Mapping and Modeling the Biogeochemical Cycling of Turf Grasses in the United States

Abstract

Turf grasses are ubiquitous in the urban landscape of the United States and are often associated with various types of environmental impacts, especially on water resources, yet there have been limited efforts to quantify their total surface and ecosystem functioning, such as their total impact on the continental water budget and potential net ecosystem exchange (NEE). In this study, relating turf grass area to an estimate of fractional impervious surface area, it was calculated that potentially 163,800 km2 (+/- 35,850 km2) of land are cultivated with turf grasses in the continental United States, an area three times larger than that of any irrigated crop. Using the Biome-BGC ecosystem process model, the growth of warm-season and cool-season turf grasses was modeled at a number of sites across the 48 conterminous states under different management scenarios, simulating potential carbon and water fluxes as if the entire turf surface was to be managed like a well-maintained lawn. The results indicate that well-watered and fertilized turf grasses act as a carbon sink. The potential NEE that could derive from the total surface potentially under turf (up to 17 Tg C/yr with the simulated scenarios) would require up to 695 to 900 liters of water per person per day, depending on the modeled water irrigation practices, suggesting that outdoor water conservation practices such as xeriscaping and irrigation with recycled waste-water may need to be extended as many municipalities continue to face increasing pressures on freshwater.

Full text

Mapping and Modeling the Biogeochemical Cycling of
Turf Grasses in the United States
CRISTINA MILESI*
STEVEN W. RUNNING
Numerical Terradynamic Simulation Group
College of Forestry and Conservation
University of Montana
Missoula, Montana 59812, USA
CHRISTOPHER D. ELVIDGE
NOAA National Geophysical Data Center
325 Broadway, Boulder, Colorado 80303, USA
JOHN B. DIETZ
Cooperative Institute for Research on the Atmosphere (CIRA)
Colorado State University
Fort Collins, Colorado 80523-1375, USA
BENJAMIN T. TUTTLE
Cooperative Institute for Research in the Environmental
Sciences (CIRES)
216 University of Colorado
Boulder, Colorado 80309
RAMAKRISHNA R. NEMANI
NASA Ames Research Center
MS-242-4, Moffett Field, California 94035, USA
ABSTRACT / Turf grasses are ubiquitous in the urban
landscape of the United States and are often associated
with various types of environmental impacts, especially on
water resources, yet there have been limited efforts to
quantify their total surface and ecosystem functioning,
such as their total impact on the continental water budget
and potential net ecosystem exchange (NEE). In this
study, relating turf grass area to an estimate of fractional
impervious surface area, it was calculated that potentially
163,800 km
2
(€ 35,850 km
2
) of land are cultivated with turf
grasses in the continental United States, an area three
times larger than that of any irrigated crop. Using the
Biome-BGC ecosystem process model, the growth of
warm-season and cool-season turf grasses was modeled
at a number of sites across the 48 conterminous states
under different management scenarios, simulating poten-
tial carbon and water fluxes as if the entire turf surface
was to be managed like a well-maintained lawn. The re-
sults indicate that well-watered and fertilized turf grasses
act as a carbon sink. The potential NEE that could derive
from the total surface potentially under turf (up to 17 Tg C/
yr with the simulated scenarios) would require up to 695 to
900 liters of water per person per day, depending on the
modeled water irrigation practices, suggesting that out-
door water conservation practices such as xeriscaping
and irrigation with recycled waste-water may need to be
extended as many municipalities continue to face
increasing pressures on freshwater.
Turf grasses are ubiquitous in the American urban
landscape, in residential, commercial, and institutional
lawns, parks, most athletic fields and golf courses, often
as monocultures, independently of the local climate
(Jenkins 1994). Existing estimates indicate that in the
early 1990s, the surface cultivated with turf was up to
three times larger than that of irrigated corn, the
largest irrigated crop in the United States (DPRA,
Incorporated 1992). As the construction of new homes,
averaging 1.6 million per year in the late 1990s (U.S.
Bureau of the Census 1999), continues to expand the
American urban landscape, the total surface under turf
is expected to further increase.
Turf grasses contribute to soil carbon (C) sequestra-
tion (Bandaranayake and others 2003, Qian and Follett
2002, Van Dersal 1936) and, as a component of urban
vegetation, to the mitigation of the urban heat island ef-
fect (Spronken-Smith and others 2000) and to enhanced
water infiltration compared to bare soil or impervious
surfaces. However, turf has also been linked with a num-
ber of negative environmental impacts. Turf grasses often
pose a neglected environmental hazard through the use
of lawn chemicals and overfertilization (Robbins and
Birkenholtz 2003, Robbins and others 2001), and, where
used, irrigation of turf grasses sharply increases the sum-
mer water consumption for residential and commercial
use, especially if grown in arid and semiarid regions,
where it can account for 75% of the total household water
consumption (Mayer and others 1999).
KEY WORDS: Turf grasses; BIOME-BGC; Impervious surface area;
Carbon budget; Carbon sequestration potential; Wa-
ter use
Published online July 19, 2005.
*Author to whom correspondence should be addressed; email: milesi
@ntsg.umt.edu
Environmental Management Vol. 36, No. 3, pp. 426–438 ª 2005 Springer Science+Business Media, Inc.
DOI: 10.1007/s00267-004-0316-2

