Multi-decadal variations in annual maximum peak flows in semi-arid and temperate regions of Chile

Abstract

In determining the possible influence of climate change, it is important to understand the temporal and spatial variability in streamflow response for diverse climate zones. Thus, the aim of this study was to determine the presence of changes in annual maximum peak flow for two climate zones in Chile over the past few decades. A general analysis, a flood frequency analysis and a trend analysis were used to study such changes between 1975 and 2008 for a semi-arid (29°S–32°S) and a temperate (36°S–38°S) climatic zone. The historic annual maxima, minima and mean flows, as well as decadal mean peak flow, were compared over the period of record. The Gumbel distribution was selected to compare the 30-year flood values of two ±15-year intervals, which showed that streamflow decreased by an average of 19.5% in the semi-arid stations and increased by an average of 22.6% in the temperate stations. The Mann-Kendall test was used to investigate the temporal changes in streamflows, with negative trends being observed in 87% of the stations analysed in the semi-arid zone, and positive trends in 57% of those analysed in the temperate zone. These differences in streamflow response between climate zones could be related to recent documented increases in altitude of the zero-degree isotherm in the Andes Mountains of Chile, since most of the significant positive and negative changes were detected in first-order rivers located closer to this mountain range.

Editor D. Koutsoyiannis; Associate editor H. Lins

Citation Pizarro, R., Vera, M., Valdés, R., Helwig, B., and Olivares, C., 2013. Multi-decadal variations in annual maximum peak flows in semi-arid and temperate regions of Chile. Hydrological Sciences Journal, 59 (2), 300–311.

Résumé

Pour déterminer l’influence possible du changement climatique, il est important de comprendre la variabilité temporelle et spatiale de la réponse en débits pour diverses zones climatiques. Ainsi, l’objet de cette étude était de déterminer l’existence de changements dans le débit de pointe maximum annuel de deux zones climatiques du Chili au cours des dernières décennies. Une analyse générale, une analyse de la fréquence des crues et une analyse de tendances ont été utilisées pour étudier les changements des valeurs du débit de pointe maximal annuel entre 1975 et 2008 pour la zone semi-aride (29°S à 32°S) et la zone tempérée (36°S à 38°S). Les maximums, minimums et moyennes historiques annuels, ainsi que le débit de pointe décennal ont été comparés sur la période d’observation. Nous avons choisi la distribution de Gumbel pour estimer les valeurs des crues tricennales (30 ans) de deux intervalles de 15 ans, ce qui a montré que la valeur de ce débit a diminué en moyenne de 19,5% dans les stations semi-arides et augmenté en moyenne de 22,6% dans les stations tempérées. Le test de Mann-Kendall a également été utilisé pour étudier les changements temporels des débits. Des tendances à la décroissance sont observées dans 87% des stations de la zone semi-aride, et des tendances à la croissance dans 57% de celles de la zone tempérée. Ces différences de la réponse en débit entre zones climatiques pourraient être liées à une élévation récemment mise en évidence de l’isotherme zéro degré dans la Cordillère des Andes du Chili, puisque la plupart des changements positifs et négatifs significatifs ont été détectés dans des rivières du premier ordre proches de cette chaîne de montagnes.

Key words:

• climate change influence
• flood frequency analysis
• probability distribution functions
• Mann-Kendall test
• peak flows
• zero-degree isotherm
• semi-arid climate
• temperate climate

Mots clefs:

• influence du changement climatique
• analyse de la fréquence des crues
• fonction de distribution de probabilité
• test de Mann-Kendall
• débit de pointe
• isotherme zéro degré
• climat semi-aride
• climat tempéré

INTRODUCTION

Streamflow is a reflection of the influence of atmospheric variables and the anthropogenic behaviour on a watershed (Hodgkins and Dudley 2006, Stewart 2009). For example, maximum or peak flows tend to increase in response to increased precipitation and snowmelt rates, and their estimation is commonly required in the risk analysis of hydraulic structures (Van Gelder et al. 2007, Eum et al. 2011). As flooding is a natural non-periodic process (Ollero 1996), an understanding of the underlying processes is important in designing infrastructure able to account for possible future extreme flood events, as well as to effectively manage water resources (Linsley et al. 1988). Flood frequency analyses are very effective in determining the probabilities of exceedence using successive years of annual maximum peak flow data (Stedinger et al. 1993).

