Sunteți pe pagina 1din 11

Catena 157 (2017) 1–11

Contents lists available at ScienceDirect

Catena
journal homepage: www.elsevier.com/locate/catena

Hydrological response to land use and climate changes in a rural hilly basin MARK
in Italy
Marco Napolia,⁎, Luciano Massettib, Simone Orlandinia
a
Dep. of Agrifood Production and Environmental Sciences, Univ. of Florence, Piazzale delle Cascine, 18, 50144 Florence, Italy
b
Institute of Biometeorology, CNR, via G. Caproni 8, 50145 Florence, Italy

A R T I C L E I N F O A B S T R A C T

Keywords: During the past sixty years, the social and economic development causes a change of traditional activities and a
Rainfall regime movement of the population towards urbanized areas in different regions of Italy. The replacement of vegetated
Streamflow land surface with impervious surface areas changed the hydrologic fluxes of a drainage basin with important
Land use change consequences on the basin resilience to rainfall events. The main objective of this paper is to present the
ArcSWAT model
methodology and the results of a study for quantifying and analysing the hydrological responses to land use and
Urbanization
climate changes by means of ArcSWAT model. The case study was the runoff dynamic of the Elsa river basin
which was analysed by climate and land use change along more than fifty years (1954–2007). Results showed
that, the model was effective in simulating daily base, peak and total runoff, scoring a very good performance.
Analysis indicated that given the same climatic conditions, changes in land use played a considerable role since
in peak and total runoff formation that were increased with the increase of artificial surfaces and specialised
agriculture. This type of analysis proved to be effective in analysing past and future hydrological dynamics of the
basin. Our findings suggest that such an approach could help policymakers involved in the land use planning in
taking into account land use change, since it can amplify the effects of climate and demographic changes on
basin hydrodynamics.

1. Introduction ization that negatively affected the extension and intensity of agricul-
tural drainage system, thus reducing the effectiveness of its function
The social and demographic changes related to economic develop- (Landi, 1989). In this context, from 1950, the reduction in ditching, tile
ment during the past sixty years have had contrasting impacts around draining and channelization activity was huge, particularly in agricul-
the world (MacDonald et al., 2000; Stonestrom et al., 2009). The hilly tural upland areas. Since the early 50s, the new olive orchard and
countryside suffered depopulation and abandonment of traditional vineyards, which in Tuscany were mainly located on terraced hillslopes
activities, while plain areas experienced an increase of population for reasons related to both tradition and quality of production, were
and industrial activities. Moreover, the socio-economic development planted up and down the slope to make machine operation more
drove land use changes such as conversion of cropland to urban area as feasible and cheaper (Napoli and Orlandini, 2015; Napoli et al., 2016).
well as changes within land use and land cover classes such as a change The consequence of this is the lack of structures controlling runoff, such
of crops or crop rotations (Stonestrom et al., 2009). Tuscany (central as the setting up of an adequate drainage system or implementing
Italy) experienced an economic growth over the past sixty years, which conservation practices can make a watershed less resilient to the intense
results in the expansion of urban, industrial and commercial areas at storms. This could be amplified by the projected increase of extreme
the expense of agricultural plains. According to the “6th General Census rainfall events in the Mediterranean area (Stocker et al., 2013).
of Agriculture” (ISTAT, 2010), Italy and Tuscany registered about 50% Because of the importance of this societal and economic change, the
reduction of farms respect to 1982 and about 25% decrease of land Soil Thematic Strategy of the European Commission pointed out that
dedicated to agriculture. Furthermore, large extension of homogeneous the land surface covering by impermeable material, such as buildings
cultivations (specialised agriculture) was preferred to small heteroge- and roads, is one of the major threats to soil quality and functionality
neous cultivation patterns. Moreover, agriculture was subjected to the (COM, 2006). In fact, urbanization determines loss of fertile soils,
transition from traditional agronomic practices to agricultural mechan- reductions in infiltration, canopy interception and groundwater re-


Corresponding author.
E-mail address: marco.napoli@unifi.it (M. Napoli).

http://dx.doi.org/10.1016/j.catena.2017.05.002
Received 26 October 2016; Received in revised form 2 May 2017; Accepted 5 May 2017
0341-8162/ © 2017 Elsevier B.V. All rights reserved.
M. Napoli et al. Catena 157 (2017) 1–11

charge of the basin (Aronica and Cannarozzo, 2000; Rose and Peters, and flows in a south–north direction with an average streamflow of
2001), and has the potential to increase runoff, resulting in changes in about 5.4 m3 s− 1.
flood frequency and intensity, as well as annual and seasonal flows The major soils in Elsa basin can be schematically summarized as
(Brath et al., 2006; Costa et al., 2003; Crooks and Davies, 2001; Huang follows (Fig. 2). Pliocene sand derived soils, mainly characterized by a
et al., 2008; Olang and Furst, 2010; Poelmans et al., 2011; Wang et al., sandy-loam texture (Calcaric Cambisols and Haplic Calcisols; FAO,
2006). Together with changes in land cover, systematic changes in the 1998) were predominant in the north-western border of the basin,
climatic variables involved in the water cycle [e.g. precipitation, while Pliocene clay derived soil were predominant in the north (Vertic
temperature and evapotranspiration (ET)] may induce notable altera- Cambisols and Calcaric Regosols). The Elsa river valley floor was
tions in the runoff released by hillslope. characterized by silty loam soil originated over alluvial deposits
In response, public policy decision-making processes are now (Calcaric Cambisols). Pliocene clay originated soil (Silti Calcaric
seeking both economic and conservation goals. More informed deci- Cambisols and Calcari Stagnic Cambisols), characterized by a texture
sions for basin planning and water allocation must rely on the better ranging from silty loam to sandy clay loam, dominate the central part of
understanding of highland basin hydrology and the relationship the basin. Sandy loam soils (Calcari Epileptic Cambisols) and while clay
between land use practices, flow generation processes, and associated soils (Chromi Profondic Luvisols) were predominant in the east of the
water distribution and use. Furthermore, the ability to evaluate basin basin in high steepness and moderate hillslope, respectively. Silty clay
hydrology beyond just stream flow is crucial for determining spatially- loam soils were found in the south-west (Calcaric Cambisols). While,
explicit relationship between landscape structure, configuration of land travertine rock originated soil (Cutani Chromic Luvisols), mainly
use change, and the hydrology across the landscape. Within this characterized by a clay texture, were found in the south.
context, managers of basin areas can be helped by tools for assessing
the impact of rainfall regimes and land use changes on runoff
generation, as well as for planning land policy to prevent and mitigate 2.2. Data for modeling
negative impact like soil erosion and floods.
Several studies investigated hydrological condition changes induced The study was conducted using meteorological and hydrological
by land use or by climate, as well as their combined effect (Ghaffari data across the period 1954–2010 provided by Regional Hydrological
et al., 2010; Juckem et al., 2008; Semadeni-Davies et al., 2008; Wang Service (SIR, 2016). Daily air temperature (Ta; °C) and daily rainfall
et al., 2008; Cuo et al., 2009; Li et al., 2009; Ma et al., 2009; Choi, 2008; depth (P; mm) were obtained from a total of twenty-five meteorological
Franczyk and Chang, 2009; Dams et al., 2008; Tu, 2009; Qi et al., 2009; stations in or near the Elsa river basin (Fig. 1). For each station, missing
Wijesekara et al., 2012). However, most of these studies used hydro- data were estimated by means of linear interpolation with nearest
logical simulation models fed by data provided by climate change or station with known data. Then, Ta and P spatialization was carried out
land use future scenarios (Dams et al., 2008; Chung et al., 2011; at daily time step by means of the regularized spline with tension
Wijesekara et al., 2012), while only few studies investigated land use (Napoli et al., 2014).
change impacts on water resources on concrete data like historical data Daily runoff data were taken by the runoff station near
series (Wagner et al., 2013). Castelfiorentino (Fig. 1). A total of 17 years of hydrological data,
Our study was focused on the Elsa river basin that was affected by having a complete daily dataset were selected. Five years (1954,
heavy rains that often create flood conditions in towns and settlements 1978, 1988, 1996, and 2007) were used for model calibration and
along the river and these occurrences were well studied by the regional land use change analysis, while twelve years were used for model
water authority (PGRA, 2016). validation (Table 1).
The objective of this study was to present a methodology to analyse Soil parameters such as soil profile, texture, and organic matter
the impacts of climate and land use changes on the runoff generation were derived from the 1:250,000 soil map of Tuscany (Gardin and
and demonstrate the potential application of such a methodology for Vinci, 2016). Stabilized infiltration rate (KSAT) and retention curve were
the improvement of land management towards a sustainable develop- derived by soil parameters using the Soil Water Characteristics Hy-
ment. draulic Properties Calculator (Saxton and Rawls, 2006). The official
For the purpose of this study land use changes in the Elsa river basin 10 m pixel Digital Elevation Model (DEM) of Tuscany Region (SITA,
were identified and quantified over a period of 54-year by analysing 5 2016) was used to derive the elevation and the slope.
orthophoto images (1954, 1978, 1988, 1996, and 2007); the ArcSWAT The study area experienced socio-economic development during the
2009·93·7b hydrologic model (Neitsch et al., 2011) was calibrated, past sixty years. In order to analyse the land use changes caused by this
validated and used to quantify impacts of land use change on runoff development, five land use classification maps (resolution of 200 m)
formation. were produced for the years 1954, 1978, 1988, 1996, and 2007 (Fig. 3).
Geo-referenced aerial ortho-photos (resolution of 1 m) of the study area
2. Material and methods (SITA, 2016) were used for visual interpretation and land use type
boundaries were on-screen digitized. The land use types were classified
2.1. Study area according to the Corine Land Cover (CLC) classification of the European
Environment Agency (©EEA, Copenhagen, 2000). Three main land use
The Elsa river is located in Province of Florence in central Tuscany classes were defined: artificial surfaces (comprising continuous urban
(Italy) (WGS84 – lat 43°27′–43°67′ N; lon 10°88′–11°36′ E) (Fig. 1). The fabric, discontinuous urban fabric, industrial or commercial units, road
study basin covers an area of about 931 km2 with elevations varying and rail networks and associated land); forest and semi natural areas
from 46 m a.s.l. at the runoff station (near Castelfiorentino, Florence) to (comprising broad-leaved forest and transitional woodland-shrub);
300 m a.s.l. at the river head. There are 18 municipalities that are fully agricultural land. Moreover, to represent the impact on the effective-
or partially within the basin area. Population estimates as of 2011 ness of agricultural drainage system determined by the transition from
(ISTAT, 2011) indicated that approximately 116,570 people lived traditional agronomic practices to agricultural mechanization, the
within the basin area with a density of 125 people/km2. The climate agricultural land use was subdivided into four land cover classes:
is typical of the European Mediterranean area, with average annual arable land (comprising field crops and not irrigated field crops);
values for rainfall and temperature of 847 mm and 13.7 °C over the last heterogeneous agricultural areas (comprising annual crops associated
50 years, respectively. The rainfall is usually concentrated mainly in with permanent crops and complex cultivation patterns); olive groves;
two periods: September to December and March to May (Napoli et al., and vineyards.
2014). The Elsa river, the main stream in the basin, is about 52 km long

