You are on page 1of 9

Journal of Environmental Management 126 (2013) 113e121

Contents lists available at SciVerse ScienceDirect

Journal of Environmental Management


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

Land use impact on water quality: Valuing forest services in terms


of the water supply sector
Julien Fiquepron a, Serge Garcia b, c, *, Anne Stenger b, c
a

CNPF e IDF, Maison de la Fort, 11 rue de la Commanderie, 54000 Nancy, France


INRA, UMR 356 Laboratoire dEconomie Forestire, 54000 Nancy, France
c
AgroParisTech, ENGREF, Laboratoire dEconomie Forestire, 14 rue Girardet, 54000 Nancy, France
b

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 6 July 2011
Received in revised form
11 February 2013
Accepted 1 April 2013
Available online 14 May 2013

The aim of this paper is to quantify the impact of the forest on raw water quality within the framework of
other land uses. On the basis of measurements of quality parameters that were identied as being the
most problematic (i.e., pesticides and nitrates), we modeled how water quality is inuenced by land uses.
In order to assess the benets provided by the forest in terms of improved water quality, we used variations of drinking water prices that were determined by the operating costs of water supply services
(WSS). Given the variability of links between forests and water quality, we chose to cover all of France
using data observed in each administrative department (France is divided into 95 dpartements),
including a description of WSS and information on land uses. We designed a model that describes the
impact of land uses on water quality, as well as the operation of WSS and prices. This bioeconomic model
was estimated by the generalized method of moments (GMM) to account for endogeneity and heteroscedasticity issues. We showed that the forest has a positive effect on raw water quality compared to
other land uses, with an indirect impact on water prices, making them lower for consumers.
2013 Elsevier Ltd. All rights reserved.

Keywords:
Water supply services
Forest
Land use
Water price
Quality
Bioeconomic model

1. Introduction
The purpose of this article is to quantify the impact of the forest
on the quality of water and its economic value. Forests have an
extensive root network and a great ability to generate porous and
ltering soils. Recycling, especially of nitrogen, is important. Under
forest cover, nitrate levels are low (Jussy et al., 2002) and similar
results are also observed for various pollutants (e.g., pesticides).
Our hypothesis is that raw water from catchment areas with a large
portion of forests is of higher quality, thus reducing the need for
treatment of drinking water and, as a result, the associated prices of
drinking water supply. In contrast, runoff from agricultural lands is
the main cause of water pollution (Hascic and Wu, 2006), and
nitrication is greater in an agricultural environment. The presence
of agricultural land in the area surrounding the water supply service (WSS) may thus lead to sophisticated and costly treatments.
While considerable research has been devoted to quantifying the
physical extent of the impact of the forest on water quality, few
studies have attempted to estimate the economic value of the impact
of forests on the quality and still fewer on the value of forests in
* Corresponding author. Laboratoire dEconomie Forestire, 14 rue Girardet,
54000 Nancy, France. Tel.: 33 (0)3 83 39 68 69; fax: 33 (0)3 39 06 45.
E-mail address: serge.garcia@nancy.inra.fr (S. Garcia).
0301-4797/$ e see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jenvman.2013.04.002

supplying water for human consumption (Nez et al., 2006; Biao


et al., 2010; Abildtrup et al., 2013). Forest land use is normally
associated with the protection of water resources from contamination and the reduced cost of drinking water supply (Abildtrup and
Strange, 2000; Willis, 2002; Ernst et al., 2004). The relative impacts of alternative land uses on water quality have already been
studied (Hascic and Wu, 2006; Langpap et al., 2008). The objective of
such studies was to analyze how water quality affects watershed
ecosystem health at different spatial scales and whether land-use
changes exacerbate these impacts. Whereas the impacts in those
studies are estimated from variations in terms of quality indicators
and the number of species at risk, our article seeks to measure the
effect of land uses on water quality indicators (for drinking water
uses) and on the price of water supply. The objective in ne is to
estimate the economic value of the ecological service provided by
forests on the quality of raw water used for drinking water supply.
The French Forest Orientation Law (Law no. 2001-602, 9/7/2001)
recently recognized the role of the forest in the supply of a number
of non-market services, including protection of water. Moreover,
with the application in France of the Water Framework Directive
(Directive, 2000/60/CE), actors involved in the water sector are now
aware of the importance of protection and prevention for water
resources intended for drinking water. Although it does not
explicitly mention forest areas, the Water and Aquatic Environment

114

J. Fiquepron et al. / Journal of Environmental Management 126 (2013) 113e121

Law (Law no. 2006-1772 30/12/2006) opens up many possibilities in


terms of collaboration with and contributions by forest owners
related to the forest service associated with water quality. A look at
some of the big cities in the world (e.g., New York) well illustrates
that the links between forest and water are real issues underlying
the success of public policies on water quality and water prices
(Johnson et al., 2000).
The objective of our study is to assess the role of land uses, and
more specically of the forest on the quality of drinking water and its
price at the national level. The interest of this approach is to estimate
whether this role is signicant throughout the French territory,
without being restricted to local effects that could challenge the
transfer of results. This approach might be complemented by more
detailed analysis at the scale of watersheds, as undertaken in the
Vosges department by Abildtrup et al. (2013). Nevertheless at this
scale, the richness of data is much lower and it is difcult to gather a
sample of representative watersheds at the national level. While
there are studies on the scale of water catchments, they are based on
catchments inclined to encounter problems. However this implies a
sample bias, with watersheds predominantly covered by very little
forest. Local features and observation levels as well are both limiting
factors when transferring results from one site to another (Ranger
et al., 2007; Gove et al., 2001; Kiersch and Tognetti, 2002).
We have thus conducted our study with departmental data from
a rich dataset but highly aggregated. It is the best way to achieve
results that are representative at national level. Effects assessed at
the national level could inspire public policy, knowing that the aggregation of data at the national level would tend to limit the effects
of land use on water quality. Given the variability of the links between the forest and water quality, we decided to cover all of France
using data observed in the different administrative departments. We
therefore collected both data related to water supply management
and data on land use and land cover, including the proportion of
woodland at the department level. The hypothesis of our model is
that land uses have a direct affect on raw water quality and an indirect effect on the price of water. The main results show an expected
positive effect of the forest on water quality with respect to other
land uses, i.e., departments with a relatively high proportion of
forest cover have better water quality and lower water prices.
In order to understand land use impact on water price and
quality, we present some elements in the following section on the
links between forest and water, focusing on the qualitative aspect of
water, followed by the economic model and the estimation results.

