Simulated changes in biogenic VOC emissions and ozone formation from habitat expansion of Acer Rubrum (red maple)

GloBEIS was developed to simulate BVOC emissions for 230 vegetation species over a wide range of temporal scales and regions (Yarwood et al 1999). The model produces emission inventories for isoprene, monoterpenes and other VOC based on input climate data (e.g., temperature, precipitation, cloud cover) with some flexibility in specifying other vegetation parameters for the purposes of sensitivity testing. GloBEIS has been evaluated versus ground-based observations for Texas (Yarwood et al 1999), central Africa (Guenther et al 1999), and North America (Guenther et al 2000).

GloBEIS was used in this study because it uses vegetation species-specific emissions factors, it requires minimal inputs, and it can be run on a desktop computer. The emission factors used in GloBEIS are the same as those reported by Guenther et al (1995) and the MEGAN model, but expressed at the leaf level and not the canopy level; although the canopy emission factors used in MEGAN are largely derived from branch and leaf enclosure data extrapolated to the whole canopy (Guenther et al 2006). Overall, isoprene emissions between GloBEIS and MEGAN are comparable (Sakulyanontvittaya et al 2010). Current modeling methods are limited by data availability of emission factors and estimations of land cover, as well as projections of future land cover change. Guenther et al (2000) estimated uncertainty in biogenic emission estimates to be around a factor of 3. Furthermore, many studies estimate vegetation cover from broad-based biomes and estimate biogenic emissions from an average flux over the land cover type, and variations within the biome can account for some uncertainty in estimated emissions. However, because most BVOC models yield similar isoprene emissions (Arneth et al 2008), GloBEIS is an appropriate model for this study.

GloBEIS calculates emissions similar to Guenther et al (2000) as well as the more recent MEGAN model (Guenther et al 2006, 2012) with

$$\begin {equation*} {\rm Emissions} = \varepsilon D\Upsilon \rho A \end {equation*}$$

where ε is the emission rate at the leaf level for a specific vegetation type at standard temperature and level of photosynthetically active radiation (PAR), D is the foliar density, and ϒ is an activity factor accounting for deviations in the standard of photon flux density, temperature, and leaf age, ρ accounts for loss within plant canopies and A is the area of the vegetation species. Additional emission factors (e.g., leaf age, prior light and temperature history) in MEGAN are not included in GloBEIS, yet these generally make minor adjustments to emissions and most isoprene models estimate similar magnitudes of emission (Arneth et al 2008). For this study we use leaf level isoprene emission factors of 0.1 μg C g⁻¹ h⁻¹ for red maple and 70 μg C g⁻¹ h⁻¹ for oak species, and monoterpene factors of 1.6 μg C g⁻¹ h⁻¹ for red maple and 0.2 μg C g⁻¹ h⁻¹ for oak species (Geron et al 1994) The options for variable LAI, variable leaf age, and drought index were not used and ρ was set to 1. Because variability of isoprene emissions is dependent on changes in temperature and PAR, our use of constant atmospheric forcing for all scenarios results in changes to the emission rate and area of the vegetation species.

GloBEIS is driven with 6-hourly climate input from the National Center for Environmental Protection (NCEP) reanalysis dataset (Kalnay et al 1996) that includes temperature, photosynthetically active radiation, wind speed, and humidity at a spatial resolution of 0.5° for the period 1990–1999. The vegetation distribution is derived from version 3 of the Biogenic Emissions Landuse Dataset (BELD3) and represents land use of the 1990s at a 1 km resolution (Kinnee et al 1997). The BELD dataset was designed for use in the EPA's BEIS model (Biogenic Emissions Inventory System), which can be used in the GloBEIS framework. Each cell is assigned a vegetation biome, with information about the fraction of forest, agriculture, and other land cover type (e.g., water, urban built-up). Species information is provided for 230 vegetation types including trees and crops as a fraction of the total area in each county (ftp://ftp.epa.gov/amd/asmd/beld3).

Our chosen domain includes 17 states (figure 1) where forest surveys show an increase in red maple and a decrease in oak (Christensen 1977, Russel 1980, Ross et al 1982, Lorimer 1984, Abrams et al 1995, Abrams 1998, Dyer 2001, Abrahamson and Gohn 2004, Abrams 2005, Galbraith and Martin 2005). The dominant tree species in the domain is oak (almost 13% of the total area) followed by red maple (less than 5%). We created a land cover file from the BELD3 dataset for each county in the domain, allowing the resolution of the model to vary depending on the size of the county. The top 20 land cover types account for 85% of the total land cover in the domain (table 1).