With this in mind, research has shown that the rate of glacier melt has increased worldwide over the last few decades. In Chile, most of the glaciers have significantly regressed in response, theoretically, to climate change and the associated increases in temperatures (Rivera et al. 2000, Rignot et al. 2003). According to the Intergovernmental Panel on Climate Change, mean annual surface air temperature has increased by 0.02–0.05°C per decade during the period 1901–2005 throughout Latin America, with the largest increase of over 0.1°C per decade occurring in southeastern Brazil (Jones et al. 2007). In this context, the objective of this study is to determine the presence of changes in annual maximum peak flow for two climate zones in Chile over the past few decades. A general analysis, a flood frequency analysis and a trend analysis were used to analyse annual streamflow data in order to clarify part of the uncertainty surrounding the possible influence of climate change on the hydrological variation found in the two zones under research.

MATERIALS AND METHODS

Study area

Semi-arid zone

The semi-arid climate zone analysed in this study is located between latitudes 29° 02′ S and 32° 16′ S in the Norte Chico natural region of Chile, and pertains to the first-order administrative region of Coquimbo. The region is bordered to the north by the region of Atacama, to the east by Argentina, to the south by the region of Valparaíso, and to the west by the Pacific Ocean. It has a surface area of 40 579.5 km² and represents 5.36% of the national territory excluding Antarctica (Instituto Nacional de Estadísticas 2008).

The following landforms are present in the region: littoral plains and the merged Andes and Coastal mountain ranges, whose east–west-oriented mountains form the Transverse Valleys that cross the region and in which are located three large river basins: the Elqui, Limarí and Choapa (Instituto Nacional de Estadísticas 2008). The Elqui, Limarí and Choapa rivers all have a mixed rain- and snow-fed runoff regime, and have drainage basin areas of 9657, 11 297 and 8239 km², respectively (Cabezas et al. 2007). Streamflow information from the Elqui and Limari basins was used for the frequency and trend analyses carried out in this study.

According to the classical worldwide climate classification developed by Köppen (1900), the region has a climate type BWk, i.e. a cold semi-arid climate with annual precipitation not exceeding 400 mm and annual temperatures of less than 18°C (Peel et al. 2007). In addition to low temperatures, evaporation is significant at high altitude in the Andes mountain range in the region, which results in a mixed melt regime with maximum streamflow occurring during the spring and summer seasons, mainly from November to December (Favier et al. 2009).

Temperate zone

The temperate climate zone analysed in this study is located to the south of the semi-arid zone between latitudes 36°00′S and 38°30′S in the Central and South natural regions of Chile, and pertains to the first-order administrative region of Biobío. The region spans a linear north–south distance of 926 km and is bordered to the north by the region of Maule, to the south by the region of La Araucanía, to the east by Argentina, and to the west by the Pacific Ocean. It has a surface area of 37 062.6 km² and represents 4.2% of the national territory excluding Antarctica (Instituto Nacional de Estadísticas 2008).

In this zone, the following landforms are present: littoral plains that are most well developed in the south of the region, where their width averages 25 km, the Coastal Range in the north that reduces in elevation to a series of rolling hills with intermontane basins in the south, the Nahuelbuta mountain range, the Intermediate Depression and the Andes mountain range. These landforms define two large river basins, the Itata and the Biobío. The Itata River drains an area of 11 204 km², is located in the north of the region, and is fed by its primary tributaries the Diguillín, Larquie and Ñuble rivers. It has a mixed snow- and rain-fed regime, with maximum streamflows occurring in the autumn and winter seasons, mainly from April to September. The Biobío River is located in the south of the region, drains 24 029 km² and is the widest river in the country. It also has a mixed regime with maximum streamflow occurring in the winter and spring seasons, most markedly in the Vergara and Laja tributaries (Instituto Nacional de Estadísticas 2008). In this case, only annual maximum streamflow data for Itata River basin were used for frequency and trend analysis.

According to Köppen (1900), the region has a Csb climate: a temperate climate with warm, dry summers where the mean temperature does not exceed 22°C during the hottest month and does not drop below 10°C for 4–5 months of the year. This climate is often found as a transitional climate between the hot-summer Mediterranean (Csa) and Oceanic (Cfb) climates, but has milder summers than the Csa climate and, unlike the Cfb climate, its dry season occurs during the hottest months of the year (Peel et al. 2007).

