2.1. Model Description
 We used an updated version of the Integrated Biospheric Simulator (IBIS 2.5) [Foley et al., 1996; Kucharik et al., 2000] to simulate the biogenic emissions of isoprene and monoterpenes. IBIS is a dynamic global vegetation model that simulates the interactions of the terrestrial biosphere with the atmosphere through land surface and hydrological processes, canopy physiology, vegetation dynamics, and terrestrial carbon balance within a single integrated framework (Figure 1).
 The land surface module of IBIS simulates the energy, water, carbon, and momentum balance of the soil-vegetation-atmosphere system. It represents two types of vegetation canopies: lower (grasses and shrubs) and upper (trees), and six soil layers to simulate soil physics. Canopy radiation transfer is simulated using the two-stream approximation following the approach of Sellers et al.  and Bonan . Photosynthesis, respiration and stomatal conductance are simulated using mechanistic approaches within the soil-vegetation-atmosphere transfer scheme, allowing for explicit coupling between vegetation canopies and the atmosphere. IBIS uses a natural vegetation map with 15 ecosystem types or biomes [Ramankutty and Foley, 1999] and each of these biomes consist of a unique combination of 12 plant functional types (PFTs). The geographical distribution of each PFT is determined by climatic constraints [Kucharik et al., 2000]. The relative abundance of the 12 PFTs in each grid cell changes in time according to their ability to photosynthesize and use water. The vegetation dynamics module predicts the transient changes in leaf area index (LAI) and biomass for the 12 PFTs, based on annual carbon balance. IBIS simulates these biophysical and biogeochemical processes at timescales ranging from 60 min to a year.
 Measurement and laboratory studies show that isoprene and monoterpene emissions are highly sensitive to temperature. Isoprene emissions show a temperature maximum and subsequent reduction at higher temperatures [Guenther et al., 1991, 1993; Monson et al., 1994], while monoterpene emissions increase exponentially with increasing temperature [Tingey et al., 1980, 1991; Guenther et al., 1991]. In addition, isoprene emissions are extremely light-dependent with noticeably different emissions among shaded and sunlit leaves [Harley et al., 1996], and no emissions under dark conditions [Guenther et al., 1991; Tingey et al., 1979]. Monoterpene emissions from most plants are regarded as light-independent. There are, however, exceptions to the influence of these environmental factors on isoprene and monoterpene emissions. For example, isoprene emissions from CAM (Crassulacean Acid Metabolism) plants may be light-independent [Lerdau and Keller, 1997], and monoterpene emissions from some plant species have been found to be light-dependent [Bertin et al., 1997].
 Within the IBIS framework, we incorporated isoprenoid emission algorithms that describe the short-term influences of temperature and light on emissions (Figure 1). These empirical emission algorithms, based on field measurements and laboratory experiments, were initially proposed by Guenther et al. [1991, 1993], and further developed for a global natural volatile organic compound emissions inventory for the International Global Atmospheric Chemistry Project (IGAC) [G95]. Isoprenoid emissions from plant canopies are estimated as follows:
where F is the emission flux (μg C m−2 h−1), ɛ is the ecosystem-specific emission factor (μg C g−1 h−1) at a standard leaf temperature (Ts) of 303.15 K and standard photosynthetically active radiation (PAR) flux of 1000 μmol m−2 s−1, Fd is the foliar biomass density (g dry weight m−2), γT and γL are dimensionless scalars that describe the response of emissions to diurnal variations in leaf temperature and incident sunlight (for isoprene only), and ρ is an escape efficiency factor that represents the fraction of gas emitted by the canopy that is released into the above-canopy atmosphere. For isoprene emissions, temperature and light dependence factors are given by
where R (8.314 J K−1 mol−1) is the gas constant, CT1 (95000 J mol−1), CT2 (230000 J mol−1), Tm (314 K), α (0.0027), and CL1 (1.066) are empirical coefficients derived from measurements [Guenther, 1997]. For monoterpene emissions, temperature dependence factor is given by
where β is an empirical coefficient equal to 0.09 K−1 [Guenther et al., 1993] and γL = 1.