In spite of the pervading presence of turf grass sys-
tems in the urban and suburban landscape and their
considerable use of water resources, a national assess-
ment of the ecological functioning of these systems is
still missing. The fragmented distribution of residential
and commercial lawns and the large variability in
management practices adopted to grow the different
types of turf surfaces certainly challenges the task of
such an assessment.
In this study, we attempt a first estimate of the po-
tential impact of turf grasses to the continental U.S.
carbon and water budgets by producing a spatially ex-
plicit estimate of their distribution within the contig-
uous 48 states and simulating their growth with an
ecosystem process model. Specific objectives of this
study are (1) to compare a remote sensing-based esti-
mate with other independent estimates of the total
surface under turf grasses, and (2) to evaluate the im-
pact of different turf management practices, such as
removal versus on-site decomposition of the grass
clippings, varying nitrogen fertilization regimes, and
alternative irrigation schedules, on the continental
carbon and water budgets.
Methods
Estimation of U.S. Turf Surface
A continental assessment of the carbon and water
balances of turf grasses requires their spatial distribu-
tion to be mapped. With the exception of some golf
courses, turf grasses are rarely cultivated on surfaces
large enough to be identifiable with moderate resolu-
tion satellite data (1 km). Due to excessive costs and
time constraints, the use of high-resolution satellite
images or aerial photography has been limited. Past
efforts to estimate the continental surface of turf grasses
used indirect approaches that provided measures of the
total surface under lawn on a state-by-state basis,
therefore lacking the spatial detail required to calculate
spatially dependent biogeochemical cycles. Vinlove and
Torla (1995), for example, estimated the national total
home lawn area using methods based on adjusted
Federal Housing Authority (FHA) average and median
lot sizes by state, without accounting for the turf sur-
faces found in golf courses, parks, schools, roadsides,
etc. DPRA, Incorporated (1992), in a report commis-
sioned by the Environmental Protection Agency, esti-
mated the total area under turf on the basis of direct
surveys in 12 states, which were extrapolated to the
remaining states in proportion to their population.
In our study, also adopting an indirect approach,
we assumed the surface of turf grasses to be inversely
related to the amount of impervious surface associ-
ated with urban development (roads, roofs, parking
lots, sidewalks, etc.). We first calculated a fractional
cover of Impervious Surface Area (ISA) for the 48
states at 1-km spatial resolution using 2001 radiance
calibrated nighttime lights, a 1-km grid of road den-
sity and Landsat-derived urban landcover classes (El-
vidge and others 2004). The road density was
calculated as the length of road per square kilometer
from 1998 TIGER (Topologically Integrated Geo-
graphic Encoding and Referencing System) road
vector data from the U.S. Census Bureau. The night-
time lights were produced using cloud-free portions
of DMSP/OLS (Defense Meteorological Satellite Pro-
gram/Operational Linear Scanner) data using meth-
ods described by Elvidge and others (1999). We also
used direct measurements of the proportion of con-
structed surface (roads, parking lots, buildings) versus
the proportion of vegetated (turf grasses and/or
trees) or other (undeveloped) surface calculated from
80 high-resolution aerial photographs collected along
development transect distributed across 13 major ur-
ban centers. The transects extended from the urban
cores out to the sparsely developed (or undeveloped)
fringes of the urban centers of Atlanta, Boston, Chi-
cago, Denver, Houston, Las Vegas, Miami, Minneap-
olis, New York, Phoenix, Portland, Sacramento, and
Seattle. The aerial photographs were from year
2000 ± 1 year (see Figures 1 and 2 for an example of
the aerial photographs over industrial and residential
areas of Chicago, respectively). The measurements
were done on square kilometer tiles extracted from
the aerial photographs to match the coverage of
specific cells in the satellite and road density grids,
and were used to develop an empirical relationship
between the fractional ISA and the radiance cali-
brated nighttime lights, road density, and Landsat-
derived urban land cover classes. The highly signifi-
cant regression model (P - value < 0.0001 and Root
Mean Square Error (RMSE) of 13.4, Figure 3) was
then applied to the conterminous United States to
produce a 1-km grid depicting the spatial distribution
of ISA in percentage terms. The total ISA for the
conterminous United States calculated with this
method was estimated to be 112,610 (±12,725) km
2
or
1.3% of the total area.
The proportions of impervious versus vegetated
surfaces derived from the high-resolution aerial pho-
tography tiles were then used to develop a predictive
relationship between the fractional ISA and the com-
bined fraction of turf and tree surface, given that turf
was present under the trees observed in the samples.
For this model, only samples over areas with more than
Biogeochemical Cycling of Turf Grasses in the US 427

10% fractional ISA were used, leaving out the sparsely
developed urban fringes, where the occurrence of very
low development density is often associated with for-
ested and other nonturf vegetated surfaces. The pre-
dictive model showed a moderately strong (R
2
= 0.69),
highly significant (P < .0001, RMSE = 11.2) relation-
ship between fractional ISA and fractional turf grass
area (Figure 4) and was subsequently applied to the
conterminous United States to produce a 1-km grid of
fractional turf area (Figure 5).
Modeling of Turf Grasses Growth
Management of turf grasses is highly variable, in
part because of the different uses for which these sur-
faces are dedicated. In order to withstand considerable
wear, golf courses and athletic fields usually receive
much higher doses of nitrogen (N) than residential
lawns (up to 490 kg/ha/yr; Sartain 1998). For resi-
dential lawns, the recommended rates range between
98 and 195 kg/ha/yr (Schultz 1999) and are lower
when the clippings are left to decompose on the turf
Figure 2. Detail of aerial photography used
to measure fractional impervious surface area
over Chicago (residential area).
Figure 1. Detail of aerial photography used to
measure fractional Impervious Surface Area
over Chicago (infrastructure and commercial
buildings).
428 C. Milesi and others

