Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
Corresponding author: E. B. Peters, Institute on the Environment, University of Minnesota, 1954 Buford Ave., Saint Paul, MN 55108, USA. (firstname.lastname@example.org)
 In a suburban neighborhood of Minneapolis–Saint Paul, Minnesota, USA, we simultaneously measured net CO2 exchange of trees using sap flow and leaf gas exchange measurements, net CO2exchange of a turfgrass lawn using eddy covariance from a portable tower, and total surface-atmosphere CO2 fluxes (FC) using an eddy covariance system on a tall tower. Two years of continuous measurements showed that net CO2exchange varied among vegetation types, with the largest growing-season (Apr–Nov) net CO2 uptake on a per cover area basis from evergreen needleleaf trees (−603 g C m−2), followed by deciduous broadleaf trees (−216 g C m−2), irrigated turfgrass (−211 g C m−2), and non-irrigated turfgrass (−115 g C m−2). Vegetation types showed seasonal patterns of CO2exchange similar to those observed in natural ecosystems. Scaled-up net CO2 exchange from vegetation and soils (FC(VegSoil)) agreed closely with landscape FC measurements from the tall tower at times when fossil fuel emissions were at a minimum. Although FC(VegSoil) did not offset fossil fuel emissions on an annual basis, the temporal pattern of FC(VegSoil) did significantly alter the seasonality of FC. Total growing season FC(VegSoil)in recreational land-use areas averaged −165 g C m−2 and was dominated by turfgrass CO2 exchange (representing 77% of the total), whereas FC(VegSoil) in residential areas averaged −124 g C m−2 and was dominated by trees (representing 78% of the total). Our results suggest urban vegetation types can capture much of the variability required to predict seasonal patterns and differences in FC(VegSoil) that could result from changes in land use or vegetation composition in temperate cities.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Cities, states, and countries have recently begun to prescribe targets for greenhouse gas reductions [United Nations, 1998; State of Minnesota, 2010] and to develop carbon markets intended to mitigate climate change (e.g., Chicago Climate Exchange). Meeting these goals requires the capacity to accurately assess the net exchange of CO2 between urban landscapes and the atmosphere, as well as to quantify its sources and sinks. Yet continuous measurements of urban net CO2 exchange have only recently begun [Grimmond et al., 2002] and our ability to quantify its main components (e.g., fossil fuel emissions, uptake from vegetation), as well as the drivers of those components, lags far behind the state-of-the-art in natural and agricultural landscapes [Reichstein et al., 2005]. In urban areas, the total net CO2 flux between the surface and the atmosphere, FC, can be defined such that:
where FC(Fossil) is CO2 emissions from combustion of fossil fuels (e.g., by motor vehicles or space heating). FC(Human) is CO2 emissions from human respiration. FC(VegSoil) is the net CO2 exchange by vegetation and soils, which is determined by the balance between CO2 uptake by photosynthesis and CO2 loss by plant and soil microbial respiration, and is analogous to the net ecosystem CO2 exchange (NEE) commonly reported for natural or agricultural ecosystems. By micrometeorological convention, fluxes toward the atmosphere are positive, representing a net release of CO2 by the ecosystem, whereas fluxes toward the surface are negative, representing a net uptake of CO2, attributable to plant photosynthesis.
 Recently, FChas been measured in a number of cities around the world, representing different climates, cultures, geographical regions, and land-use types (e.g., Montreal [Bergeron and Strachan, 2011], Baltimore [Crawford et al., 2011], Edinburgh [Nemitz et al., 2002], London [Helfter et al., 2011], Tokyo [Moriwaki and Kanda, 2004], Melbourne [Coutts et al., 2007]). While these studies collectively show that urban areas are net emitters of CO2 on an annual basis, FC varies strongly in magnitude and seasonality in different urban settings. In general, cities with dense populations, high traffic volumes, and low vegetation cover tend to be larger sources of CO2 to the atmosphere per unit area than suburban areas with low population density, low traffic volumes, and high vegetation cover [Bergeron and Strachan, 2011]. The scaling analysis reported here uses a per unit area basis, but note that dense cities generally are smaller sources of CO2 on a per capita basis [Velasco and Roth, 2010]. Differences in vegetation cover among cities have been shown to correlate strongly with FC [Velasco and Roth, 2010], explaining as much as 95% of the variation in summertime FC among eight midlatitude cities [Ramamurthy and Pardyjak, 2011]. Yet such correlations do not tell us how much of the difference among cities is directly due to CO2uptake by urban vegetation as compared to reduced fossil fuel emissions in vegetation-covered areas (e.g., fewer roads, less commercial and industrial land use). We also have relatively little information on the degree to which the seasonal cycle ofFC in urban areas is affected by the activity of vegetation and soils. During times when space heating or cooling is required, energy demand and the type of heating and cooling systems also control the seasonal differences in FC among cities.
 There is very little independent data on the continuous FC(VegSoil) of vegetation and soils in urban areas. Forest inventories and allometric growth equations have been used to estimate annual carbon sequestration rates of urban forests [Nowak et al., 2008], but this approach cannot capture the seasonal or daily carbon dynamics of urban forests and it excludes non-woody vegetation, such as turfgrass lawns, which are ubiquitous in many urban areas [Milesi et al., 2005]. Other studies have applied NEE measurements made in natural or agricultural ecosystems to estimate urban FC(VegSoil) [Moriwaki and Kanda, 2004; Helfter et al., 2011], but these extrapolations are potentially limited by the large differences in urban environmental conditions (e.g., compacted and contaminated soils, high vapor pressure deficits, and human management).
 Due to horticultural and non-native species introductions, urban areas can be species-rich ecosystems [Walker et al., 2009; Knapp et al., 2012], which complicates efforts to scale up CO2 flux measurements made on only a few individuals, species, or sites to the landscape scale. For example, tree species can vary greatly in the magnitude and seasonality of their carbon uptake rates [Catovsky et al., 2002; Givnish, 2002]. Plant functional types can provide a potentially useful approach for partitioning the seasonal patterns of urban FC(VegSoil). Not only do plant functional types represent major physiological and biophysical differences among plant species [Reich et al., 1997] and can be uniquely identified in developed landscapes using high-resolution satellite imagery [Tooke et al., 2009], but they have also been previously used to partition seasonal patterns of urban evapotranspiration [Peters et al., 2011].
