Introduction

Since the second half of the twentieth century, alterations in the space-time patterns of meteorological phenomena such as storms, hailstorms and frosts, and weather variables such as temperature, humidity and precipitation have been reported at global, regional and local levels; mainly associated with the occurrence and intensity of phenomena of Climate Variability (VC), Climate Change (CC) and the El Niño - Southern Oscillation (ENSO) (^{Forero, Hernández, & Zafra, 2014}; ^{Pinilla, Rueda, Pinzón, & Sánchez, 2012}; ^{Poveda & Álvarez, 2012}; ^{Ulloa, 2014}). Accelerated CC is generating relevant and indirect effects of great complexity, which are accentuated by the interaction with other controllers of global change (land-use changes, pollution, biotic exchange).

Water resources and coastal ecosystems, marine and terrestrial, present the most obvious impacts, which in turn affect the livelihoods and health of urban populations ( ^{Moreno, 2006}; ^{IPCC, 2014}). The scheme of consequences affecting water resources is diverse with alterations to the water cycle, mainly in the rainfall regime. While certain sites experience abundant rainfall in a short interval of time, causing flooding, other sites experience extended periods of drought (^{Crossman et al., 2013}; ^{Cantú, 2014}; ^{Schewe et al., 2014}). In both cases, water availability and quality are affected, consequently impacting upon agricultural activities, putting the livelihoods and food security of populations, particularly in developing countries, at risk; a situation which if not addressed, could worsen over time leading to an increase in global malnutrition (^{Wheeler & Braun, 2013}). Additionally, the predicted average increase in annual temperatures of 2.5 ° C for the year 2050 and the 2.5% increase in annual precipitation, are likely to lead to soil degradation and a loss of organic matter on the slopes of the Andes ((^{Lau, Jarvis, & Ramírez, 2013}.

Droughts are one of the most damaging natural hazards, and can have devastating effects on social and ecological systems. Unlike other natural hazards (floods, landslides) that are normally limited to relatively small regions and occur at well-defined time intervals, droughts are probably the most expensive and least predictable, since they develop gradually and are only identified once well established (^{Hao & Aghakouchak, 2014}; ^{Hao & Singh, 2015}; ^{Hong, Guo, Zhou, & Xiong, 2015}; ^{Ravelo, Sanz, & Douriet, 2014}). Given the wide variety of sectors affected and their diverse geographical and temporal distribution, it is difficult to establish a unique and precise definition of drought ^{Rajsekhar, Singh, & Mishra, 2015}), however in general terms, drought occurs when the availability of water fails to meet the demands of human and environmental activities for a significant period and over a large area.

Considering the number of victims and losses in disasters of hydrometeorological origin, social, economic and environmental sustainability can and should be improved, with approaches to disaster risk management and adaptation (^{Quintero-Angel, Carvajal-Escobar, & Aldunce, 2012}). It is therefore essential to carry out studies aimed at understanding and characterizing drought and modeling its components, especially in those areas where much of the economic activity depends on water use (^{Wagner, Rossi, Stahl, Bonal, & Herault, 2012}. For this, statistical tools can be used to identify the probability of occurrence, duration and intensity of a drought event (^{Vicario et al., 2015}).

Different indices exist that serve as inputs in disaster prevention programs and generate scientific bases which can be used in risk management and impact mitigation. Some require complex calculations and robust information, making them difficult to apply in areas where the latter is deficient, however, others require only one variable, such as the calculation of precipitation or flow. This work seeks to identify the minimum hydrological extremes that have occurred in the Cauca River in its high valley, using the Streamflow Drought Index (SDI). Through this index, hydrological drought events were characterized, based on their intensity, duration and magnitude; information of vital importance for the formulation of adaptation measures to these events, in one of the most important hydrographic basins of Colombia, where the agricultural sector is highlighted as one of the main economic activities, which can be seriously affected by the occurrence of extreme weather events.

