You are on page 1of 12

Environmental Modelling & Software 41 (2013) 27e38

Contents lists available at SciVerse ScienceDirect

Environmental Modelling & Software


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

A new instream flow assessment method based on fuzzy habitat suitability and
large scale river modelling
S. Marsili-Libelli a, *, E. Giusti a, A. Nocita b
a
Department of Systems and Computers, University of Florence, Via S. Marta 3, 50139 Firenze, Italy
b
Museo di Storia Naturale, University of Florence, Via Romana 17, 50125 Firenze, Italy

a r t i c l e i n f o a b s t r a c t

Article history: This paper proposes an integrated tool for the assessment of fish habitat suitability based on synthetic
Received 19 December 2011 hydraulic and water quality parameters. There are three innovative features in this study: (a) The
Received in revised form proposed approach seeks to improve the capabilities of IFIM (Instream Flow Incremental Methodology)
13 October 2012
by extending the assessment to a larger spatial scale, which can be helpful for river managers in decision
Accepted 18 October 2012
Available online 14 December 2012
making. (b) The method is based on the suitability response of target species to hydraulic and water
quality parameters through a fuzzy model, which is a novel application of a TakagieSugeno fuzzy logic.
(c) The introduction of simulated river conditions enables the generation of a wide range of scenarios and
Keywords:
Habitat suitability
the detection of potentially critical situations.
Instream flow assessment After introducing the main algorithm, a sensitivity analysis is provided for the assessment of critical
Water quality modelling river segments and for ranking the influence of each parameter on the habitat. Then, a second algorithm
Hydraulic modelling is developed to produce an instream flow assessment method by determining the range of admissible
TakagieSugeno fuzzy logic flows that preserve the habitat suitability to a prescribed degree. The combined method is demonstrated
with the application to two Italian rivers in the Tuscany region. In the case of the Arno River, the method
highlights the habitat diversity for the two target species along its course, and the critical conditions that
may develop during the summer low flow. In the case of the Serchio River, the analysis helps to assess
habitat alterations likely to be caused by a planned diversion to feed a nearby lake. In both instances,
with a minimum requirement of field data, this method shows its flexibility and seems better able to
detect critical situations than the conventional IFIM approach.
 2012 Elsevier Ltd. All rights reserved.

Software availability 1. Introduction

Name of software: FISH Fuzzy Toolbox The environmental impact of flow alterations has emerged as an
Developers: Stefano MarsilieLibelli, Elisabetta Giusti important aspect of river management. However, it is often difficult
Contact address: Department of Systems and Computers, to find a single indicator linking the flow regime to the sustain-
University of Florence, Via S: Marta, 3, 50139 Florence, Italy ability of the riverine ecosystem. For this purpose, the Instream
Email: marsili@dsi.unifi.it Flow Incremental Methodology (IFIM) (Stalnaker et al., 1995; Bovee
Availability: contact the developers et al., 1998) and related software PHABSIM (Milhous et al., 1989;
Year first available: 2012 Van Winkle et al., 1998) was introduced to assess the sustainability
Hardware required: IBM compatible PC of man-made flow changes. The complexity and the limitations of
Software required: Windows XP or later this method, however, have fostered the development of other
Programming language: Matlab 7.3 or later; Matlab Fuzzy approaches, such as the ecological impact of flow, or e-flow analysis
Toolbox required (Petts, 2009), stressing the need to extend the paradigm of
scenarios on a seasonal and climatic basis.

1.1. The scale factor

* Corresponding author. Tel./fax: þ39 (0)55 47 96 264.


The inherent limitations of the micro-scale used by IFIM have
E-mail addresses: marsili@dsi.unifi.it (S. Marsili-Libelli), giusti@dsi.unifi.it stimulated the development of a wider scale approach, and several
(E. Giusti), nocita@unifi.it (A. Nocita). authors have proposed the meso-habitat scale as a good

1364-8152/$ e see front matter  2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.envsoft.2012.10.005
28 S. Marsili-Libelli et al. / Environmental Modelling & Software 41 (2013) 27e38

Nomenclature mbarbel
NH3 degree of membership of the un-ionized ammonia
suitability index for the barbel
mbarbel
U degree of membership of the stream velocity
Symbol suitability index for the barbel
DO dissolved oxygen (mg O2 L1) mbarbel
H degree of membership of the stream depth suitability
FISH75 fourth quartile of the FISH statistical distribution index for the barbel
FISHmin minimum acceptable value of the FISH indicator mE:
DO
chub degree of membership of the dissolved oxygen
H river depth (m) suitability index for the E. chub
n manning roughness coefficient (sm1/3) mE:
NH3
chub degree of membership of the un-ionized ammonia
NH3 un-ionized ammonia (mg N L1) suitability index for the E. chub
pmax maximum value of each suitability parameter (DO, mE:
U
chub degree of membership of the stream velocity
NH3, U, H) suitability index for the E. chub
pmin minimum value of each suitability parameter (DO, mE:
H
chub degree of membership of the stream depth suitability
NH3, U, H) index for the E. chub
Qinstream sustainable instream flow (m3 s1) Fo second derivative of the incremental FISH sensitivity
QSLF summer low flow (m3 s1) computed around the nominal values
SQSLF large variations flow sensitivity to summer low flow Jp logarithmic sensitivity with respect to parameter p
(%)
U stream velocity (m s1) Abbreviations
GA genetic algorithms
Greek symbols IFIM instream flow incremental methodology
Dp incremental FISH sensitivity FIS fuzzy inference system
d fractional parameter perturbation used in computing FISH Fuzzy Index for the Suitability of the Habitat
the incremental sensitivity PHABSIMPhysical HABitat SIMulation
mbarbel
DO degree of membership of the dissolved oxygen
suitability index for the barbel

