Abstract
Persistent drought events frequently intensify the aridity of ecosystems and cause catchment runoff depletion. Here, using large and long-term data sets of meteorological and hydrologic variables (precipitation, runoff, temperature, and potential evapotranspiration) investigated the major causes that modulated catchment runoff depletion between the years 1980 and 2020 in southern-central Chile. We identified the hydrological years where aridity index intensified, analyzed its relationship with annual runoff, and evaluated the effect of the annual evaporation index and annual aridity index on water balance of 44 catchments with different precipitation regimes located between 35° and 40°S. Our results showed that observed precipitation and runoff significantly decreased between 1980 and 2020 in 64% of the catchments in the study area. Potential evapotranspiration increased significantly in 39% of the catchments. Hydrological years in which precipitation decreased showed a decreased runoff trend. This result suggests that meteorological droughts tend to significantly decrease observed runoff. The runoff value decreased as the aridity index increased from 0.3 to 6.7, and the Budyko curve captured 98.5% of the annual variability of all catchments.
HIGHLIGHTS
Significant contrast in aridity in response to decreased precipitation. Consecutive droughts lead to a moderate runoff deficit and a higher aridity index.
Extreme aridity index over 6.5, PET far exceeds mean annual Q and P.
INTRODUCTION
Droughts have been recognized as a climatic hazard (Mishra & Singh 2010; Seneviratne et al. 2010, 2012; Garreaud et al. 2017), although this depends on their duration and intensity. They can often lead to a substantial decrease in surface flows and cause soil water scarcity (Loon et al. 2014; Bianchi et al. 2016; Rivera et al. 2017; Alvarez-Garreton et al. 2021), damaging both ecological and human social systems (Bevacqua et al. 2022). Mediterranean and temperate rain ecosystems are particularly prone to droughts as they are mainly controlled by regional precipitation (Ragab & Prudhomme 2002; Rockström et al. 2010). Severity of droughts can vary significantly within a climatic zone (Van Lanen et al. 2013; Swain 2015). Since the early 1980s, Chilean Mediterranean and temperate rain ecosystems have experienced intense and prolonged droughts (Garreaud et al. 2017; Alvarez-Garreton et al. 2021), occurring simultaneously along the coast (Quintana & Aceituno 2012) and the Andes (Masiokas et al. 2016). In fact, drought severity has been accentuated by an almost uninterrupted precipitation deficit over the last 40 years.
In today's rapidly changing global environment, the degree of aridity of a region becomes a direct measure and predictive tool of the magnitude of climate change (Van Loon & Laaha 2015; Wu et al. 2022; Zomer et al. 2022) that in particular under Mediterranean and temperate rain ecosystems creates implications for catchments surface flows (runoff). Many studies have suggested that climate change will increase aridity because of changes to energy and water exchange between the Earth's surface and a warmer atmosphere (Dai 2013; Feng & Fu 2013; Fu & Feng 2014; Sherwood & Fu 2014; Scheff & Frierson 2015; Fu et al. 2016; Park et al. 2018; Marvel et al. 2019; Peli et al. 2023). Most of these studies have used the aridity index (AI) as a proxy to evaluate aridity changes, following a relationship approach between precipitation and potential evapotranspiration (PET) (Arora 2002; Nastos et al. 2013; Greve et al. 2019; Zomer et al. 2022). The AI allows the monitoring and prediction of drought (Nastos et al. 2013), and its use as a climate index has become increasingly important as many regions of the world are experiencing unprecedented long periods of drought due to changes in climate (Alvarez-Garreton et al. 2021). Increases in this index have led to a significant reduction in catchment surface flows (e.g., Griffin & Anchukaitis 2014; Schewe et al. 2014; Swain 2015; Williams et al. 2015), which has in turn negatively impacted riparian ecosystems as well as forest and agricultural production (Van Dijk et al. 2013).