Period of record

In this study, stream gauging stations with a period of record of 15 or more continuous years (n ≥ 15) were considered for analysis. In this context, Linsley et al. (1988) noted that there is an 80% probability that the design discharge will be overestimated using a period of record of 20 years and that 45% of these overestimations will surpass the observed values by over 30%. Other authors, such as Pizarro and Novoa (1986), determined that a minimum of 15 years is necessary in order to draw statistical inferences with minimally acceptable precision. With this in mind, 29 stream gauging stations with periods of record between 1975 and 2008 were analysed; 15 are located in the semi-arid zone (latitudes 29°S–32°S) and 14 in the temperate zone (latitudes 36°S–38°S). Figure 1 shows the geographical distribution of the two climate zones and the stations selected for analysis, as well as the boundaries of the analysed basins. In the same sense, Table 1 shows the periods of record, the drainage basin surface area and the geographical coordinates for each of the analysed stations.

Fig. 1 Study area and location of the selected stations (the stations are described in Table 1).

Table 1 Stations used in the study and their characteristics.

Map number	Station/location	Station code	Period of record	Basin area (km^2)	South Lat	West Long
Semi-arid (Coquimbo Region)
1	Río Toro antes junta río la Laguna	TOLL	1986–2008	560	29°58′	70°05′
2	Río Turbio en Varillar	TUVA	1979–2008	4148	29°56′	70°31′
3	Río Cochiguaz en el Peñón	COPE	1986–2008	820	30°07′	70°26′
4	Estero Derecho en Alcohuaz	EDAL	1986–2008	206	30°13′	70°29′
5	Río Elqui en Algarrobal	ELALG	1975–2008	5729	29°59′	70°35′
6	Río Elqui en Almendral	ELALM	1975–2008	6681	29°58′	70°53′
7	Río Cogotí en Fraguita	COFR	1987–2008	475	31°06′	70°53′
8	Río Grande en Cuyano	GRCU	1986–2008	1262	03°55′	70°46′
9	Río Grande en las Ramadas	GRLR	1983–2008	544	31°00′	70°34′
10	Río Grande en Puntilla San Juan	GRSJ	1987–2008	3512	30°42′	70°55′
11	Río Guatulame en él Tome	GUET	1991–2006	2.362	30°48′	70°58′
12	Río Hurtado en Angostura de Pangue	HUAP	1981–2008	1.772	30°26′	71°00′
13	Río Hurtado en San Agustín	HUSA	1981–2008	656	30°27′	70°32′
14	Río Limarí en Panamericana	LIPA	1983–2008	1261	30°40′	71°32′
15	Río Mostazal en Cuestecita	MOCU	1981–2008	353	30°48′	70°36′
Temperate (Biobío Region)
16	Río Sauces antes junta con Ñuble	SAÑU	1975–2008	595	36°39′	71°16′
17	Río Ñuble en la Punilla	ÑULP	1975–2008	1280	36°39′	71°19′
18	Río Ñuble en San Fabián	ÑUSF	1975–1998	1709	36°34′	71°33′
19	Río Cato en Puente Cato	CAPC	1975–1998	868	36°33′	72°02′
20	Río Chillán en camino a Confluencia	CHCO	1987–2008	162	36°37′	72°19′
21	Río Itata en General Cruz	ITGC	1975–2008	4731	36°56′	72°21′
22	Río Itata en Trilaleo	ITTR	1975–2008	852	37°04′	72°11′
23	Río Diguillín en San Lorenzo	DISL	1975–2008	674	36°54′	71°34′
24	Río Diguillín en Longitudinal	DILO	1978–2008	224	36°52′	72°20′
25	Río Itata en balsa Nueva Aldea	ITNA	1975–2008	1822	36°39′	72°27′
26	Río Itata en Coelemu	ITCO	1987–2008	1440	36°28′	72°45′
27	Rio Andalien camino a Penco	ANPE	1985–2008	1232	36°49′	73°01′
28	Río Itata en Cholguán	ITCH	1975–2008	1420	37°09′	72°04′
29	Río Chillán en Esperanza	CHES	1975–1994	718	36°47′	71°44′