compromise between the detailed but survey-intensive micro- Fukuda (2009) assessed the advantages of the fuzzy approach in
scale of IFIM and the coarse river basin scale used by decision describing the habitat suitability, concluding that fuzzification with
support systems. Different meso-scale models have been devel- its smoothing properties was useful in reducing the uncertainty
oped, including MesoHABSIM (Parasiewicz, 2001, 2007a,b) and caused by the habitat variability. Later, Fukuda et al. (2011) con-
MesoCASiMiR (Schneider et al., 2001), which is based on fuzzy logic structed a GA-optimized fuzzy habitat spatial preference model
and was applied at different levels of the trophic chain, from based on multiple environmental variables to predict the presence-
macroinvertebrates (Mouton et al., 2006) to fish (Mouton et al., absence of the target fish. Also, Van Broekhoven et al. (2007)
2007). Further fuzzy habitat indicators were later developed and stressed the importance of prior knowledge when using GA as
compared with knowledge-based systems (Mouton et al., 2008, optimization methods.
2009), indicating that the combination of both approaches can All these studies are based on large data-sets, allowing the use of
improve model reliability, especially in site-extrapolation of local sophisticated optimization techniques to calibrate the inference
results. system. However, the importance of expert opinion as the final step
in model adjustment, sometimes in conjunction with numerical
1.2. The fuzzy approach optimization methods, was recently underlined by Marchini et al.
(2009), who proposed a qualitative automatic rule-assessment
An important advantage of the fuzzy approach is that non- procedure in place of GA optimization. The importance of the
numerical information such as expert knowledge can be directly expert judgement in the final fuzzy model acceptability has
incorporated into its rule-base, often supplementing the scarcity of recently been reaffirmed (Mouton et al., 2008, 2011; Wieland et al.,
field data. Further, the inherent uncertainty of ecological variables 2011; Krueger et al., 2012).
can be incorporated into the definition of rules and membership
functions (Prato, 2007; Fraternali et al., 2012; Rinderknecht et al., 1.3. The novel aspects of this study
2012) so that the results can be easily grasped by non-technical
parties, such as decision makers and stakeholders (Van This study aims to fill the gap between micro-scale and macro-
Broekhoven et al., 2006). scale habitat suitability assessment methods by using a fuzzy
In addition to the previously cited MesoCASiMiR, the fuzzy inference system and large scale hydraulic and water quality
approach to the definition of habitat suitability has been amply modelling to determine a comprehensive habitat suitability indi-
documented. Young et al. (2000) developed an environmental cator for the fish population. In this sense, it extends some
decision support system for the Murray-Darling river basin using preliminary results (Marsili-Libelli et al., 2008) by proposing
fuzzy logic to incorporate the uncertainties in the definition of a fuzzy habitat suitability index based upon synthetic hydraulic and
environmental attributes. Van Broekhoven et al. (2006) proposed ecological parameters. The innovative aspects of the study are: (a)
a model for macroinvertebrate habitat suitability based on the incorporation of the expert knowledge into the fuzzy inference
a knowledge-base that included a large number of species. Fukuda system as a supplement to the scarcity of field data, (b) the
and Hiramatsu (2008) have considered a fuzzy habitat preference extension of the assessment to a wide combination of simulated
model using the TakagieSugeno fuzzy inference method and hydraulic and water quality scenarios, (c) the new definition of
Genetic Algorithms (GA) optimization. In a subsequent paper habitat sensitivity to assist in the instream flow analysis. The study
S. Marsili-Libelli et al. / Environmental Modelling & Software 41 (2013) 27e38 29

partially follows the approach based on expert judgment (Marchini young-of-the-year, were sexed, their total length and weight recorded, and then
et al., 2009; Mouton et al., 2009. 2011; Wieland et al., 2011), but released. Fish biomass was calculated as the ratio between biomass estimate and
wet stream bed area (g m2). In the sequel, biomass was used as an indirect habitat
extends the analysis by using synthetic hydraulic and water quality suitability indicator.
data, respectively computed with HEC-RAS 6 (U.S. Army Corps of
Engineers, 1991) and QUAL2Kw (Chapra and Pelletier, 2003;
2.3. Calibration of the underlying hydraulic and water quality models
Pelletier et al., 2006).
Since the suitability index critically depends on the hydraulic and water quality
2. Methods simulations, some discussion concerning the reliability of the underlying models are
in order. HEC-RAS (U.S. Army Corps of Engineers, 1991; Brunner, 2010) has been
2.1. Study area adopted by the Arno River Authority as the preferred simulation tool for basin
management (Federici, 1998; Arno River Water Authority, 2010) assuming the
The study area involves portions of two rivers in central Italy: the middle course calibrated Manning coefficients obtained by Aricò et al. (2009) and Di Baldassarre
of the Arno River and the downstream end of the Serchio River, whose location is and Montanari (2009) for similar rivers as starting values and bearing in mind the
shown in Fig. 1. For the Arno River a length of nearly 180 km in its middle-lower calibration difficulties for low flows (Pappenberger et al., 2005, 2006; Di Baldassarre
course was considered, placing the upstream end at 50 km from the spring and and Claps, 2011). The tuning of the HEC-RAS model was performed according to the
the downstream end at 30 km from the mouth. Fig. 2 shows the Arno catchment and following squared error sum criterion
the river portion considered in the study, which was divided into 32 reaches
X  2
including eleven main tributaries, five major wastewater treatment plants, and the b i ¼ arg min
n
exp
Qi  QiHECRAS ; (1)
diversion of the Florence potabilization plant. The second example considers the ni
i˛reach number
Serchio River, which flows through the northern part of Tuscany and represents
a pristine aquatic environment with a highly diversified fish population. The area where fni i˛reach numberg is the set of Manning coefficients for the reaches into
involved in this study is the terminal part, shown in Fig. 3, which was divided into 17 which the river was subdivided. The hydraulic regime produced by the calibrated
reaches. The Serchio River Water Authority is considering a diversion scheme in the HEC-RAS model was then fed into the water quality model QUAL2Kw (Chapra and
downstream of the city of Lucca to feed the nearby Massaciuccoli Lake. Some Pelletier, 2003; Pelletier et al., 2006). Though this platform has its own hydraulic
environmental characteristics of the river diversion project are treated elsewhere modelling capabilities, HEC-RAS was preferred for its higher accuracy. As to the
(Giusti et al., 2011a,b; Renzetti, 2010) whereas the aspect considered here is its water quality calibration, QUAL2Kw has a built-in GA calibration procedure that was
impact on the aquatic ecosystems downstream of the diversion. used to calibrate the model against the available data obtained from several inde-
pendent sources: the Arno River Authority, the Serchio River Authority and the
Tuscany Region Environmental Protection Agency (ARPAT). The water quality
2.2. Target species and sampling methods parameters used in the calibration were Temperature, Dissolved Oxygen,
AmmoniumeN, NitrateeN, pH. From the last three parameters the fraction of un-
The barbel (Barbus tyberinus) and the European chub (Leuciscus cephalus) were ionized ammonia (NH3) was computed by chemical equilibrium. The combined
selected as targets, being the most widespread autochthonous and reophylic species hydraulic and water quality calibration results are shown in Fig. 5 for the Arno River
in the region. Their average response to the environmental parameters is described and in Fig. 6 for the Serchio River.
by the suitability curves shown in Fig. 4. The habitat suitability is based on velocity,
depth, dissolved oxygen, and un-ionized ammonia. These parameters were selected
on the basis of practical data availability and were determined by combining liter- 2.4. Computation of the Fuzzy Index for the Suitability of the Habitat (FISH)
ature studies (McMahon, 1982; McMahon et al., 1984; Raleigh et al., 1986; Milhous
et al., 1989; Svobodova et al., 1993; Lamouroux and Capra, 2002; Lamouroux and The general structure of the algorithm to compute the Fuzzy Index for the Suit-
Souchon, 2002; Moir et al., 2005; Environment Agency, 2007; Ayllón et al., 2011; ability of the Habitat (FISH) is shown in Fig. 7. The initial knowledge-base includes
Macura et al., 2011) with direct observations obtained with the IFIM sampling river hydraulics, water quality, and the target fish response to these parameters
protocol (Nocita et al., 2005, 2010). As Fig. 4 shows the response of both species to represented by the suitability curves shown in Fig. 4. The fuzzy inference system uses
un-ionized ammonia was the same, whereas they were different for the other this information to determine the composite FISH index (Algorithm 1). Critical
parameters. transects can then be detected via a low flow sensitivity analysis (Algorithm 2) that
The sampling techniques are described in detail in Nocita et al. (2009). Electro can assist in performing an instream flow assessment. The two algorithms will now
fishing and a hand-held seine were used to collect the catch. All specimens, except be described in detail.