Recent studies in Chile have identified a precipitation deficit that has led to a 90% decrease in annual river runoff in southern-central Chile (Garreaud et al. 2017; Alvarez-Garreton et al. 2021). This decrease in precipitation has also impacted coastal ecosystems streamwater and nutrient cycles (Masotti et al. 2018), vegetation (Garreaud et al. 2017; Arroyo et al. 2020), fire regimes (González et al. 2018), and water supply (Muñoz et al. 2020). Therefore, understanding variation in the AI provides useful information about the decrease in runoff during a multi-year drought in southern-central Chile. The runoff response to the aridity signal may be due to a more disconnected river drainage network, due to low precipitation input, increased temperature, and PET. This could be further related to depleted levels of catchment storage (Eltahir & Yeh 1999; Arora 2002; Ragab & Prudhomme 2002; Van De Griend et al. 2002; Rockström et al. 2010; Saft et al. 2016a; Barrientos & Iroumé 2018; Boisier et al. 2018; Barrientos et al. 2020). Therefore, these ecosystems are threatened by future decreases in precipitation and hazards related to the severity of drought events (Hagemann et al. 2013; Schewe et al. 2014; Jiménez Cisneros et al. 2014; Garreaud et al. 2017).
Assessments of runoff decrease during a drought regime of the catchments in the Mediterranean and temperate rain ecosystems are missing to date. Our goal is to identify the catchments and specific hydrological years where runoff was maintained, intensified, or attenuated due to aridity in 44 catchments located between 35° and 40°S in central south Chile. Expected results will help us to explore the underlying factors affecting runoff, as well as the influence of natural climatic variability considering historical and future changes. These analyses may help to reduce the uncertainty associated with environmental planning in Chile and to develop better strategies for managing water supply, agriculture, forest fires, and understand changes in vegetation.
MATERIALS AND METHODS
Study area and data sources
Catchment locations considered in the analyses (n = 44) for the southern-central hydro-climatic zone of Chile between 35 and 40° S (Köppen classification updated for continental Chile) (Sarricolea et al. 2017). The area in green has a Mediterranean climate and the area in blue has a temperate rain climate. Meteorological stations are presented by black circles and fluviometric stations by red triangles.
Catchment locations considered in the analyses (n = 44) for the southern-central hydro-climatic zone of Chile between 35 and 40° S (Köppen classification updated for continental Chile) (Sarricolea et al. 2017). The area in green has a Mediterranean climate and the area in blue has a temperate rain climate. Meteorological stations are presented by black circles and fluviometric stations by red triangles.
The hydrometeorological data for the past 40 years were obtained from the hydrometric network of the Dirección General de Aguas DGA-Chile. Fluviometric and meteorological stations were selected considering a maximum of 15% missing time series recorded data. Daily time series at the catchment scale considered streamflow, precipitation, evapotranspiration, and temperature from January 1980 to December 2020. Selected information encompassed the hydrological seasonality in the study area and defined the hydrological year from April of each year to next year March.
Database analysis
), evapotranspiration (ET), and temperature (
) of 232 meteorological stations, were grouped at an annual resolution between April 1 and March 31 of each hydrological year. Gaps in time series were filled according to Boisier et al. (2016) and Alvarez-Garreton et al. (2021). Subsequently, streamflow data were standardized by surface unit to obtain runoff (
) in mm of each catchment for each hydrological year. 