We performed a baseline run (BASELINE) using present day vegetation cover from BELD3 and a set of three theoretical land cover change experiments. To isolate the isoprene emission signal resulting from the imposed land cover changes alone, all model runs were driven with the same NCEP climate data. In the first theoretical experiment, all oak species were removed and replaced with red maple (MAPLE). In the second experiment, all red maple were removed and replaced with oak (OAK) to obtain an upper limit of isoprene emissions that may have existed in the past. One additional experiment was conducted to assess how isoprene changes with fractional land cover changes (e.g., 50% replacement of oak versus 100% replacement) that have been documented in some eastern forests. In this experiment, 50% of the oaks were converted to red maple (50%OAK2MAPLE).

We tested the sensitivity of ozone formation to land cover change-driven BVOC emissions changes using a 0D photochemical box model based on Silman (1991). While 3D chemical transport modeling would be ideal to investigate the full chemical impact of these ecosystem changes, the 0D model provides a first estimate of the relative impact of the biogenic VOC changes at specified locations. The 0D model treats each location as a single 'box' and assumes a constant wind speed of 2 m s⁻¹ to represent dilution due to horizontal advection along with prescribed upwind and downwind concentrations, providing a simple estimate of transport based on current land cover. Initial and advected model concentrations were derived from July climatological (2000–2007) global chemistry model simulations (MOZART; (Emmons et al 2010); ∼2.8° by 2.8° or T42 spatial resolution) for ozone, CO, NO_x, and a suite of anthropogenic VOC species. The height of the box is prescribed with a diurnally evolving mixed height layer from 150 m at night to 1500 m at midday to mimic the diurnal dilution of pollutants in the boundary layer. Explicit vertical mixing is not included, although a top boundary condition is specified based on the global MOZART concentrations A diurnal temperature range of 20 K from daily minimum to maximum is the same for all locations. Regional ozone formation can also be affected by the formation and transport of isoprene nitrates (Fiore et al 2005) but we have not considered the transport of these compounds in this study.

Emissions input to the box model are based on theMACCity global 0.5° resolution database (Lamarque et al 2010) and include emissions of NO_x, CO and anthropogenic VOC. BVOC emissions were derived from the GloBEIS simulations as described in section 2.2. The box model uses the GEOS-Chem mechanism modified for urban photochemistry (Evans et al 2003, Ito et al 2007) and treats isoprene chemistry explicitly and monoterpene chemistry with two surrogates (one as α-pinene and one as limonene), which have OH lifetimes of 2.7 h and 51 min, respectively; (Warneck and Williams 2012). We assume a 50–50 split of monoterpene emissions to these two categories, though note that there can be significant variability of the monoterpene speciation of emissions depending on the vegetation type (Geron et al 2000). Generally, the monoterpene emission quantities and changes are relatively small compared to the isoprene and have a smaller impact on ozone formation, therefore the following sections focus on isoprene emissions and their effects on ozone.

We selected four geographic locations based on the following criteria: (1) the ozone attainment status, (2) the photochemical sensitivity, (3) the baseline biogenic VOC emissions, and (4) the emissions change due to land cover changes. The four locations are (1) the Philadelphia metropolitan area (39.9°N, 75.3°W), a suburban/urban region in non attainment for ozone with low baseline BVOC emissions and relatively low capacity for maple expansion, (2) the Charlotte metropolitan area (35.2°N, 80.9°W), a suburban/urban region in non-attainment for ozone with high baseline BVOC emissions and a high capacity for maple expansion, (3) western Virginia (37.2°N, 80.4°W), a rural site in attainment for ozone with high baseline emissions and a high capacity for maple expansion, and (4) central Pennsylvania (40.8°N, 77.9°W), a rural site in attainment for ozone with moderate BVOC emissions and a relatively low capacity for maple expansion.