General analysis of the historic streamflows

The annual streamflow data set was provided by the Dirección General de Aguas (DGA) of Chile and includes flows for 1975–2008. For each station, the percentage of annual maximum peak flow values above and below the historic annual mean was calculated. Furthermore the annual mean (Q_mean), the maximum (Q_max) and the minimum flows (Q_min) were calculated for the whole available period of record in order to establish two indexes: I_max = Q_mean/Q_max × 100 and I_min = Q_min/Q_mean × 100, with the objective to analyse the differences between the ratios of the maximum and minimum streamflows with respect to the mean for both the semi-arid and temperate zones (Fig. 2). Then the coefficient of variation (CV) was calculated in order to quantify the existing variability between the observed annual peak flows with respect to their means. This measure was calculated to represent the percentage of variation between the standard deviation and the mean flows for each selected station (Fig. 3).

Fig. 2 Annual maximum peak flows, percentage of data above/below the historic mean, and Q_max and Q_min indexes for the semi-arid and temperate zones.

Fig. 3 Coefficient of variation (CV) results for the annual maximum peak flows for all stations located in the semi-arid and temperate zones.

Then a multi-decadal analysis was carried out, dividing the period of record into the following sub-periods: 1970–1979, 1980–1989, 1990–1999 and 2000–2008, with the objective to analyse and compare trends in the streamflow values obtained for each station within each basin.

Flood frequency and trend analysis

A flood frequency analysis was performed using four frequency distribution functions: Gumbel, Goodrich, log-normal and Pearson type III, to determine which distribution most accurately fits the real data, with the objective of selecting one of them for use in the frequency analysis.

The Gumbel distribution (Gumbel 1958) was included in the group of the selected probability distribution functions (PDFs) because it had been used in Chile, showing an adequate capacity to fit precipitation and streamflow maxima data (Pizarro 1996). The Gumbel distribution is expressed as follows:

(1)

where x represents the random variable, e is the Neper constant, and μ and d are the parameters (Aparicio 1997).

The Goodrich distribution is a function that adjusts very well to mean values (Linsley et al. 1982). According to Pizarro et al. (1993), the Goodrich distribution is expressed as follows:

(2)

where x represents the random variable; α, p and x₁ are the adjustment parameters; and e is the Neper constant, as above.

The log-normal distribution (Aitchison and Brown 1957) improves on the shortcomings of the normal Gaussian distribution by being able to account for the positive skew commonly found in hydrological data (Chow et al. 1994). In this distribution, normal logarithms are used to normalize values by order of magnitude, which then permits the random variable to be analysed using standard normal distribution techniques. The log-normal distribution is expressed as follows:

(3)

where α and β are the distribution parameters (Aparicio 1997).

The classical Pearson type III distribution (Pearson 1894) is also one of the most widely used distributions in hydrology, because it is very flexible, diverse and able to normalize annual precipitation and streamflow maxima (Chow et al. 1994). It is part of the two-parameter family of continuous probability distributions referred to as Gamma distribution, which can be expressed as follows:

(4)

obtained from distribution parameters:

(5a)

(5b)

(5c)

The flood values for return periods of 10, 20, 30, 40, 50 and 100 years were calculated using these four frequency distribution functions, and goodness of fit was determined using the Kolmogorov-Smirnov (K-S) test (Canavos 1988) for a value of significance of 0.05, and the coefficient of determination (R²) (Steel and Torrie 1960).

After reviewing the results of the above, the Gumbel distribution was selected for further analysis because it has been widely used for estimating the magnitudes of different hydrological variables, and has been cited extensively in related literature due to its ability to fit extreme values (e.g. Linsley et al. 1977, Témez 1978, Pizarro and Novoa 1986, Chow et al. 1994, Pizarro et al. 2001, UNESCO 2007, Pizarro et al. 2012, 2013).

Finally, the Mann-Kendall test (Kendall 1938, 1975, Mann 1945) was performed to analyse annual streamflow trends for all the stations using a significance of α = 1%. A benefit of this test is that no assumptions about the probability distribution of the data are needed. It has been widely used in the hydro-climatological literature; therefore, details of its calculation are not presented in this study, but can be found in Hipel et al. (1988), Zetterqvist (1991), Chiew and McMahon (1993), Burn (1994), Yulianti and Burn (1998), Lins and Slack (1999), Douglas et al. (2000), Yue and Wang (2004), Yue and Pilon (2004), Burn et al. (2008), Drapela (2011) and Mondal et al. (2012). In the same sense, to estimate the true slope of an existing trend as change per year, the Sen nonparametric method was calculated (Theil 1950, Sen 1968).

