Ecological Applications, 22(1), 2012, pp. 349–360
Ó 2012 by the Ecological Society of America
Urban ecosystem services: tree diversity
and stability of tropospheric ozone removal
FAUSTO MANES,
1,3
GUIDO INCERTI,
1
ELISABETTA SALVATORI,
1
MARCELLO VITALE,
1
CARLO RICOTTA,
1
AND ROBERT COSTANZA
2
1
Department of Environmental Biology, Sapienza University of Rome, Piazzale Aldo Moro, 5, 00185 Rome, Italy
2
Institute for Sustainable Solutions, Portland State University, Portland, Oregon 97207 USA
Abstract. Urban forests provide important ecosystem services, such as urban air quality
improvement by removing pollutants. While robust evidence exists that plant physiology,
abundance, and distribution within cities are basic parameters affecting the magnitude and
efficiency of air pollution removal, little is known about effects of plant diversity on the
stability of this ecosystem service. Here, by means of a spatial analysis integrating system
dynamic modeling and geostatistics, we assessed the effects of tree diversity on the removal of
tropospheric ozone (O
3
) in Rome, Italy, in two years (2003 and 2004) that were very different
for climatic conditions and ozone levels. Different tree functi onal g roups s howed
complementary uptake patterns, related to tree physiology and phenology, maintaining a
stable community function across different climatic conditions. Our results, although
depending on the city-specific conditions of the studied area, suggest a higher function
stability at increasing diversity levels in urban ecosystems. In Rome, such ecosystem services,
based on published unitary costs of externalities and of mortality associated with O
3
, can be
prudently valued to roughly US$2 and $3 million/year, respectively.
Key words: air quality; ecophysiology; ecosystem function; GIS; Rome, Italy; sanitary benefits;
tropospheric ozone; urban forest.
INTRODUCTION
Human health and well-being is known to depend on
‘‘ecosystem goods and services’’ (Costanza et al. 1997,
MEA 2005). The concept of ecosystem services was
defined by Daily (1997) as ‘‘the conditions and processes
through which natural ecosystems, and the species that
make them up, sustain and fulfill human life.’’ Since then,
different definitions have been proposed for ‘‘ecosystem
services’’ (Boyd and Banzhaf 2007, Fisher and Turner
2008, Fisher et al. 2009). According to a recent review by
Escobedo et al. (2011), which focused on air pollution
mitigation by urban forests, ecosystem services are
considered the components (including functions) of
urban forests that are directly enjoyed, consumed, or
used to produce specific and measurable human benefits.
Ecosystem services are affected by the relationship
between ecosystem functioning, stability, and biodiversi-
ty (Balvanera et al. 2006, Costanza et al. 2007, Gamfeldt
et al. 2008). The understanding of these relationships is
needed for devising the best management and policy
tools for sustainable use of ecosystem services (Kremen
2005). Urban forests provide important ecosystem
services (Bolund and Hunhammar 1999, Jim and Chen
2008, Young 2010), such as modification of urban
microclimate by lowering temperatures (Pataki et al.
2011a), changing wind patterns, reduction of building
energy use (Akbari 2002), and improvement of local and
regional air quality by removal of atmospheric pollutants
(Nowak et al. 2006). These environmental benefits are
becoming increasingly important, as today more than
half of the world’s population (;3.3 billion people) live
in urban areas. Additionally, according to UN projec-
tions, cities are growing to unprecedented sizes, absorb-
ing nearly all of the growth in the human population over
the next three decades (United Nations Population Fund
2007), with potential implications for biodiversity
conservation issues (Dearborn and Kark 2009).
The role of urban forests in providing ecosystem
services has been investigated in many papers, consid-
ering both basic ecosystem functions, like primary
productivity (Kaye et al. 2006, Pataki et al. 2011b) and
emerging services, such as the improvement of urban air
quality (Yang et al. 2005, Nowak et al. 2006, McDonald
et al. 2007, Escobedo and Nowak 2009). The reduction
of air pollution by urban trees has been recognized as a
cost-effective component of pollution reduction strate-
gies in several urban areas, such as Washington, D.C.,
New York, Baltimore, Atlanta, and Chicago, in the
United States (Nowak et al. 2000, 2006, Yang et al.
