SCOPE 56 - Global Change: Effects on Coniferous Forests and Grasslands

 

13.1 INTRODUCTION
13.2 APPROACHES TO MODELLING
	13.2.1 Physiologically based models
		13.2.1.1 Direct effects of CO2
		13.2.1.2 Temperature
		13.2.1.3 Water
		13.2.1.4 Radiation
	13.2.2 Ecosystem/tissue models
		13.2.2.1 Plant processes
		13.2.2.2 Soil processes
	13.2.3 Other model types
	13.2.4 Choice of models
13.3 MODEL DESCRIPTIONS
	13.3.1 Comparison by process
	13.3.2 BIOMASS
	13.3.3 BIOME-BGC
	13.3.4 CENTURY
	13.3.5 HYBRID
	13.3.6 MBL-GEM
	13.3.7 PnET-CN
	13.3.8 Q
13.4 METHODS
	13.4.1 Design
		13.4.1.1 BIOMASS
		13.4.1.2 BIOME-BGC
		13.4.1.3 CENTURY
		13.4.1.4 HYBRID
		13.4.1.5 MBL-GEM
		13.4.1.6 PnET-CN
		13.4.1.7 Q
	13.4.2 Site descriptions
		13.4.2.1 Biology of forest growth experiment
		13.4.2.2 Swedish coniferous forest project
13.5 COMPARISONS WITH DATA AT BFG AND SWECON CONIFER SITES
	13.5.1 Stem biomass, foliage N, aboveground productivity
	13.5.2 Single-year carbon budgets
	13.5.3 Single-year nitrogen budgets
	13.5.4 Single-year water budgets
13.6 DISCUSSION
	13.6.1 Photosynthesis and canopy dynamics l
	13.6.2 Respiration
	13.6.3 C and N allocation
	13.6.4 Nitrogen cycle and decomposition
	13.6.5 Hydrology
13.7 RECOMMENDATIONS FOR EXPERIMENTATION AND MODELLING
	13.7.1 Experimentation
	13.7.2 Modelling
13.8 REFERENCES

13.1 INTRODUCTION

Predicting how a given forest will respond to an altered climate, increased CO₂ concentration, pollution stress, or increased nutrient input is a complex problem. If increased temperatures raise decomposition more than productivity, forests may become a source for atmospheric carbon (Woodwell1987). However, given adequate nutrient availability, forests could store more carbon (C) and slow the rate of increase in atmospheric CO₂ (cf. Rastetter et al. 1992). Understanding how climate and vegetation influence the biogeochemistry of ecosystems will allow us to make predictions about how global change will affect forests. This understanding will also allow us to test strategies for managing these ecosystems to mitigate the effects of global change.

Models are an important tool for both understanding ecosystem function and predicting responses to global change. Models can summarize the results of many experiments by incorporating hypotheses and conclusions into a quantitative framework. System-level fluxes are uncertain and very difficult to measure, and models can provide independent estimates. Conducting experiments that alter climate or atmospheric CO₂ of forest ecosystems poses special difficulties, because trees are large and growth and decomposition occur over decades to centuries. Therefore, long-term response to a changed environment may differ from the results of experimentally feasible short-term manipulations. Models can be used to simulate these long-term experiments.

There are a variety of models that simulate element cycling in forest ecosystems (see section 13.2), and consensus about modelling approaches is lacking. Models differ in time scale, linkages between C, nitrogen (N), and water, representation of heterogeneity, and detail of photosynthesis, allocation, and decomposition sub models. These simulation models can be thought of as different syntheses of our understanding of what controls element cycling within ecosystems. By comparing model predictions for the same ecosystems, we hope to make these differences more explicit, and assess the impacts of different processes and the control they exert over the productivity of temperate forests.

In this chapter, we describe the results of a comparison of seven models of ecosystem element cycling. First, we review the different approaches to modelling to provide a context for the models in the comparison. Next, we describe the individual models and compare how the models represent various processes. Then, we compare the ability of the models to predict growth under natural conditions and the response to fertilization and irrigation for young conifer forests in two contrasting climates. Finally, we discuss how the model comparisons help assess the state of our knowledge, and suggest further experimentation to reduce uncertainty in our understanding of ecosystem response to climate change.

13.2 APPROACHES TO MODELLING

An extensive review of the state of the art in modelling ecosystem response to climate changes is given by Ågren et al. (1991). The models considered here are separated into two classes by level of resolution. Within each class we discuss how models deal with the various processes likely to be affected by climate changes, and how well those processes are understood.

In addressing the consequences of climate change at the regional or global level, we are faced with the task of scaling up from knowledge of an organ's response to its micro-environment. We recognize four partially overlapping levels where models have been used. These start at the plant physiological level and progress through populations and ecosystems to regional and global scales. At the physiological level, plant processes are described in great depth, down to the biochemistry. The ecosystem scale integrates the physiology to whole-plant or major plant component properties and adds a feedback interaction to the environment, in particular the soil. Population models generally operate at the same scales as ecosystem models but deal with plants in terms of population dynamics. At the coarsest level of resolution, regional and global models are used to explain the distribution of different biomes. Here, we focus only on the physiologically based and ecosystem models, but provide a brief description of population and regional models for context.

13.2.1 Physiologically based models

In recent years significant advances have been made in our capacity to explain and predict plant responses to the environment (e.g. Mooney et al. 1987). This progress enables us to predict the response of individual plant processes to climate change. The objectives of physiologically based models of canopy processes is to integrate up from the micro-environmental scale to explain and predict the aggregate element cycling of a forest stand. A model is required to interrelate the various processes and their response to environmental variation and to integrate over space and time. In the model comparison, this class of models is represented by BIOMASS (McMurtrie et al. 1990a, 1992), BIOME- BGC (Running and Coughlan 1988; Hunt and Running 1992; Running and Hunt 1993), and HYBRID (Friend et al. 1993).

13.2.1.1 Direct effects of CO₂

Models of plant production that include CO₂ enrichment must describe how CO₂ concentration will increase photosynthetic rates at both high and low light levels. Some models do not explicitly include ambient CO₂ concentration but can simulate CO₂ effects, generally by changing parameters corresponding to the photosynthetic rate at saturating irradiance and quantum yield. Others directly incorporate Farquhar and von Caemmerer's (1982) equations for the photosynthesis of C₃ plants, though these models can be computationally demanding if they simulate quantum flux density throughout the canopy (e.g. Wang et al. 1992). Nitrogen use efficiency may also affect the response to elevated CO₂ (Jarvis 1989).

Many models calculate transpiration from the Penman-Monteith equation applied to a canopy idealized as a single, so-called big leaf (e.g. Jarvis 1985). The Penman-Monteith equation has a sound theoretical base (but see Finnegan and Raupach 1987) and is general to all vegetation types. However, equations for the dependence of stomatal conductance on environmental variables, including radiation, water status and humidity, are required, and canopy conductance is derived by summation over an idealized canopy. Reduced canopy conductance due to CO₂ enrichment will likely reduce water use for forests with aerodynamically rough canopies and prolong the growing season for forests growing in water-limited environments.

13.2.1.2 Temperature  

The biochemical basis of the temperature dependence of photosynthesis has been studied (Kirschbaum and Farquhar 1984) and included in some models (e.g. McMurtrie and Wang 1993). However, temperature is usually included in models of canopy photosynthesis through empirically derived relationships typically representing the variation of photosynthetic parameters about some temperature optimum.

Most models discriminate between growth and maintenance respiration with growth respiration assumed to be in constant proportion to tissue production, and maintenance respiration of various biomass components related to temperature by either the Arrhenius function or an exponential relationship. Approaches adopted to compute stand level maintenance respiration include scaling according to surface area (Woodwell and Botkin 1970), sapwood volume (Ryan and Waring 1992), C mass of tissue (Running and Coughlan 1988) or nutrient content (Ryan 1991b).

Vegetative growth of forests may be temperature-limited in temperate environments (e.g. Cannell et al. 1988). Temperature effects on phenology and allocation have often been incorporated in grassland models (usually with degree-day sums) but less frequently in forest models. Temperature dependence also enters models through the effect of water vapour saturation deficit on stomatal conductance which in turn affects both transpiration and evaporation from wet canopies. Increases in the vapour saturation deficit will be diminished if higher temperatures occur in conjunction with increased absolute humidity of air.

13.2.1.3 Water 

Several models reliably predict plant available water (e.g. Running et al. 1983; McMurtrie et al. 1990a; McMurtrie and Landsberg 1992). These models describe rainfall, interception by canopies, soil evaporation, runoff, drainage, tree transpiration and sometimes understorey evapotranspiration and accumulation and melt of snow packs. They require inputs of daily weather data and soil physical characteristics such as soil water retention functions and rooting depth. They should successfully predict the consequences of changes in the frequency and intensity of rainfall on soil moisture.

Various approaches have been adopted to characterize the effects of water stress on photosynthesis, stomatal conductance and tissue production. Recent evidence indicates that it may be possible to relate stomatal conductance directly to plant available water in the rooting zone (Schulze et al. 1987), thereby simplifying plant and soil water interactions.

Predicting the consequences of climate change (including effects on nutrient availability) for leaf area production and retention is particularly important because of feedbacks on both water balance and C accumulation. Improved water use efficiency, together with enhanced quantum yield, which benefit C balance of foliage at the base of the canopies, may promote leaf area development. However, in areas with poor summer rainfall, stands with higher leaf area may exhaust soil water reserves earlier in the growing season.

13.2.1.4 Radiation  

Both photosynthetic rates and transpiration depend on radiation, which is expected to change in response to altered cloudiness. The relative proportions of the direct and diffuse components of incident radiation will change, indicating the need for caution where canopy assimilation is predicted by models which do not consider the two components separately. A realistic response of simulated transpiration to changes in net radiation is expected from models incorporating the Penman-Monteith equation, with transpiration particularly sensitive to radiation flux for canopies weakly coupled to the atmosphere, such as forest understoreys and grasslands (Jarvis 1989).

13.2.2 Ecosystem/tissue models

Ecosystem/tissue models include those models that consider water, C, and N flows between the plant and soil systems. Within this group of models, the plant compartments are highly aggregated and typically lump all green plant biomass into one compartment. The objectives of the ecosystem/tissue models are quite diverse; however, they generally include the ability to simulate ecosystem responses to changes in the abiotic driving variables (light intensity, soil water, and soil temperature) and the interaction of dry matter production and nutrient cycling. The four models CENTURY (Parton et al. 1987; Sanford et al. 1991), MBL-GEM (Rastetter et al. 1991), PnET-CN (Aber and Federer 1992) and Q (Rolff and Ågren, in preparation) will illustrate this model class.

13.2.2.1 Plant processes 

Some ecosystem/tissue models describe the plant processes similarly to the physiologically based models (i.e. growth is calculated from the difference between photosynthesis and respiration and allocated between different plant compartments), but with a lower resolution. In other models, phenomenological concepts such as nutrient productivity (Ågren 1983, 1985, 1988) are used to directly estimate net production. In both cases, nutrient availability is stressed as the important growth regulator. Plants influence nutrient availability through litter quality: poor quality litter decays slowly and immobilizes more nutrients than high quality litter. In these models, soil and plant C and N are explicitly represented and linked by C:N ratios of various plant and soil components.

13.2.2.2 Soil processes  

Soil processes respond rapidly to changes in the abiotic driving variables and litter quality. The impact of litter quality on soil processes has been studied for many years (Swift et al. 1979) and models have been developed to simulate the impact of litter quality on nutrient cycling and soil organic matter dynamics (Hunt, 1977; McGill et al. 1981; Aber et al. 1982; Melillo et al. 1982; Bosatta and Ågren 1985, 1991a, b; Parton et al. 1987; Ågren and Bosatta 1987,1988; Pastor and Post 1988). The N and lignin content of vegetation are the most commonly used variables for characterizing litter quality in decomposition models. Climate change may alter litter quality through two mechanisms:

(a) the effects of increased atmospheric CO₂ levels on litter quality (Melillo 1983; Feet et al. 1985; Strain and Bazzaz 1983);

(b) with changes in plant species composition (litter N and lignin content vary with species).