 In this paper, our focus was to quantify the net CO2 exchange of urban vegetation and soils, as opposed to the fossil fuel or human respiration components of FC. We report a two-year field study in which we quantified the relative contributions of three main plant functional types (evergreen needleleaf tree, deciduous broadleaf tree, cool-season turfgrass) to the seasonal and spatial variations ofFC(VegSoil) in a suburban neighborhood of Minneapolis–Saint Paul, Minn. We simultaneously measured net CO2 exchange of the major tree species in the study area using continuous heat dissipation sap flux and leaf gas exchange measurements, and the net CO2exchange of a turfgrass lawn using the eddy covariance technique from a 1.35-m portable tower. In addition, we measuredFCover the entire suburban landscape using an eddy covariance system on a 40-m tall tower. Our objectives were to: 1) determine how suburban trees vary in their CO2 uptake rates across the growing season; 2) quantify the magnitude and seasonal patterns of net CO2 exchange among urban vegetation types; 3) evaluate the sum of all vegetation and soil components, FC(VegSoil), in relation to the tall tower measurements of FC, which include FC(Fossil) and FC(Human); and 4) examine how different vegetation types influence seasonal FC(VegSoil)patterns in recreational and residential land-use types.
2.1. Study Site
 Our study was conducted in a first-ring suburban neighborhood immediately outside the border of the city of Saint Paul, in east-central Minnesota, USA (44°59′N, 93°11′W). The neighborhood had approximately 1000 inhabitants km−2 and a housing density of 350 housing units km−2 [Radeloff et al., 2005]. As described in more detail in Peters et al. , this 7 ha study area (Figure 1) has a cold temperate climate and is influenced by the urban heat island effect [Winkler et al., 1981; Todhunter, 1996; Sen Roy and Yuan, 2009]. The relatively flat terrain made this landscape suitable for eddy covariance measurements from the 150-m tall KUOM broadcast tower. The average tall tower flux footprint area had 82% vegetation cover, which consisted primarily of open turfgrass lawns, forested patches, and isolated trees with a mean tree height of 12 m (Table 1). The land-use types within the tower footprint included residential land use (single-family, detached housing) to the northwest and northeast, and recreational land use (a golf course) to the southwest (Figure 1).
Table 1. Cover Types by Percentage of the Residential Land-Use Area, Recreational Land-Use Area, and Flux Footprint Climatology Areaa
Deciduous Tree (%)
Evergreen Tree (%)
Non-Irrigated Turfgrass (%)
Irrigated Turfgrass (%)
Open Water (%)
Impervious Surfaces (%)
Percent cover was determined from the land cover classification map shown in Figure 1 and does not include turfgrass under tree canopies. The footprint area was defined as the total area bounded by the 10% contour line (Figure 1).
 We selected four stands of trees (CTC, Grove, Lauderdale, and Saint Paul sites) suitable for sap flux measurements within our study area (Figure 1), as described in Peters et al. . Briefly, these sites were selected to include tree species and sizes that were as representative as possible of an Urban Forest Inventory and Analysis [U.S. Forest Service, 2005] conducted in 2005–2006 in the study area. Trees were grown under park-like, or backyard conditions, with open, pruned canopies and turfgrass ground cover. The turfgrass had low-maintenance management regimes, which included regular mowing and little or no irrigation or fertilizer use. These site conditions were typical of those found in residential yards and rights-of-way in this study area, as compared to street trees in a more dense urban setting.
 We made eddy covariance measurements of a turfgrass lawn using a 1.35 m high portable mast in a 1.5 ha field that was located within the footprint of the tall tower (Figure 1). The site, which is described in more detail in Hiller et al. , contained C3cool-season turfgrass species that were representative of low-maintenance lawns in the area (e.g., mowed regularly, clippings left on site, and no irrigation or fertilizer applications). These conditions were typical of first-ring suburban residential landscapes within the Minneapolis–Saint Paul metropolitan area [Fissore et al., 2011b].
2.2. Component-BasedFC(VegSoil) Approach
 In a suburban area, FC(VegSoil) represents the sum of all CO2 fluxes from vegetation and soils, such that:
where FC(T) represents net CO2 fluxes from trees and FC(G) represents net CO2 fluxes from turfgrass lawns, including CO2emissions from soil respiration. This represents a “bottom-up” approach to estimating landscapeFC(VegSoil)in which continuous measurements of each major vegetation type are scaled up using a high-resolution land cover map. Results can then be evaluated with eddy covariance measurements on a tall tower. Each of the vegetation-type fluxes included in our scaled-upFC(VegSoil) estimate was either independently measured or modeled as described below. CO2 fluxes from impervious surfaces were assumed to be negligible because such surfaces effectively prevent gas exchange from soils below. CO2emissions from bare soil areas were also assumed to be negligible because bare soil covered only ∼3% of the study area and generally much less of any given 30-min flux source area, such that its percent cover was too small to be accurately estimated when we spatially sampled land cover fractions in each footprint (seeSection 2.6). However, the CO2 emissions from all other soils (under turfgrass cover) were included explicitly in FC(G).
2.3. CO2 Fluxes From Trees
 To obtain continuous estimates of FC(T)across the 2007 and 2008 growing seasons, we combined sap flow and leaf-level gas exchange measurements followingCatovsky et al. , such that:
where, EC is tree transpiration per cover area, or projected crown area (kg H2O m−2 s−1), λ is the latent heat of vaporization (MJ kg−1), γ is the psychrometric constant (kPa °C−1), cp is the specific heat capacity of water (J kg−1 C−1), ρ is the density of air (kg m−3), D is the vapor pressure deficit between the leaf interior and the bulk air (kPa), P is air pressure (Pa), T is air temperature (K), R is the ideal gas law constant (m3 Pa K−1 mol−1), ais the slope of the linear relationship between leaf-level photosynthesis and stomatal conductance (μmol mol−1), and NPP:GPP is the ratio between net and gross primary production, or carbon use efficiency, for forests.
 First, we used sap flow techniques to obtain continuous measurements of whole-tree transpiration,EC, across the growing season [Granier, 1987; Lu et al., 2004]. Second, we used canopy-level meteorological measurements, including relative humidity and air temperature, to convert our transpiration measurements to continuous estimates of canopy conductance (GC) [Monteith and Unsworth, 1990; Tang et al., 2006]. Third, we used leaf-level gas exchange measurements of photosynthesis (AL) and stomatal conductance (gS) to obtain continuous estimates of canopy photosynthesis (AC) from GC across the growing season. Finally, to account for CO2 emissions from plant respiration, we used published estimates of the ratio between net and gross primary production in forests [DeLucia et al., 2007] to convert AC to FC(T). Our measurements of FC(T) did not account for CO2emissions from the decomposition of tree leaf litter. However, survey results show that a majority of households in the study area remove tree leaf litter from their property and compost it at off-site locations, thereby moving the source of litter decomposition outside our study area [Fissore et al., 2011a]. We also expect that any underestimate of tree leaf litter decomposition would have a minimal effect on our FC(VegSoil) estimates because, with the exception of November, our analyses were limited to growing season months before annual leaf fall occurred.
