Theor Appl Climatol

DOI 10.1007/s00704-017-2263-x

ORIGINAL PAPER

Hydroclimatic change disparity of Peruvian Pacific

drainage catchments
Pedro Rau 1 & Luc Bourrel 1 & David Labat 1 & Frédéric Frappart 1,2 & Denis Ruelland 3 &
Waldo Lavado 4 & Boris Dewitte 2,5,6 & Oscar Felipe 4

Received: 12 February 2016 / Accepted: 23 August 2017

# Springer-Verlag GmbH Austria 2017

Abstract Peruvian Pacific drainage catchments only benefit 0.2 °C per decade over all catchments was found. Also,
from 2% of the total national available freshwater while they higher values of temperature and potential and actual evapo-
concentrate almost 50% of the population of the country. transpiration were found over northern latitudes. We chose to
This situation is likely to lead a severe water scarcity and also apply the Budyko-Zhang framework that characterizes the
constitutes an obstacle to economic development. Catchment water cycle as a function of climate only, allowing the iden-
runoff fluctuations in response to climate variability and/or tification of catchments with significant climatic and anthro-
human activities can be reflected in extreme events, pogenic influence on water balance. The Budyko-Zhang
representing a serious concern (like floods, erosion, methodology revealed that 11 out of 26 initial catchments
droughts) in the study area. To document this crucial issue are characterized by low water balance disparity related to
for Peru, we present here an insightful analysis of the water minor climatic and anthropogenic influence. These 11 catch-
quantity resource variability of this region, exploring the ments were suitable for identifying catchments with contrast-
links between this variability and climate and/or anthropo- ing change in their hydroclimatic behavior using the Budyko
genic pressure. We first present a detailed analysis of the trajectories. Our analysis further reveals that six hydrological
hydroclimatologic variability at annual timescale and at ba- catchment responses can be characterized by high sensitivity
sin scale over the 1970–2008 period. In addition to corrobo- to climate variability and land use changes.
rating the influence of extreme El Niño events over precipi-
tation and runoff in northern catchments, a mean warming of
1 Introduction

Changes in climatic conditions appear as more sensitive in

* Pedro Rau dryland regions around the world. In these regions, the effects
[email protected] of climate variability and human activities on runoff are sig-
nificantly more visible than in other climate regions
1
UMR 5563 GET, Université de Toulouse - CNRS-IRD-OMP-CNES, (Mortimore 2009), resulting in a reduction or increase in water
14 Avenue Edouard Belin, 31400 Toulouse, France yield (Brown et al. 2005; Donohue et al. 2011; Chen et al.
2
UMR 5566 LEGOS, Université de Toulouse - 2013).
CNRS-IRD-OMP-CNES, 14 Avenue Edouard Belin, Catchments in the Peruvian Pacific drainage region (here-
31400 Toulouse, France after Pd) are characterized by dryland conditions (e.g., arid
3
CNRS, UMR 5569 HydroSciences Montpellier, 300 Avenue du and semiarid areas), implying water shortage problems for
Professeur Emile Jeanbrau, 34095 Montpellier, France human consumption and economical activities in major cities
4
SENAMHI, Jirón Cahuide 785, 11 Lima, Peru located at lowlands. Besides, they are likely to be affected by
5
Facultad de Ciencias del Mar, Universidad Católica del Norte, the devastating effects of floods (ANA 2012). Previous stud-
Coquimbo, Chile ies such as Lavado et al. (2012) showed evidence of lack of
6
Centro de Estudios Avanzado en Zonas Áridas (CEAZA), uniform regional trend and changes mainly in minimum run-
Coquimbo, Chile off, which they attributed to the anthropogenic influence over
 P. Rau et al.

the 29 catchments that were analyzed. Earlier studies (ANA sensitivity of catchments to changes in water balance along
2012) analyzed the water supply and demand in the main this region having a large latitudinal extent with a steep topog-
catchments as an approximation of water balance with water raphy in the longitudinal direction. Firstly, a hydroclimatic
management purposes. complete description of the Pd is achieved based on the avail-
Quantifying and deciphering the effects of climate variabil- able time series of precipitation, temperature, evapotranspira-
ity and human activities on hydrological regime represent a tion, and streamflow. Then, the Budyko-Zhang framework
major challenge, especially at short scales of time and space was applied to this dataset in order to identify catchments with
(Donohue et al. 2007; Wagener et al. 2010). In order to deci- a low (high) water balance disparity, which are associated with
pher climate variability and anthropogenic influence on water environments with less (more) climatic and human activity
balance, we based our study on the Budyko theory (Budyko influence. Lastly, the use of Budyko space is considered for
1958, 1974). This theory is widely used and is a well- assessing the level of adaptation and sensitivity to climate
established global empirical framework within the hydrologi- variability and land use change.
cal community (Donohue et al. 2011; Coron et al. 2015: Greve
et al. 2015). This method relates the interannual evaporative
index (ratio between actual evapotranspiration and precipita- 2 Study area
tion) and the interannual dryness index (ratio between poten-
tial evapotranspiration and precipitation) in a global descrip- The study area comprises the Peruvian Pacific drainage
tion called the BBudyko space.^ Thereby, all interactions region (Pd) that covers an area of ~ 280,500 km2. This
through the hydrological cycle between vegetation, soil, and area is characterized by a significant altitudinal gradient
atmosphere create an empirical equilibrium represented by the ranging from 0 to ~ 6500 m asl and includes 54 main river
Budyko curve (van der Velde et al. 2013). To emphasize the catchments covering nearly 90% of this region. The rivers
impact of other factors on the water balance such as vegeta- generally flow from east to west from the Andes toward
tion, an emerging general relationship proposed by Zhang the Pacific Ocean (see Fig. 1) with bare and steep slopes
et al. (2001) known as the Budyko-Zhang framework has been from 4 to 9% within small and medium catchment areas
used. This empirical framework comes from an evaluation of from 500 to 16,000 km2 (see values in Table 1) that favor
250 catchments worldwide including dryland regions (Zhang significant rising, flooding, and erosion during huge rain-
et al. 2001). It has been applied to single catchments and fall episodes. This region presents a large range of climate
specific areas until nowadays, considering different ap- types from arid-tropical to tundra climate which results in
proaches as the assessing of their sensitivity to climate change altitudinal and latitudinal gradients in the climate condi-
(Donohue et al. 2011; Renner and Bernhofer 2012; van der tions. While near the coast, dry climate conditions are
Velde et al. 2013). In order to answer properly the issues raised heavily constrained by oceanic conditions that are charac-
by the effects of climate change on water resources (Sivapalan terized by a permanent upwelling south of ~ 5° S, the
et al. 2011), the Budyko curve is recognized Bas a much valu- highlands experience more the influence from seasonal
able tool to back to the basics, it means, the physical basis of variations in the large-scale circulation patterns (i.e.,
catchment water balance^ (Coron et al. 2015). Intertropical Convergence Zone, ITCZ) and the Southern
The degree of anthropogenic influence can be determined Pacific Anticyclone (SPA). Also, at interannual timescale,
using two types of influence on runoff change: human activity anomalous rainfall events over the Pd are related to the El
with direct influence (soil conservation, water control works, Niño Southern Oscillation (ENSO) (Rau et al. 2016).
increasing water demand) and human activity with indirect The region under consideration concentrates more than
influence (land use and land cover changes) (Wang et al. 50% of the Peruvian population. Therefore, all catchments
2013). They constitute descriptive elements to understand of Pd are characterized by recurrent conflicts about the
the behavior of hydroclimatic data series at interannual scale water allocation between water uses located in arid low-
and to identify the catchments presenting a low level of lands ever since this region only benefits from 2% of the
anthropogenization. This selection can be performed through available freshwater in Peru. Water supply for human ac-
an analysis of water balance disparity by catchment via the tivities and consumption is about 80% of the national total
Budyko-Zhang framework, which assumes that catchments and agriculture represents the greatest demand, which is
do not present changes in basin water storage over long-term based entirely on irrigation infrastructure systems because
averages (≥ 10 years) (Zhang et al. 2001). This steady-state of little rainfall over arid lands (ANA 2012). This hydrau-
assumption is related to a closed land water balance, which is lic infrastructure was built along main catchments since
expected to maintain over catchments with a low water bal- the beginning of the 1970s. This development implies the
ance disparity. presence of ten large hydraulic systems with irrigation
This study aims at explaining the hydroclimatic behavior of purposes (see their location and names in gray italic in
the Pd catchments as a benchmark for understanding the Fig. 1b) and two systems mainly used for population
Hydroclimatic change disparity of Peruvian Pacific drainage catchments