2
M. Napoli et al. Catena 157 (2017) 1–11

Fig. 1. Location map of the study area. On the left side, a map of Tuscany with highlighted the basin of Elsa river. On the right side, a digital elevation model of the basin with the position
of the meteorological stations.

2.3. SWAT model set-up

Water flow was modeled using the ArcSWAT 2009·93·7b extension


for ArcGIS 9.3 (Neitsch et al., 2011). SWAT was a continuous-time,
spatially distributed model for simulating the hydrological cycle, the
crop growth and material transfers (soil erosion, nutrient and organic
chemical transport and fate) at the basin scale. The water balance
module took into account soil water content, precipitation, surface
runoff, evapotranspiration, percolation and bypass flow, and return
flow. SWAT divided a basin into units composed of unique combina-
tions of homogeneous slope, soil, land use and management, called
hydrological response units (HRU) (Flugel, 1995). HRU's hydrology was
simulated independently from one another. Then, the predicted runoff
flows were routed within the channel network, allowing large basin to
be simulated. The total flow (Qtot; mm) routing through channel system
to the gauges comprised surface runoff (Qsurf; mm), lateral flow
(Qlatflow; mm) and base flow generated from the shallow aquifer (Qgw;
mm) (Fig. 4). The sum of Qsurf and Qlatflow corresponds to the flow of a
stream in response to a rainfall event (Qsrre; mm). In this study, Qsurf
was estimated using the Soil Conservation Service–Curve Number
method (CN) (SCS, 1972). The CN method takes into account the soil,
the land cover, the growing season and antecedent soil moisture
conditions (AMC), incorporating them in a single Curve Number
parameter (Bo et al., 2011; Mishra and Singh, 2004; Pachpute et al.,
2009; Winnaar et al., 2007).
The CN values are differentiated in Hydrologic Soil Groups (HSG)
(Table 2), represented by categories A through D, as a function of soil
infiltration for particular antecedent soil water conditions (NRCS,
2009). The HSG–A and HSG–D represent high permeability soils with
low runoff potential and very shallow or high clay content soils with
high runoff potential, respectively. The HSG–B and HSG–C are inter-
mediate classes (Sekar and Randhir, 2007). The HSG distribution was
Fig. 2. Pedological map of the Elsa river basin (on the top) and the hydrologic soil group reported in Fig. 2.
(HSG) map according to the USDA classification (on the bottom).
AMC was determined by cumulating the rainfall depth in the
antecedent five days. Then, AMC was classified into three levels
AMC-I (dry condition), AMC-II (average condition) and AMC-III (wet

Table 1
Land use percentages for calibration and validation years.

Year Forest and semi-natural areas Heterogeneous agricultural areas Arable land Olive groves Vineyards Artificial surfaces Validation years

1954 41.7% 49.9% 7.2% 0.2% 0.7% 0.3% 1956, 1957, 1958
1978 42.0% 25.0% 28.1% 0.5% 3.3% 1.1% 1980, 1981, 1983
1988 42.4% 22.7% 24.9% 1.8% 6.6% 1.6%
1996 43.3% 20.9% 22.6% 2.2% 8.6% 2.4% 1998, 1999, 2000
2007 43.5% 18.2% 23.4% 2.8% 9.6% 2.5% 2008, 2009, 2010

3
M. Napoli et al. Catena 157 (2017) 1–11

Fig. 3. Land use classification maps (resolution of 200 m) for the years 1954, 1978, 1988, 1996, and 2007.

25,400
S= − 254
CN (2)
A revised method to compute the S parameter (Eq. (3)) was
implemented in SWAT model (Kannan et al., 2008) allowing the S
parameter to vary with plant reference evapotranspiration (PET;
mm·day− 1) and improving the runoff estimates particularly for basin
with low storage and shallow soils (Neitsch et al., 2005, 2011).

St = St−1 + PETt exp ( −BSt−1


Smax )−P t−1 + Qt−1
with St ≤ Smax (3)
where, St and St − 1 represented the S parameter at present time and at
Fig. 4. Schematic of flow components in the ArcSWAT model. P is the rainfall depth the previous time step respectively, PETt was the reference evapotran-
(mm); Qtot is the total flow (mm) routing through channel system; Qsurf is the surface
spiration at present time; B was the depletion coefficient which can
runoff (mm); Qlatflow is the lateral flow (mm); Qsrre is the flow of a stream in response to a
rainfall event (mm); Qgw is the base flow generated from the shallow aquifer (mm); vary from 0 to 2, Pt − 1 and Qt − 1 were, respectively, the rainfall depth
wrchrg.sh is the amount of recharge entering the swallow aquifer (mm). and the runoff height at the previous time step, and finally, Smax was the
retention parameter maximum value.
condition), as function of rainfall depth levels during both growing and For each HRU, CN(AMC-II) and CN(AMC-III) values were used in
dormant seasons. Eq. (2) to compute the initial S parameter and the Smax, respectively.
The CN for average condition CN(AMC-II) for each HSG categories Moreover, for each simulation, the initial Q was computed with Eq. (1)
was obtained from tabular data (Table 2), while CN under dry (CN by using the P value of the first day of the time series and the initial S
(AMC-I)) and wet conditions (CN(AMC-III)) were computed according parameter.
to Chow et al. (1988). The PET was computed by using a modified Hargreaves–Samani
The Qsurf equation (Eq. (1)) was given by: equation, accounting for local conditions (Napoli et al., 2014) (Eq. (4)).
Ra ⎛T + Tmin ⎞
⎧ P−0 .2 S2 P + 0.8S−1 PET = 0.00132∙ ∙Tmax −Tmin 0.64 ∙⎜ max + 15.2⎟
Q surf = ⎨ for P > 0.2S λ ⎝ 2 ⎠ (4)
⎩0 for P ≤ 0.2S (1)
where, Tmax and Tmin were the daily maximum air temperature (°C) and
where P is the rainfall depth (mm) and S is the potential soil moisture daily minimum air temperature (°C) respectively, and Ra is the daily
retention (mm). extraterrestrial solar radiation (MJ·m− 2·day− 1) and λ is the latent heat
S is related to CN and therefore, also related to watershed features of evaporation (kg·m− 2).
and antecedent moisture conditions (Eq. (2)): A kinematic storage model was used to compute the Qlatflow in each
soil layer (Sloan and Moore, 1984) (Eq. (5)):

Table 2
Average Curve Number values for Antecedent Moisture Condition II (CN(AMC-II)) for different land uses and Hydrological Soil Groups (HSG) categories A through D.