surface rather than composted or bagged and sent to
the landfill. Many residential lawns are managed by
homeowners who pay little attention to the amount of
resources invested for lawn maintenance and often
receive excess water and fertilizer. On the other hand,
there also are some areas cultivated with turf grasses
that are not adequately watered and fertilized, spend-
ing part of the growing season in a dormant stage.
In this study, the simulation of the impact of dif-
ferent turf grass management practices on the conti-
nental C and water budget was based on the
simplifying assumption that, under a given scenario,
the entire turf surface is managed homogeneously,
such as irrigated with the same criteria, fertilized with
the same amount of N, and mowed at the same
height, whether it would be part of a residential lawn
or a golf course. Although this assumption largely
simplifies reality, it allows developing a first estimate
of the potential national impacts of turf grasses on
ecosystem functioning by asking: how would the
continental C and water budgets be affected if all the
surface currently under turf was to be managed like a
well-maintained lawn?
We adapted the Biome-BGC ecosystems process
model to predict C and water fluxes of turf ecosystems
at 865 sites distributed across the United States, cor-
responding to populated places that, according to the
2000 U.S. Census, had a population of at least 40,000
people (the list of populated places is available online
at: http://www.census.gov/geo/www/gazetteer/pla-
ces2k.html). Biome-BGC has been extensively docu-
mented and validated (Thornton and others 2002,
White and others 2000, Kimball and others 1997, Hunt
and others 1996, Running 1994, Running and Hunt
1993, Running and Gower 1991, Running and
Coughlan 1988). Biome-BGC uses prescribed site con-
ditions, meteorology, and parameter values to simulate
daily fluxes and states of C, water, and N for coarsely
defined biomes, at areas ranging from 1 m
2
to the
entire globe. Biome-BGC can be used to simulate these
fluxes for more specifically defined ecosystems when
appropriately parameterized. Adapting Biome-BGC for
simulating the ecosystem processes of turf grasses re-
quired modifying the default parameterization for C3
(cool season) and C4 (warm season) grasses to reflect
the higher specific leaf area as well as the lower C:N
ratio of leaf, litter, and fine roots of fertilized and wa-
tered turf grasses. Leaf C:N ratio was assigned to 20 and
litter C:N ratio was assigned to 40, as suggested by
Bandaranayake and others (2003). We also modified
the lignin, cellulose, and labile portions of fine roots
to, respectively, 12%,52%, and 36% (Bandaranayake
and others 2003), while the canopy average specific
leaf area (SLA) was increased to the upper range of
SLA values observed for grasses (White and others
2000) and set to 70 m
2
/kg C.
Mowing activities were simulated as mortality pro-
cesses that would take place every time the leaf area
index (LAI) reached a critical value of 1.5. The mor-
tality event was assumed to remove 20% of LAI and the
corresponding amount of fine roots. Removal of the
clippings was simulated by removing the portion of C
and N associated with the cut leaves from the ecosys-
Figure 3. Scatter diagram of the observed values of frac-
tional Impervious Surface Area (ISA) versus the values pre-
dicted from linear regression and equation for the predictive
regression model used to estimate the 1-km grid of fractional
ISA for the conterminous United States.
Figure 4. Scatter diagram of the direct measurements of
fractional turf grass area and fractional Impervious Surface
Area (ISA) and equation for the predictive regression model
used to estimate the total U.S. surface under turf from the 1-
km grid of fractional ISA.
Biogeochemical Cycling of Turf Grasses in the US 429

tem process. In the cycling scenario, the C and N
associated with the cut leaves were left on the site to
decompose as litter.
N was added to the system at a constant rate,
simulating a slow-release fertilizer. To evaluate the
effect of clipping cycling on grasses N availability, N
was applied at two different rates in contrasting sim-
ulation runs. Clippings were either removed or cycled
in scenarios simulating an application of 146 kg N/
ha/yr and cycled in scenarios with an application of
73 kg N/ha/yr.
Irrigation during the growing season was simulated
by adding water to the precipitation field in the climate
parameterization. We assumed that the sprinkling
season of a certain location would start when the
minimum temperatures remained above 5C for 7
consecutive days in the spring, and end when mini-
mum temperatures decreased below 5C for 7 consec-
utive days in the fall. Although the start and end of the
sprinkling season are generally determined arbitrarily
and may incorporate other climatic factors, we found
the chosen temperature threshold to represent an
acceptable approximation of the growing season and,
consequently, of the evapotranspirational season of
turf grasses. The simulations assumed water to be
sprinkled following two different watering manage-
ment types. In one type of watering management, we
followed the common recommendation that during
the growing season, turf grasses require about 2.54 cm
(1 inch) of water per week (Schultz 1999). In the
simulations, in the case of rainfall, rain made up for
part of this amount. In the real world, it is common
that sprinklers, especially if automated, run also on
rainy days. The alternative watering management sce-
nario, rather than providing a fixed weekly amount of
water, modulated the irrigation based on the potential
evapotranspiration (PET) and precipitation, the for-
mer calculated according to Priestly and Taylor (1972).
In this case, irrigation was simulated to be triggered
when the PET minus precipitation, accumulated since
the last watering event, exceeded 60% of the added
water. Irrigation then replaced 20% of the PET,
bringing the water availability to nearly 80% of PET.
The effect of the two different water management
practices on the C and water balance was evaluated
comparing scenarios in which N added through fertil-
ization was constant and irrigation was either fixed at
2.54 cm of water/week or modulated according to
PET.
For the 865 selected populated places, soil texture
information was extracted from the STATSGO data-
base (Miller and White 1998) and 18 years of climate
data were obtained from the Daymet dataset (Thorn-
ton and others 1997). The simulation sites were as-
sumed to grow C3 (cool season) or C4 (warm season)
turf grasses based on adaptation zones (Beard 1973,
Time-Life Books 2000). In cities located in the C3-C4
transitional region, the grasses were assumed to be a
mixture of both photosynthetic models and the simu-
lation was run twice, once for each type of grass. The
resulting C and water fluxes in the transitional region
were determined to be an average of the two runs,
assuming that half of the surface was growing C3
grasses and half C4 grasses. All the C and water fluxes
are reported as an average of the 18-year model simu-
lation results.
Figure 5. Distribution of the
fractional turf grass area in the
conterminous United States.
430 C. Milesi and others