Models differ in the level of detail used to capture the dynamics of soil organic matter; models with several organic-matter pools may behave differently than single-pool models.

Soil moisture and temperature are the most important abiotic driving variables for the soil system and both would be altered substantially with the proposed climatic scenarios. The effect of soil moisture on decomposition has been represented in different ways in the various models. In more process-oriented models (Hunt 1977; McGill et al. 1981) the soil water effect is represented as a function of the soil water potential using non-linear functions. The more empirical models (Parton et al. 1987; Pastor and Post 1988) use variables such as relative water content of the soil and the ratio of rainfall to potential evapotranspiration rate to control decomposition.

13.2.3 Other model types

Two other classes of models (population models and regional models) were not included in the model comparison because these model types are more suited to different temporal and spatial scales than those used in this study.

Population models have been developed to simulate tree growth as affected by competition among individuals as well as climate. These models consider birth or recruitment, growth of individuals, stand structure and spacing, and mortality (e.g. Botkin et al. 1972; Shugart and West 1977; Shugart 1984). More recent versions of these models also simulate the effects of nutrient limitation on tree growth (Weinstein et al. 1982; Aber and Melillo 1982; Aber et al. 1982; Pastor and Post 1985, 1986). The HYBRID model (Friend et al. 1993) is a physiologically based population model, but the model comparison presented in this chapter concerns stands with a single species.

The regional and global coincidence of vegetation zones and climate patterns has long been noted (Trewartha 1968) and supports the presumption that climate changes will alter the global distribution of vegetation (Emanuel et al. 1985a, b ). This coincidence suggests that a global vegetation model can be derived and parameterized by overlaying vegetation maps with climate maps. Examples of this type of vegetation map are those that are based on the Holdridge(1947, 1964) life zone classification system. An alternative approach is to use empirically derived regressions between climate and variables such as net primary production (Rosenzweig 1968) or litterfall (Meentemeyer et al. 1982). These or similar relationships are assembled in several global models that produce maps of global distribution of net primary production, litterfall, and C storage from global climate data sets (Lieth 1972; Box 1978,1981; Meentemeyer et al. 1982). These regional and global classification and regression approaches may predict the future steady state distributions of vegetation and C pools but lack the ability to model the transient responses shown to be important in this chapter.

To avoid the problems associated with static approaches, simple ecosystem/tissue-type models have been used at regional scales (Raich et al. 1991; Burke et al. 1991; McGuire et al. 1992,1993; Melillo et al. 1993). These models assume uniform conditions over large areas (e.g. 0.5 x 0.5 grid cells) and simplified physiology but do link plant and soil processes and climate. Another recent approach uses plant physiology, soil, and climate to alter biome distribution (Prentice et al. 1992).

13.2.4 Choice of models

A variety of models, each with its specific domain of applicability, will be required to answer the diverse questions about effects of climate change. We want to emphasize the need to select a model that is appropriate with respect to the scale of the perturbation in question, particularly in time but also in space. Viewing the same system from different space and time perspectives makes different phenomena important or unimportant. Models based on plant physiology are ideal to analyse and interpret the detailed short-term reactions of plants to the various components of climate change. On the other hand, for the ecosystem models that emphasize longer time scales, insights from the physiologically based models are required to ensure that no major impact of climate change is neglected and that formulations of growth processes are realistic.

Physiologically based models provide a framework integrating the many direct  and indirect effects of elevated CO₂. Effects promoting growth include increased temperature (lengthening growing seasons and increasing photosynthesis in high light environments) and the beneficial direct effects of CO₂ enrichment on photosynthesis and water use efficiency. Effects reducing growth arising from greenhouse warming are increased water vapour saturation deficit and increased respiratory losses. Changes in rainfall patterns will be beneficial in some areas but detrimental in others. Depending on particular circumstances, increased leaf area development may increase or decrease productivity. Where the objective is to assess the net effect of such opposing factors, physiologically based models are appropriate tools (e.g. Booth and McMurtrie 1988).

Ecosystem/tissue models provide a framework for understanding the system- level constraints for the response of productivity to elevated CO₂, and altered temperature and moisture. For example, the response to elevated CO₂ is strongly controlled by nutrient availability (e.g. Oechel et al. 1994), but nutrient availability results from complex interactions between climate, litter quality, productivity, and soil. Therefore, information about nutrition is necessary to assess longer-term interactions between the atmosphere and climate, and ecosystem/tissue models are appropriate tools for this task.

13.3 MODEL DESCRIPTIONS

13.3.1 Comparison by process

The models used in this comparison are (Table 13.1, in alphabetical order): BIOMASS (McMurtrie et al. 1990b, 1992; McMurtrie and Landsberg 1992); BIOME-BGC (Running and Coughlan 1988; Running and Gower 1991; Hunt and Running 1992); CENTURY (Parton et al. 1987, 1988; Sanford et al. 1991); HYBRID (Friend et al. 1993); MBL-GEM (Rastetter et al. 1991); PnET-CN (Aber and Federer 1992), and Q (Rolff and Ågren, in preparation). These models were developed with different objectives; hence, different processes are emphasized. However, because most ecosystem processes are controlled in part by interactions with other processes, each model deals with these interactions, either by simulating the process directly or by using that process as a model input. For example, BIOMASS and HYBRID do not simulate decomposition and N mineralization; both use leaf N concentration as an input circumventing the need to simulate these processes. Also, the model Q uses net primary production as an input with transpiration, photosynthesis and respiration not represented. A brief description of each model, the original modelling objectives and further references are given in sections 13.3.2-13.3.8.

For some processes, the various models have converged on a given algorithm (Table 13.1). For models with a daily time step, transpiration is determined using the Penman-Monteith equation and photosynthesis is simulated using variations of a model originally developed by Farquhar et al. (1980). Most of the models distinguish between growth and maintenance respiration (Amthor 1989) but differ over the scaling metric used.

Table 13.1 Comparison of model processes

^aTime step: D = daily, M = monthly, A = annual.
^bTranspiration: PM = Penman-Monteith, I = soil moisture as input, WUE = water use efficiency, - = not simulated.
^cPhotosynthesis (model): F = Farquhar et al. (1980) biochemical model, P = phenonenological model, I = net primary production only.
^dPhotosynthesis (scaling method): B = biomass, N = foliar nitrogen, C = climate.
^eRespiration (model): G = growth respiration, M = maintenance, respiration, - = not simulated.
^fRespiration (scaling method): B = biomass, N = tissue nitrogen, - = not simulated.
^gAllocation of carbon: A = allometric equations, V = variable percentage, C = constant percentage.
^hTurnover of leaves: C = constant percentage.
ⁱLeaf nitrogen: I = input, V = variable.
^jLitter nitrogen: - = not simulated, p = constant percentage of leaf nitrogen, C = constant.
^kDecomposition: - = not simulated, AETIL = actual evapotranspiration and lignin concentration, LCI = lignocellulose index,
R = residence time for given stage, M = microbial growth rate.
^lNitrogen mineralization: - = not simulated, p = proportional to decomposition, C:N = maintains constant carbon to nitrogen ratio.

The different approaches to the allocation of C and N, leaf and root turnover, decomposition, and N mineralization reflect both differences in aggregation and a greater level of uncertainty about how to represent the processes. For example, all models represent leaf turnover as a constant percentage of leaf area index, yet Reich et al. (1992) show that leaf longevity is affected by many ecosystem processes. Given the importance of these processes on ecosystem function, the analysis of alternative hypotheses is warranted.

13.3.2 BIOMASS

The BIOMASS model is a process-based model of forest growth incorporating sub models for radiation absorption, canopy photosynthesis, partitioning of assimilate between plant organs, litterfall and stand water balance (McMurtrie et al. 1990a, 1992; McMurtrie 1993; McMurtrie and Landsberg 1992; McMurtrie and Wang 1993; Figure 13.1). Canopy C gain is calculated from a radiation interception model which uses information about canopy architecture, and a mechanistic model of leaf photosynthesis by C₃ plants (Farquhar and von Caemmerer 1982). Tree crowns are represented by truncated ellipsoids, while the plant community is represented by a two-dimensional random array of trees with a correction introduced to account for spatial nonrandomness. The foliage is divided into three horizontal layers with different photosynthetic parameters.

The radiation interception sub model that is the basis for the calculation of canopy net photosynthesis separately considers the interception of direct and diffuse radiation by foliage. Photosynthesis is evaluated for both sunlit and shaded foliage. Canopy C balance is updated daily after subtracting maintenance respiration calculated using equations from Ryan (1991 b). In BIOMASS allocation of C to foliage, branch, stem and root is represented by partitioning coefficients that vary over an annual cycle and are linearly related to foliar N concentration (McMurtrie and Landsberg 1992). Litterfall is subtracted from each biomass component with rates varying over an annual cycle.

Stand water use is calculated from the Penman-Monteith equation, with biological control of the process exerted through a model of stomatal conductance (McMurtrie 1993). Meteorological data required to drive BIOMASS are daily precipitation, short-wave radiation and maximum and minimum air temperatures. The model employs a daily time step, though the diurnal time course of several meteorological variables is incorporated to obtain daily total photosynthesis and transpiration. This is necessary because of the non-linear responses of these processes to meteorological variables.

The BIOMASS model does not incorporate a dynamic model of nutrient cycling and uptake by trees. Instead, N concentrations, which influence simulated photosynthesis, respiration and allocation, are required as model inputs.

13.3.3 BIOME-BGC

The BIOME- BGC (for biogeochemical cycles) is a generalized ecosystem process model (Hunt and Running 1992; Running and Hunt 1993) derived from a model developed for coniferous forests, FOREST-BGC (Running and Coughlan 1988; Running and Gower 1991). A biome is loosely defined as a combination of a lifeform type (conifer, broad-leaved, grass, etc.) in a given climate (Hunt and Running 1992), so BIOME-BGC is strongly controlled by climate, as is also true for FOREST-BGC. The primary objective of BIOME-BGC is to simulate the C,

Figure 13.1  Schematic representation of the BIOMASS model. The carbon and water balance models are displayed on the right-and left-hand sides, respectively. Processes are denoted by shaded boxes, meteorological variables by unshaded boxes, state variables by rounded boxes, transfers of material and energy by solid arrows, and influences by broken arrows

N and water budgets for a large homogeneous area of vegetation. The key variable is leaf area index (LAI) which can be estimated by remote sensing. As a consequence, BIOME-BGC treats processes that have a direct relationship to LAI with considerable detail (photosynthesis and transpiration), whereas processes that have an indirect relationship to LAI (e.g. decomposition) are treated with significantly less detail. For coniferous forests, processes represented on a daily basis (photosynthesis and evapotranspiration) for FOREST-BGC, and hence BIOME-BGC, are validated with experimental data on stem growth and soil water content (Nemani and Running 1989; Korol et al. 1991; Hunt et al. 1991).

The BIOME-BGC has a dual time step, daily for the hydrologic budget, canopy gas exchange and maintenance respiration, and yearly for allocation, decomposition and the N cycle (Figure 13.2). Leaf area index controls the interception and subsequent evaporation of precipitation; precipitation not intercepted is routed to a single soil water storage compartment. Based on soil texture, the amount of water in the soil determines predawn leaf water potential, which defines the amount of water stress. Predawn leaf water potential, humidity, temperature and radiation are used to calculate stomatal conductance. Conductance is multiplied by LAI, which is then used in the Penman-Monteith equation to calculate transpiration per ground area. From leaf N concentration and stomatal conductance, a maximum photosynthetic assimilation rate is determined using the model of Farquhar et al. (1980) which is coupled to phenomenological equations for light and temperature to calculate canopy photosynthesis (Hunt and Running 1992).

