Modeling soil moisture effects on net nitrogen mineralization in loamy wetland soils

MODELING SOIL MOISTURE EFFECTS ON NET NITROGEN MINERALIZATIONIN LOAMY WETLAND SOILS

Steven Sleutel1, Bram Moeskops1, Willy Huybrechts2, Annemie Vandenbossche1, Joost Salomez1,

Sara De Bolle1, David Buchan1, and Stefaan De Neve1

1Department of Soil Management and Soil Care (Ghent University)

Coupure Links 653

9000 Gent, Belgium

E-mail: [email protected]

2Research Institute for Nature and Forest (INBO)

Kliniekstraat 25

1070 Brussels, Belgium

Abstract: Nutrient dynamics in wetland ecosystems are largely controlled by soil moisture content.

Therefore, the influence of soil moisture content on N mineralization should be explicitly taken into

account in hydro-ecological models. The aim of this research was to establish relationships between N

mineralization and soil moisture content in loamy to silty textured soils of floodplain wetlands in central

Belgium. Large undisturbed soil cores were taken, incubated for 3 months under various moisture

contents, and zero order and first order N mineralization rates were calculated. We used the percentage

water-filled pore space (WFPS) as an expression of soil moisture because it is a better index for aeration

dependent biological processes than volumetric moisture content or water retention. The relationship

between the N mineralization rate and %WFPS was described by a Gaussian model. The optimum

WFPS for N mineralization ranged between 57% and 78%, with a mean of 65% 6 6% WFPS. Expected

annual net N mineralization rates at field temperature (9.7uC) and at optimal moisture content varied

between 30 and 186 kg N ha21 (0–15 cm depth) year21, with a mean of 110 6 42 kg N ha21 (0–15 cm)

year21. The mean N turnover rate amounted to 2.3 6 1.1 g N 100 g21 N year21. Multiple linear

regressions between N mineralization and general soil parameters showed that soil structure has an

overriding impact on N mineralization in wetland ecosystems.

Key Words: Belgium, N mineralization, valley soils, water-filled pore space

INTRODUCTION

The hydrology of many wetlands has been drasti-cally changed by management for agriculture, forest-

ry, and flood control. Drainage efforts lower ground-

water levels and can impact wetland areas.

Furthermore, wetlands are threatened by nutrient

enrichment from sewage treatment effluent and

agricultural inputs (Neal and Whitehead 2002). Many

wetlands are now recognized as important ecosystem

resources (e.g., Ramsar Convention, Natura 2000network). However, for effective conservation and

restoration of wetlands more knowledge is required

about how water management impacts important soil

processes and wetland vegetation. A number of

hydro-ecological models have been developed that

link hydrology, soil processes, and vegetation ecology,

and assess the ecological impact of planned policy

interventions. Examples of such models are NICHE(Meuleman et al. 1996), ITORS (Venterink and

Wassen 1997), Wetland-DNDC (Zhang et al. 2002),

and the Library of Hydro-Ecological Modules

(LHEM) (Voinov et al. 2004).

In wetland ecosystems, nutrient dynamics are

largely controlled by soil moisture content (Takatert

et al. 1999, Bai et al. 2004). Water controls microbial

activity in the soil and thus determines rates of

mineralization. Mineralization, in turn, determines

the availability of mineral N for plant growth. It is

therefore important that the influence of soil

moisture content on N mineralization be explicitly

taken into account in hydro-ecological models.

However, in many of these models this is not the

case. For example, the empirical-statistical model

ITORS (Venterink and Wassen 1997) does not

simulate nutrient turn-over processes, but predicts

species composition using regressions directly estab-

lished between aquatic and terrestrial variables and

vegetation data. The LHEM includes a decomposi-

tion module, but it does not consider the influence of

moisture content.

Wetlands wetl-28-03-18.3d 23/6/08 14:20:01 724 Cust # 07-105

WETLANDS, Vol. 28, No. 3, September 2008, pp. 724–734’ 2008, The Society of Wetland Scientists

724

Wetland soils have low bulk densities and often a

macroporous structure due to high organic matter

content. Soil structure influences mineralization

because pore size distribution and aggregation

results in heterogeneity in the distribution of soil

water and microbial activity, and because pore size

distribution determines air-filled porosity and the

aeration status of the soil (Strong et al. 1998a,b).

For example, Yoo et al. (2006) investigated the

effects of soil structure and its interaction with soil

water content on C mineralization and found that

differences in pore size distribution in aggregates

affected optimum soil water conditions for biolog-

ical activity. Especially in the wet range (i.e., at

moisture tensions below pF 3), the influence of soil

structure on biological processes is crucial, because

under these conditions biological processes are

controlled by diffusion of substrate and O2 (Skopp

et al. 1990), both of which depend on porosity and

tortuosity. In spite of the interaction between soil

moisture content and soil structure, mineralization

experiments are generally carried out using homog-

enized or sieved soil (Linn and Doran 1984, Aulakh

et al. 1996, Bridgham et al. 1998). The physical

characteristics of soil exposed to such treatments

are, however, completely different from those of

undisturbed soil. For example, Schjønning et al.

(1999) found a less continuous and more tortuous

pore system in disturbed soils than in undisturbed

samples and reported a significantly higher water-

filled pore space for undisturbed than for disturbed

samples. Our goal was to better establish the

relationship between the evolution of mineral N

and moisture content in the wet range using large

undisturbed soil columns, and to examine the

relationship between N mineralization and a number

of soil properties for typical wetland soils.