 At each grid cell, emissions are predicted every 60 min using canopy variables supplied by IBIS. Daily foliar density is estimated by dividing the LAI for each PFT by its specific leaf area [Foley et al., 1996]. γT is calculated using IBIS simulated hourly leaf temperature for trees (upper canopy), and grasses and shrubs (lower canopy). γL is calculated for sunlit and shaded leaves for each canopy layer. We assume a globally uniform value of ρ = 1 in the absence of estimates from field measurements. This should be regarded as an upper limit since canopy chemistry and physics may reduce the fraction of VOCs released into the free atmosphere [Guenther et al., 1999].
 Base emission factors are defined as the rate of emission per unit foliar biomass expected from a plant species under a given set of environmental conditions [Guenther, 1997]. They vary significantly for different plant species and are influenced by nutrient status, soil moisture content, and foliage developmental stage [Fuentes et al., 2000, and references therein]. A robust modeling scheme to predict global VOC emissions would require taking into account the influence of these factors on ɛ and the heterogeneity in plant species. Limited field measurements, however, restrict the assignment of ɛ for a wide variety of plant species on a global scale. As described above, IBIS uses a combination of 12 PFTs to define a biome on each grid cell. This approach deals with the issue of species variability to some extent. Therefore we assigned ɛ for the 12 PFTs based on G95 recommendations (Table 1). We assigned an isoprene emission factor of 0.0 for grasses contrary to recommendations of G95, as several measurement studies have since shown that grasses are not a major emitter of isoprene (see http://www.es.lancs.ac.uk/cnhgroup/download.html).
Table 1. Plant Functional Types and Their Specific Leaf Area Defined in IBIS and Emission Factors (ɛ) for Isoprene and Monoterpenes Based on Guenther et al. a
|Plant Functional Type||Isoprene, μg C g−1 hr−1||Monoterpenes, μg C g−1 hr−1||Specific Leaf Area, m2 kg−1|
|Tropical broadleaf evergreen tree||24.0||0.4||25.0|
|Tropical broadleaf drought-deciduous tree||45.0||1.2||25.0|
|Warm-temperate broadleaf evergreen tree||24.0||0.8||25.0|
|Temperate conifer evergreen tree||16.0||2.4||12.5|
|Temperate broadleaf cold-deciduous tree||45.0||0.8||25.0|
|Boreal conifer evergreen tree||8.0||2.4||12.5|
|Boreal broadleaf cold-deciduous tree||45.0||0.8||25.0|
|Boreal conifer cold-deciduous tree||8.0||2.4||25.0|
|Shrubs and Grasses|
 Initially, IBIS was driven in the dynamic mode to an equilibrium state using a monthly mean climatological data set of temperature, precipitation, relative humidity, and cloud cover for the 1961–1990 period. This data set, compiled by New et al. , is referred to as CRU05 hereafter. The model was run at a resolution of 2° longitude by 2° latitude. The simulation was initialized with an “observed” potential vegetation map [Ramankutty and Foley, 1999] that represents vegetation that would exist in the absence of human activities. Therefore croplands are not included in this vegetation map. A constant atmospheric CO2 concentration of 333.4 ppm (parts per million), which is the mean for the thirty-year period 1961–1990 [Keeling and Whorf, 2003], was used. IBIS was run for 300 years to arrive at a near equilibrium state. Beginning from this initial equilibrium state, the following simulations were performed for the 1961 to 1990 period:
 1. CON: Control using CRU05 mean 1961–1990 climatology.
 2. CLIM: Climate only, using transient changes in climate data (monthly climate anomalies relative to 1961–1990 from New et al. ).
 4. CLIM_CO2: Transient changes in both climate and CO2 concentration.
 CON was aimed at simulating ecosystem attributes resulting from mean climate and a constant atmospheric CO2 for the thirty-year period from 1961 to 1990. We only use years 1971–1990 for our analysis of simulations CLIM, CO2 and CLIM_CO2 to let the carbon pools adjust to the different climatic conditions and CO2 level. The CO2 simulation, for instance, is initialized with the results of the CON simulation that was performed with an atmospheric CO2 level of 333.4 ppmv. However, the observed CO2 concentration for 1961 used in the CO2 simulation is only 317 ppmv. In the following section, we first evaluate IBIS's ability to simulate biogenic emissions using results from CON, and then, using results from the three simulations we investigate the interannual and spatial variations in isoprenoid emissions in response to the past 20 year climate variability and increasing atmospheric CO2.