You are on page 1of 13

Sustainability of Water Quality and Ecology 5 (2015) 6476

Contents lists available at ScienceDirect

Sustainability of Water Quality and Ecology


journal homepage: www.elsevier.com/locate/swaqe

Modeling denitrication in a changing climate


Janet R. Barclay a,b,, M. Todd Walter a,
a
b

Department of Biological and Environmental Engineering, Riley Robb Hall, Cornell University, Ithaca, NY 14853, United States
Department of Natural Resources and the Environment, University of Connecticut, Storrs, CT 06269, United States

a r t i c l e

i n f o

Article history:
Received 3 September 2014
Received in revised form 18 December 2014
Accepted 19 December 2014
Available online 14 January 2015
Keywords:
Denitrication
Climate change
Modeling
Ecosystem services
Nitrogen

a b s t r a c t
Although uncertainty exists in the specics of climate change, sufcient consensus has
emerged regarding the scale and direction of changes that it is important for managers
to consider the implications of these changes for essential ecosystem services such as
denitrication. For the northeastern United States, it is expected that by the end of the
21st century, annual precipitation will increase by 10% and annual temperatures by 3 C.
Because denitrication rates are highly inuenced by soil moisture and soil temperature,
we expect the projected changes in temperature and precipitation to alter the rates and
patterns of denitrication. We developed three future weather scenarios based on the
B1, A2, and A1F1 emission scenarios and utilized them as inputs to a coupled hydrologic-denitrication model to analyze the effects of changing temperature and soil moisture
on denitrication rates, assuming nitrate availability remains unchanged. Our results suggest denitrication rates will increase by 5.111.8 kg N ha 1 yr 1 across the watershed. The
greatest projected increases are in the areas and seasons with the highest baseline rates,
with smaller increases in those with lower baseline rates. We found that changing temperature is likely to be a much stronger driver of change to denitrication rates
(7.7 kg N ha 1 yr 1) than changing precipitation (1.4 kg N ha yr 1). This study utilizes scenario modeling and a eld validated model to quantify the future benets of the ecosystem
service denitrication, addressing a lack of ecosystem service quantication studies and
providing a model for quantication studies.
2015 Elsevier B.V. All rights reserved.

1. Introduction
Much uncertainty exists in several of the specics of how global climate is changing, however some consensus has
emerged. For the northeastern USA, it is expected that by the end of the 21st century annual precipitation will increase
on the order of 10%, with larger increases in the winter and slight decreases in the summer months (Hayhoe et al., 2007).
Annual temperatures for the same period are projected to increase approximately 35 C, with greater increases in the summer months and lesser increases in the winter months (Hayhoe et al., 2007). Although the uncertainty in these estimations is
large, particularly with regard to the precipitation (Hayhoe et al., 2007), the implications of changes of this magnitude are
likely signicant and it is important that we consider the effects of projected changes in precipitation and temperature.
One important biogeochemical ecosystem service that is likely to be affected is denitrication, the microbially mediated
reduction of nitrate (NO3 ) to dinitrogen (N2), nitric oxide (NO), or nitrous oxide (N2O) gas. Denitrication plays an important
Tel.: +1 802 310 2685 (J.R. Barclay). Address: Department of Biological and Environmental Engineering, Riley Robb Hall, Cornell University, Ithaca, NY
14853, United States. Tel./fax: +1 607 255 2488 (M.T. Walter).
E-mail addresses: jrb474@cornell.edu (J.R. Barclay), mtw5@cornell.edu (M.T. Walter).
http://dx.doi.org/10.1016/j.swaqe.2014.12.006
2212-6139/ 2015 Elsevier B.V. All rights reserved.

J.R. Barclay, M.T. Walter / Sustainability of Water Quality and Ecology 5 (2015) 6476