Land use (CLC2000 classification) Source CN-II

HSG-A HSG-B HSG-C HSG-D

Road and rail networks and associated land USDA NRCS, 2004 - paved parking lots, roofs, driveways, etc. 98 98 98 98
Industrial or commercial units USDA NRCS, 2004 - industrial 81 88 91 93
Continuous urban fabric USDA NRCS, 2004 - 1/8 acre or less (town houses) 77 85 90 92
Discontinuous urban fabric Napoli et al., 2014 75 82 87 90
Non-irrigated arable land Napoli et al., 2014 67 78 85 89
Complex cultivation patterns Napoli et al., 2014 57 73 82 86
Vineyards Napoli et al., 2014 60 70 81 86
Transitional woodland-shrub USDA NRCS, 2004 - shrub - fair 55 72 81 86
Olive groves Napoli et al., 2014 45 67 77 82
Broad-leaved forest USDA NRCS, 2004 - woods - fair 36 60 73 79

4
M. Napoli et al. Catena 157 (2017) 1–11

Table 3
SWAT parameter values determined by model calibration.

Parameter Description Units Forest Heterogeneous Arable land Olive tree Vine Urban

Alpha_BF Base flow alpha factor 0.4


CH_K2 Muskingum routing coefficient 33
CH_N2 Manning's N value for the main channel 0.15
CN2 Runoff curve number for AMC-II condition Data from Table 2
cncoeff B: Depletion coefficient 0.87
EPCO Plant uptake compensation factor 0.31
ESCO Soil evaporation compensation coefficient 0.9
GW_DELAY Groundwater delay d 17
GWQMN Depth of water in the shallow aquifer required for return flow mm 552
SOL_AWC Soil layer available water content mm Set at HRU level as function of the soil characteristics
SOL_K Ksat: Soil saturated hydraulic conductivity mm h− 1 Set at HRU level as function of the soil characteristics
SURLAG Surface runoff lag time 12
ALAI_MIN Minimum LAI for plant during dormant period m2·m− 2 0.03 0.0805 0 1.6 0.01
BIO_E Biomass-energy ratio kg·ha− 1 per 15 23.33 24.2 8 23
MJ·m− 2
BIO_LEAF Tree fraction biomass converted to residue during dormancy Fraction 0.15 0.0155 0 0.01 0.3
BLAI Max potential leaf area index m2·m− 2 7 4.147 4.38 2.1 2
CANMX Maximum canopy storage mm 2.1 2.1 2.1 2.1 2.1
CHTMX Max canopy height m 10 0.822 0.58 3 3
DLAI Fraction of growing season when leaf area declines heat units per heat 0.99 0.874 0.87 0.9 0.92
units
FIMP Impervious fraction for urban land type 0.45
FCIMP Directly connected impervious fraction for urban land type 0.4
FRGRW1 Fraction of growing season corresponding to 1st point on optimal leaf Fraction 0.1 0.099 0.1 0.1 0.08
area development curve
FRGRW2 Fraction of growing season corresponding to 2nd point on optimal leaf Fraction 0.5 0.464 0.46 0.5 0.5
fraction area develop. curve
−1
GSI Maximum stomatal conductance m·s 0.0005 0.0062 0.0063 0.0056 0.0059
HVSTI Harvest index kg·ha− 1 per 0.76 0.8558 0.912 0.5 0.2
kg·ha− 1
LAIMX1 Fract. of max LAI corresp. to 1st point on optimal leaf area curve Fraction 0.1 0.119 0.1 0.5 0.08
LAIMX2 Fract. of max LAI corresp. to 2nd point on optimal leaf area curve Fraction 0.95 0.94 0.95 0.85 0.85
MAT_YEARS Number of years required to reach full development Years 15 0 0 6 3
OV_N Manning's roughness coefficient for overland flow 0.6 0.17 0.09 0.21 0.13 0.01
RDMX Max root depth m 3 0.5 0.4 0.7 1.2
T_BASE Minimum temperature for plant growth °C 9 9 10 7 7
T_OPT Optimal temperature for plant growth °C 20 24 25 18 20
VPDFR Vapor pressure deficit (2nd point on stomatal conductance curve) kPa 4 3.75 4 2 1
WAVP Rate of decline in radiation use efficiency per unit increase in vapor Rate 10 6.6 7 3 3
pressure deficit

⎛ 2SWly,excess × K sat × slp ⎞ 2.4. Model calibration and validation


Q latflow = 0.024 ⎜ ⎟
⎝ Φd × Lhill ⎠ (5)
For each year, measured weather parameters were replicated and
where SWly,excess was the drainable volume of water stored in the used as “warm-up” period (their results were not considered in this
saturated zone of the hillslope per unit area (mm); Ksat was the analysis). Then, the simulation continued to the studied year and the
saturated hydraulic conductivity (mm h− 1); Lhill and slp were the results were used for calibration and validation analysis (Di Luzio et al.,
hillslope length (m) and slope, respectively; and Φd was drainable 2002).
porosity of the soil layer. Calibration and validation were performed according to statistics
The base flow generated from the shallow aquifer was computed as and criteria recommended by Moriasi et al. (2007). The optimal value
follow (Neitsch et al., 2011) (Eq. (6)): of SWAT parameters (Table 3) and model performances were assessed
by using three statistical criteria: the mean percentage error (M%E),
⎧ Q gw,i = Q gw,i−1 exp−αgw ∆t + wrchrg,sh 1 − exp−αgw ∆t for aq sh > aq shthr,q which measures model overestimation or underestimation of measured

⎪ values (Mayer and Butler, 1993), the ratio of the root mean square error
⎨ Q gw,i = 0 to observation standard deviation (RSR) (Moriasi et al., 2007), and

⎪ for aq sh ≤ aq shthr,q Nash–Sutcliffe coefficient (NSC) (Nash and Sutcliffe, 1970). Optimal

results corresponded to the value for which the following conditions
(6) were satisfied: M%E < 10, RSC < 0.5 and NSC > 0.75.
Model error was calculated as the daily difference between actual
where Qgw,i and Qgw,i − 1 (mm) were the ground water flow in the
runoff and estimated runoff for five years with different land use of the
channel on day i and day i − 1, respectively; αgw was the base flow
basin (1954, 1978, 1988, 1996, 2007). Statistical indices were calcu-
recession constant; Δt was the time step (1 day); wrchrg.sh was the
lated for each year, for each season over the five years, and for the
amount of recharge entering the swallow aquifer on day i (mm); aqsh
entire period.
and aqshthr,q (mm) were the amount of water stored in the swallow
Model validation was performed for 4 land use maps (1954, 1978,
aquifer at the beginning of day i and the threshold water level in the
1996 and 2007). Model error was calculated as the daily difference
shallow aquifer for groundwater contribution to the main channel to
between actual and estimated runoff for three meteorological and
occur, respectively.
hydrological validation years for each land use map (Table 1). Statis-
tical indices were calculated for each three year period, for each season

5
M. Napoli et al. Catena 157 (2017) 1–11

Table 4
Annual and seasonal cumulated rainfall (P) and average, standard deviation and maximum value of daily air temperature (Ta), base runoff (Qgw), peak runoff (Qsrre) and total runoff (Qtot)
in the study area.