Figure 13.2 Compartment diagram for BIOME-BGC, an ecosystem process model based on data derived by remove sensing

In the annual part of BIOME-BGC, constant fractions of leaf, stem, and root C and N turn over and are added to the litter storage components (Figure 13.2). Litter C decomposition rate is determined by actual evapotranspiration and lignin concentration. Decomposition of soil C and mineralization of soil N are set as proportions of litter C decomposition rate. The amounts of N immobilized or lost from the ecosystem are set as proportions of the N mineralized and N added to the ecosystem through fixation, deposition and fertilization. The net amount of N mineralized and N added to the ecosystem and the amount of N retranslocated from the foliage determine the amount of available N. From the available N and water stress integral (Myers 1988), a leaf C/(leaf C + fine root C) ratio is formed that controls the allocation of C to leaves and fine roots. The difference of C available for growth and C allocated to leaves and fine roots is then allocated to stems and coarse roots.

13.3.4 CENTURY

The grassland version of the CENTURY model is fully described in Chapter 11. The adaptation of this model to forest systems is discussed by Sanford et al. (1991). Decomposition and hydrologic components of the forest model are similar to those of the grassland model, but C and nutrient pools representing above-and belowground woody debris have been added. Aboveground woody debris is divided between fine branch and large wood debris. Decomposition rates of woody debris are specified as inputs, and wood decomposition products are partitioned among soil organic matter pools according to initial wood lignin content. Soil C and nutrient components of CENTURY have been extensively validated against field observations of soil organic matter dynamics (Parton et al. 1987, 1988; Chapter 11, this volume). Values of parameters that control these routines are generally not altered in calibration of the model to a particular site. Nitrogen mineralization and immobilization are controlled by soil C dynamics and C : N ratios of soil C pools.

The forest version of CENTURY contains new plant production routines that replace those of the grassland version. The forest production model (Figure 13.3) divides trees into leaves, fine branches, large wood, fine roots, and coarse roots.

Figure 13.3 Compartment diagram for the Forest CENTURY model. The soil organic matter (SOM) component has several individual pools not shown in this figure

Carbon and nutrients are allocated among these components by a fixed scheme of C allocation fractions and tissue nutrient concentrations. Separate allocation patterns may be used for young and mature forests. Allocation patterns and nutrient concentrations are specified for the system being simulated, and are among the most important calibration parameters in this model. Monthly potential NPP is calculated from a specified maximum value by applying climate and nutrient limitations in the same manner as in the grassland model. Additionally, NPP may be reduced by maintenance respiration demands of sapwood if large wood biomass becomes sufficiently great. Live biomass is transferred to litter and woody debris pools according to component-specific death rates (in deciduous systems, 95% of leaf biomass is transferred to litter at the end of the growing season).

13.3.5 HYBRID  

The HYBRID model grew from the merging of the big-leaf model FOREST-BGC (Running and Coughlan 1988) and a version of the FORET class of gap models (Shugart 1984), ZELIG (Urban et al. 1991). All of the physiological and allocation routines from FOREST-BGC have been substantially modified and used to replace the less mechanistic growth routine of ZELIG. In HYBRID individual trees are grown in plots, C is taken up and respired and water lost by each individual on a daily time step, and at the end of each year the accumulated C is partitioned to the various structural components (Friend et al. 1993; Figure 13.4). Soil water is estimated for a plot with the hydrological and other climate routines taken directly from FOREST-BGC. Individual trees are assigned a set of physiological parameters, depending on their species (Friend et at. 1993), and these determine the growth rate of each individual in the plot. Individuals with a low growth rate will become shaded and eventually die.

Photosynthesis and stomatal conductance are calculated using PGEN (Friend 1991), a modified Farquhar et al. (1980) model. Leaf water potential is assumed to have a direct effect on photosynthesis (Figure 13.4), and so it is possible to find an optimal stomatal conductance for any set of conditions (Friend 1991). This optimal conductance and the associated rate of net photosynthesis is calculated for each day. Day and night canopy respiration is calculated from foliage N content (Ryan 1991b) and temperature. Woody maintenance respiration is calculated from the total amount of living sapwood and temperature (Ryan 1990). The total amount of living sapwood is calculated from the difference between total woody biomass and dead heartwood biomass, assuming that a species-specific fraction of the sapwood is alive. Each crown is divided into 1 m vertical sections to calculate average daily radiation down through each plot. The daily average radiation in each crown drives daily photosynthesis for that individual. Thus each crown is treated as one layer for the photosynthesis routine. However, the lowest layer of each crown is followed separately because the total annual C balance in this layer, minus the maintenance respiration required by the sapwood to support this layer, is used to calculate annual heartwood growth: if the balance is negative, the heartwood grows by the amount of sapwood that was supporting this layer.

Partitioning logic is based on the pipe-model. A constant ratio of foliage to fine root biomass is assumed. Foliage biomass is linearly related to sapwood area at breast height. Sapwood area is the result of the dynamics of total diameter growth and heartwood growth from year to year. A potential two-thirds of the living sapwood can be used for storage. This storage is used if there is insufficient C available at the end of a year to produce the full complement of foliage, with its associated fine root biomass. A minimum 10% of available C is always partitioned to stem, branch, and coarse root components (this partitioning is calculated allometrically). If the storage is below capacity, this gets preference for C over any additional stem, branch, and coarse root growth. Foliage litter production is a fixed, species-specific fraction of total foliage biomass, as is fine root litter production.

Figure 13.4 Principal physiological and hydrological features of HYBRID at the individual tree level. Temperature affects many of the daily processes, but it is not included for clarity. Narrow lines indicates; broad lines indicate fluxes; shaded boxes indicate structural components. DBH is the diameter at 1.4m

13.3.6 MBL-GEM  

The Marine Biological Laboratory's General Ecosystem Model (MBL-GEM, Figure 13.5) is a process-based, biogeochemical model designed to examine changes in the fluxes and allocation of C and N among plant tissues (foliage, fine roots, stems) and soils in response to changes in atmospheric CO₂ concentration, temperature, soil moisture, irradiance, and inorganic N inputs (Rastetter et al. 1991). The model consists of 23 simultaneous ordinary differential equations describing the temporal dynamics of 23 state variables (Figure 13.5). These variables represent the amounts of C and N in several plant tissues and in four soil organic fractions, as well as the amount of soil inorganic N. Within the vegetation, both labile (readily mobilized) and structural (including enzymatic machinery and structural framework) components of foliage, sapwood, and fine roots are represented. The amount of C and N in heartwood (structural only) is also simulated. Younger soil organic matter (including litter) is divided into 'extractives' (extractable in methylene chloride and hot water), 'cellulose' (acid soluble fraction of remaining residue), and 'lignin' (acid insoluble fraction). Older organic matter is converted to humus.

The uptake of C and N and the growth of plant tissues are simulated as enzymatically mediated processes, and the allocation of C and N to the various tissues is controlled by the gradients of labile C and N between tissues (similar to Thornley 1972a, b). This allocation scheme, in conjunction with the co-limitation of tissue growth by labile C and N in the tissues, the stoichiometric shifts in the C:N ratios of leaves and roots, and the dependence of C and N uptake on structural N in leaves and roots, form the mechanisms of N limitation on photosynthesis and C limitation on N uptake.

The turnover of soil organic matter is controlled by the rate of litterfall and the rate and efficiency of the transformations from one soil organic matter fraction to another. Litterfall is partitioned into 'cellulose', 'extractives', and 'lignin' based on the C: N ratio of the litter. Rates of transformation of soil organic matter from one organic matter fraction to another are assumed to be proportional to the amount of C in the donor fraction and are corrected for temperature, soil moisture, and microbial efficiency. Respiratory losses of C during these transformations are calculated from the rate of transformation and the microbial efficiency. Nitrogen is mineralized and immobilized during these transformations to maintain the appropriate C:N ratios of the respective fractions.

The model is implemented through an automated calibration routine. Most of the parameters in the model are set a priori using general, process-based information from the literature (e.g. half-saturation constants, Q₁₀ quantum yield) or from our own prior experience with the model (e.g. decomposition parameters have been found to be general across a wide range of litter types, Ryan et al., pers. comm.). However, the rate parameter for each of the major fluxes in the model is calibrated by back calculating its value based on the specified rates of several key fluxes. Specifically, the user must specify site-specific rates of gross primary production, litter production for foliage, wood, and fine roots, and the annual growth increment for these tissues. The user must also specify amounts of  and N in foliage, sapwood, heartwood, and fine roots, and the total amount of C and N in soil organic matter.

Figure 13.5 Compartment diagram for the MBL-GEM model. Arrows indicate fluxes.GPP is GROSS primary production

13.3.7 PnET-CN

The PnET model (Aber and Federer 1992) is a simple, lumped-parameter, monthly-time-step model of C and water balances of forests built on two principal relationships: (1) maximum photosynthetic rate is a function of foliar N concentration, and (2) stomatal conductance is a function of realized photosynthetic rate. Using these two relationships greatly simplifies the calculation of transpiration and provides a direct link between C gain and water loss. Monthly leaf area display and C and water balances are predicted by combining equations derived from these relationships with standard equations describing light attenuation in canopies and photosynthetic response to diminishing radiation intensity, along with effects of soil water stress and vapour pressure deficit. The PnET model has been successfully validated against field data from 10 well-studied temperate and boreal forest ecosystems, suggesting that the aggregation of climatic data to the monthly scale and biological data such as foliar characteristics to the ecosystem level does not cause a significant loss of information relative to long-term, mean ecosystem responses (Aber and Federer 1992).

The PnET model is similar to the CENTURY model (Parton et al. 1988) in that it operates at a monthly time step, uses a single set of parameters to define the physiology of the plant community, produces biomass only by tissue type (e.g. foliage, wood and fine roots), and predicts seasonal changes in leaf area display in response to climatic drivers. It is also similar in structure to the C and water portion of the FOREST-BGC and BIOMASS models (Running and Coughlan 1988; McMurtrie et al. 1990a) but differs from these in the methods used to link the photosynthetic and transpiration processes, and in that a monthly, rather than a daily, time step is used.

There are 5 compartments in the model, and 11 fluxes, 3 for C and 8 for water (Figure 13.6, net photosynthesis includes separation of daytime and night-time C fluxes). All fluxes are calculated monthly except C allocation to wood and fine roots, to which excess C accumulated by net photosynthesis is allocated at the end of the year.

The foliage production routine uses radiation, temperature, water stress during the previous month and N content of foliage to derive a potential gross photosynthesis rate and day and night respiration rates for leaves at the top of the canopy. These are combined with a light attenuation coefficient, a photosynthetic light response curve, and foliar longevity to determine the production and shedding of foliar mass, and a potential gross photosynthesis (in the absence of water stress) for the whole canopy.

The C and water balance routine partitions precipitation between rain and snow, calculates snowmelt, total water input to the soil, and a fast, non-Darcian, drainage volume to determine water availability over the month. It then performs a numerical integration (daily time step) of water inputs to soil and transpirational demand (a function of gross photosynthesis and water use efficiency) over the month to produce changes in water storage, the degree of actual water stress on vegetation, and realized transpiration and net photosynthesis. Water remaining beyond water holding capacity at the end of this integration is drained away. Finally, the C allocation routine accumulates net photosynthesis over the year, subtracts foliage production and allocates the remainder to wood and root tissues.

