— Natural attenuation as example from The Netherlands

Science of the Total Environment 415 (2012) 49–55
Contents lists available at ScienceDirect
Science of the Total Environment
j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / s c i t o t e n v
How to calculate the spatial distribution of ecosystem services — Natural attenuation
as example from The Netherlands
H.J. van Wijnen a,⁎, M. Rutgers a, A.J. Schouten a, C. Mulder a, D. de Zwart a, A.M. Breure a, b
National Institute for Public Health and the Environment, P.O. Box 1, 3720 BA Bilthoven, The Netherlands
Radboud University Nijmegen, P.O. Box 9010, 6500 GL Nijmegen, The Netherlands
a r t i c l e
i n f o
Article history:
Received 4 February 2011
Received in revised form 11 May 2011
Accepted 11 May 2011
Available online 2 July 2011
Soil quality
Digital Soil Mapping
a b s t r a c t
Maps play an important role during the entire process of spatial planning and bring ecosystem services to the
attention of stakeholders' negotiation more easily. As example we show the quantification of the ecosystem
service ‘natural attenuation of pollutants’, which is a service necessary to keep the soil clean for production of
safe food and provision of drinking water, and to provide a healthy habitat for soil organisms to support other
ecosystem services. A method was developed to plot the relative measure of the natural attenuation capacity
of the soil in a map. Several properties of Dutch soils were related to property-specific reference values and
subsequently combined into one proxy for the natural attenuation of pollutants. This method can also be used
to map other ecosystem services and to ultimately integrate suites of ecosystem services in one map.
© 2011 Elsevier B.V. All rights reserved.
1. Introduction
Ecosystem services are key factors for a healthy and prosperous life
of humans. Examples of the products of such services are: protection
against flooding, provision of clean drinking water, and production of
food and fibers. The economic value of these services becomes even
more clear (Balmford et al., 2008; Costanza et al., 1997; Millennium
Ecosystem Assessment, 2005).
Increasingly we see ecosystems as ‘natural resources’ to be
sustainably managed and kept healthy. As in other countries, land
use in The Netherlands (like elsewhere) is driven by economic,
technological and demographic developments. Still, one of the
major challenges remains the optimization of the land management in such a way that ecosystem services will be provided
optimally, not only according to wellbeing and human activities at
(most) suitable places, but also according to optimal chemical,
physical and environmental characteristics. All these factors are
helpful to model future land-use needs, in addition to the energy
and elemental flow across soil, water and environmental compartments (sensu Bateman et al., 1999; Eade and Moran, 1996; Farber
et al., 2002).
⁎ Corresponding author. Tel.: + 31 302743715; fax: + 31 302744413.
E-mail address: [email protected] (H.J. van Wijnen).
0048-9697/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
A way to express such characteristics is by indication of the
extent to which an ecosystem is “fit” for the performance of the
desired ecosystem services. A method for analysis and assessment of
ecosystem services is presented by Rutgers et al. (2012). Another
tool for raising awareness and environmental assessment and
planning is to express the performance of ecosystems on maps
(Haygarth and Ritz, 2009). This can be done by expressing the
actual, relative, or maximal possibilities (Jeffery et al., 2010). We
present here a procedure to make these maps in a Geographical
Information System (GIS), based on the algorithm of Rutgers et al.
(2012) and on multiple regression models of existing empirical data
of biotic and abiotic attributes from the Dutch national soil survey,
namely the Biological indicator of Soil Quality (BiSQ; Rutgers et al.,
The optimal use of ecosystem services is recognized by beneficiaries. These are the land users in the case of the soil system. Often
ecosystem services co-occur and interact in a wide range of spatial
and temporal scales. The spatial distribution of ecosystem services
can therefore play an important role in the process of spatial
planning and bring ecosystem services on the table during negotiations
(Grêt-Regamey et al., 2008; Metzger et al., 2008; Troy and Wilson,
2006). As an example we used a method to quantify and to map the
ecosystem service ‘natural attenuation of pollutants’. This is a service
that keeps the soil clean for, e.g. production of food and clean drinking
water (National Research Council, 2000; Rittmann, 2004) and which is
necessary to maintain a healthy habitat for soil organisms, which in turn
support other ecosystem services (Millennium Ecosystem Assessment,
H.J. van Wijnen et al. / Science of the Total Environment 415 (2012) 49–55
2. Methods
The US Environmental Protection Agency (EPA) defines natural attenuation as a process that includes ‘a variety of physical, chemical, or
biological processes that, under favorable conditions, act without human intervention to reduce the mass, toxicity, mobility, volume, or
concentration of contaminants in soil or groundwater’. These in-situ processes include biodegradation, dispersion, dilution, adsorption,
volatilization, decay, stabilization, transformation, or destruction of contaminants (EPA, 1999). Several soil properties influence these processes.