Fig. 1 Location of the 26 studied

catchments over the Peruvian
Pacific drainage region: a location
of hydrometeorological stations
considered and land cover derived
from MODIS data in 2008; b
location of major cities and main
large hydraulic infrastructures
(names of hydraulic systems in
italic gray) built between 1970
and 2008, and topography from
SRTM digital elevation model

supply and hydroelectricity purposes for the major cities 3 Data

of Lima and Arequipa. Figure 1b shows the large hydrau-
lic infrastructure predominant in the studied period, where 3.1 Hydroclimatic dataset and validation
one system can present only lake intakes and diversions
(not shown for a reason of scale) as in the case of BTacna^ The database covers the 1970–2008 period and includes
located around latitude 17.5° S, present diversion chan- monthly precipitation, temperature, and streamflow observa-
nels only as in the case of BChavimochic^ and BChinecas^ tions. Precipitation series were obtained from 139
located around latitude 9° S, present two mega-reservoirs pluviometric stations, temperature series from 59 meteorolog-
of 1000 and 250 Hm3 capacity as in the case of BChira- ical stations, and monthly streamflow from 35 hydrological
Piura^ located around latitude 5° S, and present only one stations (see Fig. 1a) managed by the SENAMHI (National
mega-reservoir as the general case of around 300 Hm3 Meteorological and Hydrological Service of Peru).
capacity. These mega-reservoirs are located in highlands A careful quality check of these data was performed as fol-
and midlands, and diversion channels are responsible in lows: monthly precipitation data were previously homogenized
many cases of huge water transfer from the Atlantic to the and validated in this region (see Bourrel et al. 2015 and Rau
Pacific drainage basins (see Fig. 1b for the cities of Lima et al. 2016 for details about the processing). Mean monthly
and Ica). However, in some cases, this infrastructure has temperature data were homogenized and validated following
an influence restricted to areas located downstream of the Lavado et al. (2013). Missing values were filled by monthly
hydrological stations and does not have relevant impacts average and by a multiple correlation method based on nearby
over the water balance upstream of these stations. geographical station data. Monthly streamflow data were
 P. Rau et al.

Table 1 General characteristics

of the 26 studied catchments at No. Catchment Gauging Lat Long A Min Max S Q Qsp
their outlets gauging stations for station °S °W (km2) alt (m alt (m (%) (m3/ (l/s/
the 1970–2008 period asl) asl) s) km2)

1 Piura up Pte. Ñacara 5.11 80.17 4762 119 3526 5.7 27.4 5.7
2 Piura Pte. Sanchez 5.18 80.62 7622 23 3547 3.6 51.3 6.7
Cerro
3 La Leche Puchaca 6.38 79.47 771 355 3982 9.3 6.1 7.9
4 Zaña Batan 6.8 79.29 664 260 3799 7.9 7.7 11.6
5 Chicama Salinar 7.67 78.97 3684 350 4217 8.5 24.7 6.7
6 Moche Quirihuac 8.08 78.87 1883 200 4238 9.2 8.1 4.3
7 Santa up Condorcerro 8.65 78.25 1041 450 6567 8.1 142.2 13.7
8 Santa Pte. 8.97 78.63 1166 18 6567 7.7 200.1 17.2
Carretera
9 Casma Sector 9.48 78.3 2567 71 4769 9.1 6.1 2.4
Tutuma
10 Huarmey Puente 9.97 77.87 1329 555 4742 8.3 5.4 4.0
Huamba
11 Pativilca Yanapampa 10.67 77.58 4196 800 5941 8.4 41.6 9.9
12 Huaura Sayan 11.12 77.18 2893 650 5348 7.0 28.1 9.7
13 Ch. Santo 11.38 77.05 1856 697 5122 7.7 17.8 9.6
Huaral Domingo
14 Chillon Larancocha 11.68 76.8 1207 1200 5099 8.6 6.3 5.2
15 Rimac Chosica 11.93 76.69 2352 906 5370 7.2 31.5 13.4
16 Cañete Socsi 13.03 76.2 5845 330 5632 6.7 54.5 9.3
17 San Juan Conta 13.45 75.98 3057 350 5049 6.9 11.5 3.8
18 Ica La Achirana 13.97 75.68 2119 500 4591 5.7 7.2 3.4
19 Acari Bella Union 15.48 74.63 4242 70 4620 6.1 12.4 2.9
20 Yauca Puente Jaqui 15.48 74.45 4140 214 4923 6.8 9.7 2.3
21 Majes Huatiapa 16.0 72.47 13,414 699 6300 5.4 85.5 6.4
22 Camana Pte. Camana 16.6 72.73 16,238 122 6300 5.4 70.7 4.4
23 Chili Puente del 16.42 71.87 8393 2360 5954 4.3 13.3 1.6
diablo
24 Tambo Chucarapi 16.99 71.64 13,063 281 5554 5.0 33.9 2.6
25 Moquegua Chivaya 17.13 70.83 469 2000 5279 5.8 1.0 2.0
26 Caplina Aguas 17.85 70.12 548 130 5522 7.7 0.8 1.5
Calientes

Lat: latitude; Long: longitude; A: drainage area; Min alt: minimum altitude; Max alt: maximum altitude; S: mean
slope; mean annual values of Q: streamflow; Qsp: specific runoff

homogenized and validated considering the approaches of re- data). Finally, since the only data available for calculating po-
gionalization between neighboring catchments (Lavado et al. tential evapotranspiration (PET) were temperature data, a for-
2012). Missing values (less than 5% of total) were filled by mula relying on clear monthly sky solar radiation and mean
monthly average. monthly air temperature was selected (Oudin et al. 2005):
Precipitation and temperature data were interpolated to a
5 × 5-km grid using the inverse distance weighting technique. Re T þ K 2
PET ¼ i f T þ K2 > 0
Orographic effects on precipitation and temperature were λρ K 1 ð1Þ
accounted for using the SRTM digital elevation model in a PET ¼ 0; otherwise
similar way as described in Ruelland et al. (2014). These effects
on precipitation were considered using the approach proposed
by Valéry et al. (2010) with a correction factor of 4 × 10−4 m−1 where PET is the rate of potential evapotranspiration (mm/day),
(estimated from the observed data), which corresponds to a Re is the extraterrestrial radiation (MJ/m2/day), λ is the latent
20% increase in local precipitation with an elevation of heat flux (2.45 MJ/kg), ρ is the density of water (kg/m3), T is the
500 m. Temperature was interpolated by accounting for a con- mean daily air temperature (°C), and K1 and K2 are fitted
stant lapse rate of − 6.5 °C/km (estimated from the observed parameters (for a general case: K1 ~ 100 and K2 ~ 5).
Hydroclimatic change disparity of Peruvian Pacific drainage catchments