used. Regarding the effect of the type of forest stand on nitrate


losses, results can also be seen as somewhat contradictory in the
literature. And mixing species, especially broadleaved and conifers,
may lessen the disadvantages of some mono-specic stands such as
plantations made up solely of spruce (Knoke et al., 2008).
Compared to other types of land uses, the forest differentiates
itself by its high level of aerial development, its extensive root
network and its ability to generate porous and ltering soils. The
efcient functioning of biological cycles in forest stands optimizes
the use of nutritional elements from the soil (Ranger et al., 1995),
making it possible to obtain a large biologic production, often from
poor soils, and strongly limiting mineral element leakage out of the
soil/plant system. Recycling, especially of nitrogen, is important in
forests. Nitrate levels are therefore low under forest cover
(Gundersen, 2007). Vittel, a small town in eastern France, is a good
example since it subsidizes agriculture to reduce chemical inputs
and purchases forested land around the private catchment area for
the purpose of improving water quality. Under the forest cover on
the plain near Vittel, inltration waters alone do not exceed 2 mg/l
of nitrates (Benot and Papy, 1997), whereas the quantity of nitrates
on agricultural land is signicantly higher (Table 1). This trend can
be veried at the catchment area level. An afforestation rate of 30%
in agricultural watersheds should be enough to produce water with
a nitrate content lower than the norms for drinking water in Europe
(50 mg/l) (Benot et al., 2002).
Compared with other land uses such as urbanization or arable
land, the forest generates lower runoff coefcients (Adhikari et al.,
2002; Sikka and Selvi, 2005). Cultivated land releases more than
ve times more sediment into the watercourse than wooded areas,
whereas wooded areas have a tendency to occupy the most uneven
areas (Brown and Binkley, 1994). Thus, the forest contributes to
protecting the land, tending to favor inltration and reducing rapid
ow at the surface. The forest generally makes it possible to limit
sediment ow and turbidity.
Some wooded formations clearly have a purication role:
riparian forests, alluvial forests and hedged farmland are cases in
point. The root systems of riparian forests and alluvial forests have a
ltering role and trap nutritive elements (nitrogen, potassium,
phosphorus) as well as some toxic elements (Broadmeadow and
Nisbet, 2004). However, it should be mentioned that these conditions, which are ideal for denitrication, are not specic to the
forest. Grassland can also offer very favorable conditions. Hedges
and other linear wooded areas constitute effective lters as well.

2. Links between the forest and water quality

3. Empirical approach

The inuence of the forest on the quality of the water resource is


all the more important in that it constitutes one of the main types
of land cover, at least in temperate zones like France, where the
metropolitan afforestation rate is 28.5% (Agreste, 2004). This inuence of the forest can be distinguished between effects related to
the existence of forests and those related to forest management, a
distinction that was notably made in a reference document produced by the US Forest Service (Brown and Binkley, 1994). Forest
management requires different types of operations such as creating
stands, cleanings, successive thinnings and timber transport that
can cause considerable disturbance to the soil, depending on the
precautions taken while the work is carried out. Forest management is not neutral in terms of water quality, but many factors tend
to attenuate harmful effects, particularly the fact that human
interventions are less frequent in this sector than in agriculture.
Land management in forests is less intensive than for agriculture, and interventions are less frequent due to the long-term
rotations in forests. The use of chemicals in forest areas is very
limited. Agropharmaceutical products and fertilizers are rarely

The approach used to process the data was based on an


estimation of a simultaneous equations model that included one
equation related to water price, two equations related to raw water
quality via pesticide and nitrate indicators, and one equation
related to water supply service (WSS) management regime.
Table 1
Nitrate levels in water collected by porous-cup lysimeters at a depth of 1.10 m under
different types of land cover in Lorraine.
Land cover

NO
3  in water at a
depth of 1.10 m in mg/l

Forests
Cut elds
Pastures
Temporary grassland
Winter wheat
Rape seed
Spring cereals
Maize as a fodder crop

2
19
31
28
46
62
120
126

Source: Benot and Papy, 1997.

J. Fiquepron et al. / Journal of Environmental Management 126 (2013) 113e121

We now describe the theoretical hypothesis that our model is built


on and then present the econometric method to estimate it.

115

In order to assess the impact of the forest and other land uses on
the quality of raw water and the price of drinking water P, we rst
assumed that the price applied by the WSS could be explained by
the characteristics of the service (including the delivered drinking
water volume VOL or the length of water supply network LENG),
grouped together in the vector X, and that it reects the average
cost of water supply. The quality of raw water, which is made
drinkable (if necessary) through the use of adapted water purication facilities, was also assumed to have a direct effect on P, but
was considered to be a function of land use at the same time.
A variation of the quality of raw water may increase the cost of
treatment and increase P as a result. The price of water is thus
indirectly affected by different land uses since they constitute an
essential factor underlying the quality of raw water. Variables
relating to land use are represented by the vector Z. Two quality
indicators (pesticides and nitrates), referred to as q1 and q2, are
used, and the quality equations are respectively written as:

information maximum likelihood (FIML) method can be used, or


methods with instrumental variables (IV).
Compared with IV methods, the FIML method has the advantage
of not requiring instrument specications, but instead calls for a
complete system of equations to be dened for each endogenous
variable. Moreover, like all methods that use maximum likelihood,
it supposes that the error terms have a normal multivariate distribution. In contrast, methods with IV (2SLS, two-stage least
squares, or 3SLS, three-stage least squares) produce no hypotheses
on error distribution.
In the context of our study where the potential for a high degree
of heterogeneity between departments existed, mainly because of
variables that were not observed due to the particular scale of our
observations, the hypothesis of different error variance seemed the
best adapted. For equation systems with heteroscedastic errors, the
generalized method of moments (GMM) would seem to be the best
method for efcient estimations of the parameters. The basic idea
underlying estimation by GMM consists of specifying moments
without specifying the parametric form, in order to construct
conditions of orthogonality that will be used to identify the equation system. The equation system can be rewritten in the following
compact form:

q1i a10 a1X Xi a1Z Zi 1i

(1)

Y GR

q2i a20 a2X Xi a2Z Zi 2i

(2)

3.1. A simple bioeconomic model

where i is the department index and 1i and 2i are error terms that
account for non-observed shocks on quality indicators.
It has often been demonstrated that the management regime of
the WSS has a considerable impact on the price they propose, but
there is also a problem of endogeneity associated with this variable.
The choice of management regime can itself be accounted for by
certain observable characteristics of the service as well as by many
non-observable factors (Boyer and Garcia, 2008; Carpentier et al.,
2006). In particular, the difference in price, which is unfavorable
for delegated management, can be explained by the more difcult
operating conditions (including low quality raw water), precisely
the reason why French communes (municipalities) have chosen
this management regime. In order to take this factor into account in
our model, we used the variable DEL, which represents the proportion of the population supplied by one unit of distribution (UDI)
when management is delegated. Hence, the relationship between
water price and management regime can be modeled by the
following two equations:

Pi b0 bX Xi bZ Zi b1 q1 b2 q2 bD DEL Pi

(3)