Modeling of these properties and comparing the outcome with a reference situation may predict in which areas the soil natural attenuation is
optimal and where it is not. The natural attenuation capacity that is modeled in this study must be seen as a measure that describes the
‘biodegradation capacity’ of the soil, including biodegradation of all types of contaminants.
For this, we used aspects of the ‘Digital Soil Mapping’ procedures (McBratney et al., 2003). First, we developed a model to quantify the natural
attenuation capacity of the soil relative to an optimal situation. Several biotic and abiotic parameters were selected to be incorporated in the
model (Section 2.1). Maps of some of these parameters are readily available in literature, but some others needed to be modeled. Therefore,
measurements were needed to feed these models. In Section 2.2 we will explain how measurements in BiSQ were done and in Section 2.3 is
explained how models were developed to map some of the non-available input parameters. Finally, mapping procedures are explained in Section
2.1. Development of the preliminary model for natural attenuation
The evaluation of contribution of soil organisms to ecosystem services is of conceptual and descriptive nature (Luck et al., 2009). Quantitative
relations are not yet available, and best professional judgment (BPJ) was used to select the appropriate biotic indicators, like the procedure as
described by Rutgers et al. (2012). We used three microbial indicators from the database as proxy indicators of the capacity of the soil to perform
its natural attenuation capacity (NAC): 1) functional microbial activity (FMA), 2) potential carbon mineralization rate (PotC) and 3) potential
mineralization rate of organic nitrogen (PotN). Also abiotic soil parameters may serve as indicators for ecosystem services, but again quantitative
relations are not available. BPJ was used to produce a selection of abiotic indicators in the database as proxy indicators for the NAC, i.e. soil organic
matter (SOM) content, pH of the soil and phosphorus content (PAL, i.e., P2O5 extracted with a solution of ammonium lactate).
The level of the NAC was made a function of the ‘condition of the soil’ according to one equivalent equation as used by Rutgers et al. (2012),
with the default syntax:
+ log
+ log
+ log
+ log
+ log
B log FMA
NAC = 10
j j
j j
j j
j j
j j
and in few grid cells:
+ log
+ log
+ log
− log
− log
B log pH
NAC = 10
j j
j j
j j
j j
j j
In the default syntax (Eq. 1), the more the observed and calculated values of these six soil properties differed from ‘reference values’, the lower
the NAC. In some grid cells of The Netherlands, deviations from default syntax were necessary (Eq. 2). In fact, as soon the observed functional
activity value was lower than the reference value, it was considered to be ‘better’ and therefore did not result in a lower natural attenuation
capacity. The same can occur when the observed soil organic matter was higher than the reference value, as more SOM was considered to be
‘better’ for the natural attenuation capacity. Seen the opposite trends of FMA and SOM in comparison to the other four terms, these two terms
were entered into the numerator of Eq. (2) with a negative sign, in contrast to the default syntax of Eq. (1). More information on this switching
syntax of the equation can be found in Rutgers et al. (2012).
2.2. Measurements in BiSQ
Maps of some input parameters were readily available. The soil properties pH and soil organic matter were taken from De Vries (1999) and
converted to a 100 × 100 m resolution. The PAL content in the soil was derived via pedotransfer functions from Mulder et al. (2009, 2011) and
converted to 100 × 100 m resolution. Maps of the other input parameters were derived by Generalized Linear Regression modeling (see Section
2.3). To feed these models, a dataset with observations from the Netherlands Soil Monitoring Network was used (BiSQ; Rutgers et al., 2009).
The data with abundance and diversity for several soil biota collated in the BiSQ, together with abiotic characteristics were collected from
1999 to 2005. These data comprises information from 283 unique locations predominantly found in The Netherlands, which can be grouped in 5
soil types and 5 land-use types (Table 1). The samples are geographically identified by the latitude and longitude of the sampling locations (Dutch
triangulation coordinates in meters from origin) and the date of sampling. The locations in the database are represented by farm level data,
typically 5–50 ha, or less in the case of nature areas.