2008, Morani et al. 2011), Beijing (Yang et al. 2005),
Santiago de Chile (Escobedo and Nowak 2009), London
(Tiwary et al. 2009), and Toronto (Millward and Sabir
2011). Among the air pollutants removed by urban
forests, tropospheric ozone (O
3
) is dominant in the
Manuscript received 24 March 2011; accepted 5 August
2011; final version received 23 August 2011. Corresponding
Editor (ad hoc): A. Guenther.
3
E-mail: fausto.manes@uniroma1.it
349
photochemical air pollution mixture in urban areas
during summer periods, particularly in Mediterranean
areas (Milla
´
n et al. 2000), with negative effects on public
health (Bell et al. 2006, Martuzzi et al. 2006). While
robust evidence exists that the physiology of the main
tree functional groups and their abundance and spatial
distribution within cities are the basic parameters
affecting the magnitude and efficiency of O
3
removal
from the urban environment (Escobedo and Nowak
2009), little is known about the effects of tree species
diversity on the magnitude and stability of this
ecosystem service.
The aim of this paper was to quantify and value the
effects of urban tree diversity on the O
3
removal in the
city of Rome (Italy). The underlying hypothesis is that
different tree functional groups exert a complementary
role in stabilizing this emerging ecosystem service over
time, and across different environmental conditions. A
spatial analysis integrating system dynamic modeling
and geostatistics was applied to estimate seasonal and
annual ozone removal by three functional groups of
urban trees, under two climatically different years: the
extremely dry year 2003, and the year 2004, which was
more representative of the average long-term climatic
pattern of the city of Rome (Gerosa et al. 2009).
M
ETHODS
Urban forests in the city of Rome
The city of Rome (41854
0
N, 12829
0
E) extends over an
area of ;1270 km
2
and hosts roughly 2.8 million
inhabitants. Overall, the city is characterized by high
levels of urban traffic and urban expansion, which
largely increased in the last decades. The urban
landscape in Rome is very heterogeneous in terms of
geology, soil, morphology, and land use (Fig. 1a). The
climate is mediterranean, with an average annual
temperature of 15.18C, average annual rainfall of 839
mm, and a typical hot and dry summer period favoring
high tropospheric O
3
concentrations (Manes et al. 2003).
Notwithstanding the long-lasting human impact, span-
ning over .2700 years, and the recent increase of the
urbanized surface, Rome is still considered as one of the
‘‘most green’’ Italian cities, with public green space
covering .20% of the total municipality area and
including a system of nine natural reserves for a total
cover of ;16 000 ha, hosting roughly 1200 plant species
(Celesti-Grapow et al. 2006). Residual fragments of
ancient woodlands still occur within the city boundaries,
hosting a wide set of different tree species ranging from
typical Mediterran ean ever green bro adleaf sp ecies
(Quercus ilex and Q. suber; hereafter broadleaves) to
deciduous Quercus woods (Q. cerris, Q. frainetto) and
conifer plantations (Pinus pinea). These three groups of
species show important functional differences in their
ecophysiological and phenological traits (Manes et al.
1997, Anselmi et al. 2004), providing an optimal case
study to test the diversity–stability relationship in urban
ecosystems.
As a detailed inventory of the urban forests of the
metropolitan area does not exist, a Landsat 5 TM image
(from 21 July 1999), with a spatial resolution of 30 3 30
m, was used to assess the distribution of the main tree
functional groups (evergreen broadleaves, deciduous
broadleaves, and conifers) across the city of Rome.
First, a supervised classification of the Landsat image
into 18 land use classes was performed using the TM
bands 3, 4, 5, and 7 by means of a maximum likelihood
algorithm (Anselmi et al. 2003). The overall accuracy of
the classification, calculated through an error matrix,
was 96%. Large green areas can be observed in the
suburban zones, and inside the city center (Fig. 1a). The
most important forested areas, located in the southern
coastal area, are the Castelporziano Presidential Estate,
characterized by high plant community diversity (Seu-
fert et al. 1997), and the Castel Fusano urban park.