65

role in reducing nitrogen (N) loads to streams, estuaries and coastal waters, and changes to denitrication rates, and therefore stream N loads, will have environmental, economic and human health effects (EPA, 2011).
Denitrication is one of many essential and benecial services provided to humans by our ecosystems (Daily et al., 1997;
Sandhu and Wratten, 2013). As with other ecosystem services, it can be difcult to incorporate the benets of denitrication
into management decisions because those services provide natural, rather than nancial, capital (Nelson and Daily, 2010). An
ecosystem services approach accounts for the services provided by an ecosystem, beginning with a quantication of those
benets. Because we expect the benets of denitrication to change with changes in temperature and precipitation,
exploratory modeling to consider the ability of a landscape to provide denitrication services in a changing climate is essential to sound management (Seppelt et al., 2011; Volk, 2013).
The implications of changing climate for denitrication rates are complex due to the complexity of the processes
involved. At least two mechanisms of rate changes are possible: changes in residence time and changes in reaction
conditions.
Some have suggested that climate change will reduce the residence time of soluble nitrate, thereby reducing the amount
that is denitried and increasing the nitrate load to streams (Boyacioglu et al., 2012; Howarth et al., 2006). With this change
mechanism, nitrate-rich water ows more quickly through the landscape, reducing the time spent in areas conducive to
denitrication and thereby reducing the amount of nitrate that is denitried. Others have noted the potential complications
of disconnecting zones of nitrate loading from zones of denitrication (Butterbach-Bahl and Dannenmann, 2011). This
change mechanism would decrease the residence time to zero, reducing denitrication rates to near zero.
In terms of reaction conditions, soil temperature and degree of saturation have been shown to be important controls on
denitrication rates and changes in these conditions are expected to have a direct effect on denitrication rates (Heinen,
2006; Henault and Germon, 2000; Singh et al., 2010). The increased temperatures of the climate projections would be
expected to have a rate-increasing effect on denitrication. Increased precipitation would be expected to increase the degree
of soil saturation, however, increased evapotranspiration might counteract this effect (Huntington, 2006; Walter et al.,
2004). We acknowledge that many other factors, including carbon and nitrate availability, inuence denitrication rates
and may be affected by climate change (Barnard and Leadley, 2005; Brown et al., 2012; Butterbach-Bahl and
Dannenmann 2011); however, as an initial step and for the purpose of this study we are focusing on the direct effects of
changes in soil moisture and temperature. This is consistent with work by Pinay et al. (2007) identifying soil moisture
and temperature as the most important factors in denitrication rates.
The objective of this study is to explore how the benets provided by the ecosystem service of denitrication may be
affected by changes in temperature and precipitation. Specically, we seek to answer three questions: (1) How might the
magnitude and temporal patterns of the ecosystem service of denitrication change?, (2) What are the relative strengths
of precipitation and temperature as drivers of these changes?, and (3) How might changes in the patterns of precipitation
and temperature affect denitrication rates? Answers to these questions will allow the ecosystem service of denitrication
to be factored into future management plans. Additionally, it will contribute to addressing the lack of ecosystem service
quantication studies and provide a model that subsequent quantication studies can follow.
2. Methods
2.1. Site description
To address these questions, we applied a simple denitrication model to the southern watershed draining the Cornell
University Teaching and Research Farm in Harford, NY, USA and surrounding hillslopes (Fig. 1A). This 1550 ha watershed
consists of highlands to the northeast and southwest and a central valley that drains into the East Branch of Owego Creek.
The elevation ranges from 360 to 613 m. In the valley bottom, the primary soil type is Howard gravelly loam and on the adjacent hillslopes the primary soil types are Langford channery silt loam and Volusia channery silt loam (USDA NRCS). The primary land uses are forest and agriculture (Fig. 1B) and the primary N input to the soil is manure fertilizer. Anderson et al.
(2014a) gives a detailed analysis of N inputs and exports for the farm. Seventy-eight percent of the farm is in the watershed
of this study, accounting for fty-one percent of the total watershed area.
2.2. Model overview
We used the simple denitrication model coupled with a semi-distributed hydrologic model described in Barclay et al.
(2015). Briey, this model contains a semi-distributed lumped hydrologic module based on a ThornthwaiteMather water
balance (Thornthwaite and Mather, 1955) and a simple denitrication module of the type described by Heinen (2006). This
hydrologic model describes upland soil moisture well, which is essential for spatially distributed denitrication models due
to the importance of soil saturation to denitrication rates (Heinen, 2006; Tague, 2009; Whelan and Gandol, 2002). We
selected this denitrication model because of its simple form and easily measured inputs.
The hydrologic model tracks daily soil water in 10 wetness classes from driest (wetness class 10) to wettest (wetness
class 1). Soil water above the available water capacity is routed to groundwater, lateral ow (to the next wettest class) or
to overland ow. The main inputs to the hydrologic model are daily precipitation and air temperature and the primary para-

66

J.R. Barclay, M.T. Walter / Sustainability of Water Quality and Ecology 5 (2015) 6476

Fig. 1. Maps of the study watershed from Barclay et al. (2015) (A) The watersheds location within the Finger Lake region of New York State; (B) The primary
land uses in the catchment; (C) The Soil Topographic Index (STI), binned by quantile into 10 equal area wetness classes; (D) The 30 unique combinations of
wetness class and land use which formed the calculation basis for the denitrication model.

meters concern the available water capacity and ow routing. We mapped the wetness classes onto the watershed using the
Soil Topographic Index (STI) (Lyon et al., 2004), which had been binned into 10 quantiles (Fig. 1C). The highest STI values
correspond to wetness class 1 (the wettest class) and the lowest to wetness class 10. In wetness class 1, the shallow groundwater intercepts the soil water, creating high levels of soil water that makes this class of particular interest for biogeochemical processing. The hydrologic model was calibrated using stream discharge and soil moisture to maximize the
overall NashSutcliffe efciency (NSE) (Nash and Sutcliffe, 1970). The optimization was done using the DEoptim package
in R (Ardia et al., 2012) and resulted in an overall NSE of 0.72. Details of the hydrologic model are given in Barclay et al.
(2015).
The denitrication model is a reduction function in which a potential denitrication rate is modied by scalars related to
carbon and nitrate availability, soil moisture, and soil temperature. This model was parameterized on the study watershed
using in situ pushpull denitrication measurements (Addy et al., 2002; Anderson et al., 2014b) using the DEoptim package
in R (Ardia et al., 2012) and a leave-one-out cross-validation process (Arlot and Celisse, 2010). The calibrated model compared well with the observed values (R2 = 0.77, RMSE = 272 kg N ha 1 yr, 1 NRMSE = 0.77) (Barclay et al., 2015).
We overlaid the 10 wetness classes from the hydrologic model with the three primary land uses (wetlands, cropped agriculture, and forests/pastures/meadows) to create 30 overall classes (Fig. 1D). The coupled model predicts daily denitrication rates for each overall class, based on inputs of Julian Day, nitrate (NO3N), dissolved organic matter (DOC), soil organic
matter (OM), air temperature, and precipitation. OM is based on the wetness class and NO3N and DOC vary according to