Figure 13.6 Compartment diagram for the PnET-CN model. Numbered arrows represent fluxes: (1) net photosynthesis (a function of light, temperature, water, N, and LAI); (2) fine root allocation (a function of foliage production); (3) wood allocation (by difference); (4) precipitation (input); (5) interception (constant fraction); (6) snow-rain division (a function of temperature); (7) snow melt (a function of temperature); (8) plant uptake (equals evapotranspiration); (9) evapotranspiration (a function of net photosynthesis and vapour-pressure deficit); (10) leaching (a function of soil water content and soil water-holding capacity); (11) 'fast flow' (a function of precipitation and snow melt)

13.3.8 Q

The forest growth model Q (after one of its central variables, substrate quality (q) was developed to analyze the N cycle and its relation to forest stand development. The model is based on the interaction between a soil subsystem and a plant subsystem, each being described in terms of their content of C and N, respectively. The first development and application of the model were done in a context of nutritional consequences of whole-tree harvesting (Rolff and Ågren, in preparation). The model has later been extended and adopted to answer other questions as well, including those presented by the current project.

Figure 13.7 Compartment diagram for the Q model

Plants are represented in the model by trees (separated into components) and a 'grass' component that includes all vegetation not in the dominant tree layer. Individuals are not distinguished. Each model component has one C and one N state variable (Figure 13.7). Tree biomass is separated into foliage (by age-class), branches, stems, coarse roots (including stumps), and fine roots. The primary role of the single grass component is to conserve mineralized N after clear-felling. Soil organic matter is divided into litter cohorts, thus keeping track of both the source of the material as well as the time of its appearance in the soil. The time step is one year.

Growth of all tree components (except foliage greater than one year old) is described by the N productivity concept (Ågren 1983, 1985, 1988), where potential maximum growth is reduced as a function of needle biomass and N content. Needles greater than one year old are created by transfer from the next younger year-class. Growth of grass is calculated from N uptake by grass and a constant N concentration, with an assumed maximum biomass. A constant fraction of leaf and root biomass is transferred to the soil in each time step. In addition, plant components are lost at harvests.

The behavior of a litter cohort is described by two variables, amount and quality (Ågren and Bosatta 1987; Bosatta and Ågren 1985, 1991a). Quality measures the degradability of the litter cohort, and decreases with time. Modelling of C mineralization is based on the assumption that decomposers are C limited. Both litter amount and quality decrease over time as a function of microbial utilization. Total soil C is the sum of all litter cohorts.

Uptake of N by trees depends on N availability and the capacity of trees to take it up. Nitrogen availability is the sum of the current year's N mineralization, deposition, and fertilization. Uptake capacity is a negative exponential function of fine root biomass; small trees could assimilate only a small fraction of the available N, while a fully developed stand could assimilate it all. Uptake is restricted by a maximum N concentration in needle biomass. Nitrogen required for growth of all tree compartments but needles is calculated assuming fixed N concentrations; the remainder is transferred to needle biomass. Any N deficit for a given year is drawn from the needles. Nitrogen loss from plants occurs with loss of biomass at constant N concentrations. Nitrogen concentration of senescing branches is lower than that of live ones, creating an internal recycling that is added to the N taken up.

Nitrogen mineralization is derived from changes in the total amount of organically bound N in the soil. The difference between that in the current year and in the previous year is net N mineralization. The N amount of a litter cohort is calculated from the N concentration in microbial biomass and the initial N: C ratio of the litter cohort. Microbial N concentration is an increasing function of available soil inorganic N, which provides a way of immobilizing external inputs of N.

13.4 METHODS

13.4.1 Design

We compared estimates from seven forest models (Table 13.1) against data from two intensively studied sites with contrasting climates (Biology of Forest Growth (BFG) site, 10-14-year-old Pinus radiata near Canberra, Australia; and Swedish Coniferous Biome (SWECON) site, 14-30-year-old P. sylvestris near Jädraås, Sweden). We selected these sites because:

(a) detailed information about ecosystem processes was available;
(b) both sites had similar treatments to manipulate water and nutrient availability (control, fertilization, irrigation, irrigation plus fertilization);
(c) they represented physiologically similar species growing in very different climates (Table 13.2). Detailed site descriptions are given below (section 13.4.2).

Treatments were started in 1974 for SWECON and in 1983 for BFG. Each model estimated stem biomass, foliar N content, annual foliage and stem wood production for 1984-87 (BFG) and for 1979-90 (SWECON). Additionally, each model estimated as much of a complete C, N, and water budget as possible for a single year for each site. For the time-series comparisons, cumulative errors could affect results. For the complete budgets, all models were initialized with standing crops in 1979 (SWECON) or 1986 (BFG).

Modellers were provided with a summary of the data for the BFG and SWECON sites, assembled from the sources given in Tables 13.3 and 13.4. Climate data (daily values of maximum, minimum, and average temperature, relative humidity, precipitation, and shortwave radiation) for the duration of the studies were also provided. Additional data or parameters required for model runs were the responsibility of the individual modellers. Special adjustments to the models necessary for the BFG and SWECON sites are summarized below.

Table 13.2 Long-term average temperature and precipitation of the BFG (1929-82) and SWECON (1931-60) sites.

13.4.1.1 BIOMASS  

The early development of the BIOMASS model occurred in conjunction with the BFG field experiment (Benson et al. 1992a). Because the model has been parameterized for the BFG site (McMurtrie et al. 1990b, 1992; McMurtrie 1993; McMurtrie and Landsberg 1992), comparison of simulations and BFG biomass data did not constitute an independent test of its performance. Adapting the model to describe Pinus sylvestris growing at SWECON in the northern hemisphere required modifications to assumed annual patterns of biomass production and litterfall. Aboveground production was restricted to a three-month period, June to August (S. Linder, pers. comm.), with assimilates produced outside these months allocated to either storage or roots. Needle longevity was set at three years with needle fall occurring between August and October.

13.4.1.2 BIOME-BGC  

Changes were made to the published version of the model (Running and Gower 1991), mainly to accommodate the heavy fertilization treatments. The maximum ratio of foliage to foliage + fine roots was allowed to increase to 0.67 from the published value of 0.5. Nitrogen loss for these runs increases exponentially with N availability (capped at 75% of available N), compared with a linear loss rate in the published version. Maximum photosynthesis was adjusted to data taken at the sites, by altering the parameter controlling the fraction of leaf N in RubP carboxylase. A Farquhar-type photosynthesis model was used for these simulations (Hunt and Running 1992).

13.4.1.3 CENTURY  

The forest version of CENTURY was calibrated to the BFG and SWECON sites in a similar fashion as for other forest sites. Site-specific parameters included C:N ratios of biomass components, leaf mortality, and initial C:N ratios of soil organic matter pools. Parameters controlling production and C allocation were calibrated to match the control and irrigation + fertilization treatments. Biomass C:N ratios and leaf longevity are not normally altered in response to model treatments. However, for these runs, the control parameters for biomass C:N and leaf longevity were manually adjusted to reflect the data for the fertilizer treatments. At the SWECON site, it was necessary to double maximum allowable C:N ratios for intermediate-and slow-turnover soil C pools in order for CENTURY to realistically simulate observed soil C:N ratios.

13.4.1.4 HYBRID  

Species-specific parameters for allometry and the ratio of sapwood area to leaf area were derived from the literature. Optimal temperature for electron transport was set at 31^oC, living sapwood fraction of total sapwood was set at 5.5 %, fine root turnover was set at twice biomass. Foliar respiration was calculated using Ryan (1991a), and leaf litter production was set at a constant 33% of foliage biomass. The HYBRID model was run only for the BFG site.

13.4.1.5 MBL-GEM  

Parameters for the MBL-GEM model were derived by calibrating (as described in section 13.3.6) to year 1 for the control stands at each site. Annual gross production was estimated using equations in Ryan (1991a). The MBL-GEM model was designed as an equilibrium model where results applied to mature stands. To simulate the stand development of young, rapidly aggrading stands, parameters controlling growth were calibrated to the first year's growth of the control stands, and turnover of 'humus' was set to that of a mature stand.

13.4.1.6 PnET-CN  

A simple one-box soil model was added to the published version of PnET (Aber and Federer 1992) for these model runs. The soil model uses the C:N ratio of the active pool to estimate gross and net N mineralization. Plant N availability is estimated as net N mineralization.

13.4.1.7 Q

Model runs used a constant, average climate. Differences in water availability for the irrigated and nonirrigated treatments at BFG were simulated by varying the reduction in N productivity per unit biomass between treatments. Water availability was assumed to have no effect at SWECON.

13.4.2 Site descriptions

13.4.2.1 Biology of Forest Growth experiment  

The Biology of Forest Growth (BFG) experiment is located at 630 m elevation in the Pierce's Creek Forest about 20 km west of Canberra in the Australian Capital Territory (35° 21' S, 148° 56' E). The site, climate and the experiment have been described in detail by Benson et al. (1992a), and key aspects are outlined below. A partial list of values used for modelling and sources for data are given in Table 13.3.

Table 13.3 Site and stand conditions and data sources used for simulations of the BFG site

The site was cleared of the original eucalypt woodland in 1934, broadcast burned and planted to Pinus radiata in 1935. The pine crop was clear-felled in 1972, residues burned, and the second pine stand planted in 1973. The forest was thus 10 years old when the BFG experiment commenced in winter 1983; at that time tree density was ~ 700 ha ^-1, height ~ 9.5 m, mean tree diameter (1.4 m) ~ 15 cm and stand basal area ~ 12.5 m² ha^-1 (full details given by Snowdon and Benson 1992).

The soil supporting the forest is a yellow podzolic (Stace et at. 1968) or typic albaqualf (Soil Survey Staff 1975) derived from granite. It has a duplex profile comprising a sandy A horizon extending to about 40 cm depth, and a very dense (bulk density 1.7 g cm^-3) B horizon of clay and gravel overlying a C horizon of weathering granite (> 100 cm). The soil has a very low N-supplying capacity (Raison et al. 1992a) but provides adequate P for tree growth (Raison et al. 1990). Soil water storage capacity is only, ~75 mm in the A horizon where most fine roots are concentrated, and", 250 mm to a depth of 2 m (Myers and Talsma 1992). The combination of low soil water storage and high probability of summer drought results in trees suffering severe water stress during the summer-early autumn period (Myers and Talsma 1992).

The site experiences mild winters (mean daily minimum and maximum temperatures in July are 1 and 11 °C) and warm summers (14 and 28 °C in January). Summer maxima may reach 35-40 °C when relative humidity is very low ( < 10%). Annual rainfall averages 791 mm (Table 13.2), but this is highly variable with summer droughts of 2-3 months' duration common; during the life of the stand, < 00 mm of rainfall in the combined three summer months occurred in five growing seasons (1978-79, 1981-82, 1982-83, 1984-85, and 1985-86). 

During 1983 and 1984, combinations of N fertilization and irrigation were applied to (0.25-ha plots) with the objective of markedly modifying growth rates to facilitate study of the mechanisms by which N and water interact to control productivity. The treatments were: control (C), irrigated (I), fertilized (F), irrigated-fertilized (IF) and irrigated-liquid fertilized (IL). The irrigation treatment aimed to maintain soil moisture near field capacity, and approximately doubled annual rainfall. Water was added as often as every 2-3 days, based on a simple model of water use and fortnightly adjustment based on measurement of soil water content. Irrigation was not applied during the summer of 1983-84 because the growing season was too wet. Solid fertilizer (F and IF treatments) was added in spring 1983 and contained 400 kg N, 200 kg P, 100 kg K and 10 kg B ha ^-1. The IL treatment commenced in spring 1984 and provided a balanced nutrient input weekly in the irrigation water and was varied seasonally to match tree demand; annual inputs of N totalled ~ 300 kg ha ^-1. The IL treatment aimed to remove both water and nutrients as limitations to growth.