The growth of turf grasses at the 865 sites was sim-
ulated for the following five different scenarios:
 Control: Turf grasses growth was simulated with no
management (no irrigation and no N fertilization)
except for cycling of the clippings;
 Removed-146N: The grass was irrigated during the
growing season so that a total of 2.54 cm of water per
week, rainfall included, was provided, fertilized with
146 kg N/ha/yr, and the clippings were removed
from the system after each mowing event;
 Cycled-146N: same as Removed-146N, except for the
clippings, which were left on the site after each
mowing event;
 Cycled-73N: same as Cycled-146N, except for the
amount of fertilizer, which was halved to 73 kg N/
ha/yr;
 Cycled-73N-PET: same as Cycled-73N, except for the
irrigation management, which was calculated based
on Priestly-Taylor PET.
Mann-Whitney U-test for differences was used to
evaluate whether model results under the five scenarios
differed significantly from each other.
The net accumulation of C in the ecosystem was
estimated by calculating the Net Ecosystem Exchange
(NEE), where NEE = Net Primary Productivity (NPP) –
heterotrophic respiration –fluxes of C out of the
ecosystem. Fluxes of C out of the ecosystem refer here
to the C removed with the clippings.
The simulation results were extrapolated to the
continental surface, assuming that turf areas in the
vicinity of a simulation site displayed similar C and wa-
ter fluxes. The continental United States was divided
into Thiessen polygons centered on the simulation sites
to identify individual ‘‘regions of influence’’ around
each of the 865 simulation localities. The output results
at each simulation site were then multiplied by the total
turf area estimated within the respective polygon.
Model Validation
There are only a few studies on the effect of turf
grass management on the C budget. The adaptation of
Biome-BGC to simulate the growth of turf grasses was
validated by comparing the simulated clipping yield
with published clipping yield data (C was assumed to
represent 48% of the dry yield). Two studies (Kopp and
Guillard 2002, Heckman and others 2000) presented
clipping yields under different N fertilization rates for
C3 grasses. Kopp and Guillard (2002) present yields
both for removed clippings and for recycled clippings.
Only one value of clipping yield was available for C4
grasses (Harivandi and others 1996). The measured
versus modeled yield data showed a strong and highly
significant correlation (r = 0.83, P < 0.0001) (Fig-
ure 6).
Results and Discussion
Estimation of Turf Grass Area
The total turf grass area estimated in this study
summed up to 163,800 km
2
(± 35,850 km
2
for the
Figure 6. Scatter diagram of the modeled versus observed
grass clipping yields expressed in C biomass (kg/m
2
/yr). 1*–
Observed and modeled data at Santa Clara, CA, clippings
removed, N rate 146.5 kg/ha/yr (Harivandi and others
1996); 2–Observed at Spring Manor Farm (SM), Storrs, CT,
clippings removed (bagged), N rate 0 kg/ha/yr, modeled at
Hartford, CT (Kopp and Guillard 2002); 3–Observed at SM,
Storrs, CT, clippings removed, N rate 98 kg/ha/yr, modeled
at Hartford, CT (Kopp and Guillard 2002); 4–Observed at
SM, Storrs, CT, clippings removed, N rate 146.5 kg/ha/yr,
modeled at Hartford, CT (Kopp and Guillard 2002); 5–Ob-
served at SM, Storrs, CT, clippings removed, N rate 196 kg/
ha/yr, modeled at Hartford, CT (Kopp and Guillard 2002);
6–Observed at SM, Storrs, CT, clippings removed, N rate 392
kg/ha/yr, modeled at Hartford, CT (Kopp and Guillard
2002); 7–Observed at SM, Storrs, CT, clippings cycled, N rate
0 kg/ha/yr, modeled at Hartford, CT (Kopp and Guillard
2002); 8–Observed at SM, Storrs, CT, clippings cycled, N rate
98 kg/ha/yr, modeled at Hartford, CT (Kopp and Guillard
2002); 9–Observed at SM, Storrs, CT, clippings cycled, N rate
196 kg/ha/yr, modeled at Hartford, CT (Kopp and Guillard
2002); 10–Observed at SM, Storrs, CT, clippings cycled, N rate
392 kg/ha/yr, modeled at Hartford, CT (Kopp and Guillard
2002); 11–Observed at Rutgers, NJ, clippings removed, N rate
97.6 kg/ha/yr, modeled at Edison, NJ (Heckman and others
2000); 12–Observed at Rutgers, NJ, clippings removed, N rate
195.2 kg/ha/yr, modeled at Edison, NJ (Heckman and oth-
ers, 2000). *Only C4 grass site. Points 2–12 refer to sites
growing C3 grasses.
Biogeochemical Cycling of Turf Grasses in the US 431

upper and lower 95% confidence interval bounds)
(Table 1). This estimate, intended to include all resi-
dential, commercial, and institutional lawns, parks, golf
courses, and athletic fields, accounts for approximately
1.9% of the total continental U.S. area, which com-
pares with 3.5–4.9% of the total surface estimated to be
devoted to urban development (Nowak and others
2001, National Association of Realtors 2001). Although
it is difficult to validate the estimate of total turf grass
area derived from this analysis, it reasonably compares
to the estimates of the other studies, in particular when
considering the recent growth in population and ur-
ban areas in the United States (Fulton and others
2001). DPRA, Incorporated (1992), assuming turf sur-
face to be directly related to the population, estimated
a total surface of 188,180 km
2
, among which 94,090
km
2
were of home lawns (Grounds Maintenance 1996).
A 1987 study by Roberts and Roberts (1987) estimated
a total surface of 100,000–120,000 km
2
. Another study,
focusing only on residential lawns, analyzing state-
based average lot sizes of single family homes, esti-
mated a total home lawn area ranging between 58,000
km
2
and 71,680 km
2
, considerably downsizing DPRA’s
estimate of home lawns (Vinlove and Torla 1995). One
of the earliest estimates of total turf surface dates back
to the late 1960s, when it was reported that 67,000 km
2
of lawn existed nationally (Falk 1976).
Even when the estimate of total surface is consid-
ered to be closer to the lower bound of the 95% con-
fidence interval (128,000 km
2
), it appears that turf
grasses would represent the single largest irrigated
‘‘crop’’ in the United States, occupying a total area
three times larger than the surface of irrigated corn
(43,000 km
2
according to the 1997 Census of Agricul-
ture, out of 202,000 km
2
of total irrigated cropland
area).
Water Budget
The two alternate irrigation methods produced
watering requirements that varied widely across the
climatic regions of the 48 states, with the yearly total
amount of water that needed to be provided through
irrigation at each site depending both on the total
rainfall and its distribution during the growing season
and the length of the sprinkling season. In general, a
fixed irrigation management based on turf require-
ments of 2.54 cm of water per week, including rain-
fall, resulted in a minimum of no irrigation in
Lincoln Park, Michigan (meaning that here rainfall
alone is able to satisfy the watering requirements of
the turf throughout the growing season) to a maxi-
mum of 125 cm of water per year to be added
through irrigation in Yuma, Arizona. In contrast, the
irrigation management based on PET tended to de-
crease the amount of water supplied through irriga-
tion in wet regions and increase it in arid and
semiarid regions of the United States, where it was by
far larger than 2.54 cm/week. Modulating irrigation
according to PET required a minimum of 17 cm/yr of
water to be added through irrigation in Pensacola,
Florida, to a maximum of 197 cm/yr in Yuma, Ari-
zona. The Mann-Whitney U-test for differences indi-
cated that the two irrigation methods would provide
significantly different annual amounts of water at 77%
of the 865 sites. The sites with no significant differ-
ence between the two irrigation methods were all but
three located just east of the Great Plains. The spa-
tially interpolated differences in irrigation water use
between the two irrigation managements (Figure 7)
indicates that adopting the PET-based method versus
applying constantly 2.54 cm/week would result in a
larger amount of water sprinkled in the West, with a
maximum difference of up to 72 cm/yr in the
Southwest, and less water in the southeastern United
States, with a reduction in water use of up to 38 cm/
yr in southern Florida, where the high relative
humidity reduces the evapotranspirational demand.
Extrapolating the water use for irrigation with the
two methods at each of the 865 sites to the surface of
turf grasses contained in the respective Thiessen poly-
gons yields an average total of 73,560 Mm
3
(Mega cubic
meters) of water with the constant 2.54 cm/week
method and 95,100 Mm
3
of water with the PET meth-
od, while rain contribution during the sprinkling sea-
son to the watering of the total estimated turf grass
area would amount to 99,130 Mm
3
(Figure 8).
These estimates indicate that, in the scenario that
the entire turf surface in the United States was to be
irrigated to satisfy the 2.54 cm/week water supply or at
80% of PET, domestic and commercial consumptive
water use would be, respectively, 695 to 900 liters of
water per person per day. Noteworthy is that in spite of
the elevated irrigation requirements, there appears to
be a considerable amount of water leaving the soil layer
as outflow (water in excess of field capacity) rather
than evapotranspiration (56,620 to 57,670 Mm
3
of wa-
ter, depending on the irrigation management scenar-
ios). Ninety percent of the estimated outflow takes
place in the eastern and southern United States, where
it is related to rainfall rather than sprinkling events. On
occasions of abundant rainfall, precipitation is larger
than the soil water-holding capacity and leaves the soil
before the grass can use it for evapotranspiration. In
spite of a surplus of available water during the rainy
periods, sprinkling is still required during the drier
periods.
432 C. Milesi and others