67

J.R. Barclay, M.T. Walter / Sustainability of Water Quality and Ecology 5 (2015) 6476

land use and Julian Day. Although agricultural crop models exist to model NO3N on a daily time-step using agricultural
management data (see Ferrant et al. (2011) for a comparison of TNT2 and SWAT, two such models), for simplicity we
assumed that agricultural management and current seasonal patterns of NO3N would remain unchanged. The details of these inputs and the calibration process are given in Barclay et al. (Barclay et al., 2015) Air temperature and precipitation inputs
are from the weather series described below.

2.3. Baseline Climate Series


Following the example of Hayhoe et al. (2007), we used a baseline period of 19611990, however the on-site weather
station at our study site was not established until 2004 so we generated our baseline data series using historic weather data
from the NOAA National Climate Data Center, Station GHCND: USC00304174 (ITHACA CORNELL UNIVERSITY) (NCDC), which
is in Ithaca, NY, 21 km to the west of our study site. A comparison of the concurrent weather data from Ithaca and Harford
(20052012) indicated that Harford is approximately 0.9 C (0.1 C) colder than Ithaca, and annually receives 3 cm (6 cm)
more rain and has 18 (14) more days of rain. Over the eight year comparison period, Harford experienced 3 more rainy
days month 1 in the wetter months of JanuaryApril, and 0.5 more rainy days month 1 in the months of May through
December. We assumed this same relationship held in the baseline period of 19611990. To create a historic baseline for
Harford, NY (19611990) from the historic Ithaca weather, we decreased the daily minimum, maximum, and average temperatures by 0.9 C, increased the daily precipitation depth by 5%, and randomly added 0.5 days of rain month 1 for May
December and 3 days rain month 1 for JanuaryApril.

2.4. Future climate series generation


We used three future emissions scenarios (B1, A2, and A1F1) developed by the Intergovernmental Panel on Climate
Change (IPCC, 2001) as the basis for generating our future climate series. Briey, A1F1 is a scenario of rapid economic growth
with an emphasis on fossil fuels, A2 is a scenario of slower, more fragmented economic growth focused at the regional and
local scale, and the B1 scenario involves a rapid shift in economic structures such that material resource use decreases and
efciency increases (IPCC, 2001).
Projections for annual, summer and winter shifts to mean temperature and precipitation (Table 1) are multi-model
ensemble averages taken from Hayhoe et al. (2007). We modied the daily baseline summer and winter temperature and
precipitation values by the projected shifts to the summer and winter mean values. Then, we adjusted the spring and fall
values until the new annual mean matched the projected annual mean values. This method, known as the delta method,
has the advantage of simplicity and has been shown to be relatively successful at modeling the observed climate. The drawback of the delta method is that it can mis-predict extreme events (Anandhi et al., 2011; Gleick, 1986; Hay et al., 2000;
Hayhoe, 2010).
To address the question regarding the relative strengths of projected changes in temperature and precipitation as drivers
of changes in denitrication rates, we generated two hypothetical climate series. The Precipitation Only and Temperature
Only series were generated by using the precipitation (or temperature) series from the A1F1 scenario and the temperature
(or precipitation) series from the baseline data (last two rows of Table 1).
One drawback of the delta method of downscaling is that it assumes the future weather patterns will be the same as the
current weather patterns. To explore the inuence of changing weather patterns on denitrication, we generated three
hypothetical weather series with the same annual and seasonal means as the A1F1 series. In the Random series, precipitation
amounts randomly sampled from a distribution of double the A1F1 seasonal values occur on randomly selected days from
the season in question. Average daily temperature values for this series were generated randomly using the monthly mean
and standard deviation. Daily maximum and minimum temperatures were generated from daily temperature ranges, determined randomly using the monthly mean daily temperature range. In the Mean series, daily precipitation and temperature
values were taken from the mean daily values from the A1F1 scenario. This creates a scenario in which precipitation is a daily
occurrence. In the Cyclic series, temperature and precipitation patterns occur in 5 day increments each month. Again, based
on the A1F1 series, each months precipitation is uniformly divided among ve consecutive days in the month (all other days

Table 1
Projected temperature and precipitation changes for the northeastern United States for 20702099 relative to a baseline period of 19611990 (Hayhoe et al.,
2007).
Series

Annual temp (C)

Summer temp (C)

Winter temp (C)

Annual precip. (%)

Summer precip. (%)

Winter precip. (%)

Baseline
B1
A2
A1F1
A1F1 Temp.
A1F1 Precip.