DELi d0 dX Xi dZ Zi d1 q1 d2 q2 D
i

(4)

where Pi and D
i are stochastic disturbances to the ith departments
water price and mode of management.
Finally, the simultaneous equation system consists of equations
that dene the quality level of the raw water, the price equation,
and an equation related to the management regime used by the
WSS. Assumptions concerning the various error terms in the
equation system are dened in the next section.
3.2. The estimation method
The equations in the model were estimated simultaneously in
order to account for any possible correlations between errors
in each equation associated with common non-observable factors.
In order to take the interdependence of the system and the
presence of endogenous variables into account, either a full-

(5)

where Y is the dependent variable vector, R is the explanatory


variable matrix, G is the associated parameter vector and is the
error vector.
By using the expression W to represent the instrument vector,
all conditions of orthogonality (or of moment) can be written as
follows:

EW 0

(6)

In practice, these conditions are approached by their empirical


equivalents that, by using a positive semi-dened weighted matrix,
enable us to construct the GMM criterion to be minimized in order
to obtain the estimator for G. A two-step method is used, which
consists rst of estimating the equations separately with the 2SLS
method to construct the weighted matrix from the residuals, and
then minimizing the GMM criterion constructed from the weighted
matrix calculated in the rst step. The GMM estimator based on this
method is robust for all forms of heteroscedasticity and efcient.
Since the model to be estimated contains endogenous regressors,
instrumental variables must then be provided to use the methods
presented above. The set of instrumental variables includes the
explanatory variables of the model, considered as exogenous, as well
as the variables that are external to the model. There are two
endogenous regressors in our model: the variable representing the
choice of management regime for water supply services (DEL) that is
considered to have an impact on water prices, as well as the variable
VOL, which corresponds to the demand of drinking water from users.
Population (POP) is a good instrument for this latter variable. Moreover, the variable DEL is explained by the proportion of the population with delegated management of wastewater services (WW_DEL),
which is not correlated with the drinking water prices. This is
the exclusion condition that allows us to properly identify parameters of the water price equation and of the equation determining
the choice of management mode for drinking water supply services.
Once the parameters have been estimated, it is then necessary
to ensure that the instruments chosen verify the moment conditions constructed on the basis of the hypotheses produced by the
econometrician. To do this, a test is carried out on the overidentifying constraints, based directly on the GMM criterion. If
L expresses the number of moment conditions (corresponding to
the number of instruments multiplied by the number of equations)

116

J. Fiquepron et al. / Journal of Environmental Management 126 (2013) 113e121

and K is the total number of parameters to be estimated, the criterion follows a c2 distribution with L  K degrees of freedom under
the null hypothesis for validity of the moment conditions. This is
known as the Hansen specication test with L  K overidentifying
constraints for which it is not necessary to identify the parameters
but that transmit information about the model specication. In
particular, if the instruments contain variables that are not exogenous, the Hansen test will detect them and the whole set will then
have to be modied.
4. Data
The decision to use the department as the scale of reference to
analyze the data has both advantages and limitations. The main
limitation lies in the administrative nature of the way the departments are divided up. Using the natural boundaries of the
catchment areas would be better adapted to our analysis of the
impacts of land use on water quality. However, analysis on a scale
more accurate than that of the department would have posed a
dual problem linked with the availability of data at this scale and
the homogeneity of data at the national level. The departmental
scale enabled us to assemble a rich and homogeneous set of data
(76 variables), well adapted to the analysis of variations in water
quality and price across the entire country, with just two exceptions. We decided not to use the data for Paris. The origins of the
water resources used in this area were too different and remote.
We also excluded Corsica from the study because some data were
unavailable. The variables that were used are described in Table 2
and the descriptive statistics are shown in Table 3.
4.1. Dependent variables
To establish the price of drinking water, we used data from the Eau
2004 survey carried out by IFEN and SCEES. This survey-based study
covered the calendar year 2004 and was carried out in 5183 communes in France. The sample was stratied by department and
commune size. Survey rates ranged from exhaustive sampling for
communes of over 10,000 inhabitants to 1/20 for communes with
fewer than 400 people, which were much more numerous.

Table 3
Descriptive statistics of variables (at the department level).
Variable

Unit

Mean

Standard
deviation

Min

Max

P
DEL
q1
q2
VOL
LENG
DENS
WATER
UG
POPMAX
FOREST
GRASS
CULTURE
VIARMA
MOUNT
PIG
BOVINE
POP_UDI
WW_DEL

V/m3

1.42
0.61
34.74
16.32
58.66
9.23
0.33
35.91
71.82
130.21
27.84
22.75
29.32
2.84
22.50
0.01
0.18
8.10
0.47

0.24
0.24
30.03
9.95
41.42
4.46
1.20
13.66
29.39
34.90
13.12
14.42
18.28
4.15
32.65
0.02
0.16
28.03
0.23

1.06
0.01
0
1.90
9.44
1.21
0.01
12.93
5.40
100.71
3.63
0.02
0.02
0.02
0
0
0.0004
0.11
0

2.17
1
100
45.90
230.39
18.87
8.64
74.36
100
273.27
68.42
61.25
69.73
22.10
100
0.14
0.72
204.77
0.93

%
mg/l
1000 m3
1000 km
1000 inhab./km2
cm
%
%
%
%
%
%
%
pigs/ha
Bovines/ha
1000 inhab.

Note: Number of observations 93 departments (Paris and Corsica excluded).

The sample covered 14% of the communes and 68% of the population,
and thus ensured a good level of representativity for the regions and
the departments (Coutellier, 2007). The price of water used in our
study corresponded to the fraction of the household water bill that
covered drinking water (excluding the part covering sanitation).
In France, the average water consumption of a representative
household is around 120 m3 per year. Hence, the average price per m3
was based on a typical bill with a consumption of 120 m3 per user
and per year. Concerning the data from the Eau 2004 survey, we always used data that were weighted according to the stratied sampling, which were therefore representative at the department level.
We decided to interpret quality data measured on raw water
from drinking water catchment areas, although these were fewer in
number than quality data on distributed water. Working on analyses of raw water meant that we avoided a major bias associated
with drinking water treatments that could mask the inuence of
land use on water quality. We chose to concentrate on a four-year

Table 2
Denition of variables (unit of observation department).
Group

Variable

Denition

Year(s)

Source

Y
e
e

P
DEL
q1

2004
2007
2002 to 2005

IFEN SCEESa
Ministry of Healthb
Ministry of Healthb

e
X
e
e
e
e
Z
e
e
e
e

q2
VOL
LENG
DENS
WATER
UG
FOREST
GRASS
CULTURE
VIARMA
MOUNT
BOVINE
PIG
POP_UDI
POPMAX
WW_DEL

Average price of drinking water share of bill for all communes


Proportion of population supplied by a UDI with delegated management
Proportion of raw water ow corresponding to resources that have
exceeded the pesticide level xed for drinking water
Average content of nitrates in tested raw water ows
Delivered drinking water volume
Length of water supply network
Population density
Water recharge (PeETP) monthly average from October to April
Proportion of resources from underground water
Proportion of wooded area
Proportion of permanent grassland
Proportion of cereals, oilseeds and protein crops
Proportion of vine, arboriculture and market gardening
Proportion of mountain area
Nitrogen per Livestock Unit of bovines per hectare
Nitrogen per Livestock Unit of pigs per hectare
Number of inhabitants deserved by a UDI
Ratio of maximum population to resident population
Proportion of population with delegated management of wastewater service

2002 to 2005
2004
2004
2005
1961 to 1990
2007
2004
2004
2004
2004
2007
2004
2004
2007
2005
2004

Ministry of Healthb
IFEN SCEESa
IFEN SCEESa
INSEE
LERFoBc
Ministry of Healthb
SCEESd
SCEESd
SCEESd
SCEESd
MAPe
SCEESd
SCEESd
Ministry of Healthb
Ministry of tourism
IFEN SCEESa

e
Instr
e
a
b
c
d
e

IFEN SCEES e water survey.