For the BASELINE experiment, the average total isoprene emissions over the domain from 1990 to 1999 are 3.18 Tg C yr⁻¹ (table 2). Direct comparison of our results with other studies is difficult because of the selected domain, which accounts for less than 12% of the continent. Total simulated isoprene emissions reported in this study are the same order of magnitude but slightly lower as those reported by Lamb et al (1993) covering just the Mid-Atlantic EPA Region 3 (table 2). Simulated July daily isoprene emissions range from a low of 4to8 mg C m⁻² day⁻¹ in the northern part of the domain to nearly 100 mg C m⁻² day⁻¹ in Virginia, North Carolina, and Tennessee (figure 2(a)) where oak density is highest along the Appalachian Range (figure 1). Regions with low concentrations of oak, such as Maine, New York, and Vermont do not show substantial isoprene emissions.

Zoom In Zoom Out Reset image size

GloBEIS predicts a ∼70% decrease in isoprene emissions in the MAPLE simulation (table 2). Isoprene emissions decrease by more than 50 mg C m⁻² day⁻¹, especially in the southern part of the domain where the greatest density of oak was replaced (figure 2(b)) In contrast, isoprene emissions increase in the OAK simulation by approximately 20% over the entire domain (table 2). July daily emissions increased by more than 10 mg C m⁻² day⁻¹ with the largest increases concentrated along the Appalachian Mountains and in Massachusetts, New Hampshire, Rhode Island, and Connecticut (figure 2(c)). When a less dramatic land cover change of 50% conversion of oak to maple is imposed, the model simulates a 35% decrease in isoprene emissions in the 50% OAK2MAPLE scenario (table 2).

Major metropolitan areas (identified as those with a population greater than 100 000; figure 2) located in proximity to high isoprene emissions have a greater potential for ozone production. While most major cities are located in regions of relatively low isoprene emissions, several near the Appalachian Mountains have high isoprene emissions, particularly in North Carolina and Tennessee. Eighty six percent of the cities plotted are in counties of nonattainment of the EPA's 8-h ozone standard (figure 1). Depending on the prevailing wind direction, transport of isoprene downwind of high-density oak regions could impact air quality in surrounding areas, assuming that NO_x is not a limiting factor.

Table 3 and figure 3 describe the impacts of BVOC emissions changes as simulated by the 0D photochemical box model in the four locations for the BASELINE and three land cover change scenarios. In all simulations, isoprene emissions are largest at the Western VA and Charlotte sites because they are in the southern portion of the domain where oak density is highest (figures 1 and 3). We compare the BASELINE modeled diurnal cycle of ozone with that observed at EPA Photochemical Assessment Monitoring Stations (PAMS) in each of the four locations (an average July diurnal cycle based on data from PAMS data from 1994–2010; (EPA 2011)).

In general, the box model reproduces the diurnal cycle of ozone although it tends to overestimate the daily maximum by about 5–10 ppb in urban areas, and about 15 ppb in rural areas (figure 4). While this overestimation may influence our understanding of local air quality, these biases are typical for photochemical models in the Northeastern USA (Fiore et al 2009, Reidmiller et al 2009).

The large decrease in isoprene emissions in the MAPLE scenario yields the most substantial ozone reductions at three of the four sites (figure 4). In Philadelphia, the model is not sensitive to any of the changes in BVOC emissions, highlighting the small role of BVOCs in this region and the NO_x-sensitivity of this location. In Charlotte, the 100% conversion of oak to maple leads to a 4 ppb reduction in the daily ozone maximum, and a 2 ppb reduction under the 50% conversion case (50% OAK2MAPLE). The effects on ozone are even stronger in the two rural sites, with an ozone reduction of up to 6 ppb in western VA. This suggests that the oak to maple transition may have greater impacts on rural regional ozone concentrations rather than within urban cores.

The OAK simulation predicts slight increases in isoprene emissions (figure 3(a)) accompanied by small decreases in monoterpene emissions (figure 3(b)). In Charlotte and Philadelphia, the OAK simulation shows almost no impact on ozone (<0.1 ppb increase) because of the relatively large amounts of anthropogenic VOC emissions (figure 4). In the rural areas where anthropogenic sources are much smaller, slight increases in isoprene emissions in the OAK scenario have a larger impact and lead to increases in daily maximum ozone concentrations of 1–3 ppb.

At all locations, changes in monoterpenes have little impact on ozone formation. While monoterpenes increase in the MAPLE scenario by about 20% at the southern sites and nearly double at the central PA site, the relatively low magnitude of monoterpene emissions and their lower ozone formation potential result in no significant impact on ozone production at the four sites.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size