MATERIALS AND METHODS

Sampling and Analyses

Our study targeted colluvial and alluvial soils in

valley areas of the silt region in central Belgium.

This region has a temperate maritime climate with

mild winters and cool summers with an average

precipitation of 757 mm y21 and an average yearly

temperature of 9.7uC. Bulk N deposition amounts to

about 30 kg N ha21 y21 (VMM 2004). Twenty-eight

soils with differing organic carbon (OC) content and

vegetation cover were selected in seven river valleys:

the Zuunvallei (ZUV), Silsombos (SIL), Molenbeek-

vallei (MOL), Dijlevallei (DYL), Snoekengracht

(SNO), Grypenveld (GRY), and Hoeleden (HOE)

(Figure 1). These are floodplains of streams or rivers

of varying dimensions that are embedded in a mostly

agricultural landscape. The dominant plant commu-

nities (meadow, shrub, or forest) of all sample

locations are given in Table 1. Silt loam was the

most abundant soil textural class (20 soils). The

selected 28 sampling locations were chosen out of a

total of 139 to include a range of soil OC content,

soil texture, and vegetation type. Groundwater levels

have been measured biweekly for the last 5 to 16

years using piezometers (except at ZUV003). High-

est (HGL), lowest (LGL), mean (MGL), and spring-

time (SGL) groundwater levels were calculated with

the time-series analysis software tool ‘‘Menyanthes’’

(Von Asmuth et al. 2006) (Figure 2). For ZUV003,

only a single year was monitored, so these data were

calculated manually.

In May and June 2005, undisturbed soil samples

were obtained from each location by pushing large

PVC tubes (diameter: 24 cm, height: 20 cm) into the

substrate to a depth of 15 cm. Litter and above-

ground vegetation were removed first. The tubes

were extracted carefully, and the excess soil at the

bottom of the tubes was removed and the tubes

closed at the bottom with a PVC cap. Five columns

were collected at each sampling point, one for every

moisture level to be tested. At four sampling

locations, 15 soil columns were taken for measure-

ments in triplicate to obtain information about the

variability of mineralization. As a result, 180 soil

columns were collected in total. The depth of the

humus-rich soil layer was determined visually by

observing change in color using four auguring.

Generally, soil moisture content at the time of

sampling was high. To be able to apply predeter-

mined moisture contents for the incubations, all

tubes were allowed to dry at room temperature to a

moisture content of 50% water-filled pore space

(WFPS);

WFPS %ð Þ~ 100h

1 { BD=rð Þ , ð1Þ

with h being the volumetric soil moisture content


Figure 1. Location of the seven valleys within the study

area in Belgium.

Sleutel et al., SOIL MOISTURE EFFECTS ON NET N MINERALIZATION 725

(cm3 cm23), BD being the bulk density (g cm23),

and r being the particle density (2.65 g cm23). At

high moisture content (i.e., below pF 3), WFPS is a

better index for aeration dependent biological

processes than volumetric moisture content or water

potential (Aulakh et al. 1996, De Neve and Hofman

2002). For each five sample set, exact soil moisture

content was measured, and then deionized water was

added as needed to attain following range: 95, 85,

75, 65, and 55% WFPS. Bulk density was estimated

from the volume and oven dry mass of the soil in the

PVC tubes. After addition of water, the PVC tubes

were covered with a single layer of gas-permeable

parafilm that allowed gas exchange but minimized

evaporative water loss. Tubes were incubated for 3

months in a dark room at an average temperature of

21.7uC. Precise temperatures were monitored with a

temperature logger over the whole incubation

period. At time intervals of 15–17 days, three

samples were taken per tube over the entire depth

(15 cm) using an auger (diameter: 2.5 cm). The three

samples were pooled into one composite sample of

which moisture content and mineral N content were

determined. For the determination of mineral N

content the pooled samples were extracted with 1N

KCl (soil:KCl ratio of 1:2). Extracted NO32 and

NH4+ was measured with a continuous flow auto-

analyzer (Chemlab System 4, Skalar, The Nether-

lands). At each sampling time, the water loss by

evaporation was checked by weighing the tubes and

deionized water was added accordingly. Losses of

mineral N by denitrification or N immobilization

were not considered. Rather, our aim was to

investigate the net evolution of mineral N in wetland

soils because it is this N availability that directly

governs plant productivity and eutrophication.


Table 1. Vegetation and organic C content (%OC), total N content (%TN), C:N ratio, pHKCl, texture, bulk density (BD)

of the 0–15 cm layer, and depth of the humus-rich soil layer (HL) at the 28 sampling points.