 We used the heat dissipation sap flow technique [Granier, 1987; Lu et al., 2004], to continuously measure whole-tree transpiration on a total of 37 trees (17 trees in 2007 and 20 different trees in 2008). The sampled trees included seven different genera (Fraxinus, Juglans, Picea, Pinus, Quercus, Tilia, and Ulmus), representing two plant functional types (evergreen needleleaf and deciduous broadleaf). Further details of the sap flow measurements and their environmental controls are given in Peters et al. .
 Using a simplified form of the Penman-Monteith equation that assumes an aerodynamically well-mixed canopy [Monteith and Unsworth, 1990], we converted the transpiration measurements to continuous estimates of canopy conductance in m s−1 such that:
where Dwas calculated using measurements of air temperature and relative humidity (model HMP45C, Campbell Scientific Inc., Logan, Utah) within the tree canopies at each site, as leaf and air temperature were assumed to be similar due to a well-mixed canopy. It is reasonable to assume a well-mixed canopy in this study area due to the diversity of surface types creating a high surface roughness. The mean (±SD) roughness length (z0) calculated from the 40 m eddy covariance measurements during summer was 1.07 ± 0.74 m, similar to other suburban sites [Oke, 1987]. GC was converted to mol m−2 s−1 using the Ideal Gas Law:
 Leaf-level gas exchange measurements were made in 2007 on tree canopies at the University of Minnesota Saint Paul campus and in 2008 on tree canopies at the Lauderdale and Saint Paul sites and at two Minneapolis parks, Logan Park and Windom Park (Figure 1). Logan Park and Windom Park were both located 5 and 4 km, respectively, from the tall tower in residential areas similar to the other sites. Leaves were sampled at the University of Minnesota site on one day in each of the months of August and September; at the Lauderdale site in July and September; at the Saint Paul site in June, August, and September; at the Logan Park site in August and October; and at the Windom Park site in August and September. The numbers of trees of each species that were sampled at each site are given in Table 2. To maximize the number of high quality measurements, sampling was restricted to weather conditions with clear skies and low wind speeds. On each sampling date, gas exchange measurements were made throughout the day (08:00–15:00 h LT) on leaves that were accessible from an aerial lift truck. Most leaves sampled were exposed to direct sunlight during part of the day because all trees had open canopies and were grown under park-like conditions. In addition, access restrictions due to the aerial lift truck meant that most of the measured leaves were in the outer canopy. Six to 43 leaves were sampled across the full height of each canopy on each sampling day.
Table 2. Number of Trees Measured, Site, Slope (a), and Coefficient of Determination (R2) of the Linear Regression Model, AL = a gS, Between Leaf-Level Photosynthesis (AL, μmol m–2 s–1) and Stomatal Conductance (gS, mol m–2 s–1), and Relative Abundance (as a Percentage of All Trees) by Land-Use Type for Ten Tree Speciesa
a (μmol mol–1)
Relative Abundance (%) Recreational Area
Relative Abundance (%) Residential Area
Species that did not occur within the defined land-use types were located in other areas of the footprint. Numbers in parentheses represent ±1 standard error.
Lauderdale (3), Saint Paul (2)
Logan (2), Saint Paul (2)
Logan (2), Saint Paul (3)
University of Minnesota
Lauderdale (1), Saint Paul (1), Windom (1)
 We made the leaf-level gas exchange measurements using a portable infrared gas-exchange system (LI-6400, LI-Cor, Lincoln, Neb.) that was set to match existing environmental conditions of air temperature, irradiance, and water vapor concentration. Ambient air was drawn through a dry, empty carboy to stabilize air temperature and water vapor concentrations in the LI-6400 leaf cuvette. The clear LI-6400 leaf cuvette was used to measure leaves of broadleaf species and a clear conifer chamber (LI-6400-05) was used for needleleaf trees, both providing ambient light conditions. CO2 cartridges were used to provide a constant reference CO2 concentration of 390 ppm, which was representative of local CO2concentrations at the study sites. For deciduous broadleaf species, the measuring area of the standard LI-6400 leaf cuvette determined the sampled leaf area. For evergreen needleleaf trees, the sampled needles were collected and their leaf area was determined using a flatbed scanner and image analysis software (ImageJ version 1.36b) [Abramoff et al., 2004].
 To convert GC in mol m−2 s−1 to AC in μmol m−2 s−1, we developed species-specific linear regression models betweenAL and gS, following Catovsky et al. . Linear regression models were fit through the origin in the form AL = agS. AC was calculated by multiplying GC by the slope, or parameter a, of the linear regression model for each species. AC was converted to g C m−2 s−1 by multiplying by the molecular weight of carbon. We separately tested each tree species for significant differences in the regression slopes among seasons, canopy positions, and sites to evaluate the need for specialized regression equations. For species not measured in 2007, we applied linear regression models developed in 2008 for the same species or averaged by genera.
 Finally, we estimated the net CO2 exchange of trees, FC(T), by multiplying AC by the ratio between net and gross primary production (NPP:GPP) for forests. In a review of carbon use efficiencies (NPP:GPP), DeLucia et al.  found an average ratio of 0.53 for all forests, with a range from 0.23 to 0.83. Based on this, we converted daily sums of AC to FC(T)by multiplying by −0.5 and 30-min daytimeACfluxes by −0.75. The negative sign was used to convert fluxes from the ecological to micrometeorological sign convention. For 30-min daytimeACwe increased the ratio by 50% to account for the fact that daytime fluxes do not include nighttime respiration, which is included when NPP:GPP is calculated from daily sums. We acknowledge that because NPP:GPP is typically calculated at annual or daily time scales, the extrapolation of the ratio to instantaneous daytime 30-min values represents a potential source of error in our 30-minFC(T)estimates. This approach was required only for the 30-min daytimeFC(T) values that were compared with simultaneous observations of FC from the tall tower.
 Following Peters et al. , which showed that species' differences in tree transpiration rates could be explained largely by plant functional type, we used evergreen needleleaf and deciduous broadleaf plant functional types to estimate FC(T) at the ecosystem scale, such that:
where FC(Teg) is the net CO2 exchange of evergreen needleleaf trees and FC(Tdec) is the net CO2 exchange of deciduous broadleaf trees.
 We estimated the relative error of our FC(Teg) and FC(Tdec) estimates by propagating errors associated with each term in equation (3) [Taylor, 1982]. EC was assumed to have a relative error of 29% and 23% for evergreen needleleaf and deciduous broadleaf trees, respectively, based on measured variation in growing season transpiration rates among individual trees of the same species. We assumed the measurement error associated with D and T was negligible. The slope parameter a was assumed to have a relative error of 8% and 15% for evergreen needleleaf and deciduous broadleaf trees, respectively, based on the coefficients of determination of the regression models used to predict AC. The NPP:GPP ratio was assumed to have a relative error of 28% based on the variation in published NPP:GPP ratios reported for temperate forests [DeLucia et al., 2007].