Fig. 1. Location of the two rivers used as case studies in Tuscany, western central Italy.
30 S. Marsili-Libelli et al. / Environmental Modelling & Software 41 (2013) 27e38

Fig. 2. Catchment of the Arno River. The portion of the river course considered in this study (thick line) was divided in 30 reaches. There are eleven main tributaries and five major
wastewater treatment plants, together with one large water treatment plant providing drinking water to the city of Florence.

Algorithm 1. Computation of the Fuzzy Index for the Suitability of the Habitat Step 2: Define a set of membership functions for the fuzzification of the suit-
(FISH) ability curves;
Step 3: Compute the individual fuzzy suitability responses;
Step 1: Generate the synthetic hydraulic and water quality data for the Step 4: Process the suitability indexes through the fuzzy inferential system to
prescribed flows and reaches; obtain the FISH along all the reaches for each computed flow.

Fig. 3. Outline of the Serchio River catchment, with the inset showing the study area. The envisioned diversion to the Massaciuccoli lake is in the third reach, whereas the only
tributary (Ozzeri channel) flows into the sixth reach.
barbel E. chub
a 1 b 1
0.8 0.8

0.6 0.6

NH3SI
DOSI
0.4 0.4

0.2 0.2

0 0
0 2 4 6 8 10 12 0 0.01 0.02 0.03 0.04 0.05
DO (mg L-1) NH (mg L-1)
3

c 1
d 1
0.8 0.8

0.6 0.6

HSI
USI

0.4 0.4

0.2 0.2

0 0
0 0.5 1 1.5 0 0.5 1 1.5 2 2.5
U (m s-1) H (m)

Fig. 4. Normalized E. chub and barbel suitability curves for the river parameters considered in the study. In the case of un-ionized ammonia (b), both species exhibit the same
response. The subscript SI stands for suitability index in the normalized range (0, 1).

WWTP
Tributaries Diversion
discharge
a 12 a 14.0
Flow model Flow model
10 Flow data 13.5 Flow data
Flow (m3 s-1)

Flow (m3 s-1)

8 13.0
6 12.5 Ozzeri Channel Tributary

4 12.0
2 11.5
0 11.0
b 10 b 10
Ozzeri Channel Tributary
9 9
8 8
DO (mg L-1)

DO (mg L-1)

7 7
6 6
5 5
DO saturation DO saturation
4 4 DO model
DO model
3 DO data 3 DO data
2 2
c 0.05 c 0.014
NH3 model Ozzeri Channel Tributary NH3 model
0.012
0.04 NH3 data NH3 data
NH3 (mg L-1)

0.010
NH3 (mg L-1)

0.03 0.008

0.02 0.006
0.004
0.01
0.002
0.0 0.000
0 20 40 60 80 100 120 140 160 0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0
distance from the upstream end (km) distance from the upstream end (km)

Fig. 5. Calibration of the Arno River joint Hec-Ras/QUAL2Kw model for the typical Fig. 6. Calibration of the Serchio River joint Hec-Ras/QUAL2Kw model for the typical
summer low flow regime, as explained in the text. The main variables used in this summer low flow regime, as explained in the text. The main variables used in this
study, Flow (a), DO (b), and NH3 (c) are shown. The arrows indicate the position of the study, Flow (a), DO (b), and NH3 (c) are shown, with the arrow indicates the position of
main tributaries, of the wastewater treatment plants, and of the diversion of the the only tributary, the Ozzeri Channel. The decreasing value of the saturation DO is due
Florence potabilization plant. to the increasing salinity near the mouth due to saltwater intrusion.
32 S. Marsili-Libelli et al. / Environmental Modelling & Software 41 (2013) 27e38

Fig. 7. General outline of the FISH algorithm. The prior knowledge about the river hydraulics and ecology are used as input data to the suitability curves. Algorithm 1 computes the
FISH index via the fuzzy inference system, whereas Algorithm 2 performs an instream flow assessment based on the critical reaches detected by Algorithm 1.

After generating a suitable set of hydraulic and water quality data (Step 1) and RE:
i
chub : if mE: chub and mE: chub and mE: chub and mE: chub then S ¼ S
DOSI NH3SI USI HSI i
having defined a set of membership functions (Fig. 8) to fuzzify the suitability (3)
Rbarbel
i : if mbarbel barbel barbel and mbarbel then S ¼ S ;
DOSI and mNH3SI and mUSI HSI i
curves of Fig. 4 (Step 2) the individual fuzzy suitability responses to each
parameter are then computed (Step 3). The fuzzy membership functions (High,
where the antecedent degrees of membership are those of Eq. (2) and the conse-
Medium, Low) of Fig. 4 were defined by expert judgement on the basis of past
quent singletons Si are defined for both species over the suitability support as High
experience (Nocita et al., 2005, 2009, 2010). They were used to obtain the degree
(H ¼ 1.1), Middle High (MH ¼ 0.75), Middle (M ¼ 0.5), Middle Low (ML ¼ 0.12) and
of membership of the current suitability indexes for each target species, which can
Low (L ¼ 1.5). The High and Low singletons are placed outside the (0,1) range in
be expressed as
order to obtain a final index spanning the whole (0,1) range. The FISH index is then
obtained by weighted sum
( h i
E: chub : mE: chub mE: chub mE: chub mE: chub
h DO NH3 U H
i (2) PNrules
barbel : mbarbel mbarbel mbarbel mbarbel : Sm
1 i i
DO NH3 U H FISH ¼ Pi ¼
Nrules
; (4)
i¼1
mi

The core of the algorithm (Step 4) is a TakagieSugeno fuzzy inference system where Nrules is the number of rules of each inference system and mi is the degree of
(Takagi and Sugeno, 1985) for each target species yielding the FISH index. It is membership of each rule Ri. Considering three membership functions for each of the
composed of a series of logical rules whose antecedents are the degrees of four parameters the complete inference set would consist of 34 ¼ 81 rules, but many
membership defined in Eq. (2), and whose consequents are singleton values in the of these were pruned by expert judgement or by direct inspection, deleting the
normalized suitability range (0,1). The generic rule Ri has the form inconsistent ones and those whose degree of activation remained below the

a Dissolved Oxygen b Unionized Ammonia


Degree of membership
Degree of membership

1 1
DOM NH3H
0.8 0.8 NH3M
DOL NH3L
0.6 0.6
DOH
0.4 0.4

0.2 0.2

0 0
0 0.5 1 0 0.5 1
DOSI NH3SI

c River velocity d River depth


Degree of membership
Degree of membership

1 1
UM HM HH
0.8 0.8
UL UH HL
0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 0.5 1 0 0.5 1
USI HSI

Fig. 8. The three membership functions (Low, Medium, High) used to fuzzify each suitability index of Fig. 4.
S. Marsili-Libelli et al. / Environmental Modelling & Software 41 (2013) 27e38 33

threshold of 2.5%, as described later. The logical connectives in Eq. (3) were imple- 3. Results
mented with the product and max operators.