If irrigation could just replace actual evapotran-
spirational losses, the water to be added through
sprinkling would amount to 11,070 Mm
3
in the case
of the 2.54 cm/week method and 33,300 Mm
3
with
the PET-based method. The large increase in water
requirements with the PET-based method has to be
attributed to the arid western United States, where
grasses can evaporate much more than 2.54 cm of
water per week if more irrigation is supplied. Still,
part of the water reaching the surface during the
growing season, either from precipitation when
abundant rainfall occurs, or from the sprinkler, due
to Priestly-Taylor PET overestimating actual evapo-
transpiration, would not be used by the grass and
would leave the soil layer as outflow.
Carbon Budget
Table 2 reports the ranges in C fluxes and mowing
counts for the control and the four management
scenarios. In general, the simulation results indicate
that the C fluxes of a well-watered grass increase with
the amount of available N. For a certain amount of N
input through fertilization, the C fluxes were larger
when cycling of the grass clippings was simulated,
since the onsite decomposition of the mowed grass
clippings returned a consistent amount of N to the
soil. For each scenario, differences in the maximum–
minimum ranges are related mainly to the growing
season length.
Unsurprisingly, the control scenario displays the
lowest range of C fluxes, and the lowest range of
mowing counts, which are both significantly different
from all the other scenarios at all 865 sites. The low
number of mowing counts simulated in the control
scenario let us infer the obvious: in the absence of
irrigation and fertilization, turf grasses would not be
able to grow in most of the United States. A general
guess of where turf grasses could grow with no added
resources of N and water can be inferred by calculating
the number of days between successive mowings (ratio
of yearly mowing counts to growing season length).
Assuming that turf grasses should grow back to an LAI
of 1.5 in at least 30–35 days in order not to be out-
competed by weeds, Figure 9 shows that monocultures
of turf could probably grow without managed inputs of
water and fertilizer only in a few of the modeling sites,
all but one located in the northeastern portion of the
country (the site in the western United States corre-
sponds to Flagstaff, AZ). If turf grasses reach an LAI of
1.5 only six to seven times in areas where the growing
season is as long as 300–360 days, then it is probable
that between subsequent cuts there are several
opportunities for nonturf species to invade the surface
and prosper over time. Because the LAI is reduced by
20% every time the LAI would reach the value of 1.5,
the NPP of an unmanaged turf would be considerably
lower than that of natural grasslands, which in tem-
Table 1. Estimates of turf grass area by state
Turf grass area (km
2
)
State Mean
Upper
95% C.I.
Lower
95% C.I.
Alabama 3130 3741 2520
Arizona 2559 3178 1941
Arkansas 2098 2519 1679
California 11159 13890 8434
Colorado 2478 3047 1910
Connecticut 2429 2946 1913
Delaware 533 644 422
District of Columbia 57 86 28
Florida 11570 14221 8925
Georgia 5688 6848 4530
Idaho 942 1133 751
Illinois 5729 7102 4359
Indiana 3843 4679 3008
Iowa 2227 2772 1822
Kansas 2004 2453 1555
Kentucky 2446 2935 1958
Louisiana 3377 4099 2656
Maine 975 1157 793
Maryland 2471 3013 1929
Massachusetts 4183 5054 3314
Michigan 4538 5598 3480
Minnesota 3176 3866 2487
Mississippi 1969 2362 1578
Missouri 3442 4217 2669
Montana 735 884 585
Nebraska 1149 1401 898
Nevada 928 1162 694
New Hampshire 1126 1339 913
New Jersey 3942 4885 3002
New Mexico 1545 1860 1231
New York 6320 7770 4873
North Carolina 8112 9715 6512
North Dakota 572 693 452
Ohio 6733 8213 5257
Oklahoma 2689 3294 2086
Oregon 1977 2406 1549
Pennsylvania 7293 8789 5799
Rhode Island 506 622 390
South Carolina 4034 4822 3248
South Dakota 692 829 555
Tennessee 4201 5064 3339
Texas 13187 16242 10138
Utah 1207 1493 922
Vermont 524 621 427
Virginia 4544 5510 3581
Washington 3579 4345 2814
West Virginia 1459 1731 1189
Wisconsin 3110 3764 2457
Wyoming 554 665 444
Total U.S. 163,812 199,679 128,016
Biogeochemical Cycling of Turf Grasses in the US 433