Methodology

Study area

The Cauca River is the second most important surface water source in Colombia, after the Magdalena River. It is born in the Colombian Andean Massif between the Western and Central mountain ranges in the department of Cauca and flows into the Magdalena River. Its elevation varies between 900 to 4 000 masl and it has 39 tributaries, influencing mean flow, which increases up to three times as water is transported along the river from the Salvajina dam (approximately 152 m³ / s) until it exits the department of Valle del Cauca (approx. 568 m³ / s) (^{Enciso, Carvajal, & Sandoval, 2016}). Along its route, the Cauca River passes through more than 180 municipalities.

The Cauca River Basin of approximately 63,300 km² presents various productive activities, principally involved in the sugar industry, coffee cultivation, electricity generation, mining and agricultural exploitation (^{Pérez-Valbuena, Arrieta-Arrieta, & Contreras-Anaya, 2015}). In particular, the Upper Cauca River Basin, is of special strategic importance, since it includes almost all the production chains prioritized by the Colombian Ministry of Agriculture, as well as a diversity of agricultural systems that range from small operations to industrial farms. Additionally, it is the basis of much of the technical and high-value agriculture in the country ((^{Peterson et al., 2012}). The area evaluated for this study was the Upper Cauca River Basin between La Balsa station, located at the entrance of Valle del Cauca, to the Anacaro station located at the exit of the river from the department (Figure 1).

Figure 1 Location of the study area. 

Results

The hydrometeorology of the Upper Cauca River Basin is strongly influenced by the ENSO phenomenon, and as a result of the double passage of the Intertropical Convergence Zone (ITCZ), it has a bimodal rainfall regime. The two periods with the highest rainfall are March-April-May and September-October-November and the two with the lowest rainfall are December-January-February and June-July-August. In response to the bimodality of precipitation, the Cauca river flow in its high valley also has periods of greater and lesser flow, with the July, August and September quarter characterized as by being the driest, while the quarter November, December and January has the highest mean flow.

Four stations were considered for this study: La Balsa, Juanchito, La Victoria and Anacaro, as can be seen in Figure 2, which also shows the behavior of the multi-annual mean flow. The input flow registered at La Balsa station, i.e. at the point where the river enters the department, is lower than the outflow. At Anacaro, maximum monthly values of up to 1 242 m³ / s have been registered, while at La Balsa the maximum was 838 m³ / s - a result which shows the influence of the water contributions to river effluents.

Figure 2 Hydrological stations in the study.  

For the characterization of hydrological drought events through SDI, statistical tests were initially carried out to select the most appropriate distribution for normalization of the flow series, since these did not present a normal distribution. With the Q-Q graphs, the distribution of the data was compared with the normal, log-normal, exponential and gamma distributions. The results obtained for the six-month accumulated flow series at the Anacaro station are shown in Figure 3, from which data it was identified that the distribution that presents the best fit is log-normal. Similar results were obtained in the other three stations. In addition to graphical adjustment, the KS and DA tests were performed. As an example, Table 2 shows the results obtained for the six-month accumulated flow series from Anacaro station, the results indicated for all cases that log-normal was the appropriate distribution to normalize the data.

Figure 3 Q-Q graphs for six-month accumulated flow at Anacaro station, in mm³ / sec. 

The 6 and 12 month accumulated flow data was normalized using the log-normal distribution function, therefore the methodology used by ^{Nalbantis (2008)} was followed and the index was calculated using Equation 1. Regardless of the system analyzed (hydrological, agricultural or ecological), the drought indices SWSI, SPI and SDI have been proven to have greater capacity for calculations over different time scales, since they are better correlated with temporal variability of the different variables ((^{Vicente-Serrano et al., 2012}). The SDI (by means of standard deviations) allowed the identification of the periods in which the 6 and 12 month accumulated flow was below average. The series were constructed from a mobile sum of monthly flow, therefore to facilitate the handling of the information, the month to which reference is made to describe each event, corresponds to the last month of each period. Table 3 shows the registration period for each station and the number of events calculated.