RESULTS AND DISCUSSION

General analysis of the historic streamflows

A first approximation of the behaviour of the historical streamflow is analysed from Fig. 2, which shows the differences between the maximum streamflow values for the semi-arid and temperate climatic zones. The percentage of annual maximum peak flow values above and below the historic mean flow for each analysed station is also presented. For the semi-arid zone, all the analysed gauging stations showed streamflow values below their respective historic mean; e.g. 75% of annual maximum streamflows at Rio Cogotí en Fraguita (CO_FR) station were less than the historical mean flow. In years when streamflow values above the mean occurred they were found to be mainly associated with outliers related to extreme events occurring in these semi-arid basins, which usually increase the mean flow value (Ashkar et al. 1993). Similar results were observed in the temperate zone; however, only around a 70% of the stations showed streamflow values below their respective historic means, and the percentage of annual maximum streamflows below the mean was in general lower than for stations located in the semi-arid zone, e.g. 58% of the annual maximum flows at Rio Ñuble en San Fabián (ÑU_SF) station were below the mean value. In general, the percentage of data below and above the historic mean is very comparable for the semi-arid and temperate zones, but the results are more pronounced in the former.

The calculated indexes and showed totally different results: the average value for was around 23% in the semi-arid zone, and 47% in the temperate zone, suggesting that in most cases the mean flows are closer to maximum flows in the temperate zone. The average values of were 11% and 0.63%, respectively, confirming that the ratio between minimum and mean flows is greater in the semi-arid zones.

Annual and multi-decadal streamflow analysis

Using results obtained by analysing the CV, it is possible to indicate that the semi-arid zone showed a higher variability when compared to the temperate zone. The greatest variability in the semi-arid zone was observed in the Estero Derecho en Alcohuaz (ED_AL) station (656.4%). In the temperate zone, the greatest variability was observed at the Río Diguillín en Longitudinal (DI_LO) station (77.3%) (Fig. 3).

Table 2 shows that, for 80% of stations in the semi-arid zone, the historic maximum annual peak flow occurred between 1980 and 1989, and for the remaining 20% of stations, between 1990 and 1999. In the temperate zone, the historic maximum annual peak flow occurred between 2000 and 2008 for 64% of stations, between 1990 and 1999 for 21% of stations, and between 1980 and 1989 for the remaining 15% of stations. This analysis can be considered as a first approach to identifying the streamflow trend over recent decades, and suggests a difference between the semi-arid and temperate zones. In the semi-arid zone, lower streamflow values were detected in the more recent decades, and in the temperate zone greater streamflows were detected during the same period. This is true for most of the analysed stations within the corresponding climate zones. For example, Río Elqui en Almendral station (EL_ALM, semi-arid) showed a decadal mean peak flow value of 55.22 m³ s^-1 for 1980–1989, decreasing to 40.21 m³ s^-1 for 1990–1999, and finally to 28.05 m³ s^-1 for 2000–2008. Other semi-arid stations, Río Elqui en Algarrobal (EL_ALG), Rio Grande en Puntilla San Juan (GR_SJ) and Rio Limari en Panamericana (LI_PA), also showed large negative differences when comparing the last two decades in analysis. A decrease in annual maximum peak flow values was observed in the last 8 years (2000–2008) for 87% of the semi-arid stations.

Table 2 The decade in which the historic annual maximum peak flow occurred.

Decade	Number of stations in the semi-arid zone	Number of stations in the temperate zone
1970–1979	–	–
1980–1989	12	1
1990–1999	3	3
2000–2008	–	10
Total	15	14

Most stations located in temperate zones showed that streamflows have increased over the last decades, but this trend is not as strong as the decrease observed in the semi-arid zone. Nevertheless, significant changes were also detected, e.g. decadal mean peak flow of Río Ñuble en la Punilla (ÑU_LP) stationincreased by nearly 300% when comparing the 2000–2008 period with respect to the earlier periods. This recent increase in annual maximum peak flow values was observed for 64% of temperate zone stations (Fig. 4).

Fig. 4 Decadal mean peak flow values for the semi-arid and temperate zones.