Moreover, patches of urban forests are present in the
historical villas, such as Villa Ada, Villa Borghese, and
Veio Park (Celesti-Grapow et al. 2006).
The area covered by each functional group was then
estimated by assigning the urban forest classes in Fig. 1a
to three leaf categories (Fig. 1b). The attribution of the
forest classes to the corresponding functional groups,
being scale dependent, was necessarily affected by some
degree of approximation. Stands characterized by a
dominant species were entirely attributed to the corre-
sponding leaf type: for example, the land use classes
‘‘Holm oak prevailing’’ and ‘‘Cork oak prevailing’’ were
assigned to the ‘‘evergreen broadleaf’’ functional group,
deciduous woods (oak woods dominated by Q. cerris, Q.
frainetto, Tilia cordata, Platanus x acerifo lia,and
Robinia pseudoacacia woods) were assi gned to the
‘‘deciduous broadleaf’’ group, while Italian stone pine
woods were assigned to the ‘‘conifer’’ group. Deciduous
woods with sclerophyllous species were completely
attributed to the ‘‘deciduous broadleaf’’ category,
because in these stands the evergreen species are mostly
located in the understory layers, thus giving a negligible
contribution to pollutants uptake and deposition
processes (Manes et al. 2007). The areas covered by
mixed conifers and evergreen broadleaved species were
partitioned at 50% between the two leaf categories, while
the attribution of maquis with Holm oak prevailing
areas to ‘‘evergreen broadleaf’’ was limited to 50% of the
total coverage, considering only the area covered by
trees and excluding shrubs and herbaceous species.
As a result, the area covered by the tree functional
groups totaled 7198 ha, corresponding to 5.6% of the
municipa lity area (Fig. 1b). Deciduous bro adleaves
represented the most abundant functional group (3474
ha), followed by evergreen broadleaves (2121 ha) and
conifers (1605 ha). In particular, for the Castelporziano
Presidential Estate, the Castel Fusano urban park, and
Villa Ada, the vegetation cover was: 1080.3 ha, 230.8 ha,
and 38.5 ha for evergreen broadleaves; 2348.9 ha, 215.1
ha, and 35.7 ha for deciduous broadleaves; and 735.2 ha,
429.7 ha, and 42.2 ha for conifers, respectively.
FAUSTO MANES ET AL.350
Ecological Applications
Vol. 22, No. 1
Ozone data
Ozone (O
3
) and nitrogen oxides (NO
x
) concentrations
in air, hourly recorded by 8 and 13 ai r quality
monitoring stations, respectively, were considered for
the years 2003 and 2004. All categories of air pollution
monitoring sites of the municipal network (urban traffic,
urban background, suburban background, rural back-
ground) were included in the analysis. The spatial
distribution of O
3
concentrations over the survey area
was estimated by applying a spherical co-kriging model
FIG. 1. Urban vegetation of the Municipality of Rome (city limits outlined in black): (a) Landsat 5 TM supervised classification
(modified from Manes et al. [2008]) and (b) distribution and total surface cover of the three tree functional groups (evergreen
broadleaf species, deciduous broadleaf species, and conifers) analyzed in this study within the metropolitan area. Axes report UTM
grid coordinates, WGS84 zone 33 N.
January 2012 351URBAN TREE DIVERSITY AND OZONE REMOVAL
(Isaa ks and Srivastava 1989). Dai ly maps of O
3
concentrations were produced in a Geographic Infor-
mation System (GIS) by using the geostatistical tool
Spatial Analyst in ESRI ArcGIS v. 9.2 ( ESRI 2006),
where average daily values of O
3
and NO
x
have been
considered as input data. Further, NO
x
was used as an
external drift variable in the co-kriging interpolation
technique.