Figure 4 shows the course of the SDI calculated with the six-month accumulated flow series, and Figure 5 shows the twelve-month flow of the Juanchito station. When comparing the graphs, it was found that increasing the grouping reduced the number of events, but resulted in longer events. Additionally, a relationship was observed between hydrological droughts events and the years which experienced the warm phase of the ENSO phenomenon, El Niño, such as 1958, 1977, 1992 and 2016. The effects of the ENSO phenomenon on river flow rates in the country are not negligible; in percentage terms, the main effects on the El Niño phenomenon occur in the Magdalena-Cauca river basin, with an mean reduction of 26% inflow, a mean of 38% in the Cauca river basin, in the Sogamoso and Suarez up to 30%, whilst in the Sumapaz reductions can reach up to 40% and in the Urabá of Antioquia between 30% and 40% (^{García, Piñeros, Bernal, & Ardila, 2012}).

Figure 4 The course of SDIs at Juanchito station. 

Figure 5 Course of SDI_A at Juanchito station. 

Events of longer duration are associated with greater magnitudes, however occasionally, events of short duration with large magnitudes occur. Table 4 summarizes the five semiannual events of the greatest magnitude for each season, the start and end date, the duration and magnitude, and Table 5 gives the annual grouping. The highest figures were presented at La Victoria and Anacaro stations, located downstream. In addition to reflecting possible changes in climatic factors, this finding may also be due to anthropic factors, caused by water intake from the river.

Hydrological drought is determined by the propagation of meteorological drought through the terrestrial hydrological cycle, and therefore, is influenced by the latter's properties (^{Van Loon & Laaha, 2015}). According to the results, the largest hydrological droughts occurred during 2015 and early 2016, during which time occurred one of the strongest El Niño events in history. According to the newspaper (^{El Tiempo, 2016}), this phenomenon triggered an average of 14 fires, and during the 15 months it lasted, Colombia lost 188 650 hectares due to forest fires, rainfall was on average reduced between 30 and 40%, and 80% of the areas affected by the influence of El Niño exhibited a temperature increase of up to 2.5 ° C. This drought historically produced the lowest levels of the Magdalena River and left more than 200 municipalities in calamity due to water shortages. The cost to the country in terms of prevention and emergency care was 1, 6 billion pesos. The above reflects the influence of the ENSO in Colombia, which during the El Niño warm phase, mainly generates a decrease in precipitation, and consequently, reduces river flow, soil moisture and plant activity (^{Rojo, 2011}).

Although these results provide an approximation for the characterization of hydrological droughts, it is important to consider that calculating the SDI with flow values, which in turn are estimated from Level-Flow curves, a high percentage of error may be implied. The calibration curves used to estimate flow are associated with high uncertainty and can be simple or complex, depending on the flow regime and the channel characteristics (^{Carvajal-Escobar, 2010}). Furthermore, it is important to remember that during dry periods there is an increase in temperature, and this can lead to an increase in evapotranspiration of crops, potentially increasing irrigation needs, putting greater pressure on rivers, and consequently decreasing river levels in a manner similar to flow, not only due to natural factors but also as society's demand for water increases. In small rivers, these changes significantly affect flow and calculation uncertainty is higher, which is why we worked with the flow of the Cauca River, which has larger and more constant flows concerning other smaller rivers.

Conclusions

River flow is one of the elements of the hydrological cycle with the greatest consequences for the life of human beings, therefore, the identification and characterization of temporally and spatially extreme hydrological events which allow the assessment of regional and local water availability is an essential component of water planning. Although, hydrological drought is defined as a significant decrease in water availability in all forms that appears in the hydrological cycle, the flow rate is the main variable and allows us to understand the general behavior of surface water resources. The SDI is therefore an important tool for the characterization of hydrological droughts, especially in areas with little information.

The Cauca River is the second most important source of surface water in the country and although studies on this river have typically focused on maximum events - since over the years the river has registered significant rainfall increases leading to flooding - the results of this study indicate that it has also been affected by hydrological drought, which has caused significant reductions to flow, mainly during periods coinciding with the occurrence of El Niño, with the events of 1958, 1977, 1992 and 2016 showing the largest magnitudes and/or intensities of hydrological drought.

In Colombia, most of the rivers in small basins have no records, and in many cases, the stations are located downstream of the main extractions, making it difficult to extrapolate these studies to other rivers, however, it is recommended to apply them to other rivers which are of importance to the country.