Equation 1 was applied at monthly time step. The Oudin for- (c) Analysis of the hydroclimatic change disparity of select-
mula is a temperature-based evapotranspiration model, ed catchments using trajectories in Budyko space
which is adapted to arid and semiarid regions limited by
scarcity of in situ climate data (see, e.g., Hublart et al.
2015, 2016). Indeed, Oudin et al. (2005) showed that from 4.1 Evaluation of hydroclimatic time series
an operational point of view, this model is as efficient as
more complex models such as the Penman model and its After compiling all datasets by catchment, complementary
variants. analyses were performed for detecting plausible trends and
Finally, we restricted our analysis to 26 catchments (see changes using the nonparametric Mann-Kendall trend test,
Fig. 1a) characterized by a valuable and confident complete the double mass curve method, and the Pettitt test.
dataset of monthly precipitation, temperature, and streamflow The rank-based Mann-Kendall test (Mann 1945; Kendall
series over 1970–2008. 1975) was used to detect trends in the hydroclimatic series.
The method is recommended by the World Meteorological
3.2 Land cover data and digital elevation model Organization (WMO) and widely used for assessing the sig-
nificance of monotonic trends in hydrological series. This
Four land cover maps based on remotely sensed images acquired method does not infer any distributional function for the data
between 1984 and 2008 were used in this study. They are based and has already proven its efficiency (e.g., Lavado et al. 2012;
on multispectral images from different sensors with a spatial Chen et al. 2013; Wang et al. 2013). The double mass curve
resolution increasing against time. The LBA (Large Scale consists of the plot of the accumulated values of one variable
Biosphere-Atmosphere Experiment in Amazonia) Regional versus the accumulated values of another related variable for
Land Cover products for 1984 (DeFries et al. 1998; 8 × 8 km the same period (Searcy and Hardison 1960). The plot is a
grid) and 1992/1993 (Hansen et al. 2000; 1 × 1 km grid) were straight line if the two variables are proportional, and the slope
derived from acquisitions from the Advanced Very High of this line represents the ratio between the two variables. A
Resolution Radiometer (AVHRR) for comparative purposes. change in the gradient of the curve indicates that the original
The land covers of 2001 and 2008 were derived from acquisi- relationship between variables was modified. It can be used to
tions from the Moderate Resolution Imaging Spectroradiometer check the consistency of hydrological records and has recently
(MODIS). These products, made available by the Global Land become an effective tool for detecting the changes of hydro-
Cover Facility (GLCF, http://glcf.umd.edu/data/lc) on logical regime due to anthropogenic disturbances (e.g., Zhang
10 × 10 km grid, have a spatial resolution of 5′. The land cover and Lu 2009; Wang et al. 2013). In this study, a double mass
map used in Fig. 1a is the MODIS land cover type for 2008. curve between precipitation and runoff was employed as an
They show that the Pd region is characterized mainly by four auxiliary confirmation of the change points when catchment
types of land cover: bare ground, open shrublands, grasslands, changes mainly induced by human activities exert influences
and croplands. These different types of vegetation are directly on the river.
related to the different climates observed in the Pd region. The Pettitt test (Pettitt 1979) is a nonparametric approach to
The digital elevation model shown in Fig. 1b was obtained determine the occurrence of a change point. This approach is a
from the 90 × 90-m grid SRTM data (Shuttle Radar rank-based and distribution-free test for detecting a significant
Topography Mission, NASA-NGA, USA available from change in the mean of a time series when the exact time of the
http://srtm.csi.cgiar.org). It allowed the orographic change is unknown. Pettitt test has been widely used to detect
corrections to be applied for the climate forcing interpolation changes in the hydroclimatic records (e.g., Lavado et al. 2012;
(see Section 3.1) and the 26 main river catchments located in Wang et al. 2013). We used the Pettitt test to identify the
the Pd to be delineated (catchment numbering and parameters change point of the runoff series and compare with the change
are shown in Table 1). points detected by double mass curve.

4.2 Catchment water balance disparity

4 Methods
The water balance for a catchment can be basically described
The methodology consists of three steps: in a general form at annual scale as:
P ¼ AET þ R þ ΔS ð2Þ
(a) Evaluation of hydroclimatic time series
(b) Evaluation of the catchment water balance disparity via where P is precipitation (mm/year), AET is the actual evapo-
the Budyko-Zhang framework. Selection of catchments transpiration (mm/year), R is runoff (mm/year), and ΔS is the
with low disparity change in basin water storage (mm/year). At the annual scale,
 P. Rau et al.

ΔS can be neglected especially for long periods (≥ 10 years) the Budyko curve in a predictable manner. Over these
(Zhang et al. 2001). conditions and following van der Velde et al. (2013) and
Based on the Budyko theory (Budyko 1974) which con- Jaramillo and Destouni (2014), the trajectories in Budyko
siders that the available energy and water are the primary space were defined, which are characterized by a direction
factors for determining the rate of actual evapotranspiration, (α) and magnitude (β) of change over the considered period.
we used here the approach developed by Zhang et al. (2001). The direction of change is calculated from the equation:
It was called the Budyko-Zhang curve, which estimates the  
ΔAET⋅ðPÞ−ðAETÞ⋅ΔP
AET based on a simple model (see Eq. 3) that is written as ∝ ¼ arctan ð4Þ
follows: ΔPET⋅ðPÞ−ðPETÞ⋅ΔP