All maps were produced by smooth interpolation at
30 3 30 m resolution to be interoperable with the urban
vegetation distribution map (Fig. 1b). In this way, at
each map cell (30 3 30 m pixel) where the target tree
functional group was located, daily time series of O
3
concentrations were available for analyses. Finally, to
provide a synoptic view of O
3
levels and spatial
variability during 2003 and 2004, annual maps of
‘‘SOMO35’’ (i.e., sum of daily eight-hour running means
of O
3
over 35 ppb; World Health Organization 2008)
were also produced by summing up daily map values.
Modeling tree physiological parameters
The MOCA-Flux (Modeling of Carbon Assessment
and Flux) model, implemented within the object-
oriented software package STELLA II (Costanza and
Gottlieb 1998, Isee Systems 2002), was used to simulate
dynamics of physiological parameters for the three plant
leaf types, for 2003 and 2004. MOCA-Flux is the newly
implemented version of a system dynamic, semi-empir-
ical model, previously applied to simulate functional
responses to changes in air temperature (Vitale et al.
2003) and O
3
stomatal fluxes (Vitale et al. 2005) of Q.
ilex. The MOCA-Flux model is based on the ‘‘big-leaf’’
approach, and it was conceived for estimation of plant
physiological variables including stomatal conduc-
tance (g
s
; mol H
2
O
m
2
s
1
), net photosynthesis (P
NET
;
lmol CO
2
m
2
s
1
), leaf transpiration ( E;mmol
H
2
O
m
2
s
1
), annual net primary productivity (NPP;
g C/m
2
), and leaf area index (LAI; m
2
of leaf/m
2
of
ground). All physiological variables are expressed as a
diurnal average for the photoperiod.
The MOCA-flux model was applied in the current
study due to its demonstrated ability to provide highly
fitting prediction of several physiological parameters
(including stomatal conductance) for different plant
species (Manes et al. 1999, Vitale et al. 2005).
MOCA-Flux calculates stomatal conductance to
water vapor (Eq. 1) by using the Ball et al. (1987)
algorithm, and corrected by Harley et al. (1992), which
is based on net photosynthesis (P
NET
), relative humidity
(RH), and air carbon dioxide concentration ([CO
2
]
air
),
assumed to be constant at 370 lmol/mol, as follows:
g
s
ðtÞ¼g
s0
þ m
P
NET
ðtÞRHðtÞ
CO
2
½
air
ð1Þ
where g
s0
is the minimum stomatal conductance to H
2
O
vapor when P
NET
¼ 0 and m is an empirical coefficient
that represents the composite sensitivity of conductance
to P
NET
, [CO
2
]
air
, and RH. Net photosynthesis was
calculated as a function of species-specific quantum yield
and solar irradiance by using a semi-empirical model
reported in de Wit et al. (1978; see the Appendix). It is
noteworthy that the leaf area index (LAI) is also related
to net photosynthesis, affecting, in turn, solar irradiance
(see the Appendix). The different modules constituting
the MOCA-Flux model are highly integrated to each
other, thus yielding stable functional interdependences,
minimizing the number of input parameters.
Net primary productivity (NPP) is derived from the
total of diurnal net photosynthesis values integrated in
the phenological time span for each tree species. For
further details on model equations, refer to Vitale et al.
(2003) and (2005). The model has been parameterized
using values of in put physiological and structural
variables (see the Appendix) derived from field mea-
surements collected in different sampling sites of the
survey area (Anselmi et al. 2003, Manes et al. 2007,
Vitale et al. 2007), and simulations of daily average g
s
were run for the years 2003 and 2004. The model
validation was bas ed on reference comparison of
simulated O
3
fluxes with eddy covariance measurements,
as reported in Vitale et al. (2005) for summer 2003.
Ozone removal by urban tree functional groups
Stomatal ozone fluxes (FO
3
) were calculated on a
daily time step based on estimated O
3
air concentration
and simulated stomatal conductance to water vapor,
corrected by the diffusibility ratio between O
3
and water
vapor:
FO
3
ði; pÞ¼g
s
ðiÞ3 O
3
½
i;p
3 0:613 ð2Þ
where FO
3
is expressed in nmol
m
2
s
1
and [O
3
] in parts
per billion (ppb [nmol/mol]), and the indices i and p
refer to the ith day of the reference period and to the pth
location of each tree functional group, respectively.