2.4. CO2 Fluxes From Turfgrass
 Following Peters et al. , which showed the importance of including irrigated and non-irrigated turfgrass vegetation types when scaling up ecosystem evapotranspiration in our study area, we estimatedFC(G) at the ecosystem scale, such that:
where FC(Gnirr) is the net CO2exchange of non-irrigated turfgrass lawns andFC(Girr) is the net CO2 exchange of irrigated turfgrass lawns.
FC(Gnirr) was measured using the eddy covariance method over a 1.5 ha turfgrass field located in the footprint of the tall tower [see Hiller et al., 2011], such that
where w is the vertical wind velocity, CO2 is the carbon dioxide concentration, and the primes indicate instantaneous departures from the mean [Baldocchi et al., 1988]. Briefly, we used an eddy covariance system consisting of a 3-D sonic anemometer (CSAT3, Campbell Scientific, Logan, Utah) and an open-path infrared gas analyzer (LI-7500, LI-Cor, Lincoln, Neb.) inclined at an angle of 30°, both mounted at a height of 1.35 m above the ground on a portable meteorological tripod (905, Met One Instruments, Grants Pass, Oreg.). Wind velocity, air temperature, and concentrations of carbon dioxide were recorded at 20 Hz using a data logger (CR5000, Campbell Scientific, Logan, Utah). Soil temperature and volumetric moisture content were recorded at two locations at 0.05 m below ground (STP1, Radiation and Energy Balance Systems, Inc., Seattle, Washington, and EC-10, Decagon Devices, Pullman, Washington, respectively). Fluxes were computed over 30-min periods using a version of theeth-flux program [Mauder et al., 2008] with minor modifications to accept structured ASCII raw data input files. Data processing steps, including corrections [Webb et al., 1980; Schotanus et al., 1983; Moore, 1986], are described in detail in Hiller et al. . In addition, we implemented a sensor self-heating correction for the LI-7500 followingBurba et al.  and Rogiers et al. , although this had only a small effect on the fluxes, especially during the growing season. Gaps in measurements from January 2007 to December 2008 represented 42% of the data, due to unfavorable wind directions from behind the tower (0–180°), power outages, and unsatisfied quality criteria [Foken and Wichura, 1996]. Gaps in FC(Gnirr)were filled using a temperature-driven model of ecosystem respiration [Lloyd and Taylor, 1994] and light-response curves [Ruimy et al., 1995]. Parameters of the gap-filling models were determined using an overlapping, centered, running window that started with a length of 7 days and increased by 2 days in each loop until there were at least 100 valid measurements in the window and the model parameters were statistically significant (Student's t-test, p < 0.1).
 Due to logistical constraints, it was not possible to obtain permission to install a portable eddy covariance tower in a large irrigated lawn, such as the golf course in our study area. Therefore, we estimated FC(Girr) by using our FC(Gnirr)measurements from spring and fall periods of each year to fit light-response and ecosystem respiration models for periods when water was not limiting to turfgrass growth. We then applied those models using observed solar radiation and soil temperatures during the midsummer periods when non-irrigated turfgrass growth was reduced. While this provided a more realistic basis for scalingFC(Girr) within the tall tower footprint, our values may underestimate the actual growth rate of irrigated turfgrasses, which would have benefited from fertilization and other management practices [Bowman, 2003]. This potential source of error would cause our bottom-up sums ofFC(VegSoil) to be a conservative estimate of the actual rate of CO2 uptake by turfgrass within the study area.
 Based on field surveys, approximately 50% of the tree-covered area at our site had a turfgrass understory. Because it was logistically impossible to obtain eddy covariance measurements under a representative urban tree canopy, we modeled the net CO2 exchange of understory turfgrass (FC(Gunder)) as a function of solar radiation passing through the tree canopy, using the Beer-Lambert law [Campbell and Norman, 1998]. The resulting estimates of FC(Gunder) were generally not significantly different from zero, and in some cases indicated net CO2release due to low light levels and the fact that light-response functions were obtained from an open site. In contrast, shaded areas often have a different turfgrass species composition (e.g., fine fescues) and we would expectFC(Gunder) to show small values of net CO2 uptake based on literature values [Qian et al., 2010]. Therefore, to be conservative in estimating the net CO2 uptake by vegetation within the study area, we assumed that FC(Gunder)was zero in our bottom-up sums ofFC(VegSoil).
2.5. Total Net CO2 Flux Measurements
 Total net CO2 fluxes (FC) over the suburban landscape were measured from the 40 m level on the tall tower using an eddy covariance system consisting of a 3-D sonic anemometer (CSAT3, Campbell Scientific, Inc., Logan, Utah) and a closed-path infrared gas analyzer (LI-7000, LI-Cor, Lincoln, Neb.). Further details of the measurement system are described inPeters et al. . Briefly, 30-min turbulentFC fluxes were computed using same version of eth-flux used for processing the turfgrass site [Mauder et al., 2008]. Data were screened to remove periods when instruments were being serviced and wind directions were from the southeast (90°–180°) to avoid interference from the tower structure. Gaps in measurements from January 2007 to December 2008 represented 48% of the data, due to unfavorable wind directions from behind the tower (20%) and system maintenance or power outages (28%). We performed the minimum necessary level of filtering to maximize the amount of data available for comparisons of wind sectors. Due to the complexity of the suburban land surface and our focus on evaluating scaled-up ecologicalFC(VegSoil) fluxes with measured FC, we did not gap-fill the tall tower data for the analyses presented here. To compare daily sums ofFC from different wind sectors, we constructed mean monthly diurnal composites.
2.6. Scaling Up Component Fluxes
 To calculate FC(VegSoil), we scaled up our vegetation component measurements to the tall tower footprint using a land cover map and a flux footprint model, following Peters et al. . The land cover map was produced from QuickBird (2.4 m resolution) multispectral imagery acquired on 26 July 2006 and leaf-off, true color aerial imagery (0.15 m resolution) acquired by Ramsey County on 8 April 2003. It had an overall classification accuracy of 82%, based on a per-pixel accuracy assessment against a field survey using the U.S. Forest Service Forest Inventory and Analysis (FIA) urban forest inventory protocol [U.S. Forest Service, 2005]. Details of the land cover classification are provided in Peters et al. .