PET where AET (mm/year) is the average AET over the period,
1þw
AET P and ΔAET (mm/year2) is the change of AET over the period.
¼
 ð3Þ
P PET P Also, PET (mm/year), ΔPET (mm/year2), P (mm/year), and
1þw þ
P PET ΔP (mm/year2) are the average PET and change of PET over
the period, the average P, and change of P over the period,
respectively. Here, α is calculated in degrees and counter-
where PET (mm/year) is the potential evapotranspiration and clockwise from the negative y-axis. In the same way, the mag-
w (nondimensional) is the plant-available water coefficient nitude of change is expressed as the following equation and
related to vegetation type. The details of the relationship can expressed in 1/year.
be found in Zhang et al. (2001). The very sensitive parameter vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u ! !2ffi
w is calibrated over the long-term interannual AET from Eq. u ΔAET⋅ðPÞ−ðAETÞ⋅ΔP 2
t ΔPET⋅ðPÞ−ðPETÞ⋅ΔP
3. The use of the Budyko-Zhang curve over the Pd appears as β¼ þ ð5Þ
ðPÞ2 ðPÞ2
a valuable method for the interpretation of the water balance
considering the importance of vegetation (Donohue et al.
Direction is related to change adaptations of ecosystems
2007) over arid and semiarid areas. It has been used previous-
and magnitude related to ecosystem sensitivity to climate var-
ly in comparable studies such as Yang et al. (2009), Zhao et al.
iability and land use change.
(2013), and Chen et al. (2013).
Low disparity of a water balance was evaluated in terms of
three criteria: (a) the shape of the association of points be-
tween dryness index (PET/P) and evaporative index (AET/
5 Results and discussion
P), which must follow a Budyko-Zhang curve with a positive
value of w; (b) the correlation coefficient r between the AET
5.1 Hydroclimatic time series
estimated using the Budyko-Zhang framework and estimated
using the water balance (P–R) which must be higher than 0.7;
Based on the processing of the original monthly time step
and (c) the relative standard error (%RSE) from the curve
database, a complete monthly hydroclimatic dataset of precip-
adjustment which should be less than 15%. Any catchment
itation (P), temperature (T), potential evapotranspiration
outside these three criteria falls off the Budyko curve and is
(PET), and streamflow (Q), over the 1970–2008 period, was
considered as a catchment with a high water balance disparity.
computed, for each of the 26 catchments in a lumped way. The
According to Wang and Hejazi (2011), Jones et al. (2012), and
series of annual PET, annual runoff (R) by the ratio between Q
Coron et al. (2015), such a catchment is interpreted as being
and catchment area, and annual actual evapotranspiration
strongly influenced by anthropogenization, a catchment under
(AET) by water balance (P–R) following the hydrological
strong climatic variability conditions especially droughts, a
year (September–August) were determined. Observed annual
catchment with other missing components of water balance
P, estimated PET, and R series from 11 catchments (mostly
(such as water demands, groundwater flow alteration), or in
covering the 1970–2008 period) are presented for displaying
the worst case, a catchment where there were inadequate mea-
purposes in Fig. 2 (the choice of these 11 catchments among
sures of the hydroclimatic variables.
the whole of our 26 studied catchments is discussed in
Section 5.2).
4.3 Hydroclimatic change disparity Mean annual values of hydroclimatic series are given in
Table 2. For mean annual precipitation, catchments located
Based on Jones et al. (2012), under stationary conditions or in northern areas generally present higher values above
natural climatic oscillations, catchments should fall on the 600 mm/year than southern areas with values under around
Budyko curve. Under nonstationary conditions or anthropo- 400 mm/year. This is because of the influence of the ENSO
genic climate change, catchments are likely to deviate from phenomenon over northern catchments that clearly appears in
Hydroclimatic change disparity of Peruvian Pacific drainage catchments

Fig. 2 a–k Long-term variations

of annual mean time series of
precipitation (P), potential evapo-
transpiration (PET), and runoff
(R) for 11 selected catchments
over the 1970–2008 period

peaks during 1982/1983 and 1997/1998 events known as 400 mm/year in catchments located at central areas such as
years of extreme El Niño events (see Fig. 2a–d). This influ- Santa upstream (no. 7), Santa (no. 8), and Rimac (no. 15),
ence is also present in the runoff variability, decreasing toward associated with the relationship of water availability and
southern latitudes in general, but showing high values above catchment size. For mean annual temperature, PET, and
 P. Rau et al.

Table 2 Hydroclimatic conditions over the 1970–2008 period and 2008 land cover types in the 26 studied catchments. Mean annual values of
precipitation (P), temperature (T), potential evapotranspiration (PET), actual evapotranspiration (AET), runoff (R), and aridity index (P/PET). Land cover
percentage from MODIS (2008) of open shrubland (OS), grasslands (GL), croplands (CL), and bare ground (BG)

No. Catchment Gauging station P T PET AET R P/PET OS (%) GL (%) CL (%) BG (%)
(mm/year) (°C/year) (mm/year) (mm/year) (mm/year)

1 Piura up Pte. Ñacara 613 20 1376 432 181 0.45 27 11 13 0

2 Piura Pte. Sanchez Cerro 551 22 1456 355 212 0.38 34 12 20 0
3 La Leche Puchaca 640 10 831 391 249 0.77 0 75 0 0
4 Zaña Batan 817 12 933 450 367 0.88 0 50 12 0
5 Chicama Salinar 643 14 1013 432 211 0.63 23 69 4 2
6 Moche Quirihuac 645 8 703 509 136 0.92 26 74 0 0
7 Santa up Condorcerro 936 2 400 505 431 2.34 13 86 0 1
8 Santa Pte. Carretera 854 3 473 381 541 1.80 17 76 0 7
9 Casma Sector Tutuma 430 9 769 354 75 0.56 10 63 0 27
10 Huarmey Puente Huamba 580 6 576 453 127 1.01 14 86 0 0
11 Pativilca Yanapampa 700 8 728 387 313 0.96 14 82 0 4
12 Huaura Sayan 654 3 444 348 307 1.47 9 91 0 0
13 Ch. Huaral Santo Domingo 579 1 341 286 303 1.70 9 91 0 0
14 Chillon Larancocha 555 2 370 390 165 1.50 23 77 0 0
15 Rimac Chosica 646 3 436 224 422 1.48 19 81 0 0
16 Cañete Socsi 556 1 339 261 294 1.64 20 74 0 6
17 San Juan Conta 393 4 496 274 119 0.79 32 63 0 5
18 Ica La Achirana 432 5 554 325 107 0.78 38 58 0 4
19 Acari Bella Union 486 8 715 394 92 0.68 23 51 0 26
20 Yauca Puente Jaqui 422 11 873 348 74 0.48 50 25 0 25
21 Majes Huatiapa 537 3 430 336 201 1.25 35 59 0 6
22 Camana Pte. Camana 441 6 593 303 137 0.74 28 47 1 24
23 Chili Pte. del diablo 370 5 556 320 50 0.67 73 3 0 22
24 Tambo Chucarapi 418 6 566 336 82 0.74 50 12 0 38
25 Moquegua Chivaya 369 3 437 305 64 0.84 57 0 0 43
26 Caplina Aguas Calientes 343 2 416 294 49 0.83 60 0 0 40

AET, they decrease in general toward southern latitudes. an annual precipitation around below 400 mm and are
Mean annual PET variability follows the same behavior of defined as arid areas. Catchments located in major rainy
the mean annual temperature variability along the Pd because areas (Santa up (no. 7), Santa (no. 8), Huarmey (no. 10),
of the empirical nature of the Oudin method. However, there is Huaura (no. 12), Chancay Huaral (no. 13), Chillon (no.
a slight increase over arid catchment located in the south 14), Rimac (no. 15), Cañete (no. 16), and Majes (no. 21))
where there is a predominance of bare ground and open shrub- are found in the limits between semiarid and dry subhu-
land areas. mid areas. The rest of the catchments are defined as semi-
It is worth to mention that dryland regions encompass arid areas.
hyperarid, arid, semiarid, and dry subhumid areas Changes in hydroclimatic series are presented in Table 3.
(Brouwer and Heibloem 1986; Mortimore 2009), and An increase or a decrease (i.e., positive or negative sign, re-
these results corroborate the dryland conditions of the spectively) indicates the long-term behavior of the series. All
Pd, accentuating toward southern latitudes. This can be slopes are significant at the 95% level using Student’s t test;
explained by the range of the aridity index (P/PET) pro- however, for highlighting the long-term effect of these
posed by Hassan and Dregne (1997) and contrasted with hydroclimatic series over catchments, the trends and their sig-
the annual precipitation module, a method recommended nificance (at the 95% of confidence level) were estimated
by the United Nations Environment Program (UNEP). based on the Mann-Kendall test (significant trends are
Table 2 provides the values of the aridity index in most shown in bold in Table 3).
of the catchments below B1^ and precipitation below On the one hand, changes in runoff and precipitation are
1000 mm in all catchments. Southern catchments present greater in northern catchments (mainly positive changes) and
Hydroclimatic change disparity of Peruvian Pacific drainage catchments