2.9
4.5
5.3
5.3

2.4
4.3
5.9
5.9

1.7
3.7
5.4
5.4

( 1)
( 2)
0

12
14
30

30

7
9
14

14

68

J.R. Barclay, M.T. Walter / Sustainability of Water Quality and Ecology 5 (2015) 6476

have no precipitation) and daily temperatures are selected as in the Random series, except that the daily minimum, maximum, and average temperatures remain constant for ve days before changing.
2.5. Statistical analysis
For statistical analysis of the model outputs, we used the DunnettTukeyKramer method of multi-level pairwise comparisons (Dunnett, 1980; Lau, 2013). The advantage of this method is that it does not require that the variances are equal.
3. Results and discussion
3.1. Future temperature and precipitation series
We compared the annual and seasonal precipitation and soil temperature series with the climate projections to verify the
weather series generation process (using the delta method) produced the anticipated results. Both the annual and seasonal
precipitation and temperature changes matched the expectations (Fig. 2). The number of low ow days increased (days in
which the modeled discharge was below the 18% threshold in the baseline scenario) and the annual stream discharge
decreased in the A2 scenario. The summer decrease in soil moisture came earlier and in wetness class 1 was larger (see Supporting Information for details).
3.2. Comparison of baseline climate and B1, A2, and A1F1 scenarios
With all future scenarios, our model predicts an increase in the ecosystem service of denitrication relative to the baseline conditions. The greatest increase is with scenario A1F1 (11.8 kg N ha 1 yr 1, 41% increase) and the smallest increase is
with scenario B1 (5.1 kg N ha 1 yr 1, 23% increase); however, all three future scenarios are statistically different than the
baseline (p < 0.05) (Fig. 3). In all four weather scenarios, the highest watershed-wide denitrication rates occur during the
spring season. Our model predicts increased seasonal denitrication rates for all three scenarios (Fig. 4). There is a slight shift
in the seasonal pattern of denitrication, with an increase in the spring fraction for the A2 scenario, a decrease in the summer

Fig. 2. Annual (A) and Seasonal (B) Mean Precipitation and Annual (C) and Seasonal (D) Mean Soil Temperature with the four climate scenarios. This
conrms that our weather series generation process duplicated the projected changes.

J.R. Barclay, M.T. Walter / Sustainability of Water Quality and Ecology 5 (2015) 6476

69

Fig. 3. Annual modeled denitrication rates. For this and all subsequent boxplots, boxes extend from the rst to the third quartiles of annual rate (n = 30),
the center line indicates the median rate, and the whiskers extend to 150% of the interquartile range. Means with different letters are signicantly different
(DunnettTukeyKramer test, p < 0.05).

Fig. 4. Seasonal rates and seasonal fraction of annual rates. Means from the same season with different letters are signicantly different (DunnettTukey
Kramer test, p < 0.05).

fraction for the B1 and A2 scenarios, an increase in the fall fraction for the A2 scenario, and an increase in the winter fraction
for the A2 and A1F1 scenarios (Fig. 4).

3.3. Wetness class 1 denitrication rates


In contrast to the watershed-wide rates, the denitrication rates in the wettest areas (wetness class 1) are high in the
summer season, as well as during the spring. During the summer, this wettest class remains sufciently wet to maintain
anaerobic conditions, while the rest of the watershed dries substantially (Fig. 5A). The locations where wetness class 1 overlaps with wetland vegetation and soils create unique and important biogeochemical hotspots. In the baseline scenario, 64%
of the annual denitrication from wetness class 1 occurs during the summer, with an additional 29% occurring during the
spring. With the A1 and B2 scenarios, the summer fraction of annual wetness class 1 denitrication decreases and the spring
fraction increases, and in the A2 and A1F1 the winter fraction increases (Fig. 5B). During the spring, when much of the watershed is wet enough to create anaerobic conditions, wetness class 1 contributes 6471% of the seasonal denitrication from

70

J.R. Barclay, M.T. Walter / Sustainability of Water Quality and Ecology 5 (2015) 6476

Fig. 5. Denitrication in wetness class 1 (A) Seasonal denitrication rates, (B) Seasonal fraction of annual class 1 denitrication, (C) Class 1 fraction of
watershed-wide seasonal rate, (D) Class 1 fraction of annual watershed-wide rate. Means in the same season with different letters are signicantly different
(DunnettTukeyKramer test, p < 0.05).

the watershed, with a lower fraction in the baseline weather conditions and the highest fraction associated with the highest
emissions, i.e., the A1F1 scenario. As the rest of the watershed dries in the summer season, the wetness class 1 contribution
increases to 8190% of the summer denitrication across the watershed, with the higher class 1 fractions from the future
climate scenarios (Fig. 5C). The higher summer contributions in the future scenarios are due to decreased soil moisture in
the rest of the watershed, resulting from higher temperatures. Annually, the wettest class contributes 7481% of the annual
denitrication in all scenarios (Fig. 5D).
3.4. Patterns of changes
The spatial pattern of changes in annual denitrication rates follows the same pattern as the baseline denitrication rates
(Fig. 6 and Fig. 7AC); specically, areas with lower baseline rates have small increases and areas with higher baseline
denitrication rates have larger increases. These changes range from 1 to 3 kg N ha 1 yr 1 in much of the watershed to
150 kg N ha 1 yr 1 in the wettest wetlands with the A1F1 scenario. As a percentage of the baseline rates, the greatest
increases are the in areas with low or high baseline rates; moderate baseline rates had the lowest percentage increases
(Fig. 7FH).
Temporally, the watershed-wide rates of denitrication follow the same basic pattern with all climate scenarios, though
the future scenarios have higher rates, especially during the spring season (Fig. 8). Our model suggests that most denitrication rates will scale upwards due to increased temperature, but that some will decrease due to decreased soil moisture.
3.5. Slight increase in days without denitrication
In all but the wettest and driest areas of the watershed, we project an increase in the number of days annually on which
soil conditions are sufciently unfavorable that no denitrication is predicted to occur (See Supporting Information for