Flood frequency and trend analysis

The four frequency distributions (Gumbel, Goodrich, log-normal and Pearson type III) were used to fit the streamflow data and obtain their goodness of fit through K-S and R² tests (Table 3). All four frequency distribution functions fitted the data adequately; however, the log-normal distribution showed the best performance, with a K-S value of 100% for both climatic zones, and average R² values of 95% and 93% for the semi-arid and temperate stations, respectively. The Gumbel distribution also fit the data very well, with average K-S values of 86.67% and 100%, and average R² values of 82.30% and 95% for the semi-arid and temperate stations, respectively. The fit with the Goodrich and Pearson type III distributions was significant for fewer stations. For temperate zone stations, the Gumbel PDF fits the data better than the log-normal PDF. However, it was clear that in the semi-arid zone the log-normal PDF fits more accurately because rainfall and streamflow data there have more variability (Pizarro et al. 2012).

Table 3 The goodness-of-fit test results for the four distributions for the two climate zones.

Climatic zone	PDF	K-S test	R^2
Semi-arid	Gumbel	86.7%	82.3%
	Log-normal	100%	95%
	Goodrich	53.3%	63.9%
	Pearson type III	73.3%	77.7%
Temperate	Gumbel	100%	95%
	Log-normal	100%	93%
	Goodrich	80%	92%
	Pearson type III	80%	94%

The Gumbel PDF was chosen for further analysis of the 30-year flood values because it has been used with good results in diverse hydrological studies in Chile (Pizarro et al. 2001, UNESCO 2007, 2012, 2013). A return period of 30 years was considered appropriate in the study of temporal streamflow trends in the medium term, because it is one of the periods used in the design of hydrological works such as sewage and drainage systems, and as most of the evidence of climate change has become evident in the most recent decades. For this analysis, the series of each station was divided into two sub-periods: for the semi-arid zone, the 30-year flood values were calculated using 1981–1994 and 1995–2008 data, and for the temperate zone the sub-periods 1975–1990 and 1991–2008 were selected. This difference in the division of the periods of record was based exclusively on the detail that most of the semi-arid stations presented data availability from 1981 onwards, whereas most of the temperate stations have records dating back to 1975. Table 4 shows the 30-year flood values for both semi-arid and temperate zones. The 30-year peak flow values decreased between 1981–1994 and 1995–2008 at 73% of the semi-arid stations. The most dramatic case was Río Hurtado en Agostura Pangue (HU_AP), which fell from 57.75 m³ s^-1 in 1981–1994 to 25.4 m³ s^-1 in 1995–2008, a 56% of decrease in peak flows. This general decrease in peak flow values in the semi-arid basins can be compared to results obtained by Garreaud et al. (2010), which showed that precipitation in the semi-arid climate zone has decreased regardless of the season. Additionally, they state that precipitation has become concentrated almost exclusively in winter due to the passage of storm fronts, and that evaporation has increased at high altitudes. Favier et al. (2009) indicated that the rivers in the region have a mixed hydrological regime, dominated principally by snowmelt in the summer months, with maximum contributions to streamflow occurring between the spring and summer months of November and December; consequently, decreased streamflows downstream can be attributed to decreases in snow and glacier water storage as a result of a diminished accumulation during the winter months. This can be mainly related to the altitudinal increase in the equilibrium line altitude (ELA), an approximation of the zero-degree isotherm (Carrasco et al. 2005, 2008) produced by the temperature increase. Furthermore, the land-use change in recent decades has been significantly related to an increase in agricultural uses; therefore they can be considered as another factor influencing the response of streamflow in these basins.

Table 4 30-year flood values (m³ s^-1) for stations located in the semi-arid and temperate zones.