perate ecosystems ranges between 320 and 750 g/m
2
/
yr (Saugier and others 2001, Schlesinger 1997).
The largest C fluxes are realized for scenario Cycled-
146N, in which Mann-Whitney U-test for differences in
C fluxes and mowing count indicates that this scenario
is significantly different from the other scenarios for all
the variables measured at the 865 sites (P < 0.01).
Abundant fertilization (146 kg N/ha/yr) and the re-
cycling of the N contained in the leaves left to de-
compose on the site boosts both the productivity as
well as the heterotrophic respiration. As a consequence
of the higher productivity of turf grasses in this sce-
nario, there is also an increase in mowing frequency,
which can reach 98 cuts per year in those sites where
climatic conditions favor year-round growing season,
resulting in about two cuts per week.
Scenario Removed-146N produces the second
highest NPP and clipped biomass ranges. Because the
clipped biomass is assumed to be removed from the
turf surface, very low on-site decomposition activity
results in the smallest C fluxes from heterotrophic
respiration.
In scenario Cycled-73N, the C fluxes are signifi-
cantly lower (P < 0.01) at all sites when compared to
those of Removed-146N and lower in 88–92% of the
sites when compared to Cycled-146N.
Scenario Cycled-73N-PET, which differs from Cy-
cled-73N only for the type of water management,
modulating irrigation according to PET rather than
supplying a weekly fixed amount of water, does not
produce significantly different C fluxes from Cycled-
73N at any of the 865 sites. Mann-Whitney U-test at 5%
Figure 7. Spatially interpolated differences in
irrigation water use between the two irrigation
methods.
Figure 8. Water budgets of the total
U.S. turf surface for the four
management scenarios. Error bars
indicate budget values calculated for
the 95% confidence interval lower and
upper bound estimate of total turf
surface.
434 C. Milesi and others

significance level indicates a water effect on NPP and
clipped C at 7% of the sites, and an effect on hetero-
trophic respiration at 5% of the sites.
Large differences in total C fluxes can be realized
under the same irrigation management of 2.54 cm of
water per week, all resulting in very large losses of water
through outflow (Figure 8). This result is most proba-
bly explained by the fact that in all the simulated
management scenarios water is not limiting growth,
which responds rather to increases in N availability.
The large increase in water application observed when
modulating irrigation according to PET, on the other
hand, results in an insignificant change in C fluxes,
indicating that the water is lost in luxury evapotrans-
piration.
The estimation of the total C budget for the conti-
nental U.S. turf surface under the five scenarios
examined (Figure 10) indicates that the highest NEE is
recorded for the Cycled-146N scenario, amounting to
16.7 Tg C/year. The lowest NEE is recorded for the
Removed-146N scenario, for which the removal of the
clippings from onsite decomposition reduces the C
sink to just 5.9 Tg C/yr, in spite of the fact that the
same amount of N as in Cycled-146N is added through
fertilization (a total of 2.39 Tg N/yr for the total esti-
mated surface under turf). Offsite composting of the
clippings allows recuperating part of the C. On the
other hand, the practice, nowadays less common, of
sending the clippings in trash bags to the landfill leads,
through anaerobic decomposition, to the production
of methane, a powerful greenhouse gas. Reducing the
N fertilization by half in scenarios that recycle the
clippings (Cycled-73N and Cycled-73N-PET) lowers the
NPP by 36–37% and the NEE by 45% compared to the
Cycled-146N scenario but also considerably lowers the
number of times the grass needs to be cut throughout
the growing season. The C budget for the control
scenario, on the other hand, results in a small source of
carbon, since the total heterotrophic respiration (15.5
Tg C/yr) slightly surpasses the total NPP (15.3 Tg C/
yr), bringing the NEE for the control scenario to –0.2
Tg C/yr.
Therefore, if the entire area was well watered and
fertilized, we would have a positive NEE, even when
assuming the bagging and removal of the grass clip-
pings. A positive NEE means that more carbon is
Table 2. Minimum and maximum values of the modeled C fluxes and mowing counts
Carbon fluxes (g C m
-2
yr
)1
) Control Removed-146N Cycled-146N Cycled-73N Cycled-73N-PET
NPP 22–121 257–641 281–1063 184–604 195–613
Clippings 0–34 79–207 87–348 55–195 58–198
Heterotrophic respiration 31–121 138–392 210–922 140–533 150–542
Mowing counts (cuts yr
)1
)0–716–52 22–98 14–55 16–56
Figure 9. Modeling sites where
growth of turf grasses appears to be
possible without irrigation or
fertilization.
Biogeochemical Cycling of Turf Grasses in the US 435