J.R. Barclay, M.T. Walter / Sustainability of Water Quality and Ecology 5 (2015) 6476

71

Fig. 6. Spatial patterns of baseline denitrication rates.

details). With the baseline weather, in most of the watershed, the model predicts denitrication rates of 0 kg N ha 1 yr 1 on
110200 days yr 1. This increases to 119220 days yr 1 in the future scenarios, with slightly greater increases in the A2 scenario. This increase is due to a decrease in anaerobic conditions resulting from drier soils.
3.6. Relative inuence of changes in precipitation and temperature
To examine the relative inuence of projected changes in precipitation and temperature, we created a future weather series that combined the precipitation from the A1F1 scenario with the temperature from the baseline series, and another that
combined the temperature from the A1F1 scenario with the precipitation from the baseline series. Though the median annual denitrication rate increased with both new scenarios, only the increases in the A1F1 (11.8 kg N ha 1 yr 1) and the temperature series (7.7 kg N ha 1 yr 1) are signicant at the 95% condence level (Fig. 9). Brown et al. (2012) found a bigger
effect from elevated precipitation, though this may be due to a low temperature increase (<1 C rather than 2.95.3 C in this
study) and a focus on N2O, rather than all denitrication products. Pinay et al. (2007) found soil moisture to be a more important indicator of denitrication rates than temperature; however this is based on a range of currently occurring conditions.
Because the magnitude of projected changes in temperature are much greater than for precipitation and resulting soil moisture (+5.3 C annually vs. +2% in volumetric water content annually), the denitrication rate increase from projected temperature increases is greater. At the mean soil temperature in the baseline series (6.8 C) and the mean wetness class 1
soil moisture (86% saturated), these temperature and precipitation changes result in changes to the temperature and soil
moisture denitrication rate scalars of 137% and 4% respectively.
Spatially, both the temperature and precipitation scenarios mirror the pattern of changes in the A1F1 scenario, but the
magnitude is very different (Fig. 7DE). The A1F1 scenario exhibits changes of 06 kg N ha 1 yr 1 in most of the watershed,
and increases of 144 kg N ha 1 yr 1 in wetlands of the wettest class. In the temperature scenario, these changes are more
modest, 01 and 94 kg N ha 1 yr 1 respectively, however in the wettest wetland, the precipitation scenario values are much
closer to the baseline, with increases of only 15 kg N ha 1 yr 1. As with the previously discusses scenarios, the percentage
increases in denitrication rates with the precipitation only and temperature only series are greatest in the areas of the
watershed with high or low denitrication rates in the baseline series. Moderate rates show lower percentage increases
(Fig. 7IJ).
Increases in denitrication rates with the A1F1 scenario are greater than the combined increases from the temperatureand precipitation-alone scenarios. In part, this is due to the multiplicative form of the denitrication model, but is it also due
to a counteracting effect the respective changes have on the watershed hydrology. In most seasons, increasing temperature
decreased the soil moisture, both watershed-wide and in the wettest areas, whereas increasing precipitation increased the
soil moisture, such that the combined effects are moderated (Fig. 10). This decrease was accompanied by an increase in modeled evapotranspiration (AET) (Fig. 11), which came from an energy-based (PriestleyTaylor) PET model (Archibald and
Walter, 2014) and not a temperature based approach, which have been shown to over-predict the effects of temperature
increases (Archibald and Walter, 2013; Shaw and Riha, 2011). Similarly increased temperatures result in decreased cumulative stream discharge, whereas the combined A1F1 series did not show this decrease. This is further evidence of the counteracting effects of the temperature and precipitation changes (see Supporting Information for details).
3.7. Effects of weather patterns
Our nal analysis sought to address the effects of temperature and precipitation patterns on denitrication rates. The spatial and temporal patterns of denitrication rates did not vary much among the A1F1, Random, and Cyclic scenarios. The

72

J.R. Barclay, M.T. Walter / Sustainability of Water Quality and Ecology 5 (2015) 6476

Fig. 7. Spatial patterns of denitrication rate changes The left column (AE) are the changes in the mean annual denitrication rates in the three climate
scenarios (B1, A2, A1F1) and the 2 variations of scenario A1F1 (temperature only and precipitation only). The right column (FJ) gives the mean annual
change as a percentage of the mean annual rate in the baseline series.