decrease toward southern latitudes. On the other hand, chang- regional climate influence due to the variability of the Pacific
es in T and PET are always positive (see Table 3). This can be Decadal Oscillation (PDO) and of the intensity of the South
explained by regional climate anomalous conditions along the Pacific Anticyclone (SPA). Our results support the interpreta-
Pd, in particular the well-known ENSO influence over precip- tion of warming at basin scale (i.e., the influence of either
itation and runoff in the northern areas (Lavado et al. 2012; global warming or climate variability at timescales longer than
Bourrel et al. 2015; Rau et al. 2016). Trend tests indicate that a the PDO) where these changes in temperature are expected to
long-term increase in temperature time series is significant in have important consequences for western Andean glaciers.
all catchments with a mean value of + 0.02 °C/year2 (~ a mean No significant increase or decrease in P was observed
warming of 0.2 °C per decade). There is an area covering over the study area (except only for Piura (no. 2) catch-
Moche (no. 6) and Santa up (no. 7) catchments with a mean ment). The R time series present a significant increase
warming above of 0.3 °C per decade. This is consistent with over Ica (no. 18) which is related to water transfers from
previous climatological studies for the Andes such as the one the Atlantic basin (see Fig. 1b) and over Yauca (no. 20)
by Vuille et al. (2015) that showed that most stations over the related with the no significant increase in PET. Runoff
Pd had a significant positive trend in total temperature with a also presents a significant decrease over central catch-
mean warming above of 0.2 °C per decade. They suggested ments (Santa up no. 7, Pativilca no. 11, and Huaura no.
that at inland and higher elevation locations, there is a clear 12) whose upper basins benefit from the melting of the
evidence of warming, and in coastal regions, there is still a Co rdillera Blanca glacier and surroundings. At

Table 3 Mean annual change (Δ) in precipitation (P), temperature (T), potential evapotranspiration (PET), actual evapotranspiration (AET), and
runoff (R). Values with significant trends at 95% of confidence level are shown in bold. Years of significant change point at 95% of confidence level are
indicated between parentheses

No. Catchment ΔP (mm/year2) ΔT (°C/year2 ΔPET (mm/year2) ΔAET (mm/year2) ΔR (mm/year2) w r RSE (%)

1 Piura up 2.84 0.02 (1986) 1.20 (1986) − 1.36 4.20 0.25 0.85 4
2 Piura 3.95 (1997) 0.02 (1986) 0.98 (1986) 1.11 3.28 0.03 0.79 2
3 La Leche 0.18 0.02 1.10 1.07 − 0.90 0.18 0.88 5
4 Zaña 5.19 (1982) 0.01 0.73 4.00 (1983) 1.19 0.00 0.67 7
5 Chicama 3.31 0.03 (1982) 1.71 (1982) 3.15 (1992) 0.16 0.32 0.83 5
6 Moche 1.46 0.06 3.11 0.61 0.85 2.59 0.85 28
7 Santa up 3.80 (1992) 0.04 (1991) 2.12 (1996) 7.28 (1992) − 3.48 3.61 0.84 135
8 Santa 2.43 0.03 1.41 4.15 (1992) − 1.97 − 0.58 0.56 33
9 Casma 1.31 0.03 (1990) 1.53 (1982) 1.44 (1992) − 0.13 1.16 0.95 8
10 Huarmey 4.77 0.03 (1982) 1.75 (1982) 2.76 2.01 3.12 0.85 40
11 Pativilca − 2.90 0.02 1.08 0.33 − 3.22 (1989) 0.02 0.91 6
12 Huaura − 2.70 0.02 1.19 (1996) 0.45 − 3.15 (1987) 0.78 0.71 32
13 Ch. Huaral − 0.33 0.03 (1996) 1.13 (1996) 2.40 − 2.78 0.65 0.64 48
14 Chillon − 0.43 0.03 (1996) 1.58 (1991) − 1.45 1.02 4.08 0.81 101
15 Rimac − 1.08 0.03 (1991) 1.48 (1991) − 2.71 1.63 − 0.49 0.67 11
16 Cañete 2.80 0.01 0.30 4.24 (1996) − 1.44 0.27 0.66 35
17 San Juan 2.38 0.01 0.55 3.77 (1992) − 1.39 0.73 0.77 12
18 Ica 2.43 0.02 (1979) 0.84 (1991) 1.17 1.26 (1993) 1.20 0.92 15
19 Acari 2.70 0.01 (1977) 0.75 (1977) 2.43 0.26 1.61 0.93 14
20 Yauca 2.29 0.01 0.69 0.98 1.31 0.78 0.89 9
21 Majes 1.53 0.03 (1986) 1.34 (1986) 2.72 (1995) − 1.19 1.16 0.90 39
22 Camana 1.23 0.02 (1986) 1.13 (1986) 2.75 (1995) − 1.52 0.47 0.88 9
23 Chili 0.76 0.03 (1986) 1.68 (1986) 0.23 0.53 (1996) 2.79 0.96 29
24 Tambo − 0.60 0.03 (1986) 1.57 (1986) 0.25 − 0.85 1.78 0.95 14
25 Moquegua − 1.02 (1977) 0.03 (1982) 1.47 (1982) − 0.90 − 0.12 3.55 0.98 46
26 Caplina − 0.75 0.02 (1982) 0.85 (1982) − 1.12 0.37 (1996) 4.30 0.97 53

Catchments with low disparity in water balance are italicized

w: Budyko-Zhang coefficient; r: correlation coefficient between AET estimated by Budyko-Zhang and water balance; RSE: relative standard error from
the curve adjustment
 P. Rau et al.