Ministry of Health, DDASS, Sise-Eaux database.
Laboratoire dEtude des Ressources Fort-Bois (LERFoB).
Annual Agricultural Statistics from the Service Central des Enqutes et tudes Statistiques (SCEES), part of the Ministry of Agriculture and Fisheries.
Ministry of Agriculture and Fisheries e Department of Rural Affairs e classication of communes into underprivileged agricultural zones, mountains and high mountain areas.

J. Fiquepron et al. / Journal of Environmental Management 126 (2013) 113e121

period (2002e2005) for quality data to avoid the bias of a year with
its particular climatic conditions that would affect water quality.
Moreover, this period allowed us to obtain data on small WSS since
they are not monitored every year (especially for pesticides). The
data was obtained from quality monitoring carried out by the
DDASS (French Departmental Health and Social Action Directorate,
now known as the ARS or Regional Health Agency). These data were
aggregated into the Sise-Eaux database (French Ministry of Health,
2008). This national database was used to manage health monitoring in 30,000 water catchment samples intended for human
consumption. Sise-Eaux is thus the most complete database available on drinking water quality. We selected two quality criteria
linked to pesticides and nitrates.
For pesticides, we analyzed the proportion of the raw water ow
requiring treatment to make it drinkable. More precisely, the water
catchments selected were those for which at least one reading on
the raw water had exceeded the pesticide norms set for water
distributed during the period under consideration. A Law of 11
January 2007 sets the limitations and quality references for raw
water and water intended for human consumption: no more than
0.1 mg/l per substance and no more than 0.5 mg/l for total substances. Despite the fact that Sise-Eaux is the most complete
database available on drinking water, the pesticide data had to be
interpreted with caution since less than 60% of the resources were
monitored. Nevertheless, it represents more than 80% of the ow.
Concerning nitrates, we analyzed the nitrate concentrations of
the corresponding raw water ow. This criterion is more accurate
than the former one that we used, which was the proportion of raw
water ow corresponding to the resources that exceeded 50 mg/l
(the drinkability limit for distributed water). Other studies on
drinking water have attempted to integrate qualitative questions
(Bontemps and Nauges, 2009). However, the variables used were
constructed from proxies based on the quality classes for raw surface
water, which dened the type of drinking water treatment to be
used (Council Directive 75/440/CEE of 16 June 1975). We clearly
specied that the nitrate data are not exhaustive although they cover
more than 80% of the resources and nearly 90% of the ow produced.
We analyzed the management regime in the drinking water services according to the proportion of the population supplied by a unit
of distribution, UDI, with delegated management, from the Sise-Eaux
database. A UDI is a continuous unit of distribution pipes in which
water quality is assumed to be homogeneous, managed by a single
operator and belonging to a single contracting authority (commune
or group of communes). A distribution unit cannot straddle several
departments. Maintaining the list of distribution units is the responsibility of the DDASS in each department. Since this database
does not store historical data related to the organization of water
departments, we analyzed data from 2007. This information is still
relevant since the management regime in drinking water services
was relatively stable at the department level over a three-year period.
On the other hand, working with the Sise-Eaux database made it
possible to work with exhaustive data, thus avoiding the bias associated with sampling. We assumed that the UDI that were not run by
delegated management were state-controlled.
4.2. Constraints of the water services and other variables
Data concerning the volume supplied for distribution (VOL) was
taken from the Eau 2004-IFEN SCEES survey, which provides a good
indication of production by the WSS. The length of the drinking
water network (LENG) was also taken from the Eau 2004e IFEN
SCEES survey. The length of the network provides an indication of
the size of the drinking water service. By combining this with information about the population, we obtain an idea of the density of
the network. Since the length of network per user involves variable

117

costs (renewal and maintenance), the density of the network can be


associated with economies of scale. The population density (DENS)
that we used was based on population estimates on 1st January
2005 by INSEE (French National Institute of Statistics and Economic
Studies). Population density is a constraint for water services that
must adapt their production capacity according to demand,
sometimes requiring the use of resources from further aeld or that
are of poorer quality.
We tried to include a variable linked to climate in the model to
account for water decit stress related to the availability of water
resources. We were able to gain access to climate data (precipitationeevapotranspiration) with monthly averages for a 30-year
period. As rainfall is a highly variable data, it is better to integrate this
indicator over long time series rather than a limited number of years.
We expect this data to differentiate among departments together
with respect to the availability of water, more than account for the
rainfall over the years we have studied the water quality. Evapotranspiration was estimated according toTurcs formula (Turc,1961).
The reports were calculated in considerable detail (one result per
50 m) across all of France, and then summarized by department. The
data from the climate reports provided us with a rich source of information, taking precipitation abundance and the demand for
water related to climate into account. We tested several monthly
combinations of climate data (annual, May to September, January to
April and October to December). The mean climate chart for October
to April was the most signicant, taking the potential for restocking
water resources into account, since evapotranspiration is at its
lowest during this period (vegetation not very active and there is less
solar radiation). This variable is referred to as WATER.
For the origin of the water, we were able to use the results from the
Sise-Eaux database, which provided the proportion of output produced from underground water (UG). This information was taken
from water production records, which strengthens its accuracy, since
data from the drinking water services often involve classications of
mixed origin, whereas in this case, the other output was implicitly
derived from surface water. The source of the water was our sole
information that was potentially correlated with the geological
context. This correlation had a considerable effect on water production constraints (availability and quality of the water resource). The
origin of the resource involved a cost differential for the WSS. Treatment costs are assumed to be signicantly higher for surface water.
Other factors can indirectly affect the operation of WSS, such as
population deserved by UDI (POP_UDI). Moreover, we were able to
take the seasonal population into account using the maximal
population/resident population ratio for 2005, on the basis of data
provided by the French Ministry of Tourism (Terrier, 2006). This
variable is referred to as POPMAX. In order to take the seasonal
French population into account, we had to calculate absences and
presences for each department. The survey to monitor tourist demand (SDT e Suivi de la Demande Touristique), carried out by TNSSOFRES for the French Ministry of Tourism, provided information
about monthly tourist trips by a panel of 20,000 French people. For
foreign tourists, for whom it was not possible to calculate absences,
the sources for our data were border surveys from 1996 and
monthly surveys of the number of people in hotels and campsites
(INSEE e Direction du Tourisme, 2005). The seasonal population is a
severe constraint for the WSS that must scale up their network and
their production capacity to cope with peaks in demand.
Finally, we obtained information on the proportion of the population with delegated management of wastewater services
(WW_DEL). Indeed, it is usually recognized that a municipality that
delegates its water supply service to a private operator is more
likely to also delegate its wastewater service. Thus, this variable is
assumed to be an explanatory variable in the equation DEL but not
to be correlated with the drinking water prices.