A synthesis of the major research findings from the BFG experiment has recently been published (Raison and Myers 1992a). Irrigation alone increased NPP by about 40-50% (Benson et al. 1992b; cf. Figure 13.8), and fertilization increased it by about 25-30% except when the growing season was very dry. A marked interaction between irrigation and fertilization resulted in about 200% increase in NPP. The treatments affected the rate of canopy development and stand foliage carrying capacity (Raison et al. 1992c), the period of active photosynthesis (Thompson and Wheeler 1992) and the efficiency of light utilization for biomass production (Raison and Myers 1992b). There is also evidence that a larger proportion of photosynthate may be allocated in below ground processes in those BFG stands having low N status (S. Pongracic and R.J. Raison, pers. comm., 1992).

13.4.2.2 Swedish Coniferous Forest Project

 The Swedish Coniferous Forest Project (SWECON) was started in 1972 to study the structure and function of conifer forests in Sweden. As part of this effort, extensive studies on element cycling within young Scots pine (Pinus sylvestris) stands were established at Ivantjärnsheden near Jädraås, Sweden (60° 49' N, 16° 30' E). Climate, soils, and stand characteristics of the Ivantjärnsheden site have been described by Axelsson and Bråkenhielm {1980) and many results from the study have been described in T. Persson {1980). A partial list of values used for modelling and sources for data are given in Table 13.4.

Ivantjärnsheden is located in the southern end of the 'northern coniferous' region. The site is a sandy plain and soils are sandy and podzolic with a medium sand texture (Axelsson and Bråkenhielm 1980). The A_o horizon varies from 1 to 7 cm, A₁from 0.2 to 1.6 cm, and A₂ from 2 to 7 cm; the litter layer is ~1 cm. For the Ih II stands at Ivantjärnsheden, the water table is 10 m below the surface and the soil water capacity in the 0-30 cm root zone is 72 mm. A low, dense understorey of lichen and dwarf shrubs exists in these stands.

Table 13.4 Site and stand conditions and data sources used for simulations of the Swedish Coniferous Biome site

The site experiences cool summers with long days, and long, cold winters. Mean monthly air temperature is 15.8 °C in July and -7 °C in January (Table 13.2; Axelsson and Bräkenhielm 1980). Mean annual precipitation is about 600 mm, varying between 413 and 647 mm (1931-60). Because the sandy soils have low water storage capacity, summer droughts occur in this ecosystem.

Data for the model comparison was collected at stands designated Ih II. The Ih II stand was harvested in 1957 and seed trees were felled in 1962; the naturally regenerated stand was about 14 years old at the start of treatments in 1974. In 1973, the stand had 1095 stems ha^-1, a basal area of 2.15 m2 ha^-1, and an average height of 2.1 m (Flower-Ellis and Persson 1980). Replicate plots of 20 x 20 m were established for control, irrigated (100-300 mm yr^-l), fertilized (8 9 N m^-2 yr^-1), and irrigated+fertilized (100-300 mm yr^-1; 7-20 g N m^-2 yr^-1) treatments (Aronsson and Elowson 1980). Fertilization resulted in dramatic increases in productivity (Linder and Axelsson 1982; cf. Figures 13.8 and 13.10) and treatment response persisted long after fertilization was discontinued.

13.5 COMPARISONS WITH DATA AT BFG AND SWECON CONIFER SITES

13.5.1 Stem biomass, foliage nitrogen, aboveground productivity

Most models closely estimated stem biomass at both the BFG and SWECON control stands (Figure 13.8). Models that used data values for foliar N (BIOMASS, HYBRID) or a calibration to the fertilized treatment (CENTURY) were also able to closely simulate the response of stem biomass to irrigation and fertilization. Where N fluxes were explicitly modelled (MBL-GEM, PnET-CN, BIOME-BGC, Q), model predictions of treatment response were less successful. Overestimates by Q and PnET-CN for some site and treatment combinations were partially caused by differences in initial conditions (Q grew stands from seedlings, PnET from the initial year of the study). Models differed little in their relative predictions for the two sites.

Foliar N mass is a critical variable for most models because it can affect photosynthesis, respiration, and litter decomposition and nutrient mineralization. The models that simulate N allocation (BIOME-BGC, CENTURY, MBL-GEM, PnET -CN, Q) were generally successful at estimating foliar N mass for the control and irrigated treatments (Figure 13.9), but less so for the fertilizer treatments. Modelling the fate of large fertilizer applications in a mechanistic manner appears to remain problematic.

Foliage production is critical because foliage mass partially controls transpiration and C fixation, and foliage turnover affects decomposition and nutrient mineralization. Stem production is thought to have low priority for fixed C (Waring and Pitman 1985); hence stem production should be a sensitive indicator of shifts in the annual C budget. Models were generally successful at estimating foliage and stemwood production for the control and irrigated treatments (Figures 13.10 and 13.11), but less so for the fertilizer treatments. Most models successfully predicted the twofold difference in the rate of stem production between the two sites. Where foliar N was obtained from data (BIOMASS, HYBRID), response of foliage and stem production closely matched data.

13.5.2 Single-year carbon budgets

Model estimates of an annual C budget for BFG and SWECON are given in Figure 13.12. Differences were apparent between data (where they exist) and model predictions, and also among the predictions of different models. Differences among model estimates are not the result of differences in initialization or cumulative errors: these C budgets were estimated starting from the same initial conditions (standing crops in 1979 for SWECON and 1986 for BFG) and climate. Therefore, the differences shown result from different representations of processes, their interactions, and parameterization.

Figure 13.8 Stem biomass C and model predictions for BFG and SWECON sites. For BFG, Q was initialised in 1973 from seedlings and PnET-CN in 1983; the other models started in 1984. For SWECON, Q was initialised in 1960 from seedlings and PnET-CN and GEM in 1974  (the beginning of the experiment); the other models started in 1979. Values for Q for the fertilized + irrigated (BFG) and fertilized (SWECON) treatments are off scale for the latter years of the simulations 

Figure 13.9 Foliage N mass and model predictions for BFG and SWECON sites. For BFG, Q was initialized in 1973 from seedlings and PnET-CN in 1983; the other models started in 1984. For SWECON, Q was initialized in 1960 from seedlings and PnET-CN and GEM in 1974 (the beginning of the experiment); the other models started in 1979. For CENTURY, foliage N mass is N content the foliage would have if it were fresh litter (after retranslocation). Actual foliar N for CENTURY would be roughly double the given values. Foliage N mass for Q for BFG irrigated + fertilized treatment is off scale at about 55 g m ^–2.

Figure 13.10 Foliage production ( C) and model predictions for BFG and SWECON sites. For BGF, Q was initialized in 1973 from seedlings and PnET-CN in 1983; the other models started in 1984. For SWECON, Q was initialized in 1960 from seedlings and PnET-CN and GEM in 1974 (the beginning of the experiment); the other models started in 1979. Note difference in scales for BFG and SWECON sites

Figure 13.11 Stem biomass production (as C) and model predictions for BFG and SWECON sites. For BFG, Q was initialized in 1973 from seedlings and PnET-CN in 1983; the other models started in 1984. For SWECON, Q was initialized in 1960 from seedlings and PnET-CN and GEM in 1974 (the beginning of the experiment); the other models started in 1979. Stemwood production for Q for the irrigated + fertilized treatment at SWECON is off scale

Figure 13.12 Annual estimates of C fluxes for BFG and SWECON sites. Allocation to root is fine roots for production for HYBRID, and both coarse and fine records the other models. Respiration is growth respiration for all tissues plus maintenance respiration for above-and belowground woody tissues and fine roots. Growth respiration was estimated for CENTURY as 25% of growth. Q does not estimate respiration; BFG data do not include fine root and respiration estimates. Note difference in scale for BFG and SWECON

The total height of the bars represents model estimates of annual net canopy fixation (gross C fixation minus any foliar maintenance respiration). Model estimates were generally within 50% of the mean for all treatments at both sites. Differences among models may result from differences in the photosynthesis model, extinction of light within the canopy, aggregation of fine-scale processes to the canopy, aggregation over time, or sensitivity to climate.

The irrigation + fertilization treatments are useful for comparing models, because both ecosystem/tissue and physiologically based models should agree when resources are not limiting. All models except BIOME-BGC matched data for this treatment at the BFG site (Figure 13.12), and all participating models except CENTURY matched data for this treatment at the SWECON site.

Most models assume that wood production has low priority for fixed C (see Waring and Pitman 1985). Therefore, allocation to wood should respond most to treatment and climate. Model predictions of C allocation to wood did differ within a treatment and among treatments. Allocation of net canopy fixation to wood production by the models averaged 24% at the SWECON site but 37% at the BFG site. At SWECON, wood production:net canopy fixation ranged from 6% (BIOME-BGC) to 28% (CENTURY) for the control stand and 25% (BIOME-BGC) to 33% (MBL-GEM) for the irrigated and fertilized stand. At BFG, wood production:net canopy fixation ranged from 26% (PnET-CN) to 44% (CENTURY) for the control stand and 28% (PnET-CN) to 51% (BIOME- BGC) for the irrigated and fertilized stand.

Assumptions about how climate and nutrition affect allocation to root production differ among models. The ratio of foliage production to the total of foliage plus fine root production (F: F + R) estimates shifts in root allocation. At SWECON, F:F+R varied from 22% (BIOME-BGC)to 68% (Q)for the control stand and 25% (BIOME-BGC) to 76% (Q) for the irrigated and fertilized stand. At BFG, F:F+R varied from 23% (BIOME-BGC) to 53% (Q) for the control stand and 20% (BIOME-BGC) to 58% (BIOMASS) for the irrigated and fertilized stand. The F:F + R ratio was most sensitive to treatment for BIOMASS and Q, less sensitive to treatment for MBL-GEM, BIOME-BGC, and HYBRID, and insensitive to treatment for CENTURY and PnET-CN.

13.5.3 Single-year nitrogen budgets

For those models that simulate N budgets, model estimates of net N mineraliz- ation and plant uptake were variable, and larger than field estimates (Figure 13.13). All models except Q estimated a much greater N mineralization at the BFG site. Fertilization did not greatly affect N mineralization for any model except Q at the BFG site. Differences in plant uptake in turn influenced the C budgets discussed in section 13.5.2, because tissue N concentrations affect photosynthesis and respiration in most models.

Figure 13.13 Annual estimates of N fluxes for BFG and SWECON sites. N loss includes leaching and gaseous losses. Models shown with data estimate all three components, except PnET-CN, which only provided net N mineralization. Note differences in scale for the irrigated + fertilized treatments

13.5.4 Single-year water budgets

Total annual water flux agreed closely with data for all models that simulate water balance, because total flux is nearly equal to total annual precipitation (Figure 13.14). Model predictions for the partitioning of water to transpiration, evaporation, and groundwater loss differed, particularly for the irrigated treatments.

13.6 DISCUSSION

Our comparison and validation of models had two purposes. First, we wished to assess our understanding of temperate conifer forests as captured in state-of-the- art models. Is our understanding adequate to model, from first principles, element budgets of forest ecosystems and their response to dramatic changes in resource availability? Because global change will affect resource availability, models must successfully simulate these changes to reliably predict growth in future climates. Second, we wished to determine if the different hypotheses about controls over processes and their interactions affected model response to changes in resources. Are models that explicitly link production and decomposition more robust simulators than models that consider only climate and production?

The experimental sites represented a severe challenge to the models because dramatic shifts in resources occurred over a short period of time. In general, models had difficulty (1) estimating the fate of applied fertilizer, (2) allocating C and especially N within plants, and (3) estimating actual C and N fluxes. Models that directly simulated the N cycle (all except BIOMASS and HYBRID) had serious problems determining how much N was available to plants, how much was taken up, and where in the tree N accumulated. Because most models estimate photosynthesis and respiration as some function of tissue N content, errors in plant uptake and allocation of N caused errors in other fluxes. However, the agreement between measured and modelled fluxes for the irrigation + fertilization treatments is encouraging.