Season Year Avg. daily Ta (°C) Cumulate P (mm) Daily Qgw (m3 s− 1) Daily Qsrre (m3 s− 1) Daily Qtot (m3 s− 1)

Avg. ± SD Max Avg. ± SD Max Avg. ± SD Max

Winter 1954 6.36 ± 4.09 c 156.5 b 13.96 ± 0.64 a 15.62 1.46 ± 3.64 ab 21.11 15.42 ± 3.52 a 34.76
1978 7.78 ± 3.13 ab 296.8 a 5.09 ± 1.7 c 8.71 5.32 ± 11.87 a 80.97 10.41 ± 11.76 b 83.65
1988 7.91 ± 2.88 ab 231.3 b 5.92 ± 1.67 b 10.19 4.95 ± 8.25 a 41.7 10.86 ± 8.97 b 51.88
1996 7.05 ± 2.86 bc 137.6 c 1.42 ± 0.19 e 1.86 1.14 ± 2.58 b 20.39 2.56 ± 2.59 c 21.8
2007 8.7 ± 2.93 a 182 b 2.08 ± 0.26 d 2.68 1.02 ± 3.29 b 28.82 3.1 ± 3.25 c 30.57
Spring 1954 16.72 ± 3.95 a 234.7 a 14.4 ± 0.92 a 16.61 1.62 ± 3.83 ab 29.86 16.02 ± 3.85 a 43.51
1978 15.03 ± 4.29 b 311.9 a 7.1 ± 2.18 b 11.18 5.98 ± 14.19 a 82.53 13.08 ± 14.75 b 90.99
1988 16.96 ± 3.57 a 83.8 b 4.5 ± 2.06 c 10.68 2.18 ± 5.34 b 36.63 6.68 ± 6.79 c 47.32
1996 16.78 ± 4.21 a 205 b 1.14 ± 0.4 e 1.86 0.94 ± 2.01 b 13 2.08 ± 2.19 d 14.64
2007 17.86 ± 3.84 a 140.3 b 1.8 ± 0.24 d 2.32 0.48 ± 1.36 b 11.4 2.28 ± 1.36 d 13.15
Summer 1954 22.46 ± 2.69 ab 159.4 11.2 ± 2.53 a 15.13 1.91 ± 2.23 a 6.67 13.11 ± 1.21 a 17.35
1978 21.2 ± 2.79 c 125.4 2.35 ± 0.47 b 3.39 0.88 ± 3.37 b 32.08 3.23 ± 3.44 b 34.76
1988 23.01 ± 3.61 a 127.2 2.12 ± 0.51 b 3.04 0.64 ± 1.32 bc 5.78 2.76 ± 1.51 b 8.46
1996 20.95 ± 3.58 c 101.2 0.61 ± 0.02 d 0.66 0.08 ± 0.27 c 2.43 0.68 ± 0.27 d 3.04
2007 21.59 ± 3.26 bc 133.4 1.15 ± 0.08 c 1.41 0.15 ± 0.22 bc 1.15 1.3 ± 0.23 c 2.21
Autumn 1954 11.81 ± 4.26 107.9 10.37 ± 1.73 a 14.39 2.85 ± 2.06 a 6.67 13.22 ± 1.62 a 15.13
1978 10.26 ± 4.09 134 4.3 ± 0.37 c 5.49 0.97 ± 2.12 a 12.68 5.27 ± 2.04 b 16.61
1988 10.73 ± 5.93 173.6 6.63 ± 1.52 b 9.2 2.03 ± 3.18 a 19.02 8.67 ± 2.9 c 25.01
1996 10.9 ± 4.87 88.2 0.75 ± 0.08 e 1.06 0.33 ± 0.92 b 6.44 1.07 ± 0.92 e 7.22
2007 10.49 ± 4.59 151.2 1.06 ± 0.05 d 1.18 0.09 ± 0.2 b 0.87 1.16 ± 0.19 d 1.86
Annual 1954 14.25 ± 7.2 658.5 b 12.47 ± 2.38 a 16.61 1.96 ± 3.08 a 29.86 14.43 ± 3.07 a 43.51
1978 13.23 ± 6.1 868.1 a 4.7 ± 2.21 b 11.18 3.27 ± 9.69 a 82.53 7.97 ± 10.35 b 90.99
1988 15.04 ± 6.52 615.9 b 4.79 ± 2.32 b 10.68 2.44 ± 5.4 a 41.7 7.23 ± 6.53 b 51.88
1996 14.01 ± 6.99 532 b 0.97 ± 0.38 d 1.75 0.42 ± 1.14 b 13.57 1.38 ± 1.28 d 14.98
2007 14.62 ± 7.15 606.9 c 1.52 ± 0.46 c 2.68 0.43 ± 1.81 b 28.82 1.95 ± 1.92 c 30.57

Table 5
Annual and seasonal values of mean percentage error (M%E), the ratio of the root mean square error to observation standard deviation (RSR) and Nash–Sutcliffe coefficient (NSC) of
simulated base flow, peak flow and total flow are reported for the calibration and validation dataset.

Year setup Number of records Base flow Peak flow Total flow

M%E RSR NSC M%E RSR NSC M%E RSR NSC

Calibration
1954 365 0.00% 0.19 0.96 0.06% 0.53 0.8 0.01% 0.54 0.99
1978 365 − 0.08% 0.11 0.99 1.25% 0.28 0.93 0.47% 0.26 0.96
1988 365 − 0.27% 0.16 0.98 10.31% 0.33 0.91 3.30% 0.28 0.97
1996 365 − 0.21% 0.08 0.99 9.06% 0.32 0.91 3.38% 0.29 0.95
2007 365 − 0.27% 0.12 0.99 −9.31% 0.36 0.88 − 2.27% 0.33 0.95
Winter 454 0.00% 0.06 0.99 0.45% 0.22 0.96 0.15% 0.19 0.98
Spring 455 − 0.24% 0.05 0.99 14.62% 0.33 0.9 3.90% 0.25 0.96
Summer 460 0.05% 0.07 0.99 −2.03% 0.5 0.77 − 0.31% 0.21 0.97
Autumn 456 − 0.14% 0.09 0.99 −6.07% 0.59 0.74 − 1.41% 0.26 0.97
All period 1825 − 0.09% 0.06 0.97 3.52% 0.31 0.91 0.86% 0.22 0.97

Validation
1956, 1957, 1958 1095 0.01% 0.15 0.98 6.09% 0.24 0.94 1.11% 0.22 0.95
1980, 1981, 1983 1095 0.01% 0.15 0.98 7.37% 0.23 0.95 1.63% 0.22 0.95
1998, 1999, 2000 1057 6.58% 0.28 0.92 12.29% 0.18 0.97 9.67% 0.19 0.96
2008, 2009, 2010 1075 4.79% 0.18 0.97 −14.32% 0.21 0.96 − 6.00% 0.19 0.96
Winter 1072 0.71% 0.05 0.99 0.40% 0.25 0.94 0.62% 0.17 0.97
Spring 1092 0.33% 0.04 0.99 1.95% 0.21 0.96 0.66% 0.11 0.99
Summer 1104 0.07% 0.05 0.99 5.33% 0.26 0.93 0.80% 0.10 0.99
Autumn 1054 1.03% 0.08 0.99 3.51% 0.21 0.96 1.95% 0.18 0.97
All period 4322 0.54% 0.05 0.99 2.40% 0.22 0.95 1.02% 0.16 0.98

over the twelve years, and for the entire period. validation dataset at significance level P < 0.05.
Furthermore, the model was used to estimate runoff variation due to
2.5. Climate and land use change analysis different land use situations on the calibration data set. For each
meteorological year, the model simulated runoff five times, using each
Climate differences between the years were tested. Air temperature time a different land use and the resulted were compared (e.g. the 1954
difference was tested by means of analysis of variance (ANOVA) runoff simulation was performed five times with the 1954 meteorolo-
followed by Bonferroni test and rainfall difference was tested by non- gical data each time using a different land use classification: respec-
parametric Mann–Whitney U test, both at significance level P < 0.05. tively 1954, 1978, 1986, 1996 and 2007). This procedure was applied
The relationship between base peak and total runoff and annual to the other meteorological dataset (1978, 1986, 1996 and 2007).
cumulated precipitation and average temperature was investigated by Individual contribution by each land use to base, peak and total
means of Pearson product moment correlation on the calibration and runoff for each simulation year was also derived from the model output.

6
M. Napoli et al. Catena 157 (2017) 1–11

Table 6
Average, standard deviation, minimum and maximum values of total flow, base flow and peak flow, and yearly peaks resulting from the simulations of each climate year using five
different land use. Small letters indicate the significant differences between land use year with the same climate years through ANOVA followed by Bonferroni test at P < 0.05.