118

J. Fiquepron et al. / Journal of Environmental Management 126 (2013) 113e121

4.3. Land cover and land use


We used the results from the annual agricultural statistics
(Statistiques Agricole Annuelle or SAA) from 2004, which provided
data at the department level on the use of land for agricultural
production. SAA are produced by all of the statistics services of the
Departmental Agriculture and Forests Ofces. In order to collect the
necessary information, the departments have a variety of exhaustive statistical surveys or random surveys and administrative documents (land notes, national database for the identication of
cattle, etc.) at their disposal. For our model, we selected the
following types of land cover: wooded areas (FOREST), permanent
grassland (GRASS), agricultural areas for cereals, oilseeds and protein crops (CULTURE), and areas under vine cultivation, market
gardening or arboriculture (VIARMA). VIARMA is linked with the
intensive use of pesticides. The data on pigs (UGBN/ha, Nitrogen per
Livestock Unit and per ha) also revealed land uses that were
potentially polluting for the water. We used the UGBN since
different animals generate different pollutant loads. We assimilated
all these data to the surface area criterion for each department in
order to avoid the area effect.
We took mountain zones into account based on data from the
French Ministry of Agriculture and Fisheries that classies these
communes into underprivileged agricultural zones. We rated the
surface areas of the communes classied as mountain and high
mountain. In practice, each commune is attributed a handicap
rating that takes altitude and gradient into account. In order to
benet from this classication, the average altitude criterion is
700 m, reduced to 600 m in the Vosges and increased to 800 m in the
Southern Alps. In order to benet from the classication at lower
altitudes, communes must include some slopes of more than 20%
across at least 80% of their total area. Finally, within certain limits,
communes that do not have the required handicap rating can still
benet from the classication, provided that their economy is
closely linked with that of neighboring communes that do meet the
criterion (source: ANEM, Association Nationale des Elus de Montagne). The proportion of mountain zones (MOUNT) is an indication
of areas that should, in principle, have good water quality in relation
to the nitrate and pesticide criterion, a situation that is suitable for
drinking water distribution by gravitational means, and zones

where precipitation is abundant. However, it is important to notice


that this variable could be correlated with other land use variables.
Urbanized areas were represented via data from the CORINE
Land Cover Survey, produced from satellite images from 2000.
CORINE Land Cover, acronym for Coordination of Information on
the Environment, is a database showing land use across Europe.
IFEN was responsible for production, updating and distribution for
France, and is currently directly managed by the Ministry of Sustainable Development. The surface area of the smallest unit mapped (description threshold) was 25 ha. We used information on the
proportion of urbanized areas, corresponding to the total of the
sectors classied in continuous and discontinuous urban fabric.
Using the survey on land use (Teruti) carried out by the SCEES, we
would normally have been able to use land occupation data for
2004. However, no results were available for the departments in
the inner circle around Paris. Urbanization was characterized by a
high level of soil impermeability that limited the waters capacity
for self-purication. The impact of urbanized areas on water resources depends in part on the efciency of wastewater treatment
plants and rainwater collection.
We were aware that not taking the evolution in land cover into
account could lead to uncertainty in interpretations, given, for
example, that about half of the forests in France today were formerly
agricultural land in 1850 (Jussy et al., 2002). Nevertheless, accurate
data such as those concerning former land covers were not available.
The type of soil and rock inuences water quality, particularly in
terms of the retained quality indicator, rapid inltration in karst
environments. Moreover, depending on the thickness and richness
of the soil, the ltration capacity may be different. However, we did
not have relevant data at the scale of the study to take these effects
into account. At a more local level, such variables could be introduced into the model.
5. Estimation results
5.1. Bioeconomic model results
Table 4 shows the estimation results by GMM of the best
adjusted bioeconomic model, which consists of price equations
(drinking water), the proportion of population deserved by a WSS

Table 4
Estimation results for the bioeconomic model (water price, delegated management and quality equations) e GMM estimation.
Variable

DEL

q1

q2

Constant
DENS
LENG
UG
WATER
WW_DEL
FOREST
GRASS
CULTURE
VIARMA
BOVINE
DEL
q1
q2
Adjusted R2
Test of over-identifying
restrictions (p-value)
Heteroscedasticity tests (p-value)
Whites test
BreuschePagan

1.4095 (0.0577) ***


0.0249 (0.0111) **

0.11292 (0.0394) ***


0.0960 (0.0110) ***
0.0115 (0.0039) ***

94.6321 (3.2321) ***

17.0461 (2.5213)***

0.6305 (0.0539) ***

0.0689 (0.0183)***
0.1094 (0.0415)***

0.5794 (0.1355) ***


0.3643 (0.0994) ***
0.2376 (0.0739) ***
1.1430 (0.3657) ***

0.1794 (0.0456)***
0.2850 (0.0592)***
0.2013 (0.0425)***
0.2172 (0.1038)***
27.5769 (6.6924)***

0.4362
c2 (28) 23.93 (0.6852)

0.6098

0.7064

11.27 (0.6651)
2.28 (0.3199)

44.83 (0.0012)
18.97 (0.0001)

41.62 (0.2048)
1.90 (0.3867)

0.0046 (0.0005) ***


0.0061 (0.0007) ***

0.4362 (0.0648) ***


0.0016 (0.0008) **
0.0046 (0.0012) ***
0.5217

17.56 (0.2275)
9.34 (0.0094)

Notes: q1 pesticides indicator, q2 nitrates indicator.


Standard errors in parenthesis. ***: signicant at 1%, **: at 5%, *: at 10%.
Number of observations 93.