Soil parameters included: pH, organic matter fraction and phosphor content of the topsoil (PAL, i.e., P2O5 extracted with a solution of
ammonium lactate). Table 2 gives the measured variables in the physico-chemical part of the dataset that are used in the present study, together
with their value distribution over the 283 sampling sites.
The FMA is the amount of soil sample required to reach 50% activity in the Biolog multiwell plates (Rutgers et al., 2006, 2009). When bacterial
density and diversity are low, a larger amount of soil sample will be needed to reach the 50% average coloring level in the Biolog ECOplates.
Potential carbon and nitrogen mineralization processes were determined simultaneously by incubating the samples in airtight pots. Oxygen and
carbon dioxide concentrations were regularly measured between week 1 and week 6 using gas chromatography (Bloem and Breure, 2003).
H.J. van Wijnen et al. / Science of the Total Environment 415 (2012) 49–55
Table 1
Categories of soil and land use, together with their representation in the 283 sampling sites.
Value range
Number of sites
% of sites
Fluvial clay
Silty loam
Arable farming
Semi-natural grassland
0 or 1
0 or 1
0 or 1
0 or 1
Land use
0 or 1
0 or 1
0 or 1
0 or 1
Not specified as a predictor in the regression. If 0 (zero) for all of the other categories within the same type then identified as clay or forest, respectively.
Table 2
Physico-chemical variables used in the present study, with their value distribution over 283 sampling sites.
Organic matter
(% dw)
pH (KCl)
(mg P2O5/100 g)
Carbon mineralization rate was calculated from the CO2 evolution (respiration) between week 1 and week 6 (mg C kg − 1 week − 1). The increase
in mineral nitrogen (NH4 and NO3) in the soil during 6 weeks was used to calculate the potential nitrogen mineralization rate
(mg N kg − 1 week − 1). As measurements were carried out with homogenized sieved soil at a constant temperature of 20 °C and 50% water
holding capacity, the method is considered to reflect a potential optimum for the mineralization processes.
2.3. Modeling of some input parameters
We developed models to describe soil processes that are important for the natural attenuation of pollutants. The models were restricted to the
parameters from the nationwide survey, in order to enable the mapping of these properties. It makes no sense to develop mapping procedures
without availability of relevant data with some kind of spatial representation. We used soil data from the previous paragraph to calibrate the
models (Rutgers et al., 2009). The data of BiSQ and some other soil properties were used for modeling the FMA, PotC and PotN.
Generalized Linear Regression (GLM) models of the Gaussian family (McCullagh and Nelder, 1989) are used to relate the three functional
responses of the soil microbial community to land use, soil type and abiotic characteristics. The models to be calibrated by the available data were
all formulated to be according to the syntax:
Responses ð½FMA or ½PotC or ½PotNÞ = intercept + a ½Loss + b ½fluvial clay + c ½peat + d ½sand + e ½silt loam + f ½pasture
+ g ½arable fields + h ½semi natural grassland + i ½heathland + j ½longitude
+ k ½longitude2 + l ½latitude + m ½latitude2 + o ½OM + p ½OM2 + q ½pH + r ½pH2
+ s ½PAL + t ½PAL
The quadratic terms for the scalar predictors in the regression formula allow for predicting non-linear response behavior as inflicted by
optimum and minimum conditions.
The regression models were calibrated using a stepwise procedure based on the Bayesian Information Criterion (BIC) (Schwarz, 1978). This is done in
order to restrict the addition of terms to those that have a significant (pb 0.05) contribution to the overall model, making the full model highly significant.
Calculations were conducted using S-Plus 2000 (MathSoft, Cambridge, MA). Subsequently, the calibrated regression formulae were used to generate a
continuous map of the microbial responses by substituting continuously mapped values for the model predictors in the calibrated regression formulae.
The models derived were:
FMA = 6:032•10 + −5:741•10 ½pasture + −5:494•10 ½arable farming + −5:870•10 ½semi natural grassland
+ −5:085•104 ½heathland + −8:885•101 ½PAL + 7:4•10−1 ½PAL2
PotC = 4:688•10 + −4:607•10 ½arable farming + 1:347•10 ½peat + ð6:06 ½OMÞ
PotN = 2:43 + ð5:11 ½pastureÞ + ð7:10 ½semi−natural grasslandÞ + 1:511•10 ½peat + 1:73•10
+ −1:1•10−1 ½pH2 :
H.J. van Wijnen et al. / Science of the Total Environment 415 (2012) 49–55
The explanatory capacity of all three models is relatively high, with an explained deviance of 76, 72 and 54%, respectively. Seen the variance
inflation factors (VIF), predictors' co-linearities are unlikely (VIF b 10). VIFs are calculated and interpreted according to Kline (1998) and O'Brian
2.4. Mapping
All maps were designed at a resolution of 100 × 100 m. A map with land-use types was derived from Hazeu et al. (2010). It is a 25 × 25 m
resolution map, showing 39 categories of land use. For this analysis, it was converted to a 100 × 100 m resolution map. Soil types were derived
from the 1:50,000 resolution vector soil map of The Netherlands from De Vries et al. (2003) and converted to a 100 × 100 m resolution grid map.