Climate year Land use year Qgw (m3 s− 1) Qsrre (m3 s− 1) Qtot (m3 s− 1) No. of peaks

Average Min Max Average Min Max Average Min Max

1954 1954 12.47 ± 2.19 7.47 16.4 3.04 ± 3.16 b 0.02 25.4 14.43 ± 3.37 b 7.72 39.7 236
1978 12.59 ± 2.22 7.72 16.1 3.67 ± 4.66 b 0.02 33.9 14.56 ± 4.34 b 7.72 48.3 196
1988 12.78 ± 2.32 7.72 16.6 4.89 ± 4.08 a 0.86 32.6 15.98 ± 4.5 a 7.72 47.4 239
1996 12.69 ± 2.22 7.72 16.1 4.79 ± 4.61 ab 0.14 40.4 15.81 ± 4.51 a 7.72 55 238
2007 12.78 ± 2.23 7.72 16.1 6.01 ± 6.83 a 0.06 59.8 16.59 ± 6.27 a 7.72 74.5 231
1978 1954 4.81 ± 2.15 2.09 10.7 4.1 ± 6.51 b 0.02 43.5 7.4 ± 6.26 2.09 52.5 230
1978 4.7 ± 2.17 1.98 10.4 5.62 ± 11.48 ab 0.19 82.6 8.01 ± 9.97 1.98 85.4 215
1988 4.73 ± 2.24 1.98 10.7 6.32 ± 10.84 ab 0.14 77.6 8.61 ± 9.85 1.98 83.2 224
1996 4.79 ± 2.19 1.98 10.7 6.56 ± 13.32 ab 0.19 98.6 8.92 ± 11.74 1.98 103 230
2007 4.79 ± 2.21 1.98 10.7 7.56 ± 16.54 a 0.12 121 9.47 ± 14.17 1.98 124 226
1988 1954 4.87 ± 2.18 1.64 9.2 2.83 ± 3.71 c 0.17 20.1 6.73 ± 4.51 ab 1.64 29.1 241
1978 4.77 ± 2.21 1.52 9.2 2.92 ± 4.97 cb 0.17 26.7 6.44 ± 5.14 b 1.52 35.9 209
1988 4.78 ± 2.22 1.52 9.2 4.12 ± 6.13 acb 0.02 34.5 7.46 ± 6.38 ab 1.52 43.2 238
1996 4.87 ± 2.22 1.64 9.2 4.58 ± 7.24 ab 0.03 43.3 7.8 ± 7.23 a 1.64 52 234
2007 4.88 ± 2.25 1.64 9.2 5.07 ± 8.13 a 0.05 47.3 7.91 ± 7.77 a 1.64 56 218
1996 1954 0.97 ± 0.38 0.58 1.75 0.57 ± 1.3 b 0.01 13.6 1.38 ± 1.28 b 0.58 15 266
1978 0.95 ± 0.37 0.58 1.75 0.7 ± 2.09 b 0.01 18.3 1.46 ± 1.97 b 0.58 19.7 267
1988 0.92 ± 0.36 0.55 1.64 0.8 ± 1.8 b 0.01 15.2 1.52 ± 1.76 b 0.55 16.6 274
1996 0.97 ± 0.39 0.58 1.75 0.9 ± 2.06 ab 0.02 19.3 1.65 ± 1.98 ab 0.58 20.7 274
2007 0.97 ± 0.38 0.58 1.75 1.35 ± 3.16 a 0.01 30.2 1.97 ± 2.92 a 0.58 31.6 270
2007 1954 1.6 ± 0.47 1.06 2.68 0.73 ± 1.59 b 0.02 14.3 1.92 ± 1.24 1.06 16.3 159
1978 1.49 ± 0.44 1 2.5 1 ± 1.92 ab 0 13.3 1.83 ± 1.38 1 15.1 128
1988 1.46 ± 0.45 0.93 2.5 0.65 ± 0.96 b 0.03 8.34 1.78 ± 0.96 0.93 10.1 184
1996 1.53 ± 0.46 1 2.5 1.43 ± 3.09 a 0.01 20.3 1.95 ± 1.83 1 22 108
2007 1.52 ± 0.45 1 2.5 0.81 ± 1.8 ab 0.01 18.2 1.91 ± 1.46 1 20.1 176

Runoff difference across the years and different land use within the (range: − 0.27%–0.05%), RSR (range: 0.05–0.19) and NSC (range:
same meteorological years were tested by means of analysis of variance 0.97–0.99).
(ANOVA) followed by Bonferroni test at significance level P < 0.05, The simulation of the total flow obtained slightly worse scores than
base flow for all the years and seasons (M%E = −2.27%–3.90%,
3. Results and discussion RSR = 0.19–0.54 and NSC = 0.95–0.99), however the performance
was still very good for all the data-sets except in 1954 where the RSR
3.1. Meteorological conditions and stream flows during the study period index scored 0.54, that is still considered a good performance according
to Moriasi et al. (2007).
Annual and seasonal average daily Ta, cumulate P, average and Regarding peak flow, seasonal and yearly scores were more variable
maximum daily Qgw, Qsrre and Qtot, were reported in Table 4. Regarding (M%E = −9.31%–14.62%, RSR = 0.22–0.59 and NSC = 0.74–0.96).
meteorological variables, 1978 was characterized by the significantly Overall, score was very good according all the indices and the M%E
highest amount of annual and Winter precipitation (868.1 mm, value (3.52%) showed that the model tends slightly to overestimate.
296.8 mm) while 1988 recorded the lowest amount of annual and According to M%E, the model showed very good performance in the
Winter precipitation in the study period (532.0 mm, 137.6 mm). years characterized by high flows (1954 and 1978), but was less
Significant higher spring precipitation was recorded in 1954 and 1978. satisfactory for years with low flow and in 1988 the score was slightly
According to the analyses, 1988 was also characterized by highest over the 10% limit (M%E = 10.31%). The analysis of seasonal data
annual temperature (15.04 ± 6.52 °C), while 1978 by the lowest value shows some kind of difficulty of the model in predicting peak flows
for the study period (13.23 ± 6.1 °C) but the difference was not occurring in spring. According RSR and NSC the model performance
significant. However, significant seasonal differences were found was very good in most of the cases except for RSR calculated in 1954
among the years. 2007 recorded the warmest spring and winter and RSR and NSC in autumn, but in these cases the performance was
temperature, while 1988 was characterized by the hottest summer. still good.
The year 1954 was characterized by the significantly highest annual Model validation was conducted using the calibration settings and
and seasonal total and base runoff, while 1996 was always character- land use according Table 3. In general, the differences of the model
ized by the lowest values. On the contrary, highest peak runoff value performance between calibration and validation were small in terms of
were recorded in 1978, especially in Winter (with an average value of M%E, RSR and NSC, thus indicating no model overfitting. For base
5.32 m3 s− 1 and maximum peak value of 83.7 m3 s− 1) and Spring flow, peak flow and total flow, all indices show a slight accuracy
(with an average value of 5.98 m3 s− 1 and maximum peak value of reduction but maintain the same level of precision reached for the
91.9 m3 s− 1) that were the seasons characterized by the highest calibration data set. As for the calibration, the model simulates very
seasonal precipitation values in the study period. well the years characterized by high level of water flows and the
performance was slightly lower only for M%E peak flow in 1998–2000
3.2. Model calibration and validation and 2008–2010. Results indicate that the model had the tendency to
overestimate flows during low flow periods. The weakness of SWAT in
According to the evaluation guidelines of Moriasi et al. (2007), the modeling dry conditions has also been reported by other authors
adjusted model showed a very good performance (M%E ≤ 10%, (Gebremariam et al., 2014; Molina-Navarro et al., 2015). Overall,
RSR ≤ 0.50, NSC ≥ 0.75) in simulating daily base flow for the whole model performance was very good within the study domain, and
calibration data set (Table 5). Average daily flows estimated by the therefore reliable to be extended to reconstruct the natural runoff.
model matched measurements both yearly and seasonally for M%E

7
M. Napoli et al.

Table 7
Percentage contribution of each land use to the formation of the total flow, base flow and peak flow resulting from the simulations of each climate year using five different land use.

Land Climate year Qgw (m3 s− 1) Qsrre (m3 s− 1) Qtot (m3 s− 1)


use
year Forest Heterogeneous Non- Olive Vine-yards Urban Forest Heterogeneous Non- Olive Vine-yards Urban Forest Heterogeneous Non- Olive Vine-yards Urban
and agricultural irrigated groves and agricultural irrigated groves and agricultural irrigated groves
semi- areas arable semi- areas arable semi- areas arable
natural land natural land natural land
areas areas areas

1954 1954 47.19% 40.41% 8.96% 0.50% 1.73% 1.21% 6.65% 69.54% 19.84% 0.11% 2.31% 1.55% 38.63% 49.30% 9.83% 0.11% 0.53% 1.59%
1978 48.19% 40.91% 7.72% 0.56% 1.84% 0.77% 6.84% 71.56% 17.95% 0.12% 2.43% 1.10% 39.63% 50.14% 8.52% 0.12% 0.57% 1.02%
1988 47.90% 40.84% 8.10% 0.52% 1.79% 0.85% 3.93% 81.76% 12.09% 0.06% 1.30% 0.86% 39.32% 49.97% 8.92% 0.12% 0.55% 1.12%
1996 47.99% 40.84% 7.98% 0.54% 1.81% 0.85% 6.52% 68.79% 20.83% 0.10% 2.17% 1.59% 39.41% 49.99% 8.79% 0.12% 0.56% 1.12%
2007 47.49% 40.60% 8.63% 0.52% 1.76% 1.01% 6.48% 68.56% 21.10% 0.09% 2.13% 1.63% 38.93% 49.60% 9.49% 0.11% 0.54% 1.33%
1978 1954 42.67% 22.33% 26.72% 0.25% 2.75% 5.29% 5.10% 26.68% 55.51% 0.97% 5.31% 6.43% 37.34% 19.54% 35.86% 0.22% 2.41% 4.63%
1978 45.48% 23.59% 24.05% 0.29% 3.06% 3.54% 5.74% 29.77% 52.40% 1.19% 6.20% 4.71% 40.31% 20.91% 32.69% 0.26% 2.71% 3.13%
1988 44.70% 23.29% 24.96% 0.27% 2.94% 3.84% 5.55% 28.94% 53.53% 1.08% 5.86% 5.03% 39.45% 20.56% 33.78% 0.24% 2.59% 3.39%