J. Fiquepron et al. / Journal of Environmental Management 126 (2013) 113e121

under delegated management and each of the two quality indicators (pesticides and nitrates). Several exogeneity hypotheses
had to be made in order to construct the conditions of orthogonality for the GMM criterion. Our system ultimately included 24
parameters for estimation with 52 ( 4  13) instruments
(including unity). We checked the validity of the orthogonality
conditions with the statistics of the Hansen test, which is equal to
23.93 with 28 ( 5224) degrees of freedom. The null hypothesis
(of validity) could not be rejected with a p-value of 0.6852. The
model therefore seems to be well specied. We also implemented
heteroscedasticity tests (such as those of White and Breusche
Pagan). Even if not always corroborating, these tests show forms of
heteroscedasticity of errors.
To begin with, we can observe that the model developed is very
well tted to our set of data. In particular, the adjusted R2 for the
price equation is quite high, with a value of 0.52, meaning that more
than 50 percent of the variation in price can be explained by the
explanatory variables. Moreover, the other equations (delegation,
pesticides and nitrates) also present a very satisfactory adjustment
with the adjusted R2 equal to 0.44, 0.61 and 0.71, respectively.
We can then study the factors of the quality indicators q1 and q2
(pesticides and nitrates, respectively), which are themselves
assumed to be price determinants. Land uses have a signicant effect
on q1 and q2. More precisely, when the forest cover in a department
(FOREST) increases, non-respect of the quality thresholds of pesticides and average content of nitrates decrease (with estimates
of 0.58 and 0.18, respectively). This means that an increase of one
point of the proportion of forest area in the department decreases
of 0.58 point the proportion of water resources having exceeded
the pesticides standards, and of 0.18 mg/l the content of nitrates.
Similarly but to a lesser extent, a land cover of permanent
grassland guarantees a better quality of raw water with coefcient
values of 0.36 and 0.29, respectively, for pesticides and nitrates.
Moreover, the positive values (0.24 for pesticides and 0.20 for nitrates) for the coefcient associated with the variable CULTURE
shows that a greater proportion of area cultivated with cereals, oilseeds and protein crops led to an increase in the amount of pesticides
and nitrates in the raw water. We can also observe that the presence
of vines, arboriculture or market gardening considerably increases
pesticide levels (coefcient value of about 1.14), while decreasing the
average nitrates content. In Hascic and Wu (2006), the coefcients of
urban land, cultivated cropland and pastureland are all signicantly
positive with respect to forest land, indicating that converting forests to developed land, cultivated cropland, or pastureland increases
conventional water pollution (via processes of eutrophication and
dissolved oxygen depletion). Moreover, as expected, the number of
bovine (Nitrogen per Livestock Unit/ha) was a highly determinant
factor for the presence of nitrates in the raw water, with a signicantly positive coefcient value for the BOVINE variable. It can
be noticed that the variable PIG appeared to have no signicant
effect and had been removed from the regression, probably due to
the low number of departments involved in this specic breeding.
Finally, it should be mentioned that a larger proportion of underground raw water corresponds to waters that are less laden with
pesticides but signicantly more (even if low) with nitrates. Indeed,
this is the phenomenon of leaching which is the source of nitrate
pollution of water. Leaching of a mineral is its transfer deep into the
ground, under the action of percolating. Nitrates can also be found
in surface waters during their supply by the deeper water sources.
However, there is automatically more problem of nitrates in
groundwater than in surface water. Furthermore, transfer of pesticides to the water is mostly supercial. Pesticides are found in
surface waters mainly due to dissolving in runoff.
Concerning the proportion of delegation to private companies for
water services, represented by the variable DEL, we observe that the

119

management conditions of the services were determinant, in


agreement with other studies on the subject (Carpentier et al., 2006;
Boyer and Garcia, 2008). A larger service (variable LENG) and a
higher density of population could suggest a more complex operation of the service, corresponding to a higher level of delegation. As
expected, private delegation is highly correlated for both water
supply and wastewater services. We also observe that the coefcient
of q1 signicantly differ from zero with a positive value of 0.0016,
suggesting that poorer raw water quality (in terms of pesticides) also
implies a greater delegation of the service. This result is in accordance with those of Carpentier et al. (2006) showing that the choice
of private delegation is made when operating conditions are difcult
because municipalities prefer not to directly manage their own
water supply services when they are too complex. In particular, they
observe a positive effect of the use of developed treatments of raw
waters on the choice of private delegation. Moreover, since land use
enables us to classify a department according to whether it belongs
to a rural zone or an urban zone, previous studies on water services
(Rebeix, 2001) revealed the fact that a greater proportion of urban
services are delegated.
The estimation results for the price equation P provide some
important information on the effect of the different factors on price
levels. First, regarding the water resource, estimations conrm that a
greater availability of underground raw water (UG) causes water
prices to drop with signicantly negative coefcients to a level of 1%
(0.0046). Moreover, a greater density of population implies a lower
water price, indicating cost savings due to scale economies. These
factors have a direct effect on operating costs. Lastly, it is clearly
demonstrated that the price is higher when management is delegated. It has already been shown that when operating conditions
become more difcult, the service tends to delegate management.
Furthermore, the quality indicator q2 has a signicant positive effect
on price (with a coefcient of 0.0046), indicating that the raw water
that was carrying more nitrates increased treatment costs and,
hence, the price. We have also shown that the quality indicator q1
accounted for the management regime chosen, revealing that when
raw water is polluted with pesticides, the price of water rises (via
delegation). We have thus shown that both quality indicators had
not necessarily direct effects on price and management choice.1
These results suggest that land uses have an indirect effect on
the economy of the WSS since they directly affect the quality of the
raw water. In particular, we show that a department that has more
forest cover also has better raw water quality on the basis of our
two indicators (pesticides and nitrates). We can therefore deduce
that there is a potential impact for a drop in the price of the
drinking water share of the household water bill. However, a
reduced model directly linking price and forest cover may help to
better compute the decrease of water price associated with the
forest service (see below).
It is possible to simulate the impacts of changes in land uses on
the raw water quality from the estimated bioeconomic model. As
an example, we tested the effects of a shift of the proportion of
forest from 28% (the national mean) to 33% (i.e., representing an
increase in forest area of 2.676 million hectares). This scenario was
chosen to test the models sensitivity to changes in land use. To
measure a signicant effect, it was necessary that the change was
important. As our model is parametrized at the national level with
departmental data, scenarios tested actually relate considerable
areas with high stakes in terms of technical realism and social
acceptability. However, if this model is valid at the national level,

1
We report estimation results of the initial model shown in Eqs (3) and (4) in
Table A1 in Appendix. This model includes more variables but not signicant, and
with a worse goodness of t.