As explained earlier, the maps of the soil properties pH and soil organic matter were taken from De Vries (1999) and converted to a
100 × 100 m resolution. The PAL content in the soil was derived via pedotransfer functions from Mulder et al. (2009, 2011) and converted to
100 × 100 m resolution.
Maps of the soil properties FMA, PotC and PotN were calculated from the models that are described above. Reference maps were made based
on the data in Rutgers et al. (2008). The reference data represent a soil system which can be considered as the healthiest with respect to the land
management within a given soil and a chosen land use. It may also be referred to as a ‘maximum ecological potential’ (MEP) within a given soil
and a chosen land use (Rutgers et al., 2012). For each combination of land use and soil type, a MEP value for each soil property was taken from
Rutgers et al. (2008).
3. Results and discussion
The functional microbial activity (FMA) of the soil shows the
amount of soil required to convert 50% of the total substrate. It is
generally low in agricultural areas, especially on peat, and relatively
high in nature areas (Left panel in Fig. 1). Compared to the reference
‘optimal’ situation in Fig. 1 (Right panel), it shows that in some
agricultural areas the actual functional activity is higher than desired.
Most grasslands are devoted to livestock (dairy and non-dairy
cattle) and occur on sandy soils. Among the agricultural land-use
types, the arable farms are mainly located on marine clay, whereas
grasslands can be found mainly on sandy soils in the eastern part of
the country. Arable farms mainly occur on soils that are suitable for
arable crops. To a certain extent, even the widespread livestock farms
occur on soils that are less suitable for high-productivity grasslands, as
most pastures are located on nutrient-poor soils.
Similar analyses were done for the other five biotic and abiotic soil
properties (pH, OM, PAL, PotC and PotN). These soil properties were
also compared to the reference situation (Fig. 2). The best independent predictor, soil pH, fluctuates 5 orders of magnitude, followed by
OM, whose content increases by a factor of 30. Within a particular soil
type, its variation generally is less than 2 orders, except for maritime
(Ca 2+-enriched) sand dunes, where the soil pH range is much smaller
than that of inland areas, and for river clay, where its variation is over
3 orders of magnitude. In some areas, the soil property can be lower
than the reference, whereas in other areas the same soil property can
be equal or even higher than for the reference situation. When the
deviation from the reference is high, the potential for improvement of
that soil property is high as well.
The natural attenuation of pollutants appears to be relatively low
in nature areas and in most arable fields on sandy soils, although it is
relatively high in arable fields on clay and in grasslands on sandy soils
Fig. 1. Actual soil functional activity (left-hand side) and potential soil functional activity (right-hand side).
H.J. van Wijnen et al. / Science of the Total Environment 415 (2012) 49–55
Fig. 2. Potential for improvement of six soil parameters.
(Fig. 3; Table 3). Intermediate values are found in grasslands on peat
and Löss. All together, less than 20% of the arable farming systems is
situated on unsuitable soil, but over 40% of nature is situated on soil
unsuitable for rural purposes (cf. Mulder et al., 2005).
Fig. 3 shows a remarkable lack of fit between the current land use
and the soil properties of Fig. 2 in large parts across The Netherlands
(Mulder et al., 2005). This implies that soil management regimes,
historical land use and geographical situation determine soil
H.J. van Wijnen et al. / Science of the Total Environment 415 (2012) 49–55
Fig. 3. Ecosystem service “natural attenuation of pollutants”, based on six soil properties.
properties. This situation is widely accepted in another densely
populated European country, the United Kingdom, where a flourishing
tradition of underpinned environmental planning is rapidly growing
(e.g., Haygarth and Ritz, 2009, and references therein).
An improved link between current planning systems and GIS
analyses like those shown here offers several possibilities to inform
and eventually regulate agricultural land use in The Netherlands.