8
1996 44.90% 23.35% 24.64% 0.28% 2.98% 3.84% 5.60% 29.15% 53.10% 1.12% 5.97% 5.06% 39.69% 20.64% 33.40% 0.25% 2.63% 3.40%
2007 43.55% 22.75% 26.11% 0.26% 2.84% 4.49% 5.29% 27.64% 54.75% 1.02% 5.54% 5.75% 38.23% 19.97% 35.14% 0.23% 2.49% 3.94%
1988 1954 43.35% 17.16% 26.39% 0.76% 5.83% 6.52% 7.06% 25.86% 45.87% 2.57% 10.67% 7.96% 39.90% 14.59% 31.91% 0.64% 7.42% 5.54%
1978 46.29% 18.17% 23.80% 0.90% 6.48% 4.36% 7.81% 28.36% 42.86% 3.15% 12.30% 5.52% 42.80% 15.51% 28.91% 0.76% 8.30% 3.72%
1988 45.54% 17.95% 24.72% 0.82% 6.23% 4.74% 7.61% 27.77% 44.10% 2.87% 11.71% 5.94% 42.03% 15.30% 29.97% 0.70% 7.96% 4.04%
1996 45.71% 17.99% 24.39% 0.85% 6.31% 4.74% 7.66% 27.89% 43.62% 2.97% 11.90% 5.96% 42.23% 15.34% 29.59% 0.72% 8.08% 4.04%
2007 44.31% 17.52% 25.83% 0.80% 6.02% 5.53% 7.29% 26.66% 45.35% 2.73% 11.14% 6.83% 40.81% 14.89% 31.24% 0.68% 7.67% 4.71%
1996 1954 43.03% 19.30% 19.25% 1.57% 9.91% 6.93% 6.32% 17.86% 46.17% 3.68% 15.29% 10.69% 35.75% 18.06% 25.85% 1.06% 9.98% 9.31%
1978 45.38% 20.18% 17.14% 1.84% 10.89% 4.58% 7.04% 19.73% 43.47% 4.54% 17.75% 7.47% 38.49% 19.27% 23.50% 1.26% 11.20% 6.28%
1988 44.86% 20.03% 17.89% 1.70% 10.52% 5.00% 6.87% 19.32% 44.73% 4.14% 16.91% 8.04% 37.84% 19.03% 24.39% 1.16% 10.76% 6.82%
1996 44.95% 20.04% 17.62% 1.75% 10.64% 5.00% 6.90% 19.39% 44.21% 4.28% 17.16% 8.06% 37.96% 19.06% 24.06% 1.20% 10.90% 6.82%
2007 43.85% 19.64% 18.78% 1.65% 10.21% 5.87% 6.55% 18.48% 45.82% 3.93% 16.01% 9.21% 36.69% 18.50% 25.40% 1.12% 10.35% 7.94%
2007 1954 41.62% 13.07% 20.85% 2.71% 10.07% 11.67% 3.59% 16.53% 37.46% 4.18% 23.15% 15.10% 35.77% 15.66% 25.75% 1.34% 9.95% 11.53%
1978 45.56% 14.18% 19.28% 3.29% 11.48% 6.22% 4.09% 18.68% 36.08% 5.28% 27.50% 8.38% 39.53% 17.16% 24.03% 1.64% 11.45% 6.20%
1988 44.88% 14.03% 20.05% 3.03% 11.05% 6.97% 4.00% 18.35% 37.24% 4.82% 26.27% 9.32% 38.79% 16.91% 24.89% 1.51% 10.98% 6.92%
1996 44.83% 13.99% 19.68% 3.11% 11.14% 7.24% 4.00% 18.29% 36.56% 4.96% 26.50% 9.69% 38.80% 16.88% 24.48% 1.55% 11.09% 7.21%
2007 43.30% 13.57% 20.77% 2.91% 10.58% 8.87% 3.79% 17.43% 37.88% 4.55% 24.70% 11.65% 37.30% 16.30% 25.70% 1.44% 10.48% 8.78%
Catena 157 (2017) 1–11
M. Napoli et al. Catena 157 (2017) 1–11

use impacts on base peak and total flows of the same climate data set
that is reported in Table 6. According to both analyses no significant
differences of base flows induced by land use were found in any
climatic year. However, from 1954 to 2007, the average base flow
decreased by about 10.95 m3 s− 1. Other studies found a decline in
groundwater recharge (Rose and Peters, 2001; Tripathi et al., 2005).
Zhou et al. (2013) reported a 11.2% reduction of base flow caused by an
increment of 177.1% in urbanization in the Xitiaoxi River basin
(1371 km2), from 1985 to 2008. Jinno et al. (2009) observed a
reduction of groundwater recharge to shallow aquifers and river
because of the surface flow increasing induced by land use change. In
our results, the base flow reduction could be also due to water use for
agriculture, drinking water supply, and small to medium industries
consumption, which were grown alongside with the population in-
crease in the basin. On the contrary land use of the recent years is
associated by significant higher values of annual mean and maximum
peak flow than the previous years (e.g. 2007 land use against 1954 land
use) for all the climate years except 2007 where the highest average
peak flow was obtained with 1996 land use. The response of total flow
to different land use was significant for 1954, 1988 and 1996 climate
years and this difference is due to the correspondent difference in peak
flows. In 1978 and 2007 land use didn't affect total flow because lower
values of mean peak flows were counterbalanced by a higher number of
peaks. For instance, applying 2007 climate year, mean peak flow with
1996 land use (1.43 ± 3.09 m3 s− 1) was significantly higher than
peak flow with 1954 land use (0.73 ± 1.59 m3 s− 1), while 1996
number of peaks (n = 108) was lower than 1954 (n = 159). This
means that even land use change doesn't affect total flow, it might
affected the generation and dynamic of the peak flows. The impact of
land use change, resulting from the conversion of traditional farmland
to mechanized farmland and urban area was similar to other places in
the world (Hu et al., 2004; Tripathi et al., 2005). Soil imperviousness
leads to higher runoff volumes and peak flows (Rose and Peters, 2001;
Boggs and Sun, 2011). Du et al. (2012) found that an increase of
impervious area affected more small floods than larger floods. Jinno
et al. (2009) found that land-use change determines a surface water
flow increasing along with a time to peak reduction. Zhou et al. (2013)
found that surface runoff flow, in the Xitiaoxi River basin, was sensitive
to urbanization, which had increased by 11.3% from 1985 to 2008. On
the contrary, Li et al. (2009) reported that the surface hydrology of
agricultural basin in the Loess Plateau of China was influenced more
significantly by climate variability than land use change during the
period 1981–2000, Similar results were found by Li et al. (2012) in the
Taoerhe River basin, China, where climate conditions, especially
precipitation, were more important than land use change in determin-
ing runoff variations in 1970–2000. Ma et al. (2014) followed a similar
approach to simulate sediment sourcing and transport and found that
47.8% of the decrease was due to land-use and land cover change,
19.8% to climate change, resulting in a milder rainfall regime, 26.1% to
watershed engineering measures, and the remaining 6.3% was due to
the simulation percent bias.
Looking again at Table 6, land use affected mainly peak flow and
consequently total flow. Average peak flow increase due to land use
changes occurred between 1954 and 2007 was on average 81.8%,
varying between a minimum increase of 79.2% (from 2.83 to
5.07 m3 s− 1) in the climate year 1988 and a maximum increase of
Fig. 5. Rainfall and peak runoff simulation with 1954, 1978, 1988, 1996 and 2007 land 136.8% (from 0.57 to 1.35 m3 s− 1) in the climate year 1996.
use (LU), for each year a 21-day period is shown. Moreover the increase was enhanced for extreme events, in fact
maximum peak flow would be increased of 131.1% on average between
3.3. Analysis of climate and land use changes on streamflow 1954 and 2007 land use changes, varying between a minimum increase
of 83.5% (from 8.34 to 20.27 m3 s− 1) in the climate year 2007 and a
The analysis of the influence of climate on runoff dynamics showed maximum increase of 177.7% (from 43.50 to 120.80 m3 s− 1) in the
that only annual peak runoff was negatively correlated to annual climate year 1978 that was the year with highest precipitation regime.
average temperature (R = − 0.57, P < 0.05) and positively correlated Analysis of the contribution of each land use to base, peak and total
to annual cumulated precipitation (R = 0.53, P < 0.05). flow for each climate year was reported in Table 7. Peak flow increase
These results were integrated by the analysis of variance of the land was mainly caused by urbanization (+2.2% urbanization increases the