With the exception of the handling of the N cycle, no consistent differences in model performance or response to treatment were apparent for models that differed greatly in detail, functional equations, and spatial and temporal aggregation. However, because models differed greatly, it was difficult to isolate reasons for differences in model estimates. Below, we discuss individual processes in more detail.

Figure 13.14 Annual estimates of water fluxes for BFG and SWECON sites. Loss is runoff or water transported to groundwater. BIOMASS model does not estimate loss for irrigated treatments

13.6.1 Photosynthesis and canopy dynamics

Despite starting from the same initial conditions, and using the same climate, model estimates for net canopy photosynthesis differed substantially when water and nutrients were limiting. Differences in model estimates likely resulted from differences in aggregation, plant N uptake over the year, response to climate, and whether changes in canopy (seasonality) were explicitly modelled. Seasonal changes in the canopy were quite dramatic at the BFG site (Raison et al. 1992c), and models that assumed a constant canopy may have difficulty estimating annual photosynthesis. Until we can reliably measure canopy photosynthesis, or reliably sum respiration and dry matter production, our models of photosynthesis will be uncertain. However, model agreement for simulations with non-limiting resources is encouraging, and suggests that we understand the limits to productivity but need more work to understand resource interactions.

13.6.2 Respiration

Because models generally view maintenance respiration as a fixed cost (having the first priority for fixed C), respiration can dramatically affect dry matter production. Estimates for respiration differed greatly, particularly in response to treatment. All models partitioned respiration into the components of growth and maintenance and used a similar exponential response to temperature. Therefore, differences in model estimates were caused by differences in biomass (BIOME-BGC, HYBRID) or tissue N content (BIOMASS, CENTURY, MBL-GEM, PnET-CN) that developed over the course of the year. Differences in model estimates were also caused by differences in respiration rates and the parameters controlling temperature response. No model simulated acclimation of respiration rates to climate (Amthor 1989), a process that could affect the response of productivity to global change.

13.6.3 Carbon and nitrogen allocation

Allocation of C and N to tissues (wood, roots, and foliage) from first principles remains a challenge. For C, correctly estimating allocation depends initially on correctly estimating photosynthesis and respiration. However, relative allocation is also very important because allocation can alter the amount of tissue in foliage, wood, and roots and thus affect future rates of photosynthesis, respiration, and nutrient uptake.

Models disagree about the partitioning of C to foliage and roots for the control stands, and for the response to treatment. Not surprisingly, models with flexible allocation (BIOMASS, BIOME-BGC, HYBRID, MBL-GEM, Q) decreased the relative allocation to roots for the fertilized stands. However, models with fixed allocation (PnET-CN and CENTURY) did not do noticeably worse (compared with data) than the models that altered allocation dynamically.

Allocation of the N taken up by plants to various types of plant tissue has received little experimental attention. However, this understanding is essential for successful modelling, because photosynthesis, respiration, (and perhaps C allocation to fine roots and foliage) are strongly controlled by tissue N contents.

13.6.4 Nitrogen cycle and decomposition

The lack of agreement among model estimates of N mineralization and plant uptake reflects both differences in model structure and our lack of understanding of controls over these processes. Because the N cycle is so important for the linked models of production and decomposition, differences in N mineralization or plant uptake yield differences in photosynthesis, respiration, and litter quality. Also, soil organic matter content and turnover control much of the N mineralization, but change slowly. Therefore, simulated rates of turnover depend more on assumed initial conditions than on the rates of recycling litterfall, and it is difficult to evaluate changes in these processes for short-term experiments.

Non-biological immobilization of N into soil organic matter can strongly affect N availability after fertilization (e.g. Nömmik and Vahtras 1982; Axelsson and Berg 1988). Perhaps because none of these models included an explicit representation of this process, the models had difficulty simulating both N availability in the soil and the distribution of N to plants and soil organic matter after heavy fertilization.

Models that simulate the N cycle agree that litter quality varies as some function of the content of lignin and N, and that N immobilization and mineralization from litter vary with litter quality. However, models have not reached a consensus about what controls C and N mineralization from 'humus' (older, slowly decaying soil organic matter). For example, it is uncertain whether litter quality and N in the soil solution affect decomposition rates and C:N of humus. If litter quality does affect humus quality and turnover, the changes in litter quality expected from increased atmospheric CO₂ and changes in species composition could markedly affect long-term response of ecosystems to global change.

13.6.5 Hydrology  

The hydrological cycle is better understood (Landsberg et al. 1991) than the C and nutrient cycles. Therefore, modelling hydrology has generally been more successful. Also, ecosystem estimates for precipitation and runoff, while uncertain, are readily available and constrain model predictions.

Models that simulated the hydrologic cycle agreed well for total water flux in this exercise because the total water flux was constrained by precipitation. However, differences in model estimates for the components of transpiration, evaporation, and loss to groundwater are problematic and may affect model predictions of response to a changing climate. Predictions for models that used a daily time step (BIOMASS, BIOME-BGC, HYBRID) agreed reasonably well with each other and with one of the monthly models (PnET -CN). The PnET-CN model's estimates of transpiration derived from monthly estimates of C assimilation agreed well with estimates of transpiration derived with the Pen- man-Monteith approach.

The CENTURY model showed considerable differences from the other models in the apportioning of water to transpiration versus evaporation. It does not model transpiration directly, and its C and N dynamics are not directly sensitive to it. The water budget, production, and decomposition in this model are controlled by precipitation and estimated evapotranspiration. Estimated evapotranspiration is divided between evaporation and transpiration by using linear relations between evaporation, canopy biomass, and precipitation. Evaporation is given first priority for water, with the remainder allocated to transpiration. These linear relationships produce high estimates of evaporation and low estimates of transpiration (relative to other models) when precipitation (or irrigation) and biomass are large.

13.7 RECOMMENDATIONS FOR EXPERIMENTATION AND MODELLING

13.7.1 Experimentation

Reliable estimates of fluxes at the ecosystem level are problematic for C, water, and nutrients (Landsberg et al. 1991), although much recent progress has been in estimating C and water fluxes with eddy-covariance technology. This lack of reliable flux estimates hinders evaluation of models. Additionally, ecosystem-level flux measurements are generally of net fluxes, while component fluxes are most useful for model validation. Because processes change dramatically with stand development and since most of our knowledge of these processes is derived from young stands, we also need more information from mature forests.

To evaluate models of ecosystem processes, a small number of studies where resources are manipulated and all of the components are measured can be as valuable as estimates of a few components (e.g. net primary production) at many sites. Manipulation experiments directly test hypotheses about the controls on processes such as allocation and nutrient retention and uptake. Therefore, the benefits of studies like BFG and SWECON to modelling efforts are enormous. Another imperative is the compilation and publication of integrated, whole- system data sets (such as those used for this comparison) for developing and testing models. The lack of such data, particularly for natural ecosystems, is crippling our efforts to understand the impacts of human activities on the biosphere.

On the process level, we have a critical need to understand the controls over the allocation of C and N within trees. We also need to understand the long-term controls over N mineralization from soil organic matter. Allocation of C has received attention, but we are still uncertain about the responses of below ground allocation and respiration to changes in resource availability. However, because of the importance of nutrition in defining photosynthesis (Field and Mooney 1986), respiration (Amthor 1989; Ryan 1991b) and allocation (Thornley 1972a; Field 1983; Bazzaz et al. 1987; Ingestad and Ågren 1991), we need to focus on the controls over the uptake and allocation of N as well. Finally, we need to search for and define the limits for allocation ratios, tissue N concentrations and mineralization rates.

13.7.2 Modelling

This modelling exercise helped identify the strengths and limitations of the participating models. Without the rigorous validation and exposure to new sites in this study, model limitations could not have been identified. Limitations found have generally fostered new work and searches for alternative approaches.

Because this model comparison was useful but difficult to execute, we need to develop better methods for comparing models. We need to strive harder to identify whether differences are trivial (caused by differences in rate constants), or important (caused by differences in model formulation or interactions within a model). Development of a common model framework might be one mechanism to accomplish this task. Such a framework might specify the processes to be modelled, the driving variables, and inputs and outputs to the various sub models. Formulations for the sub models, the interactions between sub models, and level of aggregation over space and time could remain flexible. Under such a system, it would be simple to determine whether another model of photosynthesis or respiration caused real differences in model estimates, or whether different hypotheses about interactions between production and decomposition affected long-term productivity.

However, if such a common model framework were implemented, the model framework should be flexible enough to allow real differences to exist. Some of the most important differences in models may not be in the simulation of individual processes, but how they are linked at a higher level. The overall strategy for assessing the effects of global change should stress a diversity of approaches. It is this diversity of approaches that will either give us confidence in our predictions (if the models agree) or indicate unknowns (if models do not agree). This diversity of approach could be incorporated by making the common structure modular and hierarchical. Thus, submodels might represent simply different formulations for a process (e.g. photosynthesis) or might represent a totally new approach to simulating ecosystems.

Despite some of the differences among model predictions, models were generally successful at estimating the large differences in fluxes resulting from the different climates in Sweden and Australia and in estimating fluxes where water and nutrients are not limiting.

13.8 REFERENCES

Aber, J. D. and Federer, C. A. (1992) A generalized, lumped-parameter model of photosynthesis, evapotranspiration and net primary production in temperate and boreal forest ecosystems. Oecologia 92, 463-474.

Aber, J. D. and Melillo, J. M.(1982) FORTNITE: a computer model of organic matter and nitrogen dynamics in forest ecosystems. Res. Bull. R3130, University of Wisconsin.

Aber,J. D., Melillo,J. M. and Federer, C. A.(1982) Predicting the effects of rotation length, harvest intensity, and fertilization on fiber yield from northern hardwood forests in New England. For. Sci. 28, 31-45.

Ågren, G. I. (1983) Nitrogen productivity of some conifers. Can. J. For. Res. 13, 494-500. Ågren, G. I. (1985) Theory for growth of plants derived from the nitrogen productivity concept. Physiol. Plant. 64, 17-28. 

Ågren, G. I. (1988) Ideal nutrient productivities and nutrient proportions in plant growth. Plant, Cell Environ. 11, 613-620.

Ågren, G. I. and Bosatta, E. (1987) Theoretical analysis of the long term dynamics of carbon and nitrogen in soils. Ecology 68, 1181-1189.

Ågren, G. I. and Bosatta, E. (1988) Nitrogen saturation of terrestrial ecosystems. Environ. Pollut. 54, 185-197.

Ågren, G. I., McMurtrie, R. E., Parton, W. J., Pastor, J. and Shugart, H. H. (1991) State-of-the-art of models of production-decomposition linkages in conifer and grassland ecosystems. Ecol. Appl. 1, 118-138.

Amthor, J. S. (1989) Respiration and Crop Productivity. Springer-Verlag, New York. 215 pp.

Aronsson, A. and Elowson, S. (1980) Effects of irrigation and fertilization on mineral nutrients in Scots pine needles. In: Persson, T. (Ed.) Structure and Function of Northern Coniferous Forests-an Ecosystem Study. Ecol. Bull. 32, 219-228. Stockholm.

Axelsson, B. and Berg, B. (1988) Fixation of ammonia (¹⁵N) to Pinus sylvestris needle litter in different stages of decomposition. Scand. J. For. Res. 3, 273-279.

Axelsson, B. and Bråkenhielm, S. (1980) Investigation sites of the Swedish Coniferous Forest Project- biological and physiographical features. In: Persson, T. (Ed.) Structure and Function of Northern Coniferous Forests-an Ecosystem Study, Ecol. Bull. 32,25-64, Stockholm.