 Footprint models have been widely used to estimate the source area of eddy covariance measurements [Schmid, 1994; Kljun et al., 2004]. Although our study area contains considerable spatial heterogeneity in surface types, which violates footprint model assumptions about extended homogeneous surfaces [Vesala et al., 2008b], it has a relatively flat topography, a dense tree canopy that overtops most of the buildings, and lacks deep street canyons caused by tall buildings. Therefore, we used a parameterized version of the Kljun et al. Lagrangian stochastic footprint model to estimate the flux source area for every 30-minFC measurement from the tall tower. As the Kljun et al. parameterization does not predict 2-D flux source areas, we used the following approach based onBarcza et al. to obtain a spatial sample of the fractional cover of land surface types within the source area. First, we used the footprint model to calculate distance from the tall tower to the maximal contribution point as well as the distances corresponding to every tenth percentile of the footprint for each 30-min measurement ofFC. Second, we plotted the 10 points of the flux footprint along the mean wind direction on the land cover map in a GIS and calculated the fractional cover of each surface type within a 1 ha circular area (radius = 56 m) around each point. Third, we calculated the average of the 10 sampled areas, each weighted by the crosswind-integrated footprint corresponding to that point, to obtain the final estimate of the fractional land cover that contributed to each 30-minFC measurement. For comparison with measured FC, the component fluxes FC(Teg), FC(Tdec), FC(Gnirr), and FC(Girr) were multiplied by the fractional land cover of each corresponding vegetation type, and summed.
 Next, to evaluate the reasonableness of our scaled up FC(VegSoil) estimates, we compared them to the total FCmeasured from the tall tower. Our comparisons included all 30-min measurements during the 2007 and 2008 growing seasons when conditions satisfied the criteria of the footprint model we used [Kljun et al., 2004], including friction velocity u* > 0.2 m s−1 and similarity parameter −200 ≤ zm/L ≤ 1, where zm is the measurement height minus the displacement height, and Lis the Monin-Obukhov length. Overall, 78% ofFCmeasurements met these footprint model criteria. We were precluded from making comparisons over time periods >30-min because gaps in the tall tower flux time series meant that we were missingFC measurements as well as the footprint information needed to scale up the component fluxes.
 To assess the relative importance of fossil fuel emissions from different wind sectors, we used continuous traffic count data (Ramsey County Department of Public Works) at the intersection closest to the tall tower (Figure 1). Daily traffic volume for the northwest sector was calculated as the sum of eastbound and westbound vehicle counts; the northeast sector as the sum of northbound, southbound, eastbound, and westbound vehicle counts; the southwest was assumed to have no traffic. We focused on traffic because other CO2 sources (e.g., natural gas used for water heating and cooking, and human respiration) that would have contributed to the measured FC could be expected to be relatively small and similar in magnitude between the northwest and northeast sectors due to the similar residential land use.
 Finally, we compared the FC(VegSoil)of the two major land-use types within the suburban landscape, a residential neighborhood and a recreational area in a golf course (Figure 1). To do this, we scaled up ecological component fluxes according to their respective land cover fractions within each land-use type (Table 1).
3.1. CO2 Exchange by Trees
 The relationship between leaf-level photosynthesis (AL) and stomatal conductance (gS) varied among the ten tree species sampled during the 2007 and 2008 growing seasons (Figure 2). This relationship was not sensitive to the measurement site, to summer versus fall sampling times, or to upper- versus lower-canopy positions, as described below. For those species measured at more than one site, there were no significant differences in the relationship betweenAL and gSamong sites (F-distribution statistics with degrees of freedom: F[1,97] = 0.84 for F. pennsylvanica; F[1,27] = 0.02 for P. glauca; F[1,73] = 2.53 for P. sylvestris; F[2,105] = 1.00 for T. americana, all P > 0.16). The relationship was also not significantly different between summer (June to August) and fall (September and October) sampling dates for all species (P > 0.10) except Fraxinus pennsylvanica, where the slope of the relationship was slightly higher in fall than in summer (F[1,97] = 29.6, P < 0.001). In addition, the relationship between AL and gS did not differ between upper and lower canopy positions (all species P > 0.15, Figure 2). The linear regression slope parameters that were multiplied with GCto estimate canopy-levelAC are given in Table 2. Although leaves were sampled across the full height of the tree canopies, access restrictions due to the aerial lift truck meant that most of the measured leaves were in the outer canopy and thus exposed to more direct sunlight. Consequently, the linear regression models may be more representative of sun-adapted than shade-adapted leaves, which could lead to a potential overestimate of gross primary production.
 Seasonal patterns of AC per cover area, or projected crown area (g C m−2 day−1) varied among tree genera. Genera measured in 2008 are shown, for example, in Figure 3. Evergreen needleleaf genera (Picea and Pinus) had higher daily AC across the entire growing season than deciduous broadleaf genera (Fraxinus, Ulmus, Tilia and Juglans), reaching maximum AC rates of >10 g C m−2 day−1 in May and June. Daily AC for three deciduous genera (Fraxinus, Ulmus, and Juglans) remained <5 g C m−2 day−1 across the entire growing season. Tilia, however, had daily AC >8 g C m−2 day−1, particularly in early summer. The main seasonal control on AC was GC (data not shown).
3.2. CO2 Exchange by Turfgrass
 Mean daily net CO2exchange from the non-irrigated turfgrass lawn (FC(Gnirr)) showed marked seasonal variations, with the highest net CO2 emissions in summer and highest net uptake in spring (Figure 4a). Throughout the measurement period, daily FC(Gnirr) ranged from a maximum of 5.8 g C m−2 day−1 in summer 2007 to a minimum of −6.5 g C m−2 day−1in spring 2008. On an annual basis, the non-irrigated lawn showed greater net CO2 uptake in 2008 (−71 g C m−2 yr−1), due to high uptake in spring and low emissions in summer, and was a net source of CO2 in 2007 (92 g C m−2 yr−1). Modeled CO2 fluxes from irrigated turfgrass (FC(Girr)) showed similar seasonal patterns (Figure 4b), but had greater CO2 uptake as compared to FC(Gnirr) during the warmer and drier midsummer period.
3.3. Vegetation Type Differences in Net CO2 Exchange
 The magnitude and seasonal patterns of net CO2 exchange varied markedly among the dominant vegetation types in the suburban landscape. Across the 2008 growing season (April to November), evergreen needleleaf trees (FC(Teg)) had greater uptake rates than all other vegetation types on a per cover area basis (Figure 5a). For perspective, we note that evergreen trees had the highest midsummer leaf area index (8.8 m2 m−2), followed by deciduous trees (5.5 m2 m−2) and the non-irrigated turfgrass lawn (1.7 m2 m−2) [Peters et al., 2010]. Seasonal patterns of net CO2 exchange by all vegetation types were consistent with expected differences in phenology, with FC(Teg) showing the highest daily rates of net CO2uptake in early spring, prior to leaf-out of deciduous trees. Deciduous broadleaf trees (FC(Tdec)) showed a pronounced seasonal carbon uptake pattern due to spring leaf-out and fall senescence, showing maximum daily net CO2 uptake from June to September. Turfgrasses (both FC(Gnirr) and FC(Girr)) showed peaks in net CO2uptake in May and October and a period of reduced growth during warm and dry conditions in midsummer, which is the characteristic seasonal pattern of cool-season C3 turfgrasses [Fry and Huang, 2004]. Throughout the growing season, FC(Teg), FC(Tdec), and FC(Girr) all represented net sinks of CO2, whereas FC(Gnirr) was a daily net source of CO2 to the atmosphere in July and August.