3.1. Computation of FISH for the Arno River


2.5. FISH sensitivity

Two sensitivity measures are now introduced to detect potentially critical


Considering the most frequent headwater summer low flow of
reaches. QSLF ¼ 0:5 m3 s1 , the river conditions were simulated as described
in Section 2.3 and the resulting FISH was compared in Fig. 9 with
2.5.1. Flow sensitivity
a set of observed fish biomass densities in the same summer
Flow sensitivity was defined around the summer low flow (QSLF) as
conditions, assumed as an indirect measure of the habitat suit-
    
þ
FISH QSLF 
 FISH QSLF  ability. This comparison should by no means be considered a cali-
 
SQSLF ¼ 100   ; (5) bration in the strict sense, but was used nonetheless as a guideline
 FISHðQSLF Þ 
to adjust the inference rules for the two species, together with
þ
where QSLF  are the flow variations around Q
and QSLF SLF (Summer Low Flow) by
expert judgement, to eliminate the rules that were never activated
a percentage of 25%, as this represents a good approximation of the central by more than 2.5% of the total, and to adjust the membership
þ  , the river functions. After this pruning, 14 rules were retained for the barbel
quartiles for the observed flow regimes. For both of the flows QSLF and QSLF
model was run and the FISH computed for the perturbed flows. High values of SQSLF (Table 1) and 20 rules for the E. chub (Table 2). A large variations
indicate that flow variations have a major impact on habitat, whereas low values sensitivity analysis, as described in Section 2.5.1, is shown in Fig. 10
denote that the habitat is little affected by flow. Reaches with a high value of SQSLF are
for the barbel, indicating that reaches 10 and 12 are the most
considered critical and may be subject to a further inflow analysis.
critical. This result is indirectly confirmed by the observed biomass
2.5.2. Suitability curves sensitivity densities shown in Fig. 9. In practice, those reaches correspond to
The incremental sensitivity, in contrast with the previous large scale one, is major changes in river morphology from pool-riffle successions to
aimed at investigating how the uncertainty associated with each suitability curve of lotic conditions. Repeated application of Algorithm 1 from
Fig. 4 affects FISH. It was computed by generating a set of perturbed suitability curves.
a minimum QSLF min ¼ 0:25Q (extreme summer low) to
Given a range of variation (pmax, pmin) for each parameter p ¼ fDO; NH3 ; U; Hg, the SLF
incremental perturbation is computed as d ¼ ðpmax  pmin Þ=500 and a sequence of max
a maximum QSLF ¼ 20QSLF (spring high flow) produced the suit-
seven perturbed parameter values are obtained by shifting the suitability range by ability surfaces of Fig. 11, confirming the higher adaptability of the
multiples of d around the reference value (d ¼ 0), i.e. f  3d; 2d; d; 0; þd; þ2d; þ3dg.
E. chub and, by contrast, the narrow barbel habitat confined to the
Being an incremental sensitivity, d is much smaller than the increment used in
Section 2.5.1. For each perturbed suitability curve, the incremental FISH sensitivity is upstream shallow reaches. A sensitivity study according to
defined as the squared difference between the perturbed ðFISHkp Þ and the reference Algorithm 2 also showed that un-ionized ammonia was the most
ðFISHop Þ indexes by repeated application of Algorithm 1, i.e. influential factor (Fig. 12). Finally, an instream flow a was conducted
for reach #10 that Fig. 10 showed to be highly critical, resulting in
 2
Dp ðkdÞ ¼ FISHkp  FISHop k ¼ 3; 2; 1; 0; 1; 2; 3: (6) the wide separation of the habitat for the two species illustrated in
Fig. 13. Given a set of flows around QSLF the cumulative FISH
The curvature of the function Dp ðkÞ around k ¼ 0 yields an indication of the distribution is shown in the (a) graph for both target species. Since,
impact of the parametric variations on FISH and can be expressed by its second
derivative. This can be numerically approximated (Abramowitz and Stegun, 1965) as
!
barbel
v2 Dp ðdÞ Dp ð2dÞ þ 16Dp ðdÞ  30Dp ð0Þ þ 16Dp ðdÞ  Dp ð2dÞ a 1 10.0

Obseved biomass density (gr m-2)


Fo ¼ 2
y 2
: (7) FISH Obs. Density
vd 12d
0.8 8.0
To take into account the large spread of Fo values, its logarithm was considered
instead
0.6 6.0
FISH

Jp ¼ logðFo Þ: (8)
0.4 4.0
According to this definition, low sensitivities result in negative amplified values of
Jp whereas large positive values are compressed, making the results graphically
more compact. Zero sensitivities are mapped as zeros to avoid infinity values in Eq. (8). 0.2 2.0

2.6. Instream analysis of critical reaches 0 0


0 50 100 150
Instream flow assessment is a part of the IFIM methodology to understand the
E. chub
impact of flow alterations on stream habitat. This section describes how FISH can
b 1 2.0
Obseved biomass density (gr m-2)

detect reaches where the habitat suitability is critically affected by flow. The
FISH Obs. Density
sustainable instream flow {Qinstream} is defined as the flow range for which FISH
remains above the fourth quartile (FISH75) of its statistical distribution, or above an 0.8
1.5
absolute minimum threshold (FISHmin), whichever is greater
0.6
FISH

fQinstream g ¼ arg½ðFISH > FISH75 Þ and ðFISH > FISHmin Þ: (9) 1.0
Q
0.4
This procedure is summarized in the Algorithm 2, where, as in Algorithm 1,
a preliminary Step 0 inherited from Algorithm 1 is assumed. 0.5
0.2
Algorithm 2. Instream flow assessment method
0 0
Step 0: Select a critical reach based on FISH; 0 50 100 150
Step 1: Run the hydraulic and water quality models for the entire flow set at the distance from the upstream end (km)
selected reach;
Step 2: Evaluate FISH according to Algorithm 1 as a function of flow; Fig. 9. FISH computation (left scale) for the Arno River in the summer low flow
Step 3: Compute the FISH cumulative distribution as a function of flow and condition ðQSLF ¼ 0:5 m3 s1 Þ for the barbel (a) and for the E. chub (b). As a compar-
determine the lower FISH limits; ison some observed fish biomass densities for the same conditions are reported (right
Step 4: Determine the instream inflow range {Qinstream} according to Eq. (9). scale).
34 S. Marsili-Libelli et al. / Environmental Modelling & Software 41 (2013) 27e38

Table 1
Rules of the fuzzy inference engine for the barbel.