interannual time step, these decreases could be mainly correlation coefficients r and acceptable %RSE are found
related to the intensive water exploitation for agricultural with w positive values in the range of 0.02 to 1.78. That
activities. AET time series present a significant increase in meant that only 11 catchments (italicized in Table 3) fol-
Zaña (no. 4), Chicama (no. 5), Santa up (no. 7), Cañete low a Bquasinatural^ shape of the Budyko-Zhang curve,
(no. 16), San Juan (no. 17), and Camana (no. 22) and a presenting a low disparity in water balance which can be
significant decrease in Rimac (no. 15) catchment. related with low climatic and anthropogenic influence.
Occurrence of change points at 95% of confidence level The average of w values is around 0.7, indicating mean
detected by the Pettitt test and double mass curve methods is values for the plant-available water coefficient related to
significant in some catchments and the year when the change predominance of grasslands over the semiarid conditions
point occurred is indicated between parentheses in Table 3. of the Pd.
Most catchments did not present change points in P (except The 15 remaining catchments are represented by gray
Piura (no. 2), Zaña (no. 4), Santa up (no. 7), and Moquegua points in Fig. 4. Their AET/P values are located very near
(no. 25)). Change points for T time series were grouped by from the energy limit line (sometimes above the line) and far
catchment proximity and regional localization along the Pd. from the water limit line, which implies a complex combina-
For the northern catchments of Piura up (no. 1) and Piura (no. tion of water losing systems (e.g., presence of open water
2), the year of 1986 meant a change point. Catchments in surfaces, water lost to ground water system) with a nonnatural
southern latitudes present significant change points in the behavior of P–R (see Eq. 2) affecting the AET calculation.
years 1977, 1982, and 1986. By the fact of presenting a Precipitation (P) time series follows a natural behavior in all
warming on such a large scale, those change points could be those catchments without strong trends or change points, sug-
related mainly by climatic influence (i.e., ENSO changes and gesting that runoff (R) has been altered. Also, those catch-
climate shifts; see the work by Bourrel et al. 2015 for ments are located mostly in an energy-limited environment
precipitation). Change points in R are mainly related to the (PET/P < 1) where changes in P and PET are likely more
development of new phases into the large hydraulic infrastruc- evident in R (van der Velde et al. 2013), highlighting the
ture as in the case of Ica (no. 18), Chili (no. 23), and Caplina strong climatic influence over R in this type of environments
(no. 26) and also with the land cover change (which is as well as the anthropogenic influence due to the points out-
discussed in the next section) as in the case of Pativilca (no. side the Budyko-Zhang curve.
11) and Huaura (no. 12). The hypothesis about the anthropogenic effect as the main
source of differentiation between catchments with high and
5.2 Catchment water balance disparity low water balance disparity could not be easy to validate
due to the difficult access or/and the scarcity of water use data.
Based on the hydroclimatic time series calculated over the 26 A complementary analysis of the evolution of the land cover
catchments, we produced and analyzed series of dryness index in the study area was performed using the LBA 1984, LBA
(PET/P) and evaporative index (AET/P). Figure 3a, c shows 1992/1993, MODIS 2001, and MODIS 2008 datasets (see
the dispersion of these two indices for two catchments (Piura Fig. 5). Even if improvements in resolution and data quality
upstream (no. 1) and Rimac (no. 15), respectively). Piura up- could impact estimates of spatial and temporal trends (from 10
stream shows the behavior of a northern catchment with to 1 km), the temporal evolution of the classes suggests that
strong climate variability as a result of ENSO influence, and catchments with high water balance disparity presented great
Rimac shows the behavior of a very anthropized catchment as changes over land cover conditions of cropland from 28 to 1%
a result of large hydraulic infrastructure to provide water to the (from 1984 to 1992/1993) and grassland from 39 to 63%
city of Lima (the capital of Peru) in the lowlands. Both catch- (from 1992/1993 to 2001) as shown in Fig. 5a. These changes
ments represent the two main types of associations found in were not observed in catchments with low water balance dis-
the study area, which were differentiated following the meth- parity (excepting croplands from 39 to 7% from 1984 to
odology explained in Section 4.2 in terms of w and B%RSE^ 1992/1993). In addition, Fig. 5b shows the evolution of land
(i.e., see Fig. 3a, c) and r (i.e., see Fig. 3b, d). cover over catchments grouped by latitudinal gradient and
Figure 4 shows the dispersion of indices for all the 26 revealed that central catchments have a large change mainly
catchments, highlighting the two types of association (the in grassland cover from 40% to 75% (from 1992/1993 to
low and high water balance disparity). The w coefficients 2001). It brings more confidence in northern and southern
theoretically take values between 0.1 and 2 (Zhang et al. catchments, which presents certain homogeneity in the
2001) which are associated with the predominance of bare last decades.
soils and forest, respectively (shown in Fig. 4 as About the influence of large hydraulic systems (see Fig.
constraints), even though some negative values on arid 1b), it is evident that some catchments with high disparity
catchments were obtained by Chen et al. (2013). Our w were influenced by those systems which transfer water from
coefficient falls in the range of − 0.58 to 4.30 and the best the Atlantic drainage (i.e., Rimac (no. 15) and Ica (no. 18))
Hydroclimatic change disparity of Peruvian Pacific drainage catchments

Fig. 3 Comparison of two

catchments with low (Piura up no.
1) and high (Rimac no. 15) water
balance disparity according to the
Budyko-Zhang framework. a, c
Black lines represent the energy
limit (diagonal) and the water
limit (horizontal). b, d Scatter plot
for correlation of annual AET ob-
tained from water balance Eq. 2
and AET estimated by using Eq. 3

and by those which present large intakes and regulations for effects according to Fig. 1b (see Piura (no. 2) and Camana
water supply in lower lands (Santa (no. 8), Majes (no. 21), and (no. 22)); however, those effects were not noticeable in terms
Chili (no. 23)). Also, there are some catchments with low of the annual water balance disparity. Finally, the methodolo-
water balance disparity that would still keep anthropogenic gy made possible the separation of catchments with low and
high water balance disparity related to low and high climatic
and anthropogenic influence, respectively. Further research is
needed to separate these two types of influences in the region.

5.3 Hydroclimatic change disparity

According to Fig. 4, most catchments move closer to the the-

oretical water and energy limits. Following the premise that 11
selected catchments have a Bquasinatural^ behavior with less
anthropogenic and climatic influence, it is expected that these
catchments move in all directions through Budyko space. The
direction and magnitude of change is presented in Fig. 6a, b,
respectively, where most of the years in all catchments are
characterized by energy (i.e., PET/P < 1) and water limitations
(i.e., PET/P > 1). Direction and magnitude values are provided
in Table 4.
Figure 6a indicates two types of change adaptations repre-
Fig. 4 Budyko space over the 1970–2008 period over the 26 studied sented mostly with green and red points in one group (i.e.,
catchments. The black lines represent the energy limit (diagonal) and toward more energy-limited environments to the left) and yel-
the water limit (horizontal). Black (gray) points represent the low and orange points in the other (i.e., toward more water-
association of yearly values from catchments showing a low (high)
water balance disparity via the Budyko-Zhang framework. Red dashed
limited environments to the right). Red and green points tend
curves repre-sent the theoretical adjustment range for w parameter to be an energy-limited environment (PET/P < 1) and those
(wmin = 0.1; wmax = 2) catchments are plotted in Fig. 7a. According to Table 3, these
 P. Rau et al.

climate) and land use changes, represented mostly with orange

and yellow points in one group (i.e., toward more energy-
limited environments to the left) and red and purple points in
the other (i.e., toward more water-limited environments to the
right). A large change in magnitude means that climate- or
human-induced changes were particularly large in this region
and that any direction from Fig. 7a having a very small mag-
nitude (see values in Table 4) can be considered as of little
relevance (van der Velde et al. 2013). The orange and yellow
point group has a low magnitude of change corresponding to
catchments with low variability on precipitation and runoff.
The red and purple point group coincides with catchments
presenting significant variability in both precipitation and run-
off. Figure 7b shows the magnitude of change of the 11 se-
lected catchments, and Table 4 shows (italicized) the six catch-
ments identified as having major sensitivity.
They can be described in the light of Table 3 as follows:
purple points represent the catchments of Piura (no. 2) and
Yauca (no. 20), which are located mostly in a water-limited
environment (PET/P > 1) and where an increase in P is likely
to increase the assessment of AET through the Budyko equa-
tion (increase in PET due to increase in T over the studied
period has the opposite effect). As our results indicate that
Fig. 5 Evolution of land cover (%) with LBA and MODIS products over there is an increase in the AET in both cases with a significant
catchments grouped by a low and high water balance disparity. b increase in P for Piura (no. 2) and a significant increase in R
Latitudinal gradient: northern catchments (no. 1 to no. 6), central
catchments (no. 7 to no. 18), and southern catchments (no. 19 to no. 26)
for Yauca (no. 20), we conclude that an increase in precipita-
tion is more likely to drive these ecosystem changes than the
increase in temperature (i.e., increase in PET). Red points
catchments present a decrease in runoff, which can be ex- represent the catchments of Piura up (no. 1), Chicama (no.
plained by the increase of both T and PET and the 5), San Juan (no. 17), and Acari (no. 19) which are located
nonalteration in precipitation, concluding that those catch- mostly in a water-limited environment. However, Piura up
ments have already the tendency to become less energy limit- (no. 1) presents a decrease in AET, suggesting that both pre-
ed. The yellow and orange point groups present high values of cipitation and temperature (i.e., increase in PET) drive these
PET/P representing strong variability of runoff and precipita- ecosystem changes. San Juan (no. 17) presents a decrease in
tion, corresponding to catchments with strong climatic influ- R, and because of its location in both water- and energy-
ence such as the ENSO and semiarid conditions. limited environment, it suggests that both precipitation and
Figure 6b shows two types of catchment sensitivity to cli- temperature (i.e., increase in PET) drive these ecosystem
mate variability (i.e., likely with high sensitivity to future changes. The remaining catchments, Chicama (no. 5) and