 The total growing season net CO2 uptake from April to November on a per cover area basis was largest for FC(Teg) (−603 g C m−2), followed by FC(Tdec) (−216 g C m−2), FC(Girr) (−211 g C m−2), and FC(Gnirr) (−115 g C m−2) (Figure 5b). The propagated relative errors for FC(Teg) and FC(Tdec)were similar at 41% and 39%, respectively. These relative errors had a large component associated with the literature-based NPP:GPP ratio (equal for all trees at 28%), with the other significant component coming from ourEC measurements (29% and 23% for evergreen and deciduous trees, respectively). This suggests that the estimates of FC(Teg) and FC(Tdec) reflect true differences in total net CO2 exchange between evergreen and deciduous trees. For perspective, the relative errors in FC(Teg) and FC(Tdec) were twice as large as the typical random error of ∼20% for eddy covariance measurements [Baldocchi et al., 1988], which were used to measure FC(Gnirr).
3.4. Total Suburban CO2 Fluxes
 The total net CO2 exchange (FC) above the suburban landscape, as measured from the 40 m level on the tall tower, is shown in Figure 6. In both 2007 and 2008, FC was largest (net CO2 emission) in winter and smallest (near zero or net CO2 uptake) from May to July. FC also varied spatially, with largest values of FC from the northeast (0°–90°) and northwest sectors (270°–360°) and smallest from the southwest sector (180°–270°). In April, we observed smaller FC in 2007, due to an earlier onset of spring, as compared to the same month in 2008. Throughout the measurement period, average daily FC per month ranged from a maximum of 2.2 g C m−2 day−1 in winter to a minimum of −0.2 g C m−2 day−1 in summer. While data gaps prevented a robust comparison between weekday and weekend daily FC, we observed similar seasonal and spatial patterns of daily FC for weekend and weekdays with only slightly greater carbon uptake on summer weekends in 2007 from the southwest sector compared to weekdays (−0.14 versus 0.08 g C m−2 day−1, data not shown).
 Our component-based totalFC(VegSoil) closely matched the tall tower measurements of FC during summer months and when winds were from the southwest or northwest, the conditions under which we would have expected FC to be dominated by FC(VegSoil) from urban vegetation and soils (Figure 7a). In contrast, FC(VegSoil) was significantly lower than FC during early spring and late fall and when winds were from the northeast, conditions under which fossil fuel emissions had the most impact on FC. The spatial variations in the difference between FC and FC(VegSoil) can be explained by a greater importance of fossil fuel emissions from higher traffic volumes in the northeast, compared to the southwest and northwest areas. During the 2007 to 2008 measurement period, mean daily traffic counts (±1 SD) at the intersection nearest the tall tower were 9440 (±2660) vehicles day−1 from the northeast wind sector, 2500 (±420) vehicles day−1 from the northwest, and 0 vehicles day−1 from the southwest. Seasonal variations in the difference between FC and FC(VegSoil) can be explained by an increased need for space heating during colder months in early spring and fall. At air temperatures below 20°C, a typical setting for home thermostats, the difference between FC and FC(VegSoil) grew larger (Figure 7b). The correlation coefficient between mean monthly air temperature and average monthly difference between FC and FC(VegSoil) was −0.82 for the northeast, −0.86 for the northwest, and −0.87 for the southwest wind sector.
3.6. Differences in FC(VegSoil)Between Suburban Land-Use Types
 Total growing season (April to November) FC(VegSoil) was lower on average (greater net CO2uptake) in recreational land-use areas (−165 g C m−2) compared to residential land-use areas (−124 g C m−2) (Figure 8a). In both land-use types, there was less net CO2uptake in 2007 as compared to 2008, due to warmer, drier conditions during midsummer in 2007 (growing-season mean air temperatures of 15.2°C in 2007 and 13.2°C in 2008). Additionally, this effect was more pronounced in the residential area than the recreational area (49% versus 29% less net CO2 uptake, respectively) because the latter had a high percent cover of irrigated turfgrasses that were less affected by the warm, dry conditions in 2007. In 2008, FC(Girr) accounted for more than twice the percentage of growing season total FC(G) in recreational compared to residential areas (91% versus 40%, respectively). In 2007, FC(Girr) represented 101% and 130% of FC(G) in recreational and residential areas, respectively, due to warmer, drier conditions in June and July that caused FC(Gnirr) to be a net source of CO2 for the growing season. Turfgrasses represented the largest contribution (77%) to total growing season FC(VegSoil) in the recreational area, while trees represented the largest contribution (78%) in the residential area. In both years, FC(Tdec) was the dominant component of growing season total FC(T) in both recreational and residential land use types (89% and 73%, respectively) because deciduous trees were much more abundant than evergreen trees in the suburban area we studied.
 In 2008, the contribution of different vegetation types to FC(VegSoil)varied seasonally in both land-use types (Figures 8b and 8c). FC(Girr) and FC(Gnirr) had the highest proportional contributions to FC(VegSoil) in spring (May) and fall (October), while the largest contribution from FC(Tdec) occurred from June to September. Although it accounted for only a small part of the total, FC(Teg) had the largest contribution in April, particularly in the residential area. The overall seasonal pattern of FC(VegSoil)in both residential and recreational land-use types was strongly influenced by turfgrasses and consequently showed maximum carbon uptake in May and October and lowest uptake in August. RecreationalFC(VegSoil)was more variable over the growing season with a large spring peak in carbon uptake, following by near-zero uptake in midsummer when turfgrass growth declined. In contrast, the residential area was a smaller, but more consistent, net sink for CO2 during the growing season because trees, which can tap into deeper water sources, provided a constant sink of carbon throughout the summer.
4.1. Comparing Net CO2 Exchange of Different Urban Vegetation Types
 Net CO2 exchange rates of the major urban vegetation types we measured fell within the range of values reported in similar urban and suburban areas. Annual productivity estimates of urban trees in the United States based on allometric equations range from −180 to −660 g C m−2 yr−1, depending on tree conditions, sizes, and estimated mortality [Nowak and Crane, 2002; Nowak et al., 2008]. This range encompasses the growing season net CO2 uptake per cover area that we measured from deciduous broadleaf trees (−216 g C m−2) and evergreen needleleaf trees (−603 g C m−2). The annual net CO2exchange that we measured in a non-irrigated lawn ranged from a sink of −71 g C m−2 yr−1in 2008, a year with near-average weather, to a net source of 92 g C m−2 yr−1 in the dry, warm year of 2007. These values bracket the range of carbon uptake (−20 to −80 g C m−2 yr−1) estimated for average years from Biome-BGC model runs for turfgrasses in Minneapolis, Minn. [Milesi et al., 2005; Fissore et al., 2011b]. Studies of soil carbon accumulation in turfgrass lawns similarly estimate net CO2 uptake ranging from −89 to −133 g C m−2 yr−1under long-term average climate conditions [Jo and McPherson, 1995; Qian and Follett, 2002; Kaye et al., 2005].