Sampling

point Vegetationa%OC

(%C)

%TN

(%N)

C:N

ratio

pHKCl

(-)

TextureBD

(g cm23)

HL

(cm)%sand %silt %clay USDA class

DYL001 Arrhenatherion elatioris M 1.69 0.16 10.6 4.0 45.2 46.9 7.9 sandy loam 1.38 40

DYL002 Alopercurion pratensis M 4.02 0.37 10.9 4.7 4.2 64.3 31.5 silty clay loam 0.80 30

DYL003 Phragmition australis S 6.47 0.58 11.2 5.4 29.5 41.5 29.1 clay loam 0.57 55

DYL004 Cynosurion cristati M 2.15 0.21 10.2 4.4 9.0 75.4 15.5 silt loam 1.17 45

DYL005 Filipendulion S 2.15 0.22 9.8 4.3 6.9 75.2 17.9 silt loam 1.19 20

DYL006 Poa trivialis - Lolium

perenne M 4.69 0.47 10.0 4.5 8.6 51.2 40.2 silty clay 0.77 20

DYL007 Alno-Padion F 5.13 0.34 15.1 5.0 61.9 26.0 12.0 sandy loam 0.70 20

DYL008 Cynosurion cristati M 4.05 0.43 9.4 4.2 5.3 50.8 43.8 silty clay 0.99 25

GRY001 Calthion palustrus M 4.48 0.36 12.4 5.9 32.0 51.3 16.7 silt loam 0.75 50

GRY002 Calthion palustrus M 4.60 0.44 10.5 6.3 10.6 71.1 18.3 silt loam 0.92 30

GRY003 Phragmites australis S 11.50 0.91 12.6 6.0 18.6 62.8 18.6 silt loam 0.34 30

GRY004 Calthion palustrus M 3.61 0.28 12.9 6.8 40.0 47.8 12.3 loam 0.87 25

HOE001 Alopecurion pratensis M 2.98 0.31 9.6 5.0 21.4 58.5 20.2 silt loam 1.08 20

HOE002 Poa trivialis - Lolium

perenne M 4.38 0.47 9.3 4.7 15.9 61.6 22.5 silt loam 1.04 20

HOE003 Poa trivialis - Lolium

perenne M 4.34 0.43 10.1 4.6 19.5 53.7 26.8 silt loam 0.82 20

MOL001 Calthion palustrus M 11.04 0.82 13.5 7.2 20.3 56.0 23.7 silt loam 0.43 50

MOL002 Caricion gracilis S 9.67 0.70 13.8 4.5 22.8 51.4 25.8 silt loam 0.49 30

MOL003 Urtica dioica (Alno-

Padion) F 3.88 0.32 12.1 5.6 17.5 62.4 20.0 silt loam 0.91 25

MOL004 Caricion gracilis S 23.19 1.69 13.7 5.1 7.5 39.7 52.8 clay 0.21 55

SIL001 Alno-Padion F 8.68 0.69 12.6 5.4 16.7 65.6 17.7 silt loam 0.55 30

SIL002 Calthion palustrus M 7.61 0.60 12.7 5.9 14.4 64.5 21.1 silt loam 0.51 55

SIL003 Alno-Padion F 5.80 0.45 12.9 6.2 16.1 63.2 20.7 silt loam 0.84 45

SNO001 Calthion palustrus M 5.99 0.57 10.5 5.5 16.7 68.3 15.0 silt loam 0.66 35

SNO002 Alnion glutinosae F 9.36 0.81 11.6 5.8 15.7 67.7 16.6 silt loam 0.52 60

SNO003 Calthion palustrus M 5.74 0.55 10.4 6.9 26.0 57.7 16.3 silt loam 0.76 40

ZUV001 Phalaris arundinacae S 4.33 0.42 10.3 4.6 6.9 66.1 27.0 silt loam 0.87 40

ZUV002 Filipendulion S 2.95 0.26 11.3 6.0 9.3 77.2 13.6 silt loam 1.15 15

ZUV003 Filipendulion S 3.36 0.30 11.2 5.4 18.8 65.6 15.6 silt loam 0.98 30a S: Shrub; F: Forest; M: Meadow.

726 WETLANDS, Volume 28, No. 3, 2008

Finally, a range of soil parameters was determined

on air-dried bulked samples obtained by mixing

subsamples from all tubes belonging to the same

sampling point (Table 1). The pHKCl was measured

in 1N KCl extracts (soil:KCl ratio of 1:5). Soil

texture was determined by the combined sieve and

pipette method (De Leenheer 1959). Total C and N

contents were measured with a Variomax CNS

analyzer (Elementar Gmbh., Germany) applying the

Dumas method. Organic carbon contents were

equated to total carbon contents for soils with pH

# 6.5. In soil with pH . 6.5, inorganic C content

was also measured by treatment of the samples with

H2SO4 and back titration with NaOH, and organic

C was determined as the difference between total

and inorganic C.

Data Processing

Mineral N contents were offset by subtracting

mineral N contents present in the soil at the

beginning of the incubation. N mineralization data

were fitted to both a zero order model,

N tð Þ~ kt, ð2Þ

with N(t) being the amount of mineral N at time t

(kg N ha21 (0–15 cm depth)) and k being the

mineralization rate (kg N ha21 (0–15 cm) day21),

and a first order model,

N tð Þ~ NA 1 { e{kt� �

, ð3Þ

with NA being the amount of mineralizable N and k

again being the mineralization rate (day21). Miner-

alization rates were expressed as kilograms of N

mineralized per ha (0–15 cm depth) instead of

mineralization per kg dry mass of soil because

organisms exploit a volume rather than a mass of

soil (Barko and Smart 1986).

The relationship between N mineralization rate

and %WFPS was described by the Gaussian

function proposed by De Neve and Hofman (2002),

k WFPSð Þ~ kopt;WFPSe{j 1 { WFPS

WFPSopt

� �2

, ð4Þ

with kopt;WFPS being the N mineralization rate at

WFPSopt (%), which is the optimal WFPS (%), and

j being a parameter describing the width of the bell-

shaped function. Non-linear regressions were carried

out using the Gauss-Levenberg-Marquardt algo-

rithm in SPSS (SPSS version 12.0, SPSS Inc.,

Chicago, IL, USA). The %WFPS-values used in

these calculations were based on the time-averaged

soil moisture contents measured during the incuba-

tion. Actual %WFPS-values sometimes deviated

considerably from the target values. In particular,

establishing soil moistures close to saturation was

hindered by the presence of large macropores in

several of the soil columns. Temporal variation in

moisture content from evaporative water loss was

minimal (average standard deviation of 4.2%

WFPS).