annual denitrication rate and seasonal fractions were similar (Fig. 12), as were the spatial patterns of annual rates (see Supporting Information), and the number of days on which no denitrication was projected (data not shown). The denitrication
rates from the Mean scenario were a bit different. The annual rate was lower (Fig. 12), the seasonal rate was higher in the
spring and lower in the fall (Fig. 12), the spatial pattern exhibits decreases in the wet areas where the other scenarios show
increases (see Supporting Information), and 80% of the watershed has more days annually (2060 days per year depending
on the class) without denitrication (data not shown). This unique pattern is likely due to the changes from the typical seasonal patterns of soil moisture. In the spring, when PET rates are typically low, soil moisture generally stays high throughout
the watershed although denitrication rates are somewhat reduced as the soil dries (albeit slowly) between rain events. In
the Mean scenario this drying does not occur because there is rain virtually every day. In the summer, on the other hand,
under baseline conditions PET rates are relatively high and the soils are generally dry, which keeps denitrication rates
low, except right after large rain storms that rewet the soils substantially; i.e., the periods right after summer rain storms
are essentially modeled as denitrication hot-moments. In the Mean scenario the periodic large events have been replaced
with continuous, low-intensity rain which evaporates almost as quickly as it falls so hot-moments are not generated.

J.R. Barclay, M.T. Walter / Sustainability of Water Quality and Ecology 5 (2015) 6476

73

Fig. 8. Mean watershed-wide denitrication averaged by Julian day. The darkest (solid) line is the baseline series, and the lighter lines are the future series
B1, A2, and A1F1 (lightest).

Fig. 9. Relative effects of temperature and precipitation changes on annual denitrication rates. Means with different letters are signicantly different
(DunnettTukeyKramer test, p < 0.05).

Fig. 10. Inuence of temperature and precipitation on seasonal soil moisture watershed-wide (A) and in wetness class 1 (B). Means from the same season
with different letters are signicantly different (DunnettTukeyKramer test, p < 0.05).

3.8. Implications of stand-alone denitrication model


In this study we chose to focus on the effects of changes in soil moisture and temperature. This simplied the analysis, but
required the assumption that nitrate and carbon availability would remain unchanged from current observations. Additionally our model structure assumes that nitrate and carbon will be replaced as they are consumed. In the baseline conditions

74

J.R. Barclay, M.T. Walter / Sustainability of Water Quality and Ecology 5 (2015) 6476

Fig. 11. Evapotranspiration averaged by Julian day. The baseline (19611990) and A1F1 precip. only series plot on the same line, as do the A1F1 and A1F1
temp only series.

Fig. 12. Annual denitrication rate (A) and seasonal fraction (B) with four weather patterns. Means from the same season with different letters are
signicantly different (DunnettTukeyKramer test, p < 0.05).

these assumptions are realistic because the DOC and NO3N inputs are based on eld measurements and the watershed
export of nitrogen via the stream and groundwater is similar in magnitude to that which is denitried (Anderson et al.,
2014a), suggesting the watershed is not nitrogen limited.
For the future scenarios, these assumptions introduce more error, but they may not be unrealistic. According to the calculations by Anderson (2013), this watershed currently exports 21 kg N ha 1 yr 1, with seasonal rates ranging from
19 kg N ha 1 yr 1 in the summer and fall to 24 kg N ha 1 yr 1 in the spring. With the exception of the summer season in
the A1F1 scenario, the current annual and seasonal NO3N export rates are greater than the modeled increases in denitrication, suggesting that the watershed currently has sufcient NO3N to support the modeled future denitrication rates. If
only stream export of NO3N is considered (assuming groundwater uxes might short-circuit denitrifying environments),
the current annual export is greater than the modeled denitrication increase for scenario B1 and for the fall and winter seasons in all three scenarios, but less than the projected annual increase for scenarios A2 and A1F1, and less than the summer
and spring increases in all three scenarios. This suggests that if only stream NO3N is considered excess, then our modeled
denitrication rates may be articially high, a question that can be explored in future work.
3.9. Implications for management
The benets of ecosystem services, such as denitrication, must be quantied if they are to be incorporated into management decisions. Unfortunately most ecosystem service quantication studies have focused on regional or national scales and
utilized secondary data (Crossman et al., 2013; Seppelt et al., 2011). We utilize a local scale, eld validated model and future
climate scenarios to quantify the ecosystem service of denitrication. These quantication results, based on a range of climate scenarios, provide managers with essential information for incorporating the ecosystem service of denitrication into
future management plans. Additionally, this study is a step towards addressing the lack of ecosystem service quantication
studies, particularly at local spatial scales and provides a model that can be used by subsequent quantication studies.
4. Summary
Our model results suggest that increasing precipitation and temperature will lead to increasing annual rates of denitrication in this watershed under projected climate changes in the Northeastern US, assuming nitrogen availability remains

J.R. Barclay, M.T. Walter / Sustainability of Water Quality and Ecology 5 (2015) 6476