Stomatal ozone fluxes were referred to unitary area of
soil surface, thus allowing a geographical representation
of the modelin g outputs, based on the loca tions
effectively covered by evergreen broadleaves, deciduous
broadleaves, and conifers within the survey area, as
reported in the vegetation map in Fig. 1b.
The annual time series of FO
3
was integrated over
time at each site to estimate the cumulative amount of
ozone yearly and seasonally taken up in 2003 and 2004
by each tree functional group:
FO
3
cumðpÞ¼
X
n
i¼1
FO
3
ði; pÞ3 Ph 3 3600
!
3 10
6
ð3Þ
where n is the number of cumulative days, Ph is the
photoperiodinhours,and10
6
is a dimensional
correction factor allowing to express the cumulated
stomatal flux in mmol O
3
m
2
yr
1
,whenFO
3
is
expressed in nmol
m
2
s
1
.
To estimate the uncertainty of yearl y cumulated
ozone fluxes, the standard deviation of daily stomatal
conductance (i.e., SD
gs
(i) for each ith day of the year)
FAUSTO MANES ET AL.352
Ecological Applications
Vol. 22, No. 1
from six model runs was considered. Consequently,
uncertainty for daily ozone flux was obtained from Eq. 2
for each vegetation type and for each day of the years
2003 and 2004
SD
FO3
i; pðÞ¼SD
g
iðÞ3 O
3
½
i; p
3 0:613: ð4Þ
Then, uncertainty for yearly ozone fluxes was obtained
by summing up daily contributions.
Significant differences of ozone removal between tree
functional groups were assessed by ANOVA (Duncan
test). Significance was evaluated in all cases at P , 0.05.
Average values of the stomatal : total flux ratio for Q.
ilex in the survey area were reported by Gerosa et al.
(2005, 2009), both for 2003 and 2004 (0.29 and 0.43,
respectively). Similar values, ranging from 0.21 and 0.33,
were reported for conifers by Mikkelsen et al. (2004),
though at higher latitude, thus allowing an estimation of
the potential cumulated flux of O
3
removed from
atmosphere by both stomatal uptake and non-stomatal
processes (FO
3t
), at each pth location, as follows:
FO
3t
ðpÞ¼FO
3
cumðpÞ3
1
R
ð5Þ
where R values were 0.29 and 0.43 for 2003 and 2004,
respectively.
To estimate the total amount of ozone removed from
the atmosphere by evergreen broadleaves, deciduous
broadleaves, and conifers in the Rome municipality, the
cumulated fluxes calculated at each lo cation were
totaled.
Given that all fluxes were referred to 1 m
2
of soil
covered by the target leaf types, and the resolution of the
vegetation map was 30 3 30 m, then the fluxes in each
pixel were calculated multiplying the flux by the pixel
area, weighted by the relative coverage of each leaf type
in each pixel:
FO
3
tot ¼
X
N
p¼1
FO
3
tðpÞ3 900 ð6Þ
where N is the number of 30 3 30 m pixels covered by
the given tree functional type (Fig. 1b), and 900 is the
area in m
2
of each pixel of the map.
The same method was used for estimating the O
3
removal that would have occurred in both years if all the
trees belonged to one single functional group. Three
configurations were considered, in each of which the
total area covered by tree vegetation within the
Municipality of Rome was attributed to one of the
functional groups in Eq. 6.
R
ESULTS
The years 2003 and 2004 were characterized by
different climatic conditions and ozone pollution levels.
Mean temperatures recorded at the air quality monitor-
ing stations in 2003 were higher than in 2004, for the
months of April (14.0861.88C vs. 12.9861.68C), May
(20.7862.08C vs. 16.1861.88C), June (26.4862.18C vs.