Although WFPS is considered as a better index

for aeration dependent biological processes than

water retention at high soil moisture contents

(Aulakh et al. 1996, De Neve and Hofman 2002),

many authors use soil water potential for describing

the effect of soil water content on mineralization

(Orchard and Cook 1983, Andren et al. 1992, Sierra

1997). Thus, we also examined the relationship

between soil water potential and N mineralization.

To this end, we substituted WFPS and WFPSopt in

equation 4 with pF and pFopt. Moisture content was

converted to a pF-value using pF-curves measured

for each soil. For high water tension (0 to

20.01 MPa), the sand box method was used, and

for low water tension (down to 21.5 MPa), the

pressure membrane method was used (Cornelis et al.

2005). The equation of Van Genuchten (1980) was

fitted to the desorption data.

Finally, two multiple linear regression models

through the origin were evaluated to establish

relationships between N mineralization rates, and a

number of soil parameters: %OC, % total N, C:N


Figure 2. Average groundwater level (in m below the soil

surface) of 27 sample sites calculated over 5–16 years. Bars

represent the spread between the averaged yearly highest

and lowest groundwater levels. Yearly mean groundwater

levels are represented by marks inside the bars.


ratio, pHKCl, %silt, %clay, bulk density (g cm23),

and depth of the humus-rich soil layer (cm).

Stepwise regression was chosen because this method

accounts for potential redundancy among indepen-

dents. Successive selection of independents in the

model took the form of a sequence of F-tests in

which independent variables were successively en-

tered (if F , P-in 5 0.05) or removed (if F . P-out

5 0.10). Regressions along with separate tests for

multicollinearity between the regression coefficients

were carried out using SPSS (Version 12.0, SPSS

Inc., Chicago, IL, USA).

RESULTS AND DISCUSSION

Generally, mineral N concentrations in soils

increased linearly with time (Figure 3), which indi-

cated that losses of mineral N from denitrification or

immobilization were limited during the incubation

period. However, for some soils (26 out of 180 soil

cores including ZUV001 and MOL003), the upward

trend during the first two months of incubation was

followed by a decrease in mineral N concentration,

which may be the result of net N immobilization. For

some other soils (24 out of 180 soil cores including

SIL002 and HOE002), a loss of mineral N was

measured after the first 20 days of incubation, which

was probably due to denitrification. As explained

previously, denitrification and N immobilization

were, however, not explicitly taken into account in

the experiment. The proportion of NO3-N to the total

mineral N increased steadily from 55% to 78% over

the 3 month incubation period. Clearly, nitrification

generally occurred concomitantly with ammonifica-

tion, and both NH4 and NO3 contributed substan-

tially to mineral N. The relatively faster nitrification

led to accumulation of NO3, which could potentially

leach or promote denitrification. Leaching of NO3

was likely minimal in these wetlands given their

shallow groundwater depths (the average lowest

depth was 20.64 m). However, denitrification may

be a relevant process in these soils. For example, total

denitrification was 5–40 kg N ha21 y21 in riparian

wetlands (Hanson et al. 1994) and 1.5–8.4 kg N

ha21 y21 in Flemish wet grasslands (Vermoesen

1999). N2O-emissions in riparian zones in Flanders

were 20.2–2.1 kg N2O-N ha21 y21 (D’Hondt et al.

2004). Not measuring denitrification seems justified

given the relatively small proportion of denitrification

compared to the net N mineralization (5%–10% for

most soils). A reasonable zero order model fit for N

mineralization was obtained for almost all soils and

moisture levels (median of R2adjusted-values 5 0.84).

The first order model could only be successfully fit to

two thirds of the soil columns because unrealistic


Figure 3. Evolution of mineral N for A) ZUV002, B)

MOL003, and C) HOE002. WFPS in the legend is the

measured time-averaged soil moisture content for each

soil moisture level.


estimations of the parameters NA and k were

obtained in 58 cases. In those cases, very high values

of NA were associated with very low values of k. The

3 month incubation period was too short in those

soils for mineralization rates to decline, as is implied

by the first order model. For this reason, only

mineralization rates estimated by the zero order

model were used in the following analysis.


Table 2. Regression parameters of the Gaussian model (standard errors are given in parentheses).