9
M. Napoli et al. Catena 157 (2017) 1–11

contribution to runoff generation from an average of 1.35% to 10.83%), Costa, M.H., Botta, A., Cardille, J.A., 2003. Effects of large-scale changes in land cover on
the discharge of the Tocantins River, Southeastern Amazonia. J. Hydrol. 283,
by vineyards (+ 8.9% vineyards increases the contribution to runoff 206–217.
generation from an average of 2.07% to 25.62%), olive groves (+2.6% Crooks, S., Davies, H., 2001. Assessment of land use change in the Thames catchment and
olive groves increases the contribution to runoff generation from an its effect on the flood regime of the river. Phys. Chem. Earth Part B 26 (7–8),
583–591.
average of 0.10% to 4.76%) and arable land (+ 16.2% arable land Cuo, L., Lettenmaier, D.P., Alberti, M., Richey, J.E., 2009. Effects of a century of land
increases the contribution to runoff generation from an average of cover and climate change on the hydrology of the Puget Sound basin. Hydrol.
18.36% to 37.04%) (Tables 1 and 7). Process. 23 (6), 907–933.
Dams, J., Woldeamlak, S.T., Batelaan, O., 2008. Predicting land-use change and its impact
This type of analyses can be useful to investigate how land use on the groundwater system of the Kleine Nete catchment, Belgium. Hydrol. Earth
variation could affect runoff generation. Furthermore, they allow to Syst. Sci. 12, 1369–1385.
analyse the different contribution to runoff formation within different Di Luzio, M., Srinivason, R., Arnold, J.R., Neitsch, S.L., 2002. Arcview Interface for
SWAT2000: User's Guide. Blackland Research and Extension Center, Grassland, Soil
land cover classes. For instance the contribution to runoff formation
and Water Research Laboratory, Texas Water Resources Institute, College Station, TX.
due to vineyards in 2007 corresponded to an average contribution to Du, J., Qian, L., Rui, H., Zuo, T., Zheng, D., Xu, Y., Xu, C.Y., 2012. Assessing the effects of
runoff of 25.62%, while the contribution due to arable land surface (2.4 urbanization on annual runoff and flood events using an integrated hydrological
times larger than vineyards) corresponded to an average contribution to modeling system for Qinhuai River basin, China. J. Hydrol. 464–465, 127–139.
FAO, 1998. World Reference Base for Soil Re-sources.
runoff of 37.04% (only 1.4 times higher than vineyard contribution). Flugel, W., 1995. Delineating hydrologic response units by geographical information
The simulation can also be used to analyse the response of different system analyses for regional hydrological modeling using PRMS/MMS in the drainage
land use to rainfall events. For instance, the response of the five basin of the River Brol, Germany. Hydrol. Proc. 9 (3–4), 423–436.
Franczyk, J., Chang, H., 2009. The effects of climate change and urbanization on the
different land use to a period of rainfall is represented in Fig. 5, where it runoff of the Rock Creek basin in the Portland metropolitan area, Oregon, USA.
is clear that the increase of artificial surfaces and specialised agriculture Hydrol. Process. 23 (6), 805–815.
corresponds to increase in peak intensity and duration. Gardin, L., Vinci, A., 2016. Carta dei suoli della Regione Toscana in scala 1:250.000.
http://159.213.57.101/pmapper/map.phtml (last accessed: 09/09/2016).
Gebremariam, S.Y., Martin, J.F., DeMarchi, C., Bosch, N.S., Confesor, R., Ludsin, S.A.,
4. Conclusion 2014. A comprehensive approach to evaluating watershed models for predicting river
flow regimes critical to downstream ecosystem services. Environ. Model. Softw. 61,
121–134.
Runoff dynamic of the Elsa river basin was analysed by considering
Ghaffari, G., Keesstra, S., Ghodousi, J., Ahmadi, H., 2010. SWAT-simulated hydrological
land use change along more than fifty years (1954–2007) by means of impact of land-use change in the Zanjanrood basin, northwest Iran. Hydrol. Process.
ArcSWAT model. Results showed that, given the same climatic condi- 24, 892–903.
Hu, Q., Willson, G.D., Chen, X., Akyuz, A., 2004. Effects of climate and landcover change
tions, land use change plays a considerable role since it affects both
on stream discharge in the Ozark highlands, USA. Environ. Model. Assess. 10, 9–19.
peak and total runoff. This result is important because highlights the Huang, H.J., Cheng, S.J., Wen, J.C., 2008. Effect of growing watershed imperviousness on
importance of land use policy. Often, policymakers focus most of their hydrograph parameters and peak discharge. Hydrol. Process. 22 (13), 2075–2085.
attention on climate change effects on the environment, while land use ISTAT, 2010. 6th General Census of Agriculture. http://dati-censimentoagricoltura.istat.
it/Index.aspx?lang=en (last accessed: 09/09/2016).
is equally important since its effect combines and can amplify that of ISTAT, 2011. 15th General Census of Population. http://dati-censimentopopolazione.
climate change. So this kind of analysis can help policymakers to plan a istat.it/Index.aspx?lang=en (last accessed: 09/09/2016).
land use management policy that is adequate to respond to climate Jinno, K., Tsutsumi, A., Alkaeed, O., Saita, S., Berndtsson, R., 2009. Effects of land-use
change on groundwater recharge model parameters. Hydrol. Sci. 54 (2), 300–315.
change. Our findings indicated that ArcSWAT model can effectively Juckem, P.F., Hunt, R.J., Anderson, M.P., Robertson, D.M., 2008. Effects of climate and
simulate the response of basin hydrology (base peak and total runoff) to land management change on streamflow in the driftless area of Wisconsin. J. Hydrol.
climate and land use changes. Moreover, model output can provide 355 (1–4), 123–130.
Kannan, N., Santhi, C., Williams, J.R., Arnold, J.G., 2008. Development of a contin-uous
useful information about the contribution of each land use to runoff soil moisture accounting procedure for curve number methodology and its behavior
formation as well as to runoff dynamics due to rainfall events. with different evapotranspiration methods. Hydrol. Process. 22, 2114–2121.
Therefore, this approach can be used to analyse future scenario Landi, R., 1989. Revision of land management systems in Italian hilly area. In:
Schhwertmann, U., Rickson, R.J., Auwerswald, K. (Eds.), Soil Erosion Protection
response to climate change and to define sustainable policy of land use. Measures in Europe. Soil Technology Series, vol. 1. Catena Werlag, Cremlingen,
Germany, pp. 175–188.
Acknowledgements Li, Z., Liu, W.Z., Zhang, X.C., Zheng, F.L., 2009. Impacts of land use change and climate
variability on hydrology in an agricultural catchment on the Loess Plateau of China.
J. Hydrol. 377 (1), 35–42.
The authors thank the unknown reviewers who carefully reviewed Li, L., Jiang, D., Hou, X., Li, J., 2012. Simulated runoff responses to land use in the middle
the paper and whose suggestions were useful in improving the manu- and upstream reaches of Taoerhe River basin, Northeast China, in wet, average and
script. dry years. Hydrol. Process. 27 (24), 3484–3494.
Ma, X., Xu, J., Luo, Y., Aggarwal, S.P., Li, J., 2009. Response of hydrological processes to
land-cover and climate changes in Kejie watershed, south-west China. Hydrol.
References Process. 23 (8), 1179–1191.
Ma, X., Lu, X., van Noordwijk, M., Li, J.T., Xu, J.C., 2014. Attribution of climate change,
vegetation restoration, and engineering measures to the reduction of suspended
Aronica, G., Cannarozzo, M., 2000. Studying the hydrological response of urban sediment in the Kejie catchment, southwest China. Hydrol. Earth Syst. Sci. 18,
catchments using a semi-distributed liner non-linear mode. J. Hydrol. 238, 35–43. 1979–1994.
Bo, X., Qing-Hai, W., Jun, F., Feng-Peng, H., Quan-Hou, D., 2011. Application of theSCS- MacDonald, D., Crabtree, J.R., Wiesinger, G., Dax, T., Stamou, N., Fleury, P., Gutierrez
CN model to runoff estimation in a small watershed with high spatial het-erogeneity. Lazpita, J., Gibon, A., 2000. Agricultural abandonment in mountain areas of Europe:
Pedosphere 21, 738–749. environmental consequences and policy response. J. Environ. Manag. 59, 47–69.
Boggs, J.L., Sun, G., 2011. Urbanization alters watershed hydrology in the Piedmont of Mayer, D.G., Butler, D.G., 1993. Statistical validation. Ecol. Model. 68 (1–2), 21–32.
North Carolina. Ecohydrology 4, 256–264. Mishra, S.K., Singh, V.P., 2004. Long-term hydrological simulation based on the soil
Brath, A., Montanari, A., Moretti, G., 2006. Assessing the effect on flood frequency of land conservation service curve number. Hydrol. Process. 18, 1291–1313.
use change via hydrological simulation (with uncertainty). J. Hydrol. 324 (1–4), Molina-Navarro, E., Hallack-Alegria, M., Martínez-Pérez, S., Ramírez-Hernández, J.,
141–153. Mungaray-Moctezuma, A., Sastre-Merlín, A., 2015. Hydrological modeling and
Choi, W., 2008. Catchment-scale hydrological response to climate-landuse combined climate change impacts in an agricultural semiarid region. Case study: Guadalupe
scenarios: a case study for the Kishwaukee River Basin, Illinois. Phys. Geogr. 29 (1), River basin, Mexico. Agric. Water Manag. 175, 29–42.
79–99. Moriasi, D.N., Arnold, J.G., Liew, M.W., Bingner, Van, Harmel, R.L., Veith, R.D., 2007.
Chow, V.T., Maidment, D.R., Mays, L.W., 1988. Applied Hydrology. McGraw-Hill, Model evaluation guidelines for systematic quantification of accuracy in watershed
NewYork. simulations. Trans. ASABE 50, 885–900.
Chung, E., Park, K., Lee, K.S., 2011. The relative impacts of climate change and Napoli, M., Orlandini, S., 2015. Evaluating the Arc-SWAT2009 in predicting runoff,
urbanization on the hydrologic response of a Korean urban watershed. Hydrol. sediment, and nutrient yields from a vineyard and an olive orchard in Central Italy.
Process. 25, 544–560. Agric. Water Manag. 153, 51–62.
COM, 2006. Communication From the Commission to the Council, the European Napoli, M., Cecchi, S., Orlandini, S., Zanchi, C.A., 2014. Determining potential rainwater
Parliament, the European Economic and Social Committee and the Committee of the harvesting sites using a continuous runoff potential accounting procedure and GIS
Regions. Brussels: Commission of the European Communities. Thematic Strategy for techniques in central Italy. Agric. Water Manag. 141, 55–65.
Soil Protection. COM (2006) 231 Final. European Commission, Brussels.