Fig. 6 Grouping of Budyko

trajectories defined by their
direction (a) and the magnitude
(b) of change for the 11 selected
catchments in the Budyko space
over the 1970–2008 period
Hydroclimatic change disparity of Peruvian Pacific drainage catchments

Fig. 7 Budyko trajectories

defined by their direction (a) and
magnitude (b) plotted across
catchments. Red/orange, yellow,
and green arrows in a show the
three regions grouped by their
hydroclimatic change adaptation.
Purple and red points in b show
the catchments with major
sensitivity to climate variability
and land use changes

Table 4 Budyko trajectories defined by direction (α) and magnitude

(β) values across the selected catchments. Catchments with major Acari (no. 19), suggest that an increase in precipitation is more
sensitivity to climate variability and land use changes are italicized. likely to drive these ecosystem changes than the increase in
Errors expressed in (%) were obtained via a 10-year moving window temperature.
No. Catchment α (deg) β (1/year)
Piura upstream (no. 1) and Piura (no. 2) catchments exhibit
a large increase in precipitation and runoff changes, and
1 Piura up 97 (2%) 0.009 (30%) Chicama (no. 5) shows a large increase only in precipitation
2 Piura 99 (1%) 0.017 (32%) change, influenced by the ENSO climate conditions, and as
3 La Leche 138 (1%) 0.002 (15%) shown in Fig. 2, those peaks are somehow inconsistent with
5 Chicama 75 (6%) 0.006 (25%) the mean dry conditions.
9 Casma 67 (13%) 0.002 (46%) It is worth pointing out that the hydrologic cycle of the
11 Pativilca 124 (3%) 0.003 (13%) 11 selected catchments is still influenced by human activ-
17 San Juan 49 (9%) 0.008 (22%) ities but only marginally (e.g., see changes in land cover
19 Acari 86 (3%) 0.007 (25%) in Fig. 5a). From an ecosystemic point of view and ac-
20 Yauca 103 (2%) 0.010 (27%) cording to the Budyko theory (see van der Velde et al.
22 Camana 16 (32%) 0.004 (16%) 2013), in those Bquasinatural^ environments, any changes
24 Tambo 73 (3%) 0.002 (2%) in vegetation are due to a co-evolution with the landscape
to optimize water and energy use for given climatic
 P. Rau et al.

conditions. It maintains a steady state. However, the an- in the Budyko-Zhang framework reaches an average value of
thropogenic influence prevents a natural adaptation of 0.7, indicating the predominance of grasslands over the semi-
vegetation species, as expected under the co-evolution. arid conditions. Besides having a climatic influence, large
Strong disturbance could modify this behavior (as found hydraulic systems and irrigation have been identified as one
with the 15 remaining catchments) in a long term, where of the main factors of human activity causing significant
it would be valid to reconsider the statement of a steady changes in runoff in most catchments. Catchments with high
state. water balance disparity presented a significant change in land
The direction and magnitude values in the Budyko relation- cover mainly with a decrease in cropland and an increase in
ship are presented in Table 4. Equations 4 and 5 were also grassland over the studied period.
validated over different periods through a 10-year moving Changes in trajectories in the Budyko space over those 11
window along the 1970–2008 period, which yields a range selected catchments revealed two groups of catchments. Six
of values expressed in error rates (%). catchments—Piura upstream (no. 1), Piura (no. 2), Chicama
(no. 5), San Juan (no. 17), Acari (no. 19), and Yauca (no.
20)—were shown to be sensitive to climate variability (i.e.,
6 Conclusions likely with high sensitivity to future climate) and land use
changes, where precipitation and temperature are the most
This study analyzes the hydroclimatic variability over 26 likely drivers of these environments change. The five other
catchments of the Peruvian Pacific drainage region for the catchments do not reveal any clear behavior in terms of sen-
1970–2008 period. We applied the Budyko-Zhang methodol- sitivity to climate variability.
ogy to analyze the water balance trends over the period and Finally, depending on the observational period of the anal-
determine if climate variability or human activities may have ysis, limitations are mostly related to limited hydroclimatic
influenced these trends. data availability and related to the steady-state hypothesis of
The study area experienced greater precipitation and runoff the Budyko framework. However, despite these limitations,
changes over the northern latitudes in relation with the ENSO our results provide a first assessment of the catchments with
influence. PET has been also studied for the first time in the less climate and anthropogenic influence on water balance and
region as an approach to relate it with temperature (and its its sensibility to climate variability and land use change which
relationship with the altitudinal gradient) using the Oudin has implication for regional water resources assessment and
method. This method is suitable for regions with nondense management.
climatic database as well as under arid and semiarid condi-
tions. Annual values of PET decrease southward with a slight Acknowledgements The authors would like to thank SENAMHI
increase over most arid catchment located in the south due to a (Meteorological and Hydrological Service of Peru) for providing com-
predominance of bare ground and open shrubland areas. plete hydrometeorological raw dataset. We thank the anonymous review-
According to significant upward trends in annual temperature er for his constructive comments that helped improve the original manu-
script. B. Dewitte acknowledges supports from FONDECYT (projects
found in all catchments and regional change points found in 1171861 and 1151185).
most catchments, our results showed a significant warming in
the study area with a mean of 0.2 °C per decade as already Funding information This work was supported by Peruvian Ministry of
identified in studies for South American Andes such as the Education (MINEDU-PRONABEC scholarship).
one by Vuille et al. (2015). Precipitation data do not exhibit
significant trends and change points in most catchments. The
runoff time series presented a significant decrease over central
catchments with a change point in the years 1987 and 1989, References
which is mainly related to the intensive water exploitation for
agricultural activities and land cover change. ANA (2012) Recursos Hídricos en el Peru, 2nd edn. Ministerio de
Following the hydroclimatic description of the region, we Agricultura. Autoridad Nacional del Agua, Lima, pp 45–189
applied the Budyko-Zhang conceptual framework to identify Bourrel L, Rau P, Dewitte B, Labat D, Lavado W, Coutaud A, Vera A,
Alvarado A, Ordoñez J (2015) Low-frequency modulation and trend
plausible annual change disparity of water balance in catch-
of the relationship between ENSO and precipitation along the north-
ments along the study area. This region is characterized most- ern to centre Peruvian Pacific coast. Hydrol Process 29(6):1252–
ly by semiarid conditions and presents strong climate variabil- 1266
ity mainly in the northern regions. We identified 11 out of 26 Brouwer C, Heibloem M (1986) Irrigation water measurement: irrigation
catchments with low disparity according to the Budyko- water needs, vol 3. United Nations Food and Agriculture
Organization, Rome, p 102
Zhang framework related to a low climatic and anthropogenic Brown AE, Zhang L, McMahon TA, Western AW, Vertessy RA (2005) A
influence and suitable for an in-depth analysis with Budyko review of paired catchment studies for determining changes in water
trajectories. The w parameter related to plant-available water yield resulting from alterations in vegetation. J Hydrol 310:28–61
Hydroclimatic change disparity of Peruvian Pacific drainage catchments