Sampling point WFPSopt (%) kopt;WFPS (kg N ha21 (0–15 cm) day21) j R2

DYL001 66 (3) 1.28 (0.20) 9.7 (6.7) 0.31

DYL002 61 (3) 1.08 (0.16) 10.1 (7.6) 0.52

DYL003a - - - -

DYL004 65 (2) 0.72 (0.18) 28.0 (27.1) 0.43

DYL005 68 (20) 0.88 (0.14) 2.6 (6.4) 0.38

DYL006a - - - -

DYL007 89 (386) 1.14 (5.27) 2.5 (4.3) 0.34

DYL008 72 (12) 0.84 (0.30) 9.9 (25.5) 0.10

GRY001 64 (5) 0.29 (0.07) 8.3 (9.7) 0.38

GRY002a - - - -

GRY003a - - - -

GRY004a - - - -

HOE001 57 (21) 0.61 (0.08) 1.7 (4.0) 0.19

HOE002a - - - -

HOE003 59 (23) 0.80 (0.26) 4.8 (13.1) 0.46

MOL001 61 (15) 0.48 (0.17) 3.9 (10.6) 0.11

MOL002 64 (2) 0.78 (0.05) 9.8 (3.0) 0.69

MOL003 62 (12) 0.89 (0.19) 7.0 (12.0) 0.60

MOL004 57 (3) 0.47 (0.03) 2.3 (1.3) 0.71

SIL001a - - - -

SIL002 64 (3) 0.38 (0.10) 39.5 (33.1) 0.54

SIL003 65 (2) 0.52 (0.13) 49.6 (24.7) 0.86

SNO001 64 (30) 0.83 (0.21) 3.3 (11.0) 0.38

SNO002 63 (7) 0.93 (0.67) 25.8 (52.2) 0.65

SNO003 78 (5) 0.55 (0.15) 10.2 (13.9) 0.20

ZUV001 69 (3) 1.36 (0.11) 4.4 (2.8) 0.61

ZUV002 76.1 (0.4) 0.89 (0.02) 28.6 (2.6) 1.00

ZUV003 74 (7) 1.07 (0.09) 4.0 (1.6) 0.90a Fitting the Gaussian model to the mineralization data resulted in unrealistic parameter estimates.

Figure 4. Relation between zero-order mineralization rate and moisture content. Error bars indicate the standard error

on the estimated mineralization rates. A) ZUV002, B) SIL001.


Equation 4 could be fitted with an acceptable R2

to most of the data (Table 2). Some of the

mineralization data, however, did not show a clear

relationship to moisture content, and fitting the

Gaussian model to these data resulted in unrealistic

estimations of model parameters (Figure 4). For

DYL003 and DYL006, this may be attributed to a

restricted range of moisture levels, i.e., about half

the range of the other soils. For GRY002, GRY003,

GRY004, and SIL001, applied moisture levels were

all situated on the right-hand side of the optimum of

the mineralization curve, complicating fit of bell-

shaped functions to the mineralization data. Only

considering the soils that fit a Gaussian curve, and

also omitting the soils with unacceptably high

standard errors (DYL007), WFPSopt ranged be-

tween 57% and 78% WFPS, with a mean of 65% 6

6% WFPS. This value is slightly higher than other

published estimations of WFPSopt for N minerali-

zation (60% WFPS, Linn and Doran 1984; 56%

WFPS, De Neve and Hofman 2002).

The graphs from soils DYL001, HOE001,

MOL002, and SIL002, which were incubated in

triplicate, show that the variability of mineral N

evolution in these soils was large (Figure 5). The

relationship between moisture content and mineral-

ization rate thus depended on which data were

selected. The influence of data selection on the

estimation of the regression parameters was ana-

lyzed as follows. For each of the soils DYL001,

HOE001, MOL002, and SIL002, equation 4 was

fitted three times to 5 randomly selected data points

(Table 3), one for each predetermined moisture

level. The robustness of Equation 4’s parameters

(i.e., their insensivity to stochastic variation in the

data) was quantified as the median coefficient of

variation in the parameter estimates of all four soils,

which was in the order: WFPSopt (4.8%) . kopt;WFPS

(27.0%) . j (45.8%). Parameter WFPSopt thus

proved to be most robust, while parameter j, which

is related to the width of the bell-shaped function,

depended strongly on data selection. In the case of

HOE001 (Table 3), the large variation in WFPSopt

can be explained by the ‘‘flatness’’ of the Gaussian

model for this soil.

In general, better fits were obtained with WFPS-

based regressions than with pF-based regressions.

Whereas the WFPS-based Gaussian model could

not be fit to seven of the 28 soils, the pF-based

model could not be used for 12 of the soils, either

because the peaty structure of the soil did not allow

taking undisturbed ring samples (and hence no pF

data could be obtained, as for MOL001), or because

the estimated minimum and maximum moisture

content of the pF-curve did not cover the full range


Figure 5. Relation between zero-order mineralization

rate and moisture content for the soils incubated in

triplicate. Error bars as in figure 4. A) DYL001, B)

HOE001, C) MOL002, and D) SIL002.



Table 3. Results of fitting equation 4 to randomly selected replicate points of DYL001, HOE001, MOL002, and SIL002

(standard errors are given in parentheses).

WFPSopt (%) kopt;WFPS (kg N ha21 (0–15 cm) day21) j R2

DYL001

66 (2) 1.61 (0.18) 13.6 (5.3) 0.88

68 (1) 1.53 (0.18) 23.1 (6.5) 0.93

63 (90) 1.04 (0.47) 21.1 (14.5) 0.01

HOE001

34 (84) 0.97 (1.00) 0.5 (3.4) 0.81

0.11 (1175) 1.16 (12.46) 0.0 (0.0) 0.33

71 (1) 0.42 (0.03) 28.7 (2.3) 0.88

MOL002

69 (1) 0.85 (0.08) 39.1 (11.1) 0.92

62 (2) 0.86 (0.16) 10.2 (4.6) 0.77

68 (1) 0.93 (0.18) 33.6 (17.3) 0.79

SIL002

67 (2) 0.54 (0.19) 84.3 (65.7) 0.83

68 (5) 0.37 (0.54) 70.9 (161.2) 0.54

65 (7) 0.29 (0.24) 38.9 (71.2) 0.48

Table 4. Annual N mineralization rate (kfield) and annual N turnover rate (kturn) at field temperature and at optimal

moisture content, and total N stock.

