), which is the physiologic process currently relying on most pre-defined parameter inputs (e.g., Table 1) and which will help to differentiate the influence of short timescale acclimatization processes of the photosynthetic pathway from long-term evolutionary trait evolution in response to climatic change (Schneider et al., 2025).