Budyko MI (1958) The heat balance of the Earth’s surface. U.S. Oudin L, Hervieu F, Michel C, Perrin C, Andreassian V, Anctil F,
Department of Commerce, Washington, p 259 Loumagne C (2005) Which potential evapotranspiration input for
Budyko MI (1974) Climate and life. International geophysics series, vol a lumped rainfall-runoff model? Part 2—towards a simple and effi-
18. Academic, New York, p 508 cient potential evapotranspiration model for rainfall-runoff model-
Chen Z, Chen Y, Li B (2013) Quantifying the effects of climate variability ing. J Hydrol 303:290–306
and human activities on runoff for Kaidu River basin in arid region Pettitt AN (1979) A non-parametric approach to the change-point prob-
of northwest China. Theor Appl Climatol 111:537–545 lem. Appl Stat 28:126–135
Coron L, Andréassian V, Perrin C, Le Moine N (2015) Graphical tools Rau P, Bourrel L, Labat D, Melo P, Dewitte B, Frappart F, Lavado W,
based on Turc-Budyko plots to detect changes in catchment behav- Felipe O (2016) Regionalization of rainfall over the Peruvian Pacific
iour. Hydrol Sci J 60(7–8):1394–1407 slope and coast. Int J Climatol. https://doi.org/10.1002/joc.4693
DeFries R, Hansen M, Townshend JRG, Sohlberg R (1998) Global land Renner M, Bernhofer C (2012) Applying simple water-energy balance
cover classifications at 8 km spatial resolution: the use of training frameworks to predict the climate sensitivity of streamflow over the
data derived from Landsat imagery in decision tree classifiers. Int J continental United States. Hydrol Earth Syst Sci 16:2531–2546
Remote Sens 19(16):3141–3168 Ruelland D, Dezetter A, Hublart P (2014) Sensitivity analysis of hydro-
Donohue RJ, Roderick ML, McVicar TR (2007) On the importance of logical modelling to climate forcing in a semi-arid mountainous
including vegetation dynamics in Budyko’s hydrological model. catchment. In: Hydrology in a changing world: environmental and
Hydrol Earth Syst Sci 11:983–995 human dimensions (Proc. 7th FRIEND-Water Int. Conf.,
Donohue RJ, Roderick ML, McVicar TR (2011) Assessing the differ- Montpellier, France, 7–10 Oct. 2014). IAHS Publ 363:145–150
ences in sensitivities of runoff to changes in climatic conditions Searcy JK, Hardison CH (1960) Double-mass curves. US Geol Survey
across a large basin. J Hydrol 406:234–244 Water-Supply Paper 1541-B:31–66
Greve P, Gudmundsson L, Orlowsky B, Seneviratne S (2015) Introducing a Sivapalan M, Thompson SE, Harman CJ, Basu NB, Kumar P (2011)
probabilistic Budyko framework. Geophys Res Lett 42(7):2261–2269 Water cycle dynamics in a changing environment: improving pre-
Hansen M, DeFries R, Townshend JRG, Sohlberg R (2000) Global land dictability through synthesis. Water Resour Res 47(10):W00J01
cover classification at 1 km resolution using a decision tree classifier. Valéry A, Andréassian V, Perrin C (2010) Regionalization of precipitation
Int J Remote Sens 21:1331–1365 and air temperature over high-altitude catchments—learning from
Hassan H, Dregne HE (1997) Natural habitats and ecosystems manage- outliers. Hydrol Sci J 55(6):928–940
ment in drylands: an overview. Environment department paper N51. van der Velde Y, Vercauteren N, Jaramillo F, Dekker S, Destouni G, Lyon
World Bank, Washington, pp 1–53 S (2013) Exploring hydroclimatic change disparity via the Budyko
Hublart P, Ruelland D, Dezetter A, Jourde H (2015) Reducing structural framework. Hydrol Process 28:4110–4118
uncertainty in conceptual hydrological modelling in the semi-arid Vuille M, Franquist E, Garreaud R, Lavado W, Caceres B (2015) Impact
Andes. Hydrol Earth Syst Sci 19:2295–2314 of the global warming hiatus on Andean temperature. J Geophys
Hublart P, Ruelland D, García de Cortázar-Atauri I, Gascoin S, Lhermitte Res Atmos 120:3745–3757
S, Ibacache A (2016) Reliability of lumped hydrological modeling Wagener T, Sivapalan M, Troch PA, McGlynn BL, Harman CJ, Gupta
in a semi-arid mountainous catchment facing water-use changes. HV, Kumar P, Rao PSC, Basu NB, Wilson JS (2010) The future of
Hydrol Earth Syst Sci 20:3691–3717 hydrology: an evolving science for a changing world. Water Resour
Jaramillo F, Destouni G (2014) Developing water change spectra and Res 46(5):W05301
distinguishing change drivers worldwide. Geophys Res Lett 41: Wang D, Hejazi M (2011) Quantifying the relative contribution of the
8377–8386 climate and direct human impacts on mean annual streamflow in the
Jones J, Creed I, Hatcher K, Warren R, Adams M, Benson M et al (2012) contiguous United States. Water Resour Res 47(10):W00J12
Ecosystem processes and human influences regulate streamflow re- Wang W, Shao Q, Yang T, Peng S, Xing W, Sun F, Luo Y (2013)
sponse to climate change at long-term ecological research sites. Quantitative assessment of the impact of climate variability and
Bioscience 62(4):390–404 human activities on runoff changes: a case study in four catchments
Kendall MG (1975) Rank correlation measures. Charles Griffin, London, of the Haihe river basin, China. Hydrol Process 27:1158–1174
p 202 Yang D, Shao W, Yeh P, Yang H, Kanae S, Taikan O (2009) Impact of
Lavado WS, Ronchail J, Labat D, Espinoza JC, Guyot JL (2012) Basin- vegetation coverage on regional water balance in the nonhumid
scale analysis of rainfall and runoff in Peru (1969–2004): Pacific, regions of China. Water Resour Res 45:W00A14
Titicaca and Amazonas drainages. Hydrol Sci J 57(4):625–642 Zhang L, Dawes WR, Walker GR (2001) The response of mean annual
Lavado WS, Labat D, Ronchail J, Espinoza JC, Guyot JL (2013) Trends evapotranspiration to vegetation changes at catchment scale. Water
in rainfall and temperature in the Peruvian Amazon-Andes basin Resour Res 37:701–708
over the last 40 years (1965–2007). Hydrol Process 27:2944–2957 Zhang S, Lu XX (2009) Hydrological responses to precipitation variation
Mann HB (1945) Non-parametric tests against trend. Econometrica 13: and diverse human activities in a mountainous tributary of the lower
245–259 Xijiang, China. Catena 77:130–142
Mortimore M (2009) Dryland opportunities. International Union for Zhao G, Mu X, Tian P, Wang F, Gao P (2013) Climate changes and their
Conservation of Nature and Natural Resources. IUCN. IIED. impacts on water resources in semiarid regions: a case study of the
UNDP, Gland-Switzerland, p 86 Wei River basin, China. Hydrol Process 27:3852–3863