accumulated in the turf grass system than is released
through respiration processes and, eventually, re-
moved with the clippings, indicating that the soils of
lawns and golf courses that are left undisturbed for a
few decades have the potential to sequester a consistent
amount of C in their soils. The sink is generally
stronger when more N is available. N availability can be
increased both through increased fertilization or, more
efficiently, by leaving the clippings to decompose on
the site after mowing. Merely applying a larger amount
of N through the use of synthetic fertilizer without
recycling the clippings would reduce the potential gain
in C sequestration because of increased discounts due
to the C costs of manufacturing, transporting, and
commercializing the fertilizer (Schlesinger 1999).
These C costs are in addition to the ones deriving from
the operation of lawn mower equipment, distributing
water for irrigation, and from transporting and
decomposing the clippings in the landfill.
In addition to the associated C costs, the current
high-input choices made by consumers and profes-
sional turf managers for maintaining monocultures of
turf grasses typical of many lawns and play fields comes
at the risk, not analyzed here, of watershed pollution
due to improper fertilization and use of pesticides
(Petrovic 1990). The input levels of herbicide and
insecticide per unit area of turf are often several times
larger than in their agricultural counterparts, inde-
pendently from the consumer’s knowledge and
understanding of personal health risks and negative
environmental effects associated with the use of these
products (Robbins and Sharp 2003a, 2003b).
Beneficial effects of turf grasses, such as a carbon
sequestration but also recreation, storm runoff reduc-
tion due to increased soil infiltration in occasion of
intense rainfall, and removal of impurities and chem-
icals during percolation of the water through the root
zone, could be sought by minimizing the application of
fertilizers and pesticides, introduction of lower input
species mixes such as clover and other so-called weeds
(Bormann and others 1993), on-site decomposition of
the grass clippings, and extending the practice of irri-
gating with waste water rather than with drinking wa-
ter.
Conclusions
In this study, we mapped the total surface of turf
grasses in the continental United States and simulated
its potential C and water budgets. We also provided a
description of how the C and water budgets can be
affected by adopting different management practices
for irrigation, fertilization, and the fate of the clip-
pings. Rather than trying to accurately quantify the
existing fluxes, we simulated scenarios in which the
entire surface was to be managed like a well-main-
tained lawn, a thick green carpet of turf grasses, wa-
tered, fertilized, and kept regularly mown. The
accuracy of the results is therefore limited by both the
uncertainty in the mapping of the total lawn area and
by the simplifying assumptions made in the modeling
of the growth of turf grasses.
The analysis indicates that turf grasses, occupying
1.9% of the surface of the continental United States,
would be the single largest irrigated crop in the
country. The scenarios described in this study also
indicate that a well-maintained lawn is a C sequestering
system, although the positive C balance has to be dis-
counted for a very large use of water and N and, not
quantified in this study, pesticides. The model simula-
Figure 10. Carbon budgets of the
total U.S. turf surface for the different
simulations. Error bars indicate budget
values calculated for the 95%
confidence interval lower and upper
bound estimate of total turf surface.
The carbon budget for the control
scenario is negligible and therefore not
displayed. NPP, Net Primary
Productivity; NEE, Net Ecosystem
Exchange.
436 C. Milesi and others

tions have assumed a conservative amount of fertiliza-
tion (a maximum of 146 kg N/ha/yr). In general, the
rates of N applications are similar to those used for row
crops, and N losses from turf surfaces can contribute to
non-point source pollution when fertilization takes
place improperly.
If the entire turf surface was well watered following
commonly recommended schedules, there would be
very large pressure on U.S. water resources, especially
when considering that drinking water is usually sprin-
kled. At the time of this writing, in most regions out-
door water use already reaches 50–75% of the total
residential use. Because of demographic growth and
because more and more people are moving towards
the warmer regions of the country, the potential exists
for the amount of water used for turf grasses to in-
crease. Several counties in the arid and semiarid re-
gions of the United States have already implemented
lawn watering restrictions, the recycling of wastewater
to replace drinking water for outdoor sprinkling, and
incentives to increase the use of xeriscaping. Although
turf grasses also provide important ecological benefits
such as slower storm runoff, improved water infiltra-
tion, and holding soil in place on sloping terrains, to
protect our water resources as further urban growth
takes place other regions will probably need to extend
the practice of recycling wastewater for outdoor use
while continuing to educate the population on the
value of water resources.
Acknowledgments
This study was supported by the NASA Earth System
Science Fellowship program to the first author and by
the NASA Land Cover Land Use Change research
program. We are grateful to Ronald Follett, Paul Rob-
bins, and Michael White, whose constructive comments
improved the quality of the manuscript. Many thanks
also to Faith Ann Heinsch, Carol Brewer, Eric Edlund,
Sarah Halvorson, David Jackson, and Stephen Siebert
at the University of Montana for interesting discussions
and insightful comments.
Literature Cited
Bandaranayake, W., Y. L. Qian, W. J. Parton, D. S. Ojima, and
R. F. Follett. 2003. Estimation of soil organic carbon
changes in turfgrass systems using the CENTURY model.
Agronomy J 95:558–563.
Beard, J. B. 1973. Turfgrass: science and culture. Prentice
Hall, Englewood Cliffs, New Jersey 658 pp.
Bormann, F. H., D. Balmori, and G. T. Geballe. 1993. Rede-
signing the American lawn: a search for environmental
harmony. Yale University Press, New Haven and London
148 pp.
DPRA, Incorporated. 1992. Benefit analysis of insecticide us
on turf: preliminary biological and economic profile re-
port. Unpublished Interim Report, Manhattan, Kansas.
Elvidge, C. D., K. E. Baugh, J. B. Dietz, T. Bland, P. C. Sutton,
and H. W. Kroehl. 1999. Radiance calibration of DMSP-
OLS low light imaging data of human settlements. Remote
Sensing of Environment 68:77–88.
Elvidge, C. D., C. Milesi, J. B. Dietz, B. J. Tuttle, P. C. Sutton,
R. Nemani, and J. E. Vogelmann. 2004. U.S. constructed
area approaches the size of Ohio. Eos, Transactions.
American Geophysical Union 85:233 .
Falk, J. H. 1976. Energetics of a suburban lawn ecosystem.
Ecology 57:141–150.
Fulton W., R. Pendall, M. Nguyen, and A. Harrison. 2001.
Who sprawls the most? How growth patterns differ across
the U.S. The Brookings Institution Survey Series. 24 pp.
Grounds Maintenance. 1996. Turf acreage. Grounds Main-
tenance 31:10.
Harivandi, M. A., W. B. Hagan, and C. L. Elmore. 1996. The
use of recycling mowers in grasscycling. California Turfgrass
Culture 46:4–6.
Heckman, J. R., H. Liu, W. Hill, M. DeMilia, and W. L.
Anastasia. 2000. Kentucky bluegrass responses to mowing
practices and nitrogen fertility management. Journal of
Sustainable Agriculture 15:25–33.
Hunt, E. R. J., S. C. Piper, R. Nemani, C. D. Keeling, R. D.
Otto, and S. W. Running. 1996. Global net carbon ex-
change and intra-annual atmospheric CO2 concentrations
predicted by an ecosystem process model and three-
dimensional atmospheric transport model. Global Biogeo-
chemical Cycles 10:431–456.
Jenkins, V. S. 1994. The lawn: a history of an American
obsession. Smithsonian Institution Press, Washington, DC
246 pp.
Kimball, J. S., M. A. White, and S. W. Running. 1997. BIOME-
BGC simulations of stand hydrologic processes for
BOREAS. Journal of Geophysical Research 102(D24):29043–
29051.
Kopp, K. L., and K. Guillard. 2002. Clipping management
and nitrogen fertilization of turfgrass: growth, nitrogen
utilization, and quality. Crop Science 42:1225–1231.
Mayer P. W., W. B. DeOreo, E. M. Opitz, J. C. Kiefer, W. Y. Davis,
B. Dziegielewski, and J. O. Nelson. 1999. Residential end uses
of water. American Water Works Association. 310 pp.
Miller, D. A., and R. A. White. 1998. A conterminous United
States multi-layer soil characteristics data set for regional
climate and hydrology modeling. Earth Interactions 2.
National Association of Realtors. 2001. Land use and land loss
in the United States: the impact of land use trends on real
estate development. The Research Division of the National
Association of Realtors. 9 pp.
Nowak, D. J., M. H. Noble, S. M. Sisinni, and J. F. Dwyer. 2001.
People & trees: assessing the US urban forest resource.
Journal of Forestry 99:37–42.
Biogeochemical Cycling of Turf Grasses in the US 437