75

unchanged. The greatest projected increases are in the areas and seasons with the highest baseline rates, with smaller
increases in those with lower baseline rates. We found that changing temperature will be a much stronger driver of change
to denitrication rates than changing precipitation, and that these changes are only moderately inuence by the daily
weather patterns we examined. Together these ndings suggest that the ability of agricultural watersheds to remove excess
nitrogen, preventing it from reaching downstream waters will increase in the coming decades, assuming wetter areas with
high denitrication rates are protected. Additionally, this study contributes to lling the gap in ecosystem service quantication studies and functions as an example that can be used to quantify other ecosystem services.
Acknowledgments
Partial funding for this project was awarded by an NSF-IGERT traineeship (Cross-scale Biogeochemistry and Climate) and
a grant funded by Federal Formula Funds distributed through the New York State Agricultural Experiment Station. We would
also like to acknowledge the following for assistance with eld measurements and analysis: Josen Klein, Helen Bergstrom,
Chelsea Morris, Megan Fitzgerald, Selene Ka-Wan Leung, Brian Buchanan, Brian Barclay, Sheila Saia, Lauren McPhillips, and
others from the Soil and Water Lab at Cornell.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.swaqe.
2014.12.006.
References
Addy, K., Kellogg, D.Q., Gold, A.J., Groffman, P.M., Ferendo, G., Sawyer, C., 2002. In situ push-pull method to determine ground water denitrication in
Riparian zones. J. Environ. Qual. 31 (3), 10171024.
Anandhi, A., Frei, A., Pierson, D.C., Schneiderman, E.M., Zion, M.S., Lounsbury, D., Matonse, A.H., 2011. Examination of change factor methodologies for
climate change impact assessment. Water Resour. Res. 47.
Anderson, T.R., 2013. Denitrication in riparian zones and other saturated soils of a northeastern agricultural landscape (PhD Dissertation). Cornell
University.
Anderson, T.R., Goodale, C.L., Groffman, P.M., Walter, M.T., 2014a. Assessing denitrication from seasonally saturated soils in an agricultural landscape: a
farm-scale mass-balance approach. Agric. Ecosyst. Environ. 189, 6069.
Anderson, T.R., Groffman, P.M., Kaushal, S.S., Walter, M.T., 2014b. Shallow groundwater denitrication in Riparian zones of a headwater agricultural
landscape. J. Environ. Qual. 43, 732744.
Archibald, J.A., Walter, M.T., 2013. Comment on Assessing temperature-based PET equations under a changing climate in temperate, deciduous forests by
Shaw and Riha. Hydrol. Process. 27 (24), 35113515.
Archibald, J.A., Walter, M.T., 2014. Do energy-based PET models require more input data than temperature-based models? An evaluation at four humid
FluxNet sites. J. Am. Water Resour. Assoc. 27, 35113515.
Ardia, D., Mullen, K.M., Peterson, B.G., Ulrish, J., 2012. DEoptim: differential Evolution in R. version 2.2-1. http://cran.r-project.org/web/packages/DEoptim/
DEoptim.pdf.
Arlot, S., Celisse, A., 2010. A survey of cross-validation procedures for model selection. Stat. Surv. 4, 4079.
Barclay, J.R., Anderson, T.R., Archibald, J.A., Walter, M.T., 2015. Modeling Denitrication in an Agricultural Catchment in Central New York. Sust. Water Qual.
Ecol. 5, 4963.
Barnard, R., Leadley, P.W., 2005. Global change, nitrication, and denitrication: a review. Global Biogeochem. Cycles 19, GB1007.
Boyacioglu, H., Vetter, T., Krysanova, V., Rode, M., 2012. Modeling the impacts of climate change on nitrogen retention in a 4th order stream. Clim. Change
113, 981999.
Brown, J.R., Blankinship, J.C., Niboyet, A., van Groenigen, K.J., Dijkstra, P., Le Roux, X., Leadley, P.W., Hungate, B.A., 2012. Effects of multiple global change
treatments on soil N2O uxes. Biogeochemistry 109, 85100.
Butterbach-Bahl, K., Dannenmann, M., 2011. Denitrication and associated soil N2O emissions due to agricultural activities in a changing climate. Curr.
Opin. Environ. Sust. 3, 389395.
Crossman, N.D., Burkhard, B., Nedkov, S., Willemen, L., Petz, K., Palomo, I., Drakou, E.G., Martin-Lopez, B., McPhearson, T., Boyanova, K., Alkemade, R., Egoh,
B., Dunbar, M.B., Maes, J., 2013. A blueprint for mapping and modelling ecosystem services. Ecosyst. Serv. 4, 414.
Daily, G.C., Alexander, S., Ehrlich, P.R., Goulder, L., Lubchenco, J., Matson, P.A., Mooney, H.A., Postel, S., Schneider, S.H., Tilman, D., Woodwell, G.M., 1997.
Ecosystem services: benets supplied to human societies by natural ecosystems. Issues Ecol. 2 (Spring 1997), 18.
Dunnett, C.W., 1980. Pairwise multiple comparisons in the unequal variance case. J. Am. Stat. Assoc. 75 (372), 796800.
EPA., 2011. Effects of Nitrogen and Phosphorus Pollution. http://water.epa.gov/scitech/swguidance/standards/criteria/nutrients/effects.cfm (accessed Nov 7,
2011).
Ferrant, S., Oehler, F., Durand, P., Ruiz, L., Salmon-Monviola, J., Justes, E., Dugast, P., Probst, A., Sanchez-Perez, J., 2011. Understanding nitrogen transfer
dynamics in a small agricultural catchment: comparison of a distributed (TNT2) and a semi-distributed (SWAT) modeling approaches. J. Hydrol. 406, 1
15.
Gleick, P.H., 1986. Methods for evaluating the regional hydrologic impacts of global climatic changes. J. Hydrol. 88, 97116.
Hay, L.E., Wilby, R.L., Leavesley, G.H., 2000. A comparison of delta change and downscaled gcm scenarios for three mountainous basins in the United States.
J. Am. Water Resour. Assoc. 36, 387397.
Hayhoe, K.A., 2010. A standardized framework for evaluating the skill of regional climate downscaling techniques (PhD Dissertation). University of Illinois at
Urbana-Champaign.
Hayhoe, K., Waker, C.P., Huntington, T.G., Luo, L., Schwartz, M.D., Shefeld, J., Wood, W., Anderson, B., Bradbury, J., DeGaetano, A., Troy, T.J., Wolfe, D., 2007.
Past and future changes in climate and hydrological indicators in the US Northeast. Clim. Dyn. 28, 381407.
Heinen, M., 2006. Simplied denitrication models: overview and properties. Geoderma 133, 44463.
Henault, C., Germon, J.C., 2000. NEMIS, a predictive model of denitrication on the eld scale. Eur. J. Soil Sci. 51, 257270.
Howarth, R.W., Swaney, D.P., Boyer, E.W., Marino, R., Jaworski, N., Goodale, C., 2006. Inuence of climate on average nitrogen export from large watersheds
in the Northeastern United States. Biogeochemistry 79, 163186.
Huntington, T.G., 2006. Evidence for intensication of the global water cycle: review and synthesis. J. Hydrol. 319, 8395.