120

J. Fiquepron et al. / Journal of Environmental Management 126 (2013) 113e121

the scenarios of land use projects should inspire projects at the


scale of catchment areas or watershed. At this scale, a 5% change of
land use is quite realistic. Furthermore, the French area of forest is
likely to evolve in signicant proportions. For instance on the
period 1990e2000, the increase in forest area was more than 8%
(source: SSP, Teruti).
In the case where forest would replace areas cultivated with
cereals, oilseeds and protein crops, this would result in a decrease
in nitrate content of 1.9 mg/l and in a drop of 4.1 points in terms of
the need to treat raw water for pesticides. These values include
both benecial effects of increasing the forest area and the
contraction of agricultural areas. It is important to stress that these
simulations dene a lower limit of the forest service. They do not
take the possibility to target lands that are more sensitive in terms
of water catchment into account, and are based on domestic water
consumption alone.

department, they found that the direct value of one hectare of forest
planted in the place of one hectare of agricultural land was valued at
V63. Nez et al. (2006) calculated the economic value of the input
(stream water) associated with the reduction in the market value of
the output (drinking water). They did not model the relationship
between quality and land uses, but instead used a formula on the
inverse relationship between the percentage of native forest cover
and stream ow, and found gures varying between $61.20 and
$162.40 (USD). One reason for these high values is that they measure a value of the ecosystem service that includes several services
and not just the forest service for raw water quality. Biao et al.
(2010) also found a very high economic benet per hectare of water conservation ($688). As specied by the authors, the term of
water conservation includes three main services: rainfall interception, soil water storage and fresh water provision.
6. Conclusion

5.2. Reduced form results


In order to evaluate the forest service in terms of water quality,
we directly assessed the effect of forest land use on the price of
drinking water. Using structural equations [1] to [4], it is possible to
derive a corresponding reduced-form representation:

Pi g0 gX Xi gZ Zi uPi

(7)

DELi q0 qX Xi qZ Zi uD
i

(8)

Since we observed that a relationship between water prices and


the choice of WSS management existed, we considered that the
error terms uPi and uD
i could be correlated. All explanatory variables
being independent, a Zellners seemingly unrelated regression
(SUR) can be used. The BreuschePagan test is used to test the
assumption that the errors across equations are contemporaneously correlated. The null hypothesis of no contemporaneous correlation is rejected at the 1% level, validating the adequacy of the
SUR approach. Estimation results are reported in Table 5.
From estimation results of this reduced model, an increase of
one point in the proportion of forest would imply a decrease of
V0.0034 per m3 of invoiced drinking water. This effect on price
seems to be low but would represent a potential savings of V11.707
million per year for French domestic users. The afforestation of one
hectare would generate a savings of around V22 per year on
household water bills. This savings is not calculated per household
but from the aggregated water bill for all domestic users. This gure
is signicant with respect to monetary ows currently generated by
forest lands. Abildtrup et al. (2013) showed the importance of
considering spatial interactions when computing the value of forest
ecosystem services. In their empirical study on a French
Table 5
Estimation results for the reduced form (water price and management regime
equations) e SUR estimation.
Variable

DEL

Constant
DENS
LENG
UG
WW_DEL
FOREST
GRASS
VIARMA
Adjusted R2
BreushePagan test of

1.3181 (0.0860)***

45.9441 (10.029)***
7.4218 (1.7672)***
0.5578 (0.4670)
0.1047 (0.0756)
0.3669 (0.0887)***

0.0043 (0.0007)***

0.0034 (0.0017)**
0.0011 (0.0013)
0.0080 (0.0051)
1.3892 (0.5158)***
0.2750
0.2672
independence c2 (1) 16.498, p-value 0.0000

Notes: Standard errors in parentheses. ***: signicant at 1%; **: at 5%; *: at 10%.
Number of observations 93.

Our aim in this article was to clearly identify and quantify the
impacts of the forest on water quality at the national level in France
by evaluating the presence of potential benets to be derived from
the forest. Our results conrm the positive inuence of the forest on
raw water quality. We have also shown the benets that the forest
cover can bring to the economy of drinking water services, with a
potential drop in water prices. This value, inherent to woodlands in
comparison to other land uses, justies and strengthens the merits
of woodlands and forestry activities in the areas concerned.
These results will make it possible to initiate discussions on
public policies at the national level that could focus either on landuse choices or on incentives that could potentially be used by forest owners who provide services. Real issues are at stake today in
countries where pressures on forest resources can be high in terms
of deforestation or species conservation, or in terms of bioenergy,
climate change and carbon storage. Very strong incentives exist in
these cases to develop payment systems for environmental forest
services, avoiding or minimizing deforestation, for example. In
developed countries, in addition to nancial savings for water services and consumers, a major issue is now emerging concerning
forest owners who claim payment for their good management as
well as their contribution to water quality. However, such an action
may be deeply challenged if links between forests and water quality,
for example, are not clearly demonstrated due to a lack of data or
because of the presence of conifers that may create additional nitrate
losses. This important issue concerns the support and involvement
of stakeholders in the forest environment in terms of the implementation of a management system that takes water resources into
account. The question of payment and incentives for forest owners
clearly arises. For example, payment schemes for water supply (in
quality and/or in quantity) could be developed if a clear link can be
demonstrated between forest land use and management and water
supply both at the hydrological and economic levels.
Acknowledgements
We would like to thank Galle Chateau from the French Ministry
of Health, Henri Davezac from the ARS Midi-Pyrnes, Vincent
Perez from LERFoB, Jacques Plaisir from the French Ministry of
Agriculture, and Annie Coutellier and Martine Wichmann from
IFEN, for their invaluable contribution to data collection and
transmission. Our thanks also to Olivier Picard for supervising the
joint study involving INRA and IDF Fort et Eau, and to Andr
Granier, Sophie Nicola, Julien Tournebize and Jean-Pierre Butault
for their comments on an earlier version of the article. This work
was carried out with nancial support from the French Ministry of
Agriculture.

J. Fiquepron et al. / Journal of Environmental Management 126 (2013) 113e121

121

Appendix
Table A1Estimation results for the bioeconomic model (water price, delegated management and quality equations) e Estimation by GMM e Initial model shown in Eqs. (3) and
(4)
Variable

DEL

q1

q2

Constant
VOL
DENS
LENG
UG
WATER
WW_DEL
FOREST
GRASS
CULTURE
VIARMA
MOUNT
BOVINE
PIG
DEL
q1
q2
Adjusted R2
Test of over-identifying restrictions (p-value)
Heteroscedasticity tests (p-value)
Whites test
BreuschePagan

1.4795 (0.0944)***
0.0011 (0.0007)
0.0145 (0.0126)

0.0499 (0.1042)

97.2075 (3.4440)***

15.3326 (2.1909)***

0.4892 (0.1634)***
0.0676 (0.2730)

0.0350 (0.0361)
0.0237 (0.0974)

0.8729 (0.2519)***
0.7087 (0.6634)
0.0450 (0.1751)
1.0070 (0.7254)
0.3071 (0.1295)**
124.704 (78.3664)
483.931 (507.0)

0.1906 (0.0631)***
0.2901 (0.1419)**
0.2594 (0.0529)***
0.1089 (0.2012)
0.0067 (0.0339)
28.5158 (15.9699)*
103.723 (124.3)

0.0058 (0.0012)***

0.0888 (0.0153)***
0.0133 (0.0048)***
0.0019 (0.0013)
0.0056 (0.0010)***

0.6433 (0.1424)***
0.0024 (0.0020)
0.0065 (0.0030)**
0.4414
c2(28) 19.01 (0.3912)

0.0041 (0.0021)*
0.0026 (0.0030)
0.4040

0.4696

0.6335

59.79 (0.0003)
10.83 (0.0044)

25.72 (0.5343)
2.62 (0.2695)

81.50 (0.0092)
18.84 (0.0001)

73.63 (0.0391)
0.59 (0.7462)

Notes: q1 pesticides indicator, q2 nitrates indicator.