1. If (DO is M) and (NH3 is H) and (U is M) and (H is L) then (S is ML)


2. If (DO is L) and (NH3 is M) and (U is M) and (H is L) then (S is L)
3. If (DO is H) and (NH3 is H) and (U is M) and (H is M) then (S is MH)
4. If (DO is M) and (NH3 is M) and (U is H) and (H is H) then (S is H)
5. If (DO is L) and (NH3 is L) and (U is M) and (H is M) then (S is ML)
6. If (DO is M) and (NH3 is M) and (U is M) and (H is H) then (S is MH)
7. If (DO is M) and (NH3 is M) and (U is M) and (H is L) then (S is MH)
8. If (DO is M) and (NH3 is H) and (U is M) and (H is M) then (S is MH)
9. If (DO is L) and (NH3 is H) and (U is L) and (H is L) then (S is L)
10. If (DO is M) and (NH3 is M) and (U is H) and (H is M) then (S is H)
11. If (DO is H) and (NH3 is H) and (U is H) and (H is H) then (S is H)
12. If (DO is H) and (NH3 is H) and (U is M) and (H is H) then (S is MH)
13. If (DO is L) and (NH3 is M) and (U is L) and (H is H) then (S is L)
14. If (DO is M) and (NH3 is M) and (U is M) and (H is M) then (S is MH)

in this case, the 75th percentile for the barbel is lower than a variety of simulated scenarios involving differing hydraulic and
FISHmin ¼ 0.7, the latter value was used to compute the minimum water quality data. Given the scarce data availability often
instream flow. The (b) graph shows the flow range satisfying these encountered in instream flow assessment studies, this paper makes
conditions for both species, whereas the corresponding FISH values minimum use of field data and emphasizes the role of the expert in
are represented by the thick lines. This example shows that, for low supplementing the missing data, in specifying the habitat charac-
flows, the reach is better suited for the barbel, whereas the E. chub teristics, and in the interpretation of results. Its most relevant
would be favoured by higher flows, as confirmed by field obser- aspects are: (a) the use of fuzzy logic to incorporate expert
vations (Nocita et al., 2005, 2009). knowledge as a supplement to the scarcity of field data, (b) two
sensitivity methods for assessing the habitat suitability, (c) the
3.2. Application to the Serchio River extension to a larger spatial scale by generating a large number of
simulated hydraulic and water quality scenarios for a what-if
The Serchio River represents a high quality riverine ecosystem, analysis. These features are now discussed in detail.
which may be threatened by a planned water diversion to feed the
nearby Massaciuccoli Lake. The downstream impact of such 4.1. The use of fuzzy logic
a scheme has not yet been considered, and for this reason an
instream flow assessment was performed by applying Algorithm 2 The fuzzy approach significantly improves the flexibility and
to the river portion shown in the inset of Fig. 3. After calibrating transparency of the method, allowing the fusion between field data
a combined HEC-RAS/QUAL2Kw model, as already shown in Fig. 5, and non-numerical information, which is often associated with
FISH analysis was applied. Assuming a conservative QSLF ¼ 13 habitat description. It also adds flexibility in the definition of suit-
(m3 s1) reach #5 upstream of the Ozzeri channel appeared to be ability curves, for which a general consensus is still lacking. The
the most critical and Algorithm 2 showed that the planned diver- fuzzy inference rules can be altered by the expert to adapt them to
sion significantly widens the barbel habitat, whereas the optimal the case under study. In terms of transparency, the consistency of
flow range for the E. chub is shifted towards higher flows, as shown the rule set can be checked by inspecting the percentage of acti-
in Fig. 14 and Table 3. vation of each rule along the reaches. Finally, describing the habitat
with a single numerical index simplifies its comprehension by non-
4. Discussion technical managers and stakeholders.
The introduction of fuzzy logic influences the calibration
The goal of this paper was to extend the conventional habitat process. The comparison between FISH and relative fish abundance
assessment method from the analysis of an existing situation to in Fig. 9 is an example of indirect calibration. Though no

Table 2
Rules of the fuzzy inference engine for the E. chub.

1. If (DO is H) and (NH3 is M) and (U is M) and (H is H) then (S is H)


2. If (DO is H) and (NH3 is M) and (U is M) and (H is L) then (S is M)
3. If (DO is L) and (NH3 is H) and (U is M) and (H is L) then (S is ML)
4. If (DO is H) and (NH3 is L) and (U is H) and (H is L) then (S is MH)
5. If (DO is H) and (NH3 is L) and (U is M) and (H is L) then (S is H)
6. If (DO is M) and (NH3 is L) and (U is M) and (H is L) then (S is ML)
7. If (DO is L) and (NH3 is M) and (U is M) and (H is L) then (S is L)
8. If (DO is H) and (NH3 is H) and (U is M) and (H is M) then (S is MH)
9. If (DO is M) and (NH3 is M) and (U is H) and (H is H) then (S is H)
10. If (DO is H) and (NH3 is M) and (U is L) and (H is L) then (S is ML)
11. If (DO is H) and (NH3 is H) and (U is H) and (H is L) then (S is ML)
12. If (DO is M) and (NH3 is M) and (U is M) and (H is H) then (S is MH)
13. If (DO is M) and (NH3 is M) and (U is L) and (H is M) then (S is MH)
14. If (DO is H) and (NH3 is M) and (U is M) and (H is M) then (S is ML)
15. If (DO is M) and (NH3 is H) and (U is M) and (H is M) then (S is MH)
16. If (DO is M) and (NH3 is M) and (U is H) and (H is M) then (S is H)
17. If (DO is H) and (NH3 is H) and (U is H) and (H is H) then (S is H)
18. If (DO is H) and (NH3 is H) and (U is M) and (H is H) then (S is MH)
19. If (DO is M) and (NH3 is M) and (U is M) and (H is M) then (S is MH)
20. If (DO is H) and (NH3 is L) and (U is M) and (H is H) then (S is MH)
S. Marsili-Libelli et al. / Environmental Modelling & Software 41 (2013) 27e38 35

8
6
4
Ψp 2
0
-2
-4 barbel
-6

5
10
U H
15 DO NH3
20
25
5
30
Fig. 10. FISH sensitivity for the barbel along the Arno River in summer low flow
0
conditions, corresponding to a headwater flow of 0.5 (m3 s1). Reaches #10 and #12
appear to be the most sensitive to flow variations. Ψp -5
-10 E. chub
-15
-20

U H
5 DO NH3
10
15
20
25
30

Fig. 12. FISH incremental sensitivity for the Arno River. Un-ionized ammonia is the
most sensitive factor, whereas DO has the least influence. Sensitivity in the down-
stream reaches approaches zero as FISH also tends to zero.

experimental FISH values were available, the qualitative corre-


spondence between these two parameters can be used as a guide-
line for rule refinement, regarded as an expert knowledge educated
adjustment (Mouton et al., 2011). Future research will consider the
optimization techniques that can further improve the fuzzy rule-
based modelling approach (Van Broekhoven et al., 2007; Mouton
et al., 2009) though this can be affected by the scarcity of data.