Sampling Point

kfield kg N ha21

(0–15 cm) year21) N stock (0–15 cm) (kg N ha21) kturn (g N 100 g21 N year21)

DYL001 175 3312 5.3

DYL002 148 4440 3.3

DYL003 133a 4959 2.7

DYL004 98 3686 2.7

DYL005 120 3927 3.1

DYL006 156a 5429 2.9

DYL007 156 3570 4.4

DYL008 115 6386 1.8

GRY001 40 4050 1.0

GRY002 168a 6072 2.8

GRY003 62a 4641 1.3

GRY004 30a 3654 0.8

HOE001 83 5022 1.7

HOE002 149a 7332 2.0

HOE003 109 5289 2.1

MOL001 66 5289 1.2

MOL002 107 5145 2.1

MOL003 122 4368 2.8

MOL004 64 5324 1.2

SIL001 85a 5693 1.5

SIL002 52 4590 1.1

SIL003 71 5670 1.3

SNO001 113 5643 2.0

SNO002 127 6318 2.0

SNO003 75 6270 1.2

ZUV001 186 5481 3.4

ZUV002 122 4485 2.7

ZUV003 146 4410 3.3a Calculated from highest measurement of k instead of k at WFPSopt.


of moisture contents measured during the incuba-

tions (e.g., ZUV003), or because the Gaussian model

yielded unrealistic estimates of the parameter pFopt

(e.g., GRY002, HOE002). Skopp et al. (1990)

indicated that the use of soil water potential tends

to emphasize the driest conditions. Moreover, at high

moisture contents N mineralization is not controlled

by water availability but by diffusion of substrate

and O2, and substrate and O2 diffusion coefficients

are typically characterized by soil water content.

Nevertheless, kopt;pF-estimates and kopt;WFPS-esti-

mates corresponded very well (kopt;WFPS 5

0.98*kopt;pF, r 5 0.96, N 5 16, P , 0.01).

Incubations were carried out at an average

temperature of 21.7uC. However, to assess mineral-

ization rates for field conditions kopt;WFPS estimates

were converted to the 9.7uC mean annual temper-

ature in Belgium using an equation developed by De

Neve et al. (1996):

k Tð Þ~ kopt;T e{ 1 { T=Toptð Þ2 , ð5Þ

with T being the temperature (uC), kopt;T being the N

mineralization rate at Topt, the optimal temperature

of 37uC, and k (2.63) being a parameter describing

the width of the bell-shaped function. kopt;T-values

were determined by inserting kopt;WFPS-estimates at

21.7uC in the left term of equation 5 and T 5 21.7uCin the right term. The ratio of kopt;WFPS at T 5

21.7uC and at T 5 9.7uC was 2.02.

Annual net N mineralization rates at field temper-

atures kfield and optimum moisture content (or at

maximum measured k when no relation between

moisture content and N mineralization could be

established) ranged from 30 to 186 kg N ha21 (0–

15 cm) year21, and averaged 110 6 42 kg N ha21 (0–

15 cm) year21 (Table 4). N turnover rate ranged

from 0.8–5.3 g N 100 g21 N year21, and averaged

2.3 6 1.1 g N 100 g21 N year21. Those rates are in

the same order of magnitude as other published

mineralization rates for wetland soils. Bridgham et

al. (1998) found a N mineralization rate of 97 kg N

ha21 (0–15 cm) year21 and a turnover rate of 1.8 g N

100 g21 N year21 in a Minnesota meadow soil with a

22.5% total carbon content (converted here to 9.7uCusing equation 5). Best and Jacobs (2001) reported a

N mineralization rate of 97.3 kg N ha21 (0–15 cm)

year21 for a wet clayey fen peat meadow in The

Netherlands with an organic matter content of 28%,

and Andersen (2004) estimated the N mineralization

in the root zone of a seasonally inundated Danish

floodplain wetland at 74.1 kg N ha21 year21.

Stepwise multiple regression models relating the

annual mineralization rate kfield and annual turnover

rate kturn to soil characteristics (Table 5) showed

that bulk density and %clay are the most important

predictors, which suggests a strong influence of soil

pore structure on the N mineralization process. We

found no indications of collinearity between predic-

tors. Other studies confirm the importance of bulk

density for predicting mineralization (Bridgham et

al. 1998). The effect of bulk density, however, also

includes an OC effect because of the strong negative


Table 5. Regression coefficients retained by stepwise multiple linear regression through the origin for kfield and kturn

(standard errors are given in parentheses).

Variable

Unstandardized

coefficients

Standardized

coefficients t Sig. R2adjusted

kfield 0.89

Bulk Density 97.9 (15.0) 0.701 6.529 0.000

%clay 1.39 (0.52) 0.286 2.661 0.013

kturn 0.86

Bulk Density 2.77 (0.21) 0.930 13.151 0.000

Table 6. Pearson’s coefficients of correlation (R) between Highest (HGL), lowest (LGL), mean (MGL), and spring-time

(SGL) groundwater levels and selected soil parameters (%OC, soil organic carbon percentage; %TN, soil N percentage;

BD, Bulk Density) and the annual N mineralization rate (kfield) and annual N turnover rate (kturn) at field temperature and

at optimal moisture content.