Station	30-year peak flow (m^3 s^-1)
	1981–1994	1995–2008	Difference (%)
Semi-arid (Coquimbo region)
TOLL	12.14	9.9	–18.46
TUVA	68.58	48.98	–28.59
COPE	36.69	39	6.29
EDAL	20.67	22.41	8.4
ELALG	292.03	141.98	–51.38
ELALM	120.87	121.9	0.86
COFR	38.65	32.12	–16.9
GRCU	54.28	58.34	7.49
GRLR	34.16	18.97	–44.46
GRSJ	179.42	161.62	–9.92
GUET	141.11	93.35	–33.85
HUAP	57.75	25.4	–56.01
HUSA	11.42	9.9	–13.33
LIPA	188.83	138.83	–26.48
MOCU	9.27	7.76	–16.33
Average	84.39	62.03	–19.51
	1975–1990	1991–2008	Difference (%)
Temperate (Biobío region)
SAÑU	810.18	910.88	12.43
NULP	1554.6	3627.94	133.37
NUSF	3212.77	2579.46	–19.71
CAPC	1376.35	1832.39	33.13
CHCO	1017.13	1032.71	1.53
ITGC	1655.42	1774.23	7.18
ITTR	1416.91	1681.28	18.66
DISL	668.75	849.37	27.01
DILO	1733.37	2196.14	26.7
ITNA	3379.61	3509.3	3.84
ITCO	11006.52	8724.12	–20.74
ANPE	269.06	324.62	20.65
ITCH	1055.63	1280.26	21.28
CHES	404.93	611.98	51.13
Average	2111.52	2209.62	22.6

In contrast, in the temperate basins the 30-year flood values between 1975–1990 and 1991–2008 increased at 86% of the analysed stations, most notably at the Río Ñuble en la Punilla (NU_LP) station, whose values increased by 133.4% (Table 4). Similar observations have been made in comparable highly forested regions around the world with the same climate characteristics as the Biobío Region; e.g. an analysis of surface runoff for both large and small watercourses in the Cascade Mountains in western Oregon, USA, showed that peak flow values have gradually increased over the last few years in 90% of drainage basins devoid of vegetation, and in 40% of significantly forested drainage basins (Thomas and Megahan 1998). According to Aguayo et al. (2009), the main transformations in land use or change in the Biobío Region can be mostly attributed to the development of forestry, but also to agricultural and urban changes. The same authors indicate that the forest plantation area increased by eight times from 1979 to 2000; and during the same period the urban area increased by around 50%, mainly by encroachment into agricultural areas. It has been estimated that positive land use/cover changes have accounted for over 50% of the reduction in mean annual streamflow for catchments analysed in China by Zhang et al. (2008). Therefore, in the Biobío Region, negative trends should be expected if the hypothesis is based on this statement; however, most of the stations presented positive trends during recent decades. This suggests that the influence of the altitudinal increase in the ELA in these climate zones can be attributed to a combination of a reduction in snowfall and increase in temperatures during the winter months (Burn et al. 2008), which produces greater runoff in the areas closer to the mountain ranges. Therefore, the contrasting observed changes in streamflow in the two regions could be explained by the same phenomenon, though it is more accentuated in the semi-arid zone because less forest coverage is present and the agricultural land use has resulted in an increased use of water.

Finally, the Mann-Kendall analysis was carried out to detect possible significant trends in the past decades. The results showed decreasing trends in streamflow in 87% of the analysed stations in semi-arid basins, and increasing trends in 57% of the analysed stations in temperate basins. Furthermore, in most cases, significant negative or positive trends (α ≤ 0.1) were detected only in stations located in first-order streams, i.e. those located closer to the Andes Mountains (Table 5). Therefore, it can be noted that these stations are not greatly affected by land-use change, and so it is suggested that the behaviour of the streamflows could be associated with the altitudinal displacement of the zero-degree isotherm. An example of the trends stated can be observed in Fig. 5, which shows the behaviour of the annual maximum peak flow vs the historic mean flow for the stations Río Turbio en el Varillar (TU_VA) in the semi-arid zone, and Río Itata en General Cruz (IT_GC) station in the temperate zone. In general, the results were consistent at all the levels of analysis, suggesting a decrease in peak flows in the semi-arid basins and an increase in temperate basins.

Table 5 Mann-Kendall test results for stations located in the semi-arid and temperate zones.