Petrovic, A. 1990. The fate of nitrogenous fertilizers applied
to turfgrass. Journal of Environmental Quality 19:1–14.
Priestley, C. H. B., and R. J. Taylor. 1972. On the assessment
of surface heat flux and evaporation using large-scale
parameters. Monthly Weather Review 100:81–92.
Qian, Y., and R. F. Follett. 2002. Assessing soil carbon
sequestration in turfgrass systems using long-term soil
testing data. Agronomy Journal 94:930–935.
Roberts, E. C., and B. C. Roberts. 1987. Lawn and sports turf
benefits. The Lawn Institute, Pleasant Hill, Tennessee, 3 pp.
Robbins, P., and T. Birkenholtz. 2003. Turfgrass revolution:
measuring the expansion of the American lawn. Land Use
Policy 20:181–194.
Robbins, P., A. Polderman, and T. Birkenholtz. 2001. Lawns
and toxins: an ecology of the city. Cities 18:369–380.
Robbins, P., and J. Sharp. 2003a. The lawn-chemical economy
and its discounts. Antipode 35:955–979.
Robbins, P., and J. Sharp. 2003b. Producing and consuming
chemicals: the moral economy of the American lawn. Eco-
nomic Geography 79:425–451.
Running, S. W. 1994. Testing forest-BGC ecosystem process
simulations across a climatic gradient in Oregon. Ecological
Applications 4:238–247.
Running, S. W., and J. C. Coughlan. 1988. A general model of
forest ecosystem processes for regional applications.
I. Hydrological balance, canopy gas exchange and primary
production processes. Ecological Modeling 42:125–154.
Running, S. W., and S. T. Gower. 1991. FOREST BGC, A
general model of forest ecosystem processes for regional
applications. II. Dynamic carbon allocation and nitrogen
budgets. Tree Physiology 9:147–160.
Running, S. W., and R. E. J. Hunt. 1993. Generalization of a
forest ecosystem process model for other biomes, BIOME-
BGC, and an application for global-scale models. Pages
141–157 in J. R. Ehleringer, C. B. Field (eds.), Scaling
physiological processes: leaf to globe. Academic Press, Inc,
San Diego, California.
Sartain, J. B. 1998. Fertilize bermudagrass greens smartly and
safely. Grounds Maintenance, September 1, 1998. Available
Online at http://grounds-mag.com/mag/grounds_main-
tenance_fertilize_bermudagrass_greens/.
Saugier, B., J. Roy, and H. A. Mooney. 2001. Estimations of
global terrestrial productivity: converging toward a single
number?. Pages 543–557 in J. Roy, B. Saugier, H. A. Moo-
ney (eds.), Terrestrial global productivity. Academic Press,
San Diego, California.
Schlesinger, W. H. 1997. Biogeochemistry: an analysis of
global change. Academic Press, San Diego, California 678
pp.
Schlesinger, W. H. 1999. Carbon and agriculture: carbon
sequestration in soils. Science 284:2095 .
Schultz, W. 1999. A man’s turf: the perfect lawn. Clarkson N.
Potter, New York 180 pp.
Spronken-Smith, R. A., T. R. Oke, and W. P. Lowry. 2000.
Advection and the surface energy balance across an irri-
gated urban park. International Journal of Climatology
20:1033–1047.
Thornton, P. E., S. W. Running, and M. A. White. 1997.
Generating surfaces of daily meteorological variables over
large regions of complex terrain. Journal of Hydrology
190:214–251.
Thornton, P. E., B. E. Law, H. L. Gholz, K. L. Clark, E. Falge,
D. S. Ellsworth, A. H. Goldstein, R. K. Monson, D. Hollinger,
M. Falk, J. Chen, and J. P. Sparks. 2002. Modeling and
measuring the effects of disturbance history and climate on
carbon and water budgets in evergreen needleleaf forests.
Agricultural and Forest Meteorology 113:185–222.
Time-Life Books. 2000. Complete garden guide: A compre-
hensive reference for all your garden needs, 992 pp.
U.S. Bureau of the Census. 1999. American Housing Survey for
the United States: 1999. Available at: http://www.
census.gov/hhes/www/housing/ahs/ahs99/ahs99.html
(last update: February 26, 2003).
Van Dersal, W. 1936. The ecology of a lawn. Ecology 17:515–
527.
Vinlove, F. K., and R. F. Torla. 1995. Comparative estimations
of U.S. home lawn area. Journal of Turfgrass Management
1:83–97.
White, M., P. Thornton, S. Running, and R. Nemani. 2000.
Parameterization and sensitivity analysis of the BIOME-
BGC terrestrial ecosystem model: net primary production
controls. Earth Interactions 4:1–85.
438 C. Milesi and others