4.2. Scaling Up Net CO2 Exchange of Vegetation and Soils
 Our comparison of measured total CO2 fluxes (FC) and scaled-up estimates ofFC(VegSoil)showed that component-based, or “bottom up,” approaches can capture much of the seasonal and spatial variation in suburbanFC(VegSoil). We found close agreement between measured FCand component-basedFC(VegSoil) when fossil fuel emissions were at a minimum: during the summer, when heating emissions were lowest, and from wind sectors associated with low traffic volumes. The seasonal differences between FCand scaled-upFC(VegSoil)were consistent with increased fossil fuel combustion for space heating in winter, as reported in other cold-climate cities [Bergeron and Strachan, 2011]. Based on data from a household energy survey [Fissore et al., 2011a], 98% of the households within our study area use natural gas for space heating. The spatial discrepancies we observed among wind sectors were consistent with studies that have found greater CO2 emissions from areas with higher traffic volumes [Coutts et al., 2007; Vesala et al., 2008a; Bergeron and Strachan, 2011; Christen et al., 2011; Crawford et al., 2011].
 Aside from these expected fossil fuel-caused differences between our measuredFCfluxes and component-basedFC(VegSoil), other potential sources of these discrepancies include remote point sources of CO2 and unaccounted for ecological components and processes. Point sources of CO2(e.g., traffic, non-residential buildings) outside the 90% contribution footprint area of the tall tower could have a disproportionately large influence on measuredFC measurements [Hiller et al., 2011]. Therefore, it is likely some of the difference we observed between FC and FC(VegSoil) in the southwest sector, which had little local influence from traffic and houses, could be attributed to remote point sources of CO2. Additionally, there were several ecological processes that we assumed to be negligible in scaling up FC(VegSoil) from different vegetation types, but that nonetheless represent potential sources of error in our landscape FC(VegSoil) estimates. For example, although open water (e.g., ponds) was a small fraction of the land cover in our study area (2%), it does represent a potential ecological source of CO2. Relatively little information exists on net CO2 exchange from urban water bodies specifically, yet freshwater lakes worldwide have generally been shown to be annual net sources of CO2 to the atmosphere [Cole et al., 1994]. We also did not account for biophysical and physiological differences between street and park-grown vegetation that are due to the different ground cover and microclimates associated with these two environments [Kjelgren and Montague, 1998; Montague and Kjelgren, 2004; Leuzinger et al., 2010]. Because these differences can be species-specific, it would be desirable for future urbanFC(VegSoil)studies to obtain measurements across a broader set of micro-environments, including street trees and lawns that are exposed to more severe environmental conditions than the relatively park-like or backyard conditions of the trees we measured.
4.3. Comparison to Surrounding Natural Ecosystems
 Compared to NEE in natural ecosystems in the mid-continental United States,FC(VegSoil) at our suburban site had peak rates of daily net CO2 uptake (−2.8 g C m−2 day−1) that fell at the low end of the range of peak net CO2 uptake reported for hardwood forests (−2 to −7 g C m−2 day−1) [Davis et al., 2003; Noormets et al., 2007] and tallgrass prairies in the Midwest (−5.4 to −8.4 g C m−2 day−1) [Suyker and Verma, 2001; Bremer and Ham, 2010]. While peak net CO2 uptake occurred in June or July at these forested and C4 tallgrass prairie sites, at our suburban site peak uptake occurred in May. The high cover of C3turfgrasses, which grow most actively in spring and fall, played a large role in the early green-up and altered seasonal patterns ofFC(VegSoil) at our site compared to surrounding natural ecosystems. Natural forest and tallgrass prairie ecosystems typically show an increase in CO2uptake at spring green-up, followed by a single peak in net CO2 uptake at the time of full leaf expansion, and then a decline in CO2 uptake with leaf senescence in the fall. In contrast, FC(VegSoil) at our suburban site had two peaks of net CO2 uptake, a large peak in the spring and a smaller one in the fall, due to the strong seasonal variability in C3 turfgrass growth.
4.4. Contributions of Trees and Turfgrasses to Suburban FC(VegSoil)
 Differences in the relative contributions of trees and turfgrass to landscape FC(VegSoil) at our study site were driven by their fractional cover and by differences among vegetation types in daily CO2 exchange throughout the growing season. Despite having the highest rates of carbon uptake on a cover area basis, evergreen needleleaf trees had a nearly negligible contribution to FC(VegSoil) because their fractional cover on the landscape was small. Turfgrass, however, due to its high cover and high rates of carbon uptake in spring and fall, represented the largest proportional contribution (77%) to growing season FC(VegSoil)in the recreational land-use portion of our study area. In residential areas, trees represented the largest contribution (78%) to growing seasonFC(VegSoil) due to their high cover and high rates of carbon uptake in midsummer.
 The proportional contribution of different vegetation types to FC(VegSoil)varied seasonally due to different patterns in physiological activity among plant functional types. While evergreen needleleaf and deciduous broadleaf trees both have maximum physiological function in midsummer in temperate ecosystems, the evergreen leaf habit and high cold tolerance allows evergreen needleleaf trees to remain physiologically active over a longer growing season than either deciduous broadleaf trees or cool-season turfgrasses [Havranek and Tranquillini, 1995; Catovsky et al., 2002; Givnish, 2002]. Correspondingly, we observed the highest proportional contribution to suburban FC(VegSoil) from deciduous broadleaf trees in summer and, albeit small, the highest contributions from evergreen needleleaf trees in April when deciduous trees were leafless. Our observation of a midsummer release of CO2 and negative contribution to FC(VegSoil)from turfgrass is consistent with the midsummer dormancy of cool-season turfgrasses [Feldhake et al., 1984; Fry and Huang, 2004; Zhang et al., 2007].