%OC %TN C/N-ratio BD kfield kturn

HGL 0.450* 0.451* 0.487** 20.678** 20.389* 20.602**

LGL 0.591** 0.572** 0.616** 20.788** 20.491** 20.647**

MGL 0.436** 0.426* 0.496** 20.611** 20.443* 20.634**

SGL 0.484** 0.475* 0.562** 20.717** 20.402** 20.591**

*Significant at the 0.05 level.**Significant at the 0.01 level.


relationship between bulk density and %OC, which

in turn, is positively related to groundwater levels

(Table 6). Significant positive correlations between

groundwater levels (HGL, LGL, SGL, and MGL)

and %OC, N, and C:N-ratio, and negative correla-

tions with BD indicate that hydrology has a

dominant effect on soil OM and indirectly on BD

(Table 6). The strongest correlations were found for

LGL (Figure 6), which showed the largest variation

among sampling locations (Figure 2). While neither

HGL, LGL, SGL, nor MGL were significant

predictors for kfield and kturn in the stepwise multiple

regression model, the indirect control of hydrology

on observed N-dynamics is confirmed by these

correlations and by significant negative correlations

with the N-dynamics (Table 6). The pronounced

negative relation between groundwater levels and N-

mineralization rates suggests that in soils with

shallow groundwater depths, more soil OM accu-

mulates with higher C:N-ratio, and N-mineraliza-

tion rates slow in wetter soils.

Finally, it should be recognized that decomposi-

tion of residual dead roots in the soil cores may have

contributed to the N-mineralization. McJannet et al.

(1995) reported a range of 2.5–20.1 g N kg21 dry

matter in tissues from 41 wetland species. Gusewell

(2004) reported 13.4, 14.8, and 11.5 g N kg21 dry

matter for woody species, forbs, and graminoids

respectively, in European wetlands. Decomposition

of N-poor plant litter should result in a net N

immobilization. Immobilization probably occurred

in a limited number of incubated soil cores, but at

low rates. While interference from dead roots is

inherent to N mineralization experiments with

undisturbed soil columns, and are thus unavoidable,

they should not be ignored. However, there were no

significant differences between the vegetation groups

in either N mineralization expressed on a gravimet-

ric (kturn) or per hectare (kfield) basis (Table 4,

ANOVA, P . 0.05), suggesting that differences in

decomposition rates of dead roots did not signifi-

cantly bias results. Instead the vegetation effect was

dominated by a large within-vegetation-group var-

iability in net N mineralization, attributable to

inherent differences in soil properties.

CONCLUSION

In this study, 180 soil columns from 28 different

wetland locations were incubated to determine the

effect of variable, but high, soil moisture content on

mineral N concentrations. For most soils, the

relationship between soil moisture and evolution of

soil mineral N was described well by a Gaussian

model when using WFPS as a measure of soil water

status. Bulk density and %clay were the most

important predictors of N mineralization, which

indicates that soil pore structure is an important

influence on biological processes in wet ecosystems.

The specific experimental approach adopted here,

large undisturbed soil cores with periodic sampling,

was especially useful because the original soil

aggregate structure was preserved, and should be

considered whenever studying soil OM and N


Figure 6. Relation between the yearly lowest groundwa-

ter level (LGL) and A) organic carbon content (%OC) (0–

15 cm depth) and B) bulk density (BD) (5–10 cm depth).


cycling in wetland soils. Finally, we suggest that

unequivocal assessment of the individual impacts of

soil hydrology, organic matter content, structure,and texture on N mineralization is not realistic

because they are all interrelated. Our data seem to be

reasonable estimates of the net evolution of mineral

N in wetland soils, and relationships developed are

probably suitable for simulating mineral N dynam-

ics using semi-empirical models of wetland biogeo-

chemistry. For fully mechanistic models, additional

experimental data will be needed to separate netmineralization from gaseous N losses, and to

calculate gross N mineralization and N immobiliza-

tion rates.

LITERATURE CITED

Andersen, H. E. 2004. Hydrology and nitrogen balance of aseasonally inundated Danish floodplain wetland. HydrologicalProcesses 18:415–34.

Andren, O., E. Steen, and K. Rajkai. 1992. Modelling the effectsof moisture on barley straw and root decomposition in thefield. Soil Biology and Biochemistry 24:727–36.

Aulakh, M. S., Kuldip-Singh, Bijay-Singh, and J. W. Doran.1996. Kinetics of nitrification under upland and flooded soils ofvarying texture. Communications in Soil Science and PlantAnalysis 27:2079–89.

Bai, J. H., W. Deng, Y. M. Zhu, and Q. G. Wang. 2004. Spatialvariability of nitrogen in soils from land/inland water ecotones.Communications in Soil Science and Plant Analysis 35:735–49.

Barko, J. W. and R. M. Smart. 1986. Sediment-relatedmechanisms of growth limitation in submersed macrophytes.Ecology 67:1328–40.

Best, E. P. H. and F. H. H. Jacobs. 2001. Production, nutrientavailability, and elemental balances of two meadows affectedby different fertilization and water table regimes in TheNetherlands. Plant Ecology 155:61–73.

Bridgham, S. D., K. Updegraff, and J. Pastor. 1998. Carbon,nitrogen, and phosphorous mineralization in northern wet-lands. Ecology 79:1545–61.