Bazzaz, F. A., Chiariello, N. R., Coley, P. D. and Pitelka, L. F. (1987) Allocating resources to reproduction and defense. BioScience 37, 58-67.

Beets, P. N. and Pollock, D. S. (1987) Uptake and accumulation of nitrogen in Pinus radiata stands as related to age and thinning. New Zeal. J. For. Sci. 17, 353-371.

Benson, M. L., Landsberg, J. J. and Borough, C. J. (1992a) The biology of forest growth experiment: an introduction. For. Ecol. Mgmt 52, 1-16.

Benson, M. L., Myers, B. J. and Raison, R. J. (1992b) Dynamics of stem growth of Pinus radiata as affected by water and nitrogen supply. For. Ecol. Mgmt. 52, 117-137.

Booth, T. H. and McMurtrie, R. E. ( 1988) Climatic change and Pinus radiata plantations in Australia. In: Pearman, G.I. (Ed.) Greenhouse: Planning for Climate Change, pp. 534-545. E.J. Brill, Leiden.

Bosatta, E. and Ågren, G. I. (1985) Theoretical analysis of decomposition of heterogeneous substrates. Soil Bioi. Biochem. 17, 601-610.

Bosatta, E. and Ågren, G. I. (1991a) Dynamics of carbon and nitrogen in the organic matter of the soil: a generic theory. Am. Nat. 138, 227-245.

Bosatta, E. and Ågren, G.I. (1991b) Theoretical analysis of carbon and nutrient interactions in soils under energy-limited conditions. Soil Sci. Soc. Am. J. 55, 728-733.

Botkin, D. B., Janak, J. F. and Wallis, J. R. (1972) Some ecological consequences of a computer model of forest growth. J. Ecol. 60, 849-872.

Box, E. (1978) Geographical dimensions of terrestrial net and gross primary production. Rad. Environ. Biophysics 15, 305-322.

Box, E. (1981) Macroclimate and Plant Forms: An Introduction to Predictive Modeling in Phytogeography. Dr W. Junk Publishers, The Hague, Netherlands.

Bringmark, L. (1980) Ion leaching through a podsol in a Scots pine stand. In: Persson, T. (Ed.) Structure and Function of Northern Coniferous Forests-an Ecosystem Study. Ecol. Bull. 32, 341-361, Stockholm.

Burke, I. C., Kittel,T. G. F., Lauenroth,W. K., Snook, P., Yonker,C. M. and Parton,W.J. (1991) Regional analysis of the central Great Plains. BioScience 41, 685-692.

Cannell, M. G. R. (1982) World Forest Biomass and Primary Production Data. Academic Press, London.

Cannell, M. G. R., Grace, J. and Booth, A. (1988) Possible Impacts of Climatic Warming on Trees and Forests in U K: A Review. Report to the Department of Environment and Natural Environment Research Council, Penicuik, United Kingdom.

Crane, W. J. B. and Banks, J. C. G. (1992) Accumulation and retranslocation of foliar nitrogen in fertilised and irrigated Pinus radiata. For. Ecol. Mgmt 52, 201-223.

Emanuel, W. R., Shugart, H. H. and Stevenson, M. P. (1985a) Climatic change and the broad-scale distribution of terrestrial ecosystem complexes. Clim. Change 7, 29-43.

Emanuel, W. R., Shugart, H. H. and Stevenson, M. P. (1985b) Response to comment: climatic change and the broad-scale distribution of terrestrial ecosystem complexes. Clim. Change 7, 457-460.

Farquhar, G. D. and von Caemmerer, S. (1982) Modelling of photosynthetic response to environment. In: Lange, 0. L. Nobel, P. S., Osmond, C. B. and Ziegler, H. (Eds) Encyclopaedia of Plant Physiology, New Series Vol. 12B, pp. 547-587. Springer-Verlag, Berlin.

Farquhar, G. D., von Caemmerer, S. and Berry, J. A. (1980) A biochemical model of photosynthetic CO₂ as simulation in leaves of C₃ species. Planta 149, 78-90.

Field, C. (1983) Allocating leaf nitrogen for the maximization of carbon gain: leafage as a control on the allocation program. Oecologia 56, 341-347.

Field, C. and Mooney, H. A. (1986) The photosynthesis-nitrogen relationship in wild plants. In: Givnish, T.J. (Ed.) On the Economy of Plant Form and Function, pp. 25-55. Cambridge University Press, Cambridge.

Finnegan, J. J. and Raupach, M. R. (1987) Transfer processes in plant canopies in relation to stomatal characteristics. In: Zeiger, E., Farquhar, G.D. and Cowan, I.R. (Eds) Stomatal Function, pp. 385-429. Stanford University Press, Stanford.

Flower-Ellis, J. G. K. and Persson, H. (1980) Investigation of structural properties and dynamics of Scots pine stands. In: Persson, T. (Ed.) Structure and Function of Northern Coniferous Forests-an Ecosystem Study. Ecol. Bull. 32, 125-138. Stockholm.

Friend, A. D. (1991) Use of a model of photosynthesis and leaf microenvironment to predict optimal stomatal conductance and leaf nitrogen partitioning. Plant, Cell Environ. 14, 895-905.

Friend, A. D., Shugart, H. H. and Running, S. W.(1993) A physiology-based gap model of forest dynamics. Ecology 74, 792-797.

Holdridge, L. R. ( 1947) Determination of world plant formations from simple climate data. Science 105, 367-368.

Holdridge, L. R. (1964) Life Zone Ecology. Tropical Science Center, San Jose, Costa Rica.

Hunt, E. R. Jr, Martin, F. C. and Running, S. W. (1991) Simulating the effects of climatic variation on stem carbon accumulation of a ponderosa pine stand: comparison with annual growth increment data. Tree Physiol. 9, 161-171.

Hunt, E. R. Jr and Running, S. W. (1992) Simulated dry matter yields for aspen and spruce stands in the North American boreal forest. Can. J. Remote Sens. 18, 126-133.

Hunt, H. W. (1977) A simulation model for decomposition in grasslands. Ecology 58, 469-484.

Ingestad, T. and Ågren, G. I. (1991)The influence of plant nutrition on biomass allocation. Ecol. Appl. 1, 168-174.

Jarvis, P. G. (1985) Transpiration and assimilation of tree and agricultural crops: the 'omega factor'. In: Cannell, M. G. R. and Jackson, J. E. (Eds) Attributes of Trees as Crop Plants, pp. 460-480. Institute of Terrestrial Ecology, Huntington.

Jarvis, P. G. (1989) Atmospheric carbon dioxide and forests. Phil. Trans.R.Soc.Lond.B 324, 369-392.

Khanna, P. K., Raison, R. J., Falkiner, R. A., Willett, I. R. and Connell, M. J. (1992) Effects of N P K fertilisation on the chemistry of a yellow podzolic soil under Pinus radiata. For. Ecol. Mgmt 52, 65-85.

Kirschbaum, M. U. F. and Farquhar, G. D. (1984) Temperature dependence of whole-leaf photosynthesis in Eucalyptus pauciflora Sieb. ex Spreng. Aust. J. Plant Physiol. II, 519-538.

Korol, R. L., Running, S. W., Milner, K. S. and Hunt, E. R. (1991) Testing a mechanistic carbon balance model against observed tree growth. Can. J. For. Res. 21, 1098-1105.

Landsberg, J. J., Kaufmann, M. R., Binkley, D., Isebrands, J. and Jarvis, P. G. (1991) Evaluating progress toward closed forest models based on fluxes of carbon, water, and nutrients. Tree Physiol. 9, 1-15.

Lieth, H. (1972) Modeling the primary productivity of the world. Nat. Resour. 8, 5-10.

Linder, S. and Axelsson, B. (1982) Changes in carbon uptake and allocation patterns as a  result of irrigation and fertilization in a young Pinus sylvestris stand. In: Waring, R.H. (Ed.) Carbon Uptake and Allocation in Subalpine Ecosystems as a Key to Management: Proceedings of an IUFRO Workshop, pp. 38-44. Forest Research Laboratory, Oregon State University, Corvallis.

Linder, S. and Troeng, E. (1980) Photosynthesis and transpiration of 20-year-old Scots pine. In: Persson, T. (Ed.) Structure and Function of Northern Coniferous Forests-an Ecosystem Study. Ecol. Bull. 32, 165-181, Stockholm.

McGill, W. B., Hunt, H. W., Woodmansee, R. G. and Reuss, J. O. (1981) PHOENIX, a model of the dynamics of carbon and nitrogen in grassland soils. Ecol. Bull. 33, 49-115, Stockholm.

McGuire, A.D., Joyce, L.A., Kicklighter, D.W., Melillo, J.M., Esser, G. and Vörösmarty, C.J. (1993) Productivity response of climax temperate forests to elevated temperature and carbon dioxide: a North American comparison between two global models. Clim. Change 24, 287-310.

McGuire, A. D., Melillo, J. M., Joyce, L. A., Kicklighter, D. W., Grace, A. L., Moore, B. III and Vörösmarty, C. J.(1992) Interactions between carbon and nitrogen dynamics in estimating net primary productivity for potential vegetation in North America. Global Biogeochem. Cyc. 6, 101-124.

McMurtrie, R.E. (1993) Modelling canopy carbon and water balance. In: Hall, D.O., Scurlock, J.M.O., Bolar, H., Leegood, R.C. and Long, S.P. (Eds) Photosynthesis and Production in a Changing Environment: A Field and Laboratory Manual, pp. 220-231. Chapman & Hall, London.

McMurtrie, R.E., Benson, M.L., Linder, S., Running, S.W.,Talsma, T., Crane,W.J.B. and Myers, B.J. (1990a) Water/nutrient interactions affecting productivity of stands of Pinus radiata. For. Ecol. Mgmt 30, 15-423.

McMurtrie, R.E. and Landsberg, J.J. (1992) Using a simulation model to evaluate the effects of water and nutrients on the growth and carbon partitioning of Pinus radiata. For. Ecol. Mgmt 52, 243-260.

McMurtrie, R.E., Leuning, R., Thompson, W.A. and Wheeler, A.M. (1992) A model of canopy photosynthesis and water use incorporating a mechanistic formulation of leaf CO₂ exchange. For. Ecol. Mgmt 52, 261-278.

McMurtrie, R.E., Rook, D.A. and Kelliher, F.M. (1990b) Modelling the yield of Pinus radiata on a site limited by water and nitrogen. For. Ecol. Mgmt 30, 381-413.

McMurtrie, R.E. and Wang, Y.-P. (1993) Mathematical models of photosynthetic response of plant stands to rising CO₂ concentrations and temperatures. Plant, Cell Environ. 16, 1-16.

Meentemeyer, V., Box, E.O. and Thompson, R. (1982) World patterns and amounts of terrestrial plant litter production. BioScience 32, 125-128.

Melillo, J.M. (1983) Will increases in atmospheric CO₂ concentrations affect decay rates? In: The Ecosystems Center 1983 Annual Report, pp. 10-11, Marine Biological Laboratory, Woods Hole, Mass.

Melillo, J.M., Aber, J.D. and Muratore, J.F.(1982) Nitrogen and lignin control of hardwood leaf litter decomposition. Ecology 63, 621-626.

Melillo, J.M., McGuire, A.D., Kicklighter, D.W., Moore, B.III, Vörösmarty, C.J. and Schloss, A.L. (1993) Global climate change and terrestrial net primary production. Nature, 363, 234-240.

Mooney, H.A., Pearcy, R.W. and Ehrlinger, J. (1987) Plant physiological ecology today. BioScience 37, 18-20.