4.5. Suburban Total CO2 Fluxes
 Total CO2 fluxes (FC) measured at this suburban site were midway between values reported for other North American suburban areas with at least 50% vegetation cover [Bergeron and Strachan, 2011; Crawford et al., 2011; Ramamurthy and Pardyjak, 2011]. For example, average daily summer FC in suburban Salt Lake City was 3.49 g C m−2 day−1 [Ramamurthy and Pardyjak, 2011] and in suburban Baltimore was −1.25 g C m−2 day−1 [Crawford et al., 2011], both cities with 67% vegetation cover. Vegetation types were relatively similar between these cities and our study area, due to modifications of native plant communities that homogenize vegetation among cities in different climates [McKinney, 2006]. Although the total vegetation cover was higher at our suburban site (82%), the average daily summer FC of 0.25 g C m−2 day−1 was higher than Baltimore, mainly due to the influence of motor vehicle CO2emissions when the tall tower footprint intersected a northeast section of road with relatively high traffic volume. Overall, we expect that our results are representative of suburban areas in the mid-continental and northeastern United States where precipitation is normally adequate for urban trees and turfgrasses [Milesi et al., 2005], and where winter temperatures are low enough to exclude a broader range of tree species and C4warm-season turfgrasses that are common in more southerly climates.
 Although FC(VegSoil) clearly did not offset fossil fuel emissions at our suburban site on an annual basis, the seasonal patterns of FC(VegSoil) significantly altered the seasonality of FC. During the summer, net CO2uptake by vegetation approximately equaled fossil fuel emissions within the study area, resulting in near-zeroFC. However, it is important to recognize that this study measured only direct exchanges of CO2 within the neighborhood immediately surrounding the tall tower. Other carbon bookkeeping approaches that include CO2 emitted outside the study area due to activities of local residents (e.g., electric power generation, wastewater treatment, air travel) would find a larger magnitude of net CO2 emissions [Fissore et al., 2011a]. In addition, our study did not separately calculate the indirect effects that vegetation can have to reduce CO2 emissions from heating and cooling of buildings [Akbari, 2002].
 The significant magnitude and strong seasonality of FC(VegSoil)within this first-ring suburb suggests that municipal carbon budgets that do not account forFC(VegSoil) of urban vegetation and soils could overestimate the total CO2 emissions from urban areas. This is important because cities are increasingly interested in using urban tree planting and vegetation management to mitigate their greenhouse gas emissions and to generate offsets in carbon markets [McHale et al., 2007; Poudyal et al., 2010]. This will depend on quantifying FC(VegSoil)across the annual cycle, which our results suggest will require attention to the seasonality of different plant functional types and species, in addition to management practices. We note that in regional climates different from our study area (e.g., semi-arid cities) it will be important to quantitatively evaluate vegetation management in light of tradeoffs between ecosystem services, such as carbon sequestration, versus costs, such as water use [Pataki et al., 2011]. In addition, it is important to recognize that net CO2 exchange represents only a part of the total greenhouse gas balance associated with urban vegetation, and lawns in particular [Townsend-Small and Czimczik, 2010].
 Continuous measurements over two growing seasons showed significant variations in net CO2exchange among the major vegetation types in a suburban ecosystem, with the largest growing-season (Apr–Nov) net CO2 uptake on a per cover area basis from evergreen needleleaf trees (−603 g C m−2), followed by deciduous broadleaf trees (−216 g C m−2), irrigated turfgrass (−211 g C m−2), and non-irrigated turfgrass (−115 g C m−2). Vegetation types showed seasonal patterns of CO2 exchange similar to those observed in natural ecosystems. To our knowledge, this is the first study to report continuous measurements of net CO2 exchange of urban vegetation and soils (FC(VegSoil)) over one or more annual cycles. Scaled-upFC(VegSoil) agreed closely with total landscape CO2 flux (FC) measurements from the tall tower at times when fossil fuel emissions were at a minimum: during the summer, when heating emissions were lowest, and from wind sectors associated with low traffic volumes. This suggests that, in temperate zone cities similar to the one we studied, a relatively simple classification of vegetation types could capture much of the variability required to predict seasonal patterns and differences in FC(VegSoil) that result from changes in land use or vegetation composition.
 Although FC(VegSoil) did not offset fossil fuel emissions on an annual basis, the temporal pattern of FC(VegSoil) did significantly alter the seasonality of FC. During the summer, net CO2 uptake by FC(VegSoil)approximately equaled fossil fuel emissions within the study area, resulting in near-zeroFC. The total growing season FC(VegSoil)in recreational land-use areas averaged −165 g C m−2 and was dominated by turfgrass CO2 exchange (representing 77% of the total), whereas FC(VegSoil) in residential areas averaged −124 g C m−2 and was dominated by trees (representing 78% of the total). Our results suggest that the net CO2 exchange by urban vegetation and soils has significant effects on the magnitude and seasonality of total CO2emissions from developed land areas that have relatively high vegetation cover, such as the first-ring suburb we studied.
leaf-level photosynthesis,μmol m−2 s−1.
canopy-level photosynthesis, g C m−2 day−1.
total net CO2 exchange from the suburban landscape, g C m−2 day−1.
net CO2 emissions from combustion of fossil fuels, g C m−2 day−1.
net CO2 exchange from turfgrass, g C m−2 day−1.
net CO2 exchange from irrigated turfgrass, g C m−2 day−1.
net CO2exchange from non-irrigated turfgrass, g C m−2 day−1.
net CO2 exchange from understory turfgrass.
net CO2 emissions from human respiration, g C m−2 day−1.
net CO2 exchange from trees, g C m−2 day−1.
net CO2 exchange from deciduous broadleaf trees, g C m−2 day−1.
net CO2 exchange from evergreen needleleaf trees, g C m−2 day−1.
net CO2 exchange from vegetation and soils, g C m−2 day−1.
stomatal conductance, mol m−2 s−1.
canopy conductance, mol m−2 s−1.
net ecosystem CO2 exchange, g C m−2 day−1.
 We thank 350 homeowners in Falcon Heights, Lauderdale, Roseville, and Saint Paul, as well as the City of Lauderdale and the University of Minnesota for granting us permission to sample vegetation on their property. We thank the cities of Minneapolis and Saint Paul, Rainbow Tree Care, and S & S Tree Specialists for generously providing aerial lift trucks and operators for tree canopy measurements; Rebecca Montgomery for providing leaf gas-exchange instrumentation; Brian Horgan for advice and access to the turfgrass site at the University of Minnesota Turfgrass Research, Outreach, and Education Center; Larry Oberg for technical advice and access to the KUOM broadcast tower; Dave Ruschy for data from the University of Minnesota climate station; Natascha Kljun for her footprint model code; Matthias Zeeman for his footprint mapping script; and Rebecca Hiller for data processing and analysis. We are grateful for assistance with field measurements, software development, and image processing from Ahmed Balogun, Ben Freeman, Vikram Gowreesunker, Cheyne Hadley, Vicki Kalkirtz, Rebecca Koetter, Mark McPhail, Julia Rauchfuss, Scott Shatto, and Yana Sorokin. This work was supported by a grant to J. P. M. from NASA Earth Science Division (NNG04GN80G).