Policy directed towards specific regions, which is a growing field in
Table 3
Summary of relative values (0 b proxy b 1) of natural attenuation capacity (NAC) for all
occurring combinations of land-use and soil type (NO = not occurring). A value of 1
indicates that the natural attenuation capacity is equal to the reference, or MEP value.
Land use
Sandy soils
Clay soils
Agricultural grasslands
5332 km2
4132 km2
2963 km2
190 km2
415 km2
4003 km2
4358 km2
1665 km2
98 km2
Arable fields
Semi-natural grasslands
the environmental policy, and active regulation of key ecosystem
services, offers many possibilities to integrate or to differentiate land
uses, with cost-benefit opportunities to match occurring soil quality
(with related ecosystem services), see with the policy-demanded land
use and policy-required tradeoffs (Daily et al., 2009; Egoh et al., 2008;
Naidoo and Ricketts, 2006; Nelson et al., 2009; Swetnam et al., 2011).
There are several uncertainties and limitations related to the use of
BPJ, reference values, GLM models and the NAC model. These uncertainties need to be addressed in order to appreciate the possibilities of
this methodology. Also, validation is needed for the final data. The maps
in this study must be seen as a first attempt to show which areas have a
relatively suboptimal ecosystem service capacity and which areas have
an optimal capacity.
The method shown in this paper to map the natural attenuation
may also be used for mapping other ecosystem services. These indexes of ecosystem-service performance (EPX) can be mapped in a
similar way. In addition, a map of an integration of all EPX values is
possible, showing integrated soil quality (Rutgers et al., 2012). The
challenge remains to select the most appropriate soil properties and
develop concomitant models to estimate a realistic NAC and other
ecosystem services using BPJ. This is both true for using existing data
such as present in the BiSQ database as for new proxy indicators.
Balmford A, Rodrigues ASL, Walpole M, Ten Brink P, Kettunen M, Braat L, et al. The
economics of biodiversity and ecosystems: scoping the science. Cambridge, UK:
European Commission; 2008.
H.J. van Wijnen et al. / Science of the Total Environment 415 (2012) 49–55
Bateman IJ, Ennew C, Lovett AA, Rayner AJ. Modelling and mapping agricultural output
values using farm specific details and environmental databases. J Agr Econ
Bloem J, Breure AM. Microbial indicators. In: Markert BA, Breure AM, Zechmeister HG,
editors. Bioindicators and biomonitors. Principles, concepts and applicationsAmsterdam, The Netherlands: Elsevier; 2003. p. 259–82.
Costanza R, D'Arge R, De Groot R, Farber S, Grasso M, Hannon B, et al. The value of the
world's ecosystem services and natural capital. Nature 1997;387:253–60.
Daily GC, Polasky S, Goldstein J, Kareiva PM, Mooney HA, Pejchar L, et al. Ecosystem
services in decision making: time to deliver. Front Ecol Environ 2009;7:21–8.
De Vries F. Karakterisering van Nederlandse gronden naar fysisch-chemische
kenmerken. Report 654 (In Dutch), Staring Centrum, Wageningen, The Netherlands; 1999.
De Vries F, De Groot WJM, Hoogland T, Denneboom J, De Bodem van Nederland digitaal.
Toelichting bij inhoud, actualiteit en methodiek en korte beschrijving van
additionele informatie. Report 811 (In Dutch), Alterra, Wageningen, The Netherlands; 2003.
Eade JDO, Moran D. Spatial economic valuation: benefits transfer using Geographical
Information Systems. J Environ Manage 1996;48:97–110.
Egoh B, Reyers B, Rouget M, Richardson DM, Le Maitre DC, Van Jaarsveld AS. Mapping
ecosystem services for planning and management. Agr Ecosyst Environ 2008;127:
EPA. Use of monitored natural attenuation at superfund, RCRA corrective action, and
underground storage of tank sites. US Environmental Protection, Office of solid
waste and emergency response; 1999 http://www.epa.gov/swerust1/directiv/
d9200417.pdf. Directive 9200. 4-17P.
Farber SC, Costanza R, Wilson MA. Economic and ecological concepts for valuing
ecosystem services. Ecol Econ 2002;41:375–92.
Grêt-Regamey A, Bebi P, Bishop ID, Schmid WA. Linking GIS-based models to value
ecosystem services in an Alpine region. J Environ Manage 2008;89:197–208.
Haygarth PM, Ritz K. The future of soils and land use in the UK: soil systems for the provision
of land-based ecosystem services. Land Use Policy 2009;26:S187–97. (Suppl.).