10
M. Napoli et al. Catena 157 (2017) 1–11

Napoli, M., Cecchi, S., Orlandini, S., Mugnai, G., Zanchi, C.A., 2016. Simulation of field- sewer system. J. Hydrol. 350, 100–113.
measured soil loss in Mediterranean hilly areas (chianti, Italy) with RUSLE. Catena SIR, 2016. http://www.sir.toscana.it/ricerca-dati (last accessed: 09/09/2016).
145, 246–256. SITA, 2016. http://www502.regione.toscana.it/geoscopio/cartoteca.html (last accessed:
Nash, J.E., Sutcliffe, J.E., 1970. River flow forecasting through conceptual models: part 1. 09/09/2016).
A discussion of principles. J. Hydrol. 10, 282–290. Sloan, P.G., Moore, I.D., 1984. Modeling subsurface stormflow on steeply sloping forested
Neitsch, S.L., Arnold, J.G., Kiniry, J.R., Williams, J.R., 2005. Soil and Water Assessment watersheds. Water Resour. Res. 20 (12), 1815–1822.
Tool: Theoretical Documentation. Ver. 2005. USDA–ARS Grassland, Soiland Water Stocker, T.F., Qin, D., Plattner, G.K., Tignor, M., Allen, S.K., Boschung, J., Nauels, A., Xia,
Research Laboratory, Temple, TX. Y., Bex, V., Midgley, P.M., 2013. The Physical Science Basis. Contribution of Working
Neitsch, S.L., Arnold, J.G., Kiniry, J.R., Williams, J.R., 2011. Soil and Water Assessment Group I to the Fifth Assessment Report of the Intergovernmental. Cambridge
Tool Theoretical Documentation Version 2009. Texas Water Resources Institute, University Press, Cambridge, United Kingdom and New York, NY, USA, pp. 1535.
College Station, TX, pp. 647. Stonestrom, D.A., Scanlon, B.R., Zhang, L., 2009. Introduction to special section on
NRCS, 2009. Chapter 7 – hydrologic soil groups. In: NRCS – National Engineering Impacts of Land Use Change on Water Resources. Water Resour. Res. 45, W00A00.
Handbook (NEH), Part 630 – Hydrology. USDA NRCS, Washington, DC, pp. 7.1–7.5. http://dx.doi.org/10.1029/2009WR007937.
Olang, L.O., Furst, J., 2010. Effects of land cover change on flood peak discharges and Tripathi, M.P., Panda, R.K., Raghuwanshi, N.S., 2005. Development of effective
runoff volumes: model estimates for the Nyando River Basin, Kenya. Hydrol. Proc. 25, management plan for critical subwatersheds using SWAT model. Hydrol. Process. 19,
80–89. 809–826.
Pachpute, J.S., Tumbo, S.D., Sally, H., Mul, M.L., 2009. Sustainability of Tu, J., 2009. Combined impact of climate and land use changes on stream-flow and water
rainwaterharvesting systems in rural catchment of Sub-Saharan Africa. Water Resour. quality in eastern Massachusetts, USA. J. Hydrol. 379 (3–4), 268–283.
Manag. 23, 2815–2839. Wagner, P.D., Kumar, S., Schneider, K., 2013. An assessment of land use change impacts
PGRA, 2016. Piano di Gestione del Rischio Alluvioni del bacino del fiume Arno. http:// on the water resources of the Mula and Mutha Rivers catchment upstream of Pune,
www.adbarno.it/adb/?page_id=4830 (last accessed: 09/09/2016). India. Hydrol. Earth Syst. Sci. 17, 2233–2246.
Poelmans, L., Van Rompaey, A., Ntegeka, V., Willems, P., 2011. The relative impact of Wang, G.X., Zhang, Y., Liu, G.M., Chen, L., 2006. Impact of land-use change on
climate change and urban expansion on peak flows: a case study in central Belgium hydrological processes in the Maying River basin, China. Sci. China Ser. D Earth Sci.
Hydrol. PRO 25 (8), 2846–2858. 49 (10), 1098–1110.
Qi, S., Sun, G., Wang, Y., McNulty, S.G., Myers, J.A.M., 2009. Streamflow response to Wang, S., Kang, S., Zhang, L., Li, F., 2008. Modelling hydrological response to different
climate and landuse changes in a coastal watershed in North Carolina. Trans. ASABE land-use and climate change scenarios in the Zamu River basin of northwest China.
52 (3), 739–749. Hydrol. Process. 22, 2502–2510.
Rose, S., Peters, N.E., 2001. Effects of urbanization on streamflow in the Atlanta area Wijesekara, G.N., Gupta, A., Valeo, C., Hasbani, J.-G., Qiao, Y., Delaney, P., Marceau,
(Georgia, USA): a comparative hydrological approach. Hydrol. Process. 15, D.J., 2012. Assessing the impact of future land-use changes on hydrological processes
1441–1457. in the Elbow River watershed in southern Alberta, Canada. J. Hydrol. 412–413,
Saxton, K.E., Rawls, W.J., 2006. Soil water characteristics estimates. Soil Sci. Soc. Am. J. 220–232.
70, 1569–1578. Winnaar, G., Jewitt, G.P.W., Horan, M., 2007. A GIS-based approach for identifying
SCS, 1972. National Engineering Handbook, Section 4, Hydrology. US Government potential runoff harvesting sites in the Thukela River basin, South Africa. Phys.
Printing Office, Washington, DC, pp. 544. Chem. Earth 34, 767–775.
Sekar, I., Randhir, T.O., 2007. Spatial assessment of conjunctive water harvesting Zhou, F., Xu, Y., Chen, Y., Xu, C.-Y., Gao, Y., Du, J., 2013. Hydrological response to
potential in watershed systems. J. Hydrol. 334, 39–52. urbanization at different spatio-temporal scales simulated by coupling of CLUE-S and
Semadeni-Davies, A., Hernebring, C., Svensson, G., Gustafsson, L.-G., 2008. The impacts the SWAT model in the Yangtze River Delta region. J. Hydrol. 485, 113–125.
of climate change and urbanisation on drainage in Helsingborg, Sweden: combined

11

S-ar putea să vă placă și