4.2. Sensitivity analysis

Two sensitivity indicators were developed: the first can be


considered as a preliminary detection of critical reaches prior to
instream flow assessment, whereas the second indicates which
parameter has the highest influence on the index. For example,
Fig. 10 indicates that reach #10 is critical for the barbel habitat as
a consequence of an upstream wastewater treatment plant
discharge (see Fig. 2). Likewise, reach #12 is critical for the
confluence of the Sieve River, whose characteristics were analysed
elsewhere (Marsili-Libelli and Giusti, 2008), causing an abrupt
transition of flow regime. In both cases the critical spots detected by
the algorithm were confirmed in practice (Nocita et al., 2009).
The second sensitivity indicator, described in Section 2.5.2, is
aimed at detecting the primary factor influencing the habitat
Fig. 11. FISH computation along the course of the Arno River for the whole set of
typical headwater flows. The barbel habitat (a) is confined to the upstream reaches,
suitability. It can be useful when assessing the ecological resilience
whereas the more tolerant E. chub (b) can adapt to the lower reaches. In all cases of river systems subject to large seasonal variability, both in
a flow increase is detrimental. quantity and quality. Applying this analysis to the Arno River, as
36 S. Marsili-Libelli et al. / Environmental Modelling & Software 41 (2013) 27e38

Table 3
a 1 Effect of the Serchio River diversion on the suitable flow range for the target species.

Target species Qinstream (m3 s1)


0.8 75th percentile Without diversion With diversion
Barbel 8.8e16.9 6.6e16.7
Cum. Freq.

FISHmin=0.7 limit
E. chub 22.6e31.7 24.0e32.5
0.6

0.4 shown in Fig. 12, reveals that un-ionized ammonia is the most
limiting factor, followed by depth, especially for the E. chub in the
0.2 upper reaches. In fact, this species prefers lotic habitats, whereas
the opposite is true for the barbel, thriving in pool-riffle habitats.
The separation of habitats for the two species indicated by the
0 algorithm is also confirmed by field observations (Nocita et al.,
0.5 0.6 0.7 0.8 0.9
FISH 2005).
b 0.9
Barbel E. chub 4.3. Simulated scenarios
0.85
This algorithm is aimed at analysing a series of what-if cases by
0.8 producing a variety of hydraulic and water quality combinations
ΔQE . chub and computing the resulting habitat index. For example, Fig. 11
0.75
shows how flow variations affect the habitat of both target
FISH

0.7 species in the Arno River along its course. This analysis is important
for these kinds of rivers, which are subject to wide seasonal flow
0.65 changes, from almost drought in summer and autumn to very high
flows in spring and winter. Fig. 11 shows that the barbel is more at
0.6 risk than the E. chub, though both species do suffer from an
0.55 ΔQbarbel excessive flow increase.

0.5 4.4. Instream flow assessment method


0 2 4 6 8 10 12
Flow (m3 s-1 )
Algorithm 2 extends the capabilities of Algorithm 1 by deter-
Fig. 13. Instream flow assessment at reach #10 of the Arno River around QSLF. (a) mining the flow values between which the suitability never dete-
Cumulative FISH distribution and selection of the suitability limits for the two target riorates below a given threshold. Repeated application of Algorithm
species. (b) Computation of the admissible flow range (thick lines in the lower graph) 1 for a suitable set of flows, and the subsequent statistical distri-
by considering the last quartile (FISH75) and the absolute minimum FISHmin of 0.7, bution of FISH, yields the flow intervals (DQ) in which the suitability
whichever is greater (in this example, the latter holds for the barbel). The dots and
is above a prescribed level. This analysis can be useful in gaining
diamonds refer to the actual flow values used for the interpolation (thin solid lines).
further insight into critical reaches or for assessing diversions in

a Normal barbel E. chub With diversion c


1 1
0.75 0.75
Cum. Freq.
Cum. Freq.

0.5 0.5

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
FISH FISH
b d
1 1
FISH
FISH

0.5 0.5
ΔQbarbel Δ Q E . chub Δ Qbarbel ΔQE . chub
0 0
0 10 20 30 40 50 0 5 10 15 20 25 30 35 40
Flow (m3 s-1) Flow (m3 s-1)

Fig. 14. Instream flow assessment at reach #5 of the Serchio River, upstream of the Ozzeri channel. Comparing the normal situation, graphs (a) and (b), to the diversion, graphs (c)
and (d), it appears that this is beneficial for the barbel, but slightly detrimental for the E. chub (see also Table 3).
S. Marsili-Libelli et al. / Environmental Modelling & Software 41 (2013) 27e38 37