Hazeu GW, Schuiling C, Dorland GJ, Oldengarm J, Gijsbertse HA. Landelijk Grondgebruiksbestand Nederland versie 6 (LGN6). Vervaardiging, nauwkeurigheid en
gebruik. Report 2012 (In Dutch). Wageningen, The Netherlands: Alterra; 2010.
Jeffery S, Gardi C, Jones A, Montanarella L, Marmo L, Miko L, et al, editors. European atlas
of soil biodiversity. Luxembourg, Luxembourg: European Commission, Publications
Office of the European Union; 2010.
Kline RB. Principles and practice of structural equation modeling. New York, NY, USA:
Guilford Press; 1998.
Luck GW, Harrington R, Harrison PA, Kremen C, Berry PM, Bugter R, et al. Quantifying
the contribution of organisms to the provision of ecosystem services. Bioscience
McBratney AB, Mendonça Santos ML, Minasny B. On digital soil mapping. Geoderma
McCullagh P, Nelder JA. Generalized linear models. 2nd edition. London, UK: Chapman
and Hall; 1989.
Metzger MJ, Schröter D, Leemans R, Cramer W. A spatially explicit and quantitative
vulnerability assessment of ecosystem service change in Europe. Reg Environ
Change 2008;8:91–107.
Millennium Ecosystem Assessment. Ecosystems and human well-being: biodiversity
synthesis. Washington, DC: World Resources Institute; 2005.
Mulder C, Den Hollander HA, Vonk AJ, Rossberg AG, Jagers op Akkerhuis GAJM, Yeates
GW. Soil resource supply influences faunal size-specific distributions in natural
food webs. Naturwissenschaften 2009;96:813–26.
Mulder C, Van Wezel AP, Van Wijnen HJ. Embedding soil quality in the planning and
management of land use. Int J Biodiv Sci Manage 2005;1:77–84.
Mulder C, Boit A, Bonkowski M, De Ruiter PC, Mancinelli G, Van der Heijden MGA, et al.
A belowground perspective on Dutch agroecosystems: how soil organisms interact
to support ecosystem services. Adv Ecol Res 2011;44:1–82.
Naidoo R, Ricketts TH. Mapping the economic costs and benefits of conservation. PLoS
Biol 2006;4:e360.
National Research Council. Natural attenuation for groundwater remediation. Washington, DC, USA: National Academy Press; 2000.
Nelson E, Mendoza G, Regetz J, Polasky S, Tallis H, Cameron DR, et al. Modeling multiple
ecosystem services, biodiversity conservation, commodity production, and tradeoffs at landscape scales. Front Ecol Environ 2009;7:4–11.
O'Brian RM. A caution regarding rules of thumb for variance inflation factors. Qual
Quant 2007;41:673–90.
Rittmann BE. Definition, objectives, and evaluation of natural attenuation. Biodegradation 2004;15:349–57.
Rutgers M, Breure AM, Insam H. Substrate utilization in Biolog™ plates for analysis of CLPP.
In: Bloem J, Benedetti A, Hopkins DW, editors. Microbiological Methods for Assessing
Soil Quality. Wallingford, Oxfordshire, UK: CAB International; 2006. p. 212–27.
Rutgers M, Mulder C, Schouten AJ, Bloem J, Bogte JJ, Breure AM, et al. Soil ecosystem
profiling in the Netherlands with ten references for biological soil quality. Report
607604009. Bilthoven, The Netherlands: National Institute for Public Health and
the Environment, RIVM; 2008.
Rutgers M, Schouten AJ, Bloem J, Van Eekeren N, De Goede RGM, Jagers op Akkerhuis
GAJM, et al. Biological measurements in a nationwide soil monitoring network. Eur
J Soil Sci 2009;60:820–32.
Rutgers M, Van Wijnen HJ, Schouten AJ, Mulder C, Kuiten AMP, Brussaard L. A method
to assess ecosystem services developed from soil attributes with stakeholders and
data of four arable farms. Sci Total Environ 2012;415:39–48. (this issue).
Schwarz G. Estimating the dimension of a model. Ann Stat 1978;6:461–4.
Swetnam RD, Fisher B, Mbilinyi BP, Munishi PKT, Willcock S, Ricketts T, et al. Mapping
socio-economic scenarios of land cover change: a GIS method to enable ecosystem
service modeling. J Environ Manage 2011;92:563–74.
Troy A, Wilson MA. Mapping ecosystem services: practical challenges and opportunities in linking GIS and value transfer. Ecol Econ 2006;60:435–49.