Myers, B.J. (1988) Water stress integral-a link between short-term stress and long-term growth. Tree Physiol. 4, 315-323.

Myers, B.J. and Talsma, T. (1992) Site water balance and tree water status in irrigated and fertilized stands of Pinus radiata. For. Ecol. Mgmt 52, 17-42.

Nemani, R.R. and Running, S.W. (1989) Testing a theoretical climate-soil-leaf area hydrologic equilibrium of forests using satellite data and ecosystem simulation. Agric. For. Met. 44, 245-60.

Nömmik, H. and Vahtras, K. (1982) Retention and fixation of ammonium and ammonia in soils. In: Stevenson, F.I.(Ed.) Nitrogen in agricultural soils, Agronomy Series 22, pp. 123-171. University of Wisconsin, Madison.

Oechel, W.C., Cowles, S., Grulke, N., Hastings, S.J., Lawrence, B., Prudhomme, T., Reichers, G., Strain, B., Tissue, D. and Vourlitis, G. (1994) Transient nature of CO₂ fertilization in Arctic tundra. Nature 371, 500-503.

Parton, W.J., Schimel, D.S., Cole, C.V. and Ojima, D.S. (1987) Analysis of factors controlling soil organic matter levels in Great Plains grasslands. Soil Sci. Soc. Amer. J. 51, 1173-1179.

Parton, W.J., Stewart, J.W.B. and Cole, C.V. (1988) Dynamics of C, N, P, and S in grassland soils: a model. Biogeochemistry 5, 109-131.

Pastor, J. and Post, W.M. (1985) Development of a linked forest productivity-soil process model. ORNL/TM-9519, Oak Ridge National Laboratory, Oak Ridge, Tenn.

Pastor, J. and Post, W.M. (1986) Influence of climate, soil moisture, and succession on forest carbon and nitrogen cycles. Biogeochemistry 2, 3-27. 

Pastor, J. and Post, W.M. (1988) Response of northern forests to CO₂-induced climate change. Nature, 334, 55-58.

Peet, M.M., Huber, S.C. and Patterson, D.T. (1985) Acclimation to high CO₂ in monoecious cucumber. II. Carbon exchange rates, enzyme activities, and starch and nutrient concentrations. Plant Physiol. 80, 63-67.

Persson, H. (1980) Structural properties of the field and bottom layers at Ivantjärnsheden. In: Persson, T. (Ed.) Structure and Function of Northern Coniferous Forests-an Ecosystem Study. Ecol. Bull. 32, 153-163. Stockholm.

Persson, T. (Ed.) (1980) Structure and Function of Northern Coniferous Forests-an Ecosystem Study, Ecol. Bull. 32, 609 pp. Stockholm.

Prentice, I.C., Cramer, W., Harrison, S.P., Leemans, R., Monserud, R.A. and Solomon, A.M. (1992) A global biome model based on plant physiology and dominance, soil properties and climate. J. Biogeog. 19, 117-134.

Raich, J.W., Rastetter, E.B., Melillo, J.M., Kicklighter, D.W., Steudler, P.A., Peterson, B.J., Grace, A.L., Moore, B. III and Vörösmarty, C.J.(1991) Potential net primary productivity in South America: application of a global model. Ecol. Appl. 1, 399-429.

Raison, R.J., Connell, M.J. and Khanna, P.K. (1987) Methodology for studying fluxes of soil mineral-N in situ. Soil Biol. Biochem. 19, 521-530.

Raison, R.J., Connell, M.J., Khanna, P.K. and Falkiner, R.A. (1992a) Effects of irrigation and nitrogen fertilization on fluxes of soil mineral nitrogen in a stand of Pinus radiata. For. Ecol. Mgmt 52, 43-64.

Raison, R.J., Khanna, P.K., Benson, M.L., Myers, B.J., McMurtrie, R.E. and Lang, A.R.G. (1992b) Dynamics of Pinus radiata foliage in relation to water and nitrogen stress. II. Needle-loss and temporal changes in total foliage mass. For. Ecol. Mgmt 52, 159-178.

Raison, R.J., Khanna, P.K., Connell, M.J. and Falkiner, R.A. (1990) Effects of water availability and fertilization on N cycling in a stand of Pinus radiata. For. Ecol. Mgmt 30, 31-43.

Raison, R.J. and Myers, B.J.(Eds)(1992a)The Biology of Forest Growth Experiment. For. Ecol. Mgmt 52, 1-311.

Raison, R.J. and Myers, B.J. (1992b) The Biology of Forest Growth Experiment: linking water and nitrogen availability to the growth of Pinus radiata. For. Ecol. Mgmt 52, 279-308.

Raison, R.J., Myers, B.J. and Benson, M.L. (1992c) Dynamics of Pinus radiata foliage in relation to water and nitrogen stress. I. Needle production and properties. For. Ecol. Mgmt 52, 139-158.

Rastetter, E.B., McKane, R.B., Shaver, G.R. and Melillo, J.M. (1992) Changes in C storage by terrestrial ecosystems: how C-N interactions restrict responses to CO₂ and temperature. Water, Air, Soil Pollut. 64, 327-344.

Rastetter, E.B., Ryan, M.G., Shaver, G.R., Melillo, J.M., Nadelhoffer, K.J., Hobbie, J.E. and Aber, J.D. (1991) A general biogeochemical model describing the responses of the C and N cycles in terrestrial ecosystems to changes in CO₂, climate, and N deposition. Tree Physiol. 9, 101-126.

Reich, P.B., Walters, M.B. and Ellsworth, D.S. (1992) Leaf life-span in relation to leaf, plant, and stand characteristics among diverse ecosystems. Ecol. Monog. 62, 365-92.

Rosenzweig, M.L. (1968) Net primary production of terrestrial communities: prediction from climatological data. Am. Nat. 102, 67-74.

Running, S. W.and Coughlan, J .C. ( 1988) A general model of forest ecosystem processes for regional applications. I. Hydrologic balance, canopy gas exchange, and primary production processes. Ecol. Mod. 42, 125-154.

Running, S.W. and Gower, S.T. (1991) FOREST-BGC, a general model of forest  ecosystem processes for regional applications. II. Dynamic carbon allocation and nitrogen budgets. Tree Physiol. 9, 147-160.

Running, S.W. and Hunt, E.R. (1993) Generalization of a forest ecosystem process model for other biomes, BIOME-BGC, and an application for global-scale models. In: Ehleringer, J.R. and Field, C.(Eds) Scaling Physiological Processes from Leaf to Globe, pp. 141-158. Academic Press, Orlando.

Running, S.W., Knight, D.H. and Fahey, T.J. (1983) Description and application of H20TRANS, a stand level hydrologic model for western coniferous forests. In: Lauenroth, W.K., Skogerboe, G. V. and Flug, M. (Eds) Analysis of Ecological Systems: State-of-the-Art in Ecological Modelling. Developments in Environmental Modelling 5, pp. 489-495. Elsevier Scientific Publishing, New York.

Ryan, M.G. (1990) Growth and maintenance respiration in stems of Pinus contorta and Picea engelmannii. Can. J. For. Res. 20, 48-57.

Ryan, M.G. (1991a) A simple method for estimating gross carbon budgets for vegetation in forest ecosystems. Tree Physiol. 9, 255-266.

Ryan, M.G. (1991b) Effects of climate change on plant respiration. Ecol. Appl. 1, 157-167.

Ryan, M.G. and Waring, R.H.(1992) Maintenance respiration and stand development in a subalpine lodgepole pine forest. Ecology 73, 2100-2108.

Sanford, R.L. Jr, Parton, W.J., Ojima, D.S. and Lodge, D.J. (1991) Hurricane effects on soil organic matter dynamics and forest production in the Luquillo Experimental Forest, Puerto Rico: results of simulation modelling. Biotropica 23, 364-372.

Santantonio, D. and Santantonio, E. (1987) Seasonal changes in live and dead fine roots during two successive years in a thinned plantation of Pinus radiata in New Zealand. New Zeal. J. For. Sci. 17, 315-328.

Schulze, E.-D., Turner, N.C., Gollan, T. and Shackel, K.A. (1987) Stomatal responses to air humidity and to soil drought. In: Zeiger, E., Farquhar, G.D. and Cowan, I.R. (Eds) Stomatal Function, pp. 311-321. Stanford University Press, Stanford.

Shugart, H.H. (1984) A Theory of Forest Dynamics. Springer-Verlag, New York. 278 pp. Shugart, H.H. and West, D.C. (1977) Development of an Appalachian deciduous forest succession model and its application to assessment of the impact of the chestnut blight. J. Environ. Mgmt. 5, 161-179.

Snowdon, P. and Benson, M.L.(1992) Effects of combinations of irrigation and fertilization on the growth and above-ground biomass production of Pinus radiata. For. Ecol. Mgmt 52, 87-116.

Soil Survey Staff (1975) Soil Taxonomy. Agric. Handb. No.436, US Dept. Agric., Soil Conserv. Serv., Washington DC, USA.

Stace, H.C.T., Hubble, G.D., Brewer, R., Northcote, K.H., Sleeman, J.R., Mulcahy, M.J. and Hallsworth, E.G.(1968) A Handbook of Australian Soils. Rellim Technical Publications, Glenside, South Australia. 429 pp.

Strain, B.R. and Bazzaz, F.A. (1983) Terrestrial plant communities. In: Lemon, E.R. (Ed.) CO 2 and Plants. The Response of Plants to Rising Levels of Atmospheric Carbon Dioxide. American Assoc. Advancement of Science Selected Symposium 84, pp. 177-222. Westview Press, Inc., Boulder, Colo.

Swift, M.J ., Heal, O. W .and Anderson, J .M.( 1979) Decomposition in Terrestrial Ecosystems. Blackwell Scientific Publications, Oxford. 372 pp.

Thompson, W.A. and Wheeler, A.M. (1992) Photosynthesis by mature needles of field grown Pinus radiata. For. Ecol. Mgmt 52, 225-242.

Thornley, J.H.M. (1972a) A model to describe the partitioning of photosynthate during vegetative growth. Ann. Bot. 36, 419-430.

Thornley, J.H.M. (1972b) A balanced quantitative model for root:shoot ratios in vegetative plants. Ann. Bot. 36, 431-441.

Trewartha, G.T. (1968) An Introduction to Climate. McGraw-Hill, New York.

Troeng, E. and Linder, S. (1982) Gas exchange in a 20-year-old stand of Scots pine. Physiol. Plant. 54, 7-14.

Urban, D.L., Bonan, G.B., Smith, T.M. and Shugart, H.H. (1991) Spatial applications of gap models. For. Ecol. Mgmt 42, 95-100.

Wang, Y.-P., McMurtrie, R.E. and Landsberg, J.J. (1992) Modelling canopy photosynthetic productivity. In: Baker, N.R. and Thomas, H. (Eds) Crop Photosynthesis: Spatial and Temporal Determinants, Topics in Photosynthesis, Vol. 12, pp. 43-67. Elsevier, Amsterdam.

Waring, R.H. and Pitman, G.B. (1985) Modifying lodgepole pine stands to change susceptibility to mountain pine beetle attack. Ecology 66, 889-897.

Weinstein, D.A., Shugart, H.H. and West, D.C. (1982) The Long-term Nutrient Retention Properties of Forest Ecosystems: a Simulation Investigation. ORNL/TM-8472, Oak Ridge National Laboratory, Oak Ridge, Tenn.

Woodwell, G.M. (1987) Forests and climate: surprises in store. Oceanus 29, 71-75.

Woodwell, G.M. and Botkin, D.B. (1970) Metabolism of terrestrial ecosystems by gas exchange techniques: the Brookhaven approach. In: Reichle, D.E.(Ed.) Analysis of Temperate Forest Ecosystems, pp. 73-85. Springer-Verlag, New York.