Cornelis, W. M., M. Khlossi, R. Hartmann, M. Van Meirvenne,and B. De Vos. 2005. Comparison of unimodal analyticalexpressions for the soil-water retention curve. Soil ScienceSociety of America Journal 69:1902–11.

De Leenheer, L. 1959. Procedures of analyses at the Centre forSoil Research (in Dutch). Centre for Soil Research, Ghent,Belgium.

De Neve, S. and G. Hofman. 2002. Quantifying soil water effectson N mineralization from soil organic matter and from freshcrop residues. Biology and Fertility of Soils 35:379–86.

De Neve, S., J. Pannier, and G. Hofman. 1996. Temperatureeffects on C and N mineralization from vegetable crop residues.Plant and Soil 181:25–30.

D’Hondt, K., P. Boeckx, G. Hofman, and O. Van Cleemput.2004. Temporal and spatial patterns of denitrification enzymeactivity and nitrous oxide fluxes in three adjacent vegetatedriparian buffer zones. Biology and Fertility of Soils 40:243–51.

Gusewell, S. 2004. N : P ratios in terrestrial plants: variation andfunctional significance. New Phytologist 164:243–66.

Hanson, G. C., P. M. Groffman, and A. J. Gold. 1994.Denitrification in riparian wetlands receiving high and lowgroundwater nitrate inputs. Journal of Environmental Quality23:917–22.

Linn, D. M. and J. W. Doran. 1984. Effect of water-filled porespace on carbon dioxide and nitrous oxide production in tilledand nontilled soils. Biology and Fertility of Soils 25:89–94.

McJannet, C. L., P. A. Keddy, and F. R. Pick. 1995. Nitrogenand phosphorous concentrations in 41 wetland plants - acomparison across habitats and functional-groups. FunctionalEcology 9:231–38.

Meuleman, A. F. M., R. A. Kloosterman, W. Koerselman, M.Den Besten, and A. J. M. Jansen. 1996. NICHE: a newinstrument for hydro-ecological effect prediction (in Dutch).H2O 5:137–39.

Neal, C. and P. G. Whitehead. 2002. Water quality functioning oflowland permeable catchments: inferences from an intensivestudy of the River Kennet and upper River Thames: anintroduction. Science of the Total Environment 282:3–7.

Orchard, V. A. and F. J. Cook. 1983. Relationship between soilrespiration and soil-moisture. Soil Biology & Biochemistry15:447–53.

Schjønning, P., I. K. Thomsen, J. P. Møberg, H. de Jonge, K.Kristensen, and B. T. Christensen. 1999. Turnover of organicmatter in differently textured soils. I. Physical characteristics ofstructurally disturbed and intact soils. Geoderma 89:177–98.

Sierra, J. 1997. Temperature and soil moisture dependence of Nmineralization in intact soil cores. Soil Biology and Biochem-istry 29:1557–63.

Skopp, J., M. D. Jawson, and J. W. Doran. 1990. Steady-stateaerobic microbial activity as a function of soil water content.Soil Science Society of America Journal 54:1619–25.

Strong, D. T., P. W. G. Sale, and K. R. Heylar. 1998a. Theinfluence of the soil matrix on nitrogen mineralisation andnitrification: I. Spatial variation and a hierarchy of soilproperties. Australian Journal of Soil Research 36:429–47.

Strong, D. T., P. W. G. Sale, and K. R. Heylar. 1998b. Theinfluence of the soil matrix on nitrogen mineralisation andnitrification: II. The pore system as a framework for mappingthe organisation of the soil matrix. Australian Journal of SoilResearch 36:855–72.

Takatert, N., J. M. Sanchez-Perez, and M. Tremolieres. 1999.Spatial and temporal variations of nutrients concentration inthe groundwater of a floodplain: effect of hydrology, vegetationand substrate. Hydrological Processes 13:1511–26.

Van Genuchten, M. T. 1980. A closed-form equation forpredicting the hydraulic conductivity of unsaturated soils. SoilScience Society of America Journal 44:892–98.

Venterink, H. O. and M. J. Wassen. 1997. A comparison of sixmodels predicting vegetation response to hydrological habitatchange. Ecological Modelling 101:347–61.

Vermoesen, A. 1999. Aspects of nitrogen losses from grassland.PhD-thesis. Ghent University, Gent, Belgium.

VMM. 2004. MIRA Achtergronddocument 2004 Vermesting.Milieu- en Natuurrapport Vlaanderen (in Dutch). FlemishEnvironmental Agency, Mechelen, Belgium.

Voinov, A., C. Fitz, R. Boumans, and R. Costanza. 2004.Modular ecosystem modeling. Environmental Modelling &Software 19:285–304.

Von Asmuth, J. R., C. Maas, and M. Knotters. 2006.Menyanthes Manual, Version 1.5. Kiwa Water Research,Nieuwegein, The Netherlands.

Yoo, G., L. A. Spomer, and M. M. Wander. 2006. Regulation ofcarbon mineralization rates by soil structure and water in anagricultural field and a prairie-like soil. Geoderma 135:16–25.

Zhang, Y., C. S. Li, C. C. Trettin, H. Li, and G. Sun. 2002. Anintegrated model of soil, hydrology, and vegetation for carbondynamics in wetland ecosystems. Global Biogeochemical Cycles16:1061.

Manuscript received 4 June 2007; accepted 22 April 2008.



Modeling soil moisture effects on net nitrogen mineralization in loamy wetland soils

Documents