76

J.R. Barclay, M.T. Walter / Sustainability of Water Quality and Ecology 5 (2015) 6476

Intergovernmental Panel on Climate Change (IPCC), 2013. Working Group I: The Scientic Basis. http://www.ipcc.ch/ipccreports/tar/wg1/029.htm (accessed
May 15, 2013).
Lau, M.K., 2013. DTK: DunnettTukeyKramer pairwise multiple comparison test adjusted for unequal variances and unequal sample sizes. R package
version 3.5. http://CRAN.R-project.org/package=DTK.
Lyon, S.W., Water, M.T., Gerard-Marchant, P., Steenhuis, T.S., 2004. Using a topographic index to distribute variable source area runoff predicted with the
SCS curve-number equation. Hydrol. Process. 18, 27572771.
Nash, J.E., Sutcliffe, J.V., 1970. River ow forecasting through conceptual models part I A Discussion of principles. J. Hydrol. 10, 282290.
NCDC, 2013. National Climate Data Center. http://www.ncdc.noaa.gov/cdo-web/datasets/GHCND/stations/GHCND:USC00304174/detail (accessed April 18,
2013).
Nelson, E.J., Daily, G.C., 2010. Modelling ecosystem services in terrestrial systems. F1000 Biol. Rep. 2 (53).
Pinay, G., Gumiero, B., Tabacchi, E., Gimenez, O., Tabacchi-Planty, A.M., Hefting, M.M., Burt, T.P., Black, V.A., Nilsson, C., Iordache, V., Bureau, F., Vought, L.,
Petts, G.E., Decamps, H., 2007. Patterns of denitrication rates in European alluvial soils under various hydrological regimes. Freshwater Biol. 52, 252
266.
Sandhu, H., Wratten, S., 2013. Ecosystem services in farmland and cities. In: Wratten, S., Sandhu, H., Cullen, R., Costanza, R. (Eds.), Ecosystem Services in
Agricultural and Urban Landscapes, First ed. John Wiley & Sons, p. 218.
Seppelt, R., Dormann, C.F., Eppink, F.V., Lautenback, S., Schmidt, S., 2011. A quantitative review of ecosystem services studies: approaches, shortcomings and
the road ahead. J. Appl. Ecol. 48 (3), 630636.
Shaw, S.B., Riha, S.J., 2011. Assessing temperature-based PET equations under a changing climate in temperate, deciduous forests. Hydrol. Process. 25, 1466
1478.
Singh, B.K., Bardgett, R.D., Smith, P., Reay, D.S., 2010. Microorganisms and climate change: terrestrial feedbacks and mitigation options. Nat. Rev. Microbiol.
8, 779790.
Tague, C., 2009. Modeling hydrologic controls on denitrication: sensitivity to parameter uncertainty and landscape representation. Biogeochemistry 93,
7990.
Thornthwaite, C.W., Mather, J.R., 1955. The Water balance. Pub. Climatol., 1104
USDA NRCS, 2013. Soil Survey Staff, Natural Resources Conservation Service, United States Department of Agriculture. http://websoilsurvey.nrcs.usda.gov/
(accessed Jan 31, 2013).
Volk, M., 2013. Modelling ecosystem services Challenges and promising future directions. Sust. Water Qual. Ecol. 12, 39.
Walter, M.T., Wilks, D.S., Parlange, J.-Y., Schneider, R.L., 2004. Increasing evapotranspiration from the conterminous United States. J. Hydrometeorol. 5 (3),
405408.
Whelan, M.J., Gandol, C., 2002. Modelling of spatial controls on denitrication at the landscape scale. Hydrol. Process. 16, 14371450.

You might also like