Standard errors in parenthesis. ***: signicant at 1%, **: at 5%, *: at 10%.
Number of observations 93.

References
Abildtrup, J., Strange, N., 2000. The option value of non-contaminated forest watersheds. Forest Policy and Economics 1, 115e125.
Abildtrup, J., Garcia, S., Stenger, A., 2013. The Effect of Forest Land Use on the Cost of
Drinking Water Supply: a Spatial Econometric Analysis. Ecological Economics,
http://dx.doi.org/10.1016/j.ecolecon.2013.01.004.
Adhikari, R.N., Rao, M.S.R.M., Selvi, V., Math, S.K.N., Husenappa, V., Chandrappa, M.,
Reddy, K.K., 2002. Studies on runoff coefcient of rational formula. Indian
Journal of Soil Conservation 30 (1), 106e108.
Agreste, 2004. Rpartition physique du territoire de 1992 2004. (Teruti Lucas).
Benot, M., Papy, F., 1997. Pratiques agricoles sur le territoire et qualit de leau
alimentant un captage. In: Leau dans lespace rural. INRA, pp. 323e338.
Benot, M., Fizaine, G., Bernard, P.-Y., 2002. Qualit nitrique des eaux en bassins
forestiers dalimentation: fonctionnement stable et effets post-tempte 26/12/
1999. In: Combe, J., Rosselli, W. (Eds.), Leau qui sort des bois e quand fort
durable rime avec eau potable. Actes de la Journe thmatique de lAntenne
romande du WSL, pp. 29e36. Lausanne.
Biao, Z., Wenhua, L., Gaodi, W., Yu, X., 2010. Water conservation of forest ecosystem
in Beijing and its value. Ecological Economics 69 (7), 1416e1426.
Bontemps, C., Nauges, C., 2009. Carafe ou bouteille? Le rle de la qualit de lenvironnement dans la dcision du consommateur. Economie et Prvision 2009/2
(188), 61e79.
Boyer, M., Garcia, S., 2008. Rgulation et mode de gestion: une tude conomtrique sur les prix et la performance dans le secteur de leau potable. Annales
dconomie et de statistique 90, 35e74.
Broadmeadow, S., Nisbet, T.R., 2004. The effects of riparian forest management on
the freshwater environment: a literature review of best management practice.
Hydrology and Earth System Sciences 8 (3), 286e305.
Brown, T.C., Binkley, D., 1994. Effect of Management on Water Quality in North
American Forests. USDA Forest Service, p. 26.
Carpentier, A., Nauges, C., Reynaud, A., Thomas, A., 2006. Effets de la Dlgation sur
le Prix de lEau Potable en France: Une Analyse partir de la Littrature sur les
effets de traitement. conomie et Prvision 174 (3), 1e19.
Coutellier, A., 2007. Les services publics de leau en 2004, p. 30. Eau potable Dossier
n 7, Ifen.
Ernst, C., Gullick, R., Nixon, K., 2004. Protecting the source e conserving forests to
protect water. American Water Works Association 30 (5), 3e7.
French Ministry of Health, 2008. Leau potable en France, 2005e2006. (Report.
Paris).
Gove, N.E., Edwards, R.T., Loveday, L., 2001. Conquest effects of scale on land use and
water quality relationships: a longitudinal basin-wide perspective. Journal of
the American Water Resources Association 37, 1721e1734.

Gundersen, P., 2007. The impact of forest management on water quality. In: Raulund-Rasmussen, K., Hansen, K. (Eds.), Synthesis Report on Impact of Forest
Management on Environmental Services EFORWOOD Report, pp. 73e91.
Hascic, I., Wu, J.J., 2006. Land use and watershed health in the United States. Land
Economics 82 (2), 214e239.
INSEE e Direction du Tourisme, 2005. Le tourisme en France, p. 122. Insee
Rfrences.
Johnson, N., White, A., Perrot-Maitre, D., 2000. Developing Markets for Water Services from Forests: Issues and Lessons for Innovators. Forest Trends, World
Resource Institute, The Katoomba group, p. 19.
Jussy, J.H., Koerner, W., Dambrine, E., Dupouey, J.-L., Benot, M., 2002. Inuence of
former agricultural land use on net nitrate production in forest soils. European
Journal of Soil Science 53, 367e374.
Kiersch, B., Tognetti, S.S., 2002. Land-water linkages in rural watersheds: results
from the FAO electronic workshop. Land Use and Water Resources Research 2,
1.1e1.6.
Knoke, T., Ammer, C., Stimm, B., Mosand, R., 2008. Admixing broadleaved a review
on yield, ecological stability and economics. European Journal of Forest
Resource 127, 89e101.
Langpap, C., Hascic, I., Wu, J.J., 2008. Protecting watershed ecosystems through
targeted local land use policies. American Journal of Agricultural Economics 90
(3), 684e700.
Nuez, D., Nahuelhual, L., Oyarzun, C., 2006. Forest and water: the value of native
temperate forests in supplying water for human consumption. Ecological Economics 58 (3), 606e616.
Ranger, J., Colin-Belgrand, M., Nys, C., 1995. Le cycle biogochimique des lments
majeurs dans les cosystmes forestiers. Importance dans le fonctionnement
des sols. Etude et Gestion des Sols 2 (2), 119e134.
Ranger, J., Loyer, S., Gelhaye, D., Pollier, B., Bonnaud, P., 2007. Effects of the clearcutting of a Douglas-r plantation (Pseudotsuga menziesii F.) on the chemical
composition of soil solutions and on the leaching of DOC and ions in drainage
waters. Annals of Forest Science 64 (2), 183e200.
Rebeix, G., 2001. Eau potable: diversit des services. grand cart des prix, p. 4. Les
donnes de lenvironnement e Eau, n 65, Ifen.
Sikka, A.K., Selvi, V., 2005. Experimental examination of rational runoff coefcient
for small agricultural and forest watersheds in the Nilgiris. IE Journal 86, 24e
27.
Terrier, C., 2006. Mobilit touristique et population prsente e Les bases de lconomie prsentielle des dpartements. Edition Direction du Tourisme, p. 128.
Turc, L., 1961. Evaluation des besoins en eau dirrigation, vapotranspiration
potentielle. Annales Agronomiques 12 (1), 13e49.
Willis, K., 2002. Benets and Costs of Forests to Water Supply and Water Quality.
Report to the Forestry Commission. Forestry Commission, Edinburgh, p. 24.

You might also like