regulated rivers. Fig. 13 shows the application of Algorithm 2 to Aricò, C., Nasello, C., Tucciarelli, T., 2009. Using unsteady-state water level data to
estimate channel roughness and discharge hydrograph. Advance in Water
reach #10 of the Arno River, in which the habitat of the two species
Resources 32, 1223e1240.
are widely separated. It also shows that the E. chub would be Arno River Water Authority, 2010. Hydrogeological Management Plan, Part II:
greatly disadvantaged by a flow decrease. A similar analysis was Hydrogeological Risk, Florence, pp. 32 (in Italian).
applied to the Serchio River in order to assess the impact of Ayllón, D., Almodóvar, A., Nicola, G.G., Elvira, B., 2011. The influence of variable
habitat suitability criteria on PHABSIM Habitat index results. River Research
a planned diversion downstream of reach #5. The results are shown and Applications. http://dx.doi.org/10.1002/rra.1496.
in Fig. 14, where the instream flow analysis in the two settings are Bovee, K.D., Lamb, B.L., Bartholow, J.M., Stalnaker, C.B., Taylor, J., Henriksen, J., 1998.
compared. The general effect of a flow decrease improves the Stream Habitat Analysis Using the Instream Flow Incremental Methodology.
Information and Technology Report USGS/BRD/ITR-1998-0004. U.S. Geological
habitat for the barbel and worsens it for the E. chub. In fact Survey-BRD, Fort Collins, CO, 130 pp.
comparing the graphs of the current situation, (a) and (b), to the Brunner, G.W., 2010. HEC-RAS, River Analysis System Hydraulic Reference Manual
ones of the diversion, (c) and (d), it appears that this moderately (Version 4.1). Report CPD-69. US Army Corps of Engineers, Hydrologic Engi-
neering Center (HEC), Davis Calif., USA, pp. 411.
expands the barbel habitat, but is slightly detrimental for the E. Chapra, S., Pelletier, G., 2003. QUAL2K: a Modeling Framework for Simulation River
chub (see also Table 3). and Stream Water Quality: Documentation and User Manual. Civil and Envi-
ronmental Engineering Dept., Tufts University, Medford, MA.
Di Baldassarre, G., Claps, P.L., 2011. A hydraulic study on the applicability of flood
5. Conclusions rating curves. Hydrology Research 42 (1), 10e19.
Di Baldassarre, G., Montanari, A., 2009. Uncertainty in river discharge observations:
a quantitative analysis. Hydrology and Earth System Sciences 13, 913e921.
This paper has presented an instream habitat indicator based on
Environment Agency, 2007. Proposed EQS for Water Framework Directive Annex
fuzzy logic and large scale environmental hydraulic and water VIII: Substances: Ammonia (Un-ionised). Science Report SC040038/SR2, 2007.
quality modelling. Its innovative aspects are: (a) the use of Federici, P.R., 1998. The highest risk in Tuscany: the flooding of the Arno River. In:
Leone, U. (Ed.), Risk and Environmental Deterioration in Italy. Patron, Bologna,
a TakagieSugeno fuzzy system to define the habitat suitability in
pp. 189e213.
terms of field data and expert knowledge, (b) the extension of the Fraternali, P., Castelletti, A., Soncini-Sessa, R., Vaca Ruiz, C., Rizzoli, A.E., 2012.
analysis to a wide combination of synthetic hydraulic and water Putting humans in the loop: social computing for Water Resources Manage-
quality scenarios, (c) its use in the instream flow assessment study ment. Environmental Modelling & Software 37, 68e77.
Fukuda, S., Hiramatsu, B., 2008. Prediction ability and sensitivity of artificial
through a new definition of habitat sensitivity. intelligence-based habitat preference models for predicting spatial distribution
Regarding this last aspect, two different sensitivities were of Japanese medaka (Oryzias latipes). Ecological Modelling 215, 301e313.
derived (Algorithm 2) to assist in the detection of critical reaches Fukuda, S., 2009. Consideration of fuzziness: Is it necessary in modelling fish habitat
preference of Japanese medaka (Oryzias latipes)? Ecological Modelling 220,
for further instream flow investigation. Combining a large scale and 2877e2884.
a micro-scale perturbation analysis this second algorithm can Fukuda, S., De Baets, B., Mouton, A.M., Waegeman, W., Nakajima, J., Mukai, T.,
detect the river reaches where the habitat is highly influenced by Hiramatsu, K., Onikura, N., 2011. Effect of model formulation on the optimiza-
tion of a genetic TakagieSugeno fuzzy system for fish habitat suitability eval-
variations of anyone of the four parameters. uation. Ecological Modelling 222, 1401e1413.
Two case studies in Italy were presented. In the case of the Arno Giusti, E., Marsili-Libelli, S., Mattioli, S., 2011a. A fuzzy quality index for the envi-
River, the analysis was extended to its middle course and revealed ronmental assessment of a restored wetland. Water Science & Technology 63
(9), 2061e2070.
several critical spots, especially for the valued barbel species, which
Giusti, E., Marsili-Libelli, S., Gualchieri, A., 2011b. Modelling a coastal lake for flood
were verified in practice. In the case of the Serchio River, the and quality management. In: Proc. Watermatex 2011, 8th IWA Symposium on
analysis was motivated by an envisioned diversion to feed a nearby Systems Analysis and Integrated Assessment, San Sebastian (SP) 19e22 June
2011, pp. 396e403.
lake, showing that this would expand the barbel habitat, while
Krueger, T., Page, T., Hubacek, K., Smith, L., Hiscock, K., 2012. The role of expert
shifting the habitat of the E. chub towards higher flows. opinion in environmental modelling. Environmental Modelling & Software 36,
The proposed Algorithm 1, though developed for specific situ- 4e18.
ations, is highly portable to other contexts by adapting the suit- Lamouroux, N., Capra, H., 2002. Simple predictions of instream habitat model
outputs for target fish populations. Freshwater Biology 47, 1543e1556.
ability curves and some rules of the fuzzy inferential systems, being Lamouroux, N., Souchon, Y., 2002. Simple predictions of instream habitat model
otherwise quite general, whereas the instream flow assessment outputs for fish habitat guilds in large streams. Freshwater Biology 47, 1531e
method (Algorithm 2) is quite general and can be applied without 1542.

Macura, V., Skrinár, A., Kaluz, K., Jal 
covíková, M., Skrovinová, M., 2011. Influence of
modifications to any river system. Extending the analysis to mac- the morphological and hydraulic characteristics of mountain streams on fish
roinvertebrate species is currently being considered, in order to habitat suitability curves. River Research and Applications. http://dx.doi.org/
widen the notion of habitat suitability to the entire food chain. 10.1002/rra.1518.
Marchini, A., Facchinetti, T., Mistri, M., 2009. F-IND: a framework to design fuzzy
indices of environmental conditions. Ecological Indicators 9, 485e496.
Acknowledgements Marsili-Libelli, S., Giusti, E., 2008. Water quality modelling for small river basins.
Environmental Modelling & Software 23, 451e463.
Marsili-Libelli, S., Nocita, A., Giusti, E., Saccà, M., 2008. A new definition of
This research was partly supported by the Serchio River Water minimum sustainable flow based on water quality modelling and fuzzy pro-
Authority in the framework of the Massaciuccoli Lake remediation cessing. In: Proc.. iEMSs 2008: International Congress on Environmental
Modelling & Software, Barcelona (SP), vol. 1, pp. 288e294.
project. The authors also wish to acknowledge the collaboration of
McMahon, T.E., 1982. Habitat Suitability Index Models: Creek Chub. FWS/OBS-82/
the Tuscany Region Environmental Protection Agency (ARPAT) and 10.4, Biological Services Program, Fish and Wildlife Service. U.S. Department of
the Arno River Water Authority in providing data and information Interior, Washington DC, Fort Collins, Colorado, pp. 34.
useful for the development of the Arno River water quality model. McMahon, T.E., Gebbart, G., Maughan, O.E., Nelson, P.C., 1984. Habitat Suitability
Index Models and Instream Flow Suitability Curves: Warmouth. FWS/OBS-82/
The preliminary work on the Serchio river performed by Dr. Fran- 10.67. Division of Biological Service, U.S. Department of the Interior, Wash-
cesca Renzetti in her MS thesis (Renzetti, 2010) is also gratefully ington DC, pp. 33.
acknowledged. The authors also express their gratitude to the Milhous, R.T., Updike, M.A., Schneider, D.M., 1989. Physical Habitat Simulation
System Reference Manual e Version II. Biological Report 89(16). U.S. Fish and
anonymous reviewers, who helped improve the paper with their Wildlife Service, Washington, D.C.
valuable suggestions. Moir, H.J., Gibbins, C.N., Soulsby, C., Youngson, A.F., 2005. PHABSIM modelling of
Atlantic salmon spawning habitat in an upland stream: testing the influence of
habitat suitability indices on model output. River Research and Applications 21,
References 1021e1034.
Mouton, A.M., Schneider, M., Kopecki, I., Goethals, P.L.M., De Pauw, N., July 2006.
Abramowitz, M., Stegun, I.A., 1965. Handbook of Mathematical Functions with Application of MesoCASiMiR: assessment of Baetis rhodani habitat suitability.
Formulas, Graphs, and Mathematical Tables. Dover Publications Inc, New York, In: Voinov, A., Jakeman, A.J., Rizzoli, A.E. (Eds.), Proceedings of the iEMSs Third
pp. 1046. Biennial Meeting: “Summit on Environmental Modelling and Software”.
38 S. Marsili-Libelli et al. / Environmental Modelling & Software 41 (2013) 27e38