Station/location	Period	n	Test Z	Significance α	Sen’s slope estimate	Sen’s slope confidence interval α = 0.01
Semi-arid (Coquimbo Region)
Río Toro antes junta la Laguna^*	1975–1997	23	–1.67	0.1	–0.06	–0.3	0.02
Río Turbio en Varillar^*	1979–2008	30	–1.78	0.1	–0.31	–1.08	0.16
Río Cochiguaz en el Peñon	1987–2008	22	–0.25	n/s	–0.07	–1.49	0.32
Estero Derecho en Alcohuaz	1986–2008	23	–0.37	n/s	–0.03	–0.52	0.11
Río Elqui en Algarrobal	1975–2008	34	0.3	n/s	0.09	–1.43	1.47
Río Elqui en Almendral	1975–2007	33	0.54	n/s	0.12	–1.02	1.23
Río Cogoti en Fraguita	1981–2008	28	–0.73	n/s	–0.06	–0.53	0.19
Río Grande en Cuyano	1986–2008	23	–0.21	n/s	–0.05	–0.96	0.68
Río Grande en las Ramadas^*	1983–2008	26	–1.72	0.1	–0.2	–0.81	0.1
Río Grande en Puntilla San Juan	1987–2008	22	–0.51	n/s	–0.26	–2.08	1.06
Río Hurtado en Angostura de Pangue	1991–2006	16	–1.13	n/s	–1.17	–3.75	1.49
Río Hurtado en Angostura de Pangue	1981–2008	28	–0.69	n/s	–0.08	–0.61	0.18
Río Hurtado en San Agustín	1981–2008	28	–1.23	n/s	–0.06	–0.25	0.08
Río Limarí en Panamericana	1983–2008	26	–1.59	n/s	–0.91	–4.09	0.27
Río Mostazal en Cuestecita	1981–2008	28	–1.36	n/s	–0.08	–0.19	0.05
Temperate (Biobío Region)
Río Sauce antes junta con Ñuble	1979–2008	30	1.5	n/s	8.26	–6.66	21.1
Río Ñuble en la Punilla*	1975–2008	34	2.33	0.05	32.3	–2.83	66.11
Río Ñuble en San Fabián	1975–1998	24	–0.32	n/s	–3.4	–81.9	64.16
Río Cato en puente Cato	1975–2008	34	1.63	n/s	13.64	–5.79	32.95
Río Chillán en camino a Confluencia	1987–2008	22	–0.11	n/s	–1.01	–33.64	30.32
Río Itata en general Cruz	1975–2008	34	2.05	0.05	14.26	–4.85	38.13
Río Itata en Trilaleo	1982–2008	27	–0.67	n/s	–7.68	–37.1	25.89
Río Diguillin en San Lorenzo	1975–2008	34	0.8	n/s	2.04	–6.33	10.19
Río Diguillin en Longitudinal^*	1981–2008	28	1.84	0.1	22.1	–9.21	51.08
Río Itata en Balsa Nueva Aldea	1975–2008	34	–1.04	n/s	–13.02	–53.27	25.94
Río Itala en Coelemu	1987–2008	22	0.34	n/s	21.72	–250.53	254.58
Río Andalien camino a Penco	1985–2008	24	–1.22	n/s	–3.32	–8.44	4.58
Río Itata en Cholguan	1975–2008	33	–0.33	n/s	–1.33	–17.54	9.17
Río Chillán en Esperanza	1975–1994	20	0.55	n/s	3.86	–10.53	19.41

Notes: n/s is α < 0.1 level of significance;

* Station located closer to the Andes Mountains.

Fig. 5 Mann-Kendall trends for annual maximum peak flows at Río Turbio en el Varillar (RT_VA) station in the semi-arid zone and Río Ñuble en la Punilla (ÑU_LP) in the temperate zone.

CONCLUSIONS

Streamflows in the semi-arid and temperate climate basins of Chile have presented different behaviours over the past few decades. Our analyses show that, in general, peak flow values have decreased in the semi-arid zone, and have increased in the temperate zone. These alterations in peak flow are likely due to underlying changes in the atmospheric processes that govern streamflow, possibly as a result of climate change. In this study, the changes in annual maximum peak flow values are suggested to have been caused by recent changes in land use and/or are due to the altitudinal increase in the zero-degree isotherm in the Andes, which can affects streamflow differently depending on the characteristics of the watercourses in question. A 30-year flood analysis was carried out; 30 years is the design period of hydrological works for sewage and drainage systems in Chile, and most of the evidence of climate change has become apparent in the most recent decades. The implications of the streamflow variations described are a reduction in the availability of water resources in the semi-arid zone, and a need for changes in the statistical analysis supporting the design of hydraulic works in the temperate zone. Hydrological models should be used in future studies to thoroughly quantify the possible influence of snowmelt in the streamflow response.

Acknowledgements

The authors would like to express their gratitude for the Dirección General de Aguas de Chile (DGA), a directorate of the Ministry of Public Works, for providing the data used in this study.