22.0862.28C), July (27.1 862.08C vs. 24.1861.98C),
and August (27.8862.18C vs. 24.1861.88C), whereas
total precipitation in 2003 was much lower than in 2004,
especially for April (51 6 12 mm vs. 104 6 16 mm), May
(6 6 3 mm vs. 85 6 4 mm), June (0 6 0 mm vs. 22 6 4
mm), and July (3 6 1 mm vs. 41 6 5 mm) (Fig. 2a, b).
Very different spatial (Fig. 3a, b) and temporal (Fig.
2c, d) patterns were observed in ozone concentrations
across the city of Rome during the two years. Mean
monthly values frequently exceeded the threshold of 70
lg/m
3
proposed by the World Health Organization to
quantify O
3
impact on human health (WHO 2008). In
particular, O
3
concentrations were equal to 74 6 5 lg/
m
3
in April, 79 6 17 lg/m
3
in May, 95 6 17 lg/m
3
in
June, 97 6 20 lg/m
3
in July, 94 6 23 lg/m
3
in August,
and 74 6 42 lg/m
3
in September, 2003, with corre-
sponding values for 2004 being fairly lower (51 6 12 lg/
m
3
in April, 67 6 13 lg/m
3
in May, 74 6 13 lg/m
3
in
June, 81 6 15 lg/m
3
in July, 75 6 13 lg/m
3
in August,
and 58 6 13 lg/m
3
in September (Fig. 2c, d).
Potential stomatal ozone uptake in 2003 and 2004 was
considerably different among evergreen broadleaves,
deciduous broadleaves, and conifers, showing peculiar
patterns both in time (Fig. 2e–j) and space ( Fig. 3c–f ).
In spring of both years, deciduous broadleaves showed
the highest, and conifers showed the lowest, potential
stomatal O
3
fluxes (Fig. 2g, i). In summer 2003,
deciduous broadleaves showed a reduced potential
stomatal O
3
flux due to the reduction of stomatal
conductance under limiting environmental conditions
(Fig. 2g, h), while evergreen broadleaves were able to
maintain high levels of potential stomatal O
3
fluxes (Fig.
2e, f), and conifers showed an increased O
3
uptake (Fig.
2i, j). In fall, the contribution of the three functional
groups was again different, with higher values estimated
for deciduous broadleaves and lower values for ever-
green broadleaves and conifers (Fig. 2e–j).
The seasonal cumulated stomatal O
3
fluxes (g/m
2
)
maps depicted in Fig. 3c–f, highlighted that evergreen
broadleaves showed cumulated stomatal flux values
fairly constant from spring to summer 2003 (;0.8 g/m
2
),
as slightly increased in 2004 (from 0.4 to 0.5 g/m
2
), as it
can be observed in the Villa Ada urban park and in the
Castelporziano Estate (Fig. 1b).The uptake values of
deciduous broadleaves, shown for example in the large
dec iduous forest dominated by Quercus cerrisand
Quercus frainetto, located in the Castelporziano Estate
(southern coastal area; see Fig. 1b), were on average
similar in the spring–summer of 2004 (0.6 g/m
2
) but, in
2003, were higher during spring (0.7 g/m
2
) than during
summer (0.4 g/m
2
). In the same area, conifers showed
ozone uptake values largely increasing from spring to
summer in both years (from 0.2 to 0.8 g/m
2
in 2003, and
from 0.4 to 0.8 g/m
2
in 2004).
Total (stomatal and non-stomatal) ozone uptake by
urban trees was 311.1 Mg in 2003 and 306.9 Mg in 2004,
with an interannual fluctuation between the two years of
January 2012 353URBAN TREE DIVERSITY AND OZONE REMOVAL
![FIG. 1. Urban vegetation of the Municipality of Rome (city limits outlined in black): (a) Landsat 5 TM supervised classification (modified from Manes et al. [2008]) and (b) distribution and total surface cover of the three tree functional groups (evergreen broadleaf species, deciduous broadleaf species, and conifers) analyzed in this study within the metropolitan area. Axes report UTM grid coordinates, WGS84 zone 33 N.](/figures/fig-1-urban-vegetation-of-the-municipality-of-rome-city-3oqqdn4y.webp)