International Environmental Modelling and Software Society, Burlington, USA. Prato, T., 2007. Assessing ecosystem sustainability and management using fuzzy
CD ROM. Internet: http://www.iemss.org/iemss2006/sessions/all.html. logic. Ecological Economics 61, 171e177.
Mouton, A.M., Schneider, M., Depestele, J., Goethals, P.L.M., De Pauw, N., 2007. Fish Raleigh, R.F., Zuckermann, L.D., Nelson, P.C., 1986. Habitat suitability index models
habitat modelling as a tool for river management. Ecological Engineering 29, and instream flow suitability curves: brown trout, revised. U.S. Fish and Wildlife
305e315. Service Biology Reports 82 (10.124), 36e43.
Mouton, A.M., Schneider, M., Peter, A., Holzer, G., Müller, R., Goethals, P.L.M., De Renzetti, F., 2010. Instream flow assessment of the terminal part of the Serchio River
Pauw, N., 2008. Optimisation of a fuzzy physical habitat model for spawning based on hydraulic and quality parameters. MS thesis in Environmental Engi-
European grayling (Thymallus thymallus L.) in the Aare river (Thun, neering, Faculty of Engineering, University of Florence, pp. 80 (in Italian).
Switzerland). Ecological Modelling 215, 122e132. Rinderknecht, S.L., Borsuk, M.E., Reichert, P., 2012. Bridging uncertain and ambig-
Mouton, A.M., De Baets, B., Goethals, P.L.M., 2009. Knowledge-based versus data- uous knowledge with imprecise probabilities. Environmental Modelling &
driven fuzzy habitat suitability models for river management. Environmental Software 36, 122e130.
Modelling & Software 24, 982e993. Schneider, M., Jorde, K., Zoellner, F., Kerle, F., 2001. Development of a user-friendly
Mouton, A.M., Alcaraz-Hernández, J.D., De Baets, B., Goethals, P.L.M., Martínez- software for ecological investigations on river systems, integration of a fuzzy
Capel, F., 2011. Data-driven fuzzy habitat suitability models for brown trout rule-based approach. In: Proceedings of Environmental Informatics 2001, 15th
in Spanish Mediterranean rivers. Environmental Modelling & Software 26, International Symposium, Informatics for Environmental Protection, ETH Zur-
615e622. ich, Switzerland.
Nocita, A.M., Pini, G., Bartali, S., Pini Prato, E., Schweizer, S., 2005. Progetto pilota per Stalnaker, C., Lamb, B.L., Henriksen, J., Bovee, K., Bartholow, J., 1995. The Instream
la determinazione e verifica del Deflusso Minimo Vitale su base biologica per Flow Incremental Methodology: a Primer for IFIM. National Ecology Research
l’asta del Fiume di Sieve e (BioDeMiV). In: Proc. 15th Meeting of the Italian Centre, Internal. Publication. U.S. Department of the Interior, National Biological
Society of Ecology (in Italian). Service, Fort Collins, Colorado, pp. 43.
Nocita, A., Massolo, A., Vannini, M., Gandolfi, G., 2009. The influence of calcium Svobodova, Z., Lloyd, R., Machova, J., Vykusova, B., 1993. Water Quality and fish
concentration on the distribution of the river bullhead Cottus gobio L. (Tele- health. EIFEC Technical Paper, FAO, Rome, n.54.
ostes, Cottidae). Italian Journal of Zoology 76 (4), 348e357. Takagi, T., Sugeno, M., 1985. Fuzzy identification of systems and its applications to
Nocita, A.M., Marsili-Libelli, S., Giusti, E., Saccà, M., 2010. Una nuova definizione di modeling and control. IEEE Transactions on System, Man, and Cybernetics 15,
Deflusso Minimo Vitale basata su parametri di qualità dell’acqua e sulla logica 116e132.
fuzzy. Studi Trentini di Scienze Naturali 87, 229e231 (in Italian). U.S. Army Corps of Engineers, 1991. HEC-6, Scour and Deposition in Rivers and
Pappenberger, F., Beven, K., Horritt, M., Blazkova, S., 2005. Uncertainty in the cali- Reservoirs, User’s Manual, Davis, CA.
bration of effective roughness parameters in HEC-RAS using inundation and Van Broekhoven, E., Adriaenssens, V., De Baets, B., Verdonschot, P.F.M., 2006. Fuzzy
downstream level observations. Journal of Hydrology 302, 46e69. rule-based macro-invertebrate habitat suitability models for running waters.
Pappenberger, F., Matgen, P., Beven, K.J., Henry, J.B., Pfister, L., de Fraipont, P., 2006. Ecological Modelling 198, 71e84.
Influence of uncertain boundary conditions and model structure on flood Van Broekhoven, E., Adriaenssens, V., De Baets, B., 2007. Interpretability-preserving
inundation predictions. Advanced Water Resources 29, 1430e1449. genetic optimization of linguistic terms in fuzzy models for fuzzy ordered
Parasiewicz, P., 2001. MesoHABSIM: a concept for application of instream flow classification: an ecological case study. International Journal of Approximate
models in river restoration planning. Fisheries 26 (9), 6e13. Reasoning 44, 65e90.
Parasiewicz, P., 2007a. The MesoHABSIM model revisited. River Research and Van Winkle, W., Jager, H.I., Railsback, S.F., Holcomb, B.D., Studley, T.K., Baldrige, J.E.,
Applications 23 (8), 893e903. 1998. Individual-based model of sympatric populations of brown and rainbow
Parasiewicz, P., 2007b. Overcoming the limits of scales. River Research and Appli- trout for instream flow assessment: model description and calibration.
cations 23 (8), 891e892. Ecological Modelling 110, 175e207.
Pelletier, G.J., Chapra, S.C., Tao, H., 2006. QUAL2Kw, A framework for modeling Wieland, R., Mirschel, W., Groth, K., Pechenick, A., Fukuda, K., 2011. A new method
water quality in streams and rivers using a genetic algorithm for calibration. for semi-automatic fuzzy training and its application in environmental
Environmental Modelling & Software 21, 419e425. modeling. Environmental Modelling & Software 26, 1568e1573.
Petts, G.E., 2009. Instream flow science for sustainable river management. Journal of Young, W.J., Lam, D.C.L., Ressel, V., Wong, I.W., 2000. Development of an environmental
American Water Research Association 45 (5), 1071e1086. flows decision support system. Environmental Modelling & Software 15, 257e265.

You might also like