was calculated using Equation (1), as proposed by Hargreaves & Samani (1985). Daily maximum and minimum temperature (
) data series were considered to calculate 
according to Barrientos & Iroumé (2018). Estimates of PET considered the following equation:
represents the potential evapotranspiration for a single day (mm/day), 
represents the average daily temperature in °C. Incident solar radiation 
was estimated from extraterrestrial solar radiation data using Equation (2) according to Samani (2000). Estimates of 
were obtained using the following equation:
is the extraterrestrial solar radiation proposed by Allen et al. (1998) and is represented in mm/day of evaporated water, the KT coefficient is an empirical coefficient that can be calculated from atmospheric pressure data. However, Hargreaves (cited in Samani 2000) recommends 
for interior areas and 
for coastal areas. tmax and tmin represent the maximum and minimum daily temperature (°C). Subsequently, the 
data were aggregated to an annual resolution between April 1 and March 31 of each hydrological year.Subsequently, the time series of precipitation (
) and 
were used to calculate the fractions of precipitation and PET at the catchment scale. A database, organized by hydrological year, was developed in raster format with an output cell size of 500 × 500 m in ArcGIS®. Reflecting the approaches of Heine (1986) and Barrientos et al. (2020), Kriging interpolation was used to interpolate values for unsampled points across the catchments that extend throughout the study area between 35° and 40°S.
) was determined using (Budyko's 1951, 1974) approach and was calculated using Equation (3) (Wouter et al. 2020). The Budyko method states that there is a functional relationship between the 
, the precipitation (
) and the 
(Wouter et al. 2020; Ni et al. 2022; Dai et al. 2023):
Years when the AI was >1 are classified as dry because precipitation does not fulfill PET; otherwise, years with an AI <1 are classified as humid according to Arora (2002).
Runoff trends and progressive aridity
Runoff trends and the AI were calculated on a yearly basis from January 1980 to December 2020. This methodology aligns with the natural hydrological cycle of the study area which spans from a year April 1 to next year March 31. The nonparametric Mann–Kendall test (Mann 1945; Theil 1950; Kendall 1975) and Sen's slope (Sen 1968) were used to quantify the significance and slope of the trend. The Mann–Kendall test assumes that the data series, here the sum and the median of the hydrological year, continuously increase or decrease. This applies to the detection of a monotonic trend of a time series without seasonality or other annual patterns.
Analyses considered a p-value threshold of 0.05 to assess whether the null hypothesis may be rejected. Therefore, when the p-value was < 0.05, the null hypothesis was rejected and a significant change in runoff was explained by the AI of the hydrological year. If p-value was > 0.05, the null hypothesis was not rejected and no relationship was found between runoff and the AI. Sen's slope is the median value calculated from estimates of the trend slope between hydrological years. The slope is an estimate that indicates the magnitude at which a trend increases or decreases per unit of time (hydrological year) considering the long-term runoff and index of each catchment (Sen 2012, 2017; Waikhom et al. 2023).
Years with lower precipitation standard deviation and greater shift toward more negative values were selected as drought events. A Pearson correlation test (r coefficient) was used to test the relationship between runoff and the AI during 13 drought years. The Pearson correlation was considered statistically significant when p-value ≤ 0.05. All statistical analyses were performed using R (version 4.1.1) and Rstudio (version 2021.09.0; R Core Team 2018).
RESULTS
Hydrometeorological trend of catchments
Trend of precipitation (a), runoff (b), and potential evapotranspiration (c) of the catchments. Sen's Slope estimates: red line, *(p < 0.05), **(p < 0.01), and ***(p < 0.001) indicate significant trend, and + not indicate significant trend. Names of catchments available in Supplementary information, Table S1.
Trend of precipitation (a), runoff (b), and potential evapotranspiration (c) of the catchments. Sen's Slope estimates: red line, *(p < 0.05), **(p < 0.01), and ***(p < 0.001) indicate significant trend, and + not indicate significant trend. Names of catchments available in Supplementary information, Table S1.
Hydrological years trend
Precipitation (a) and runoff (b) trends for all evaluated hydrological years. Sen's slope estimates adjusted considered: red line, period 1980–1981 to 2019–2020; blue line, period 1980–1981 to 1998–1999; green line, period 1999–2000 to 2019–2020.
Precipitation (a) and runoff (b) trends for all evaluated hydrological years. Sen's slope estimates adjusted considered: red line, period 1980–1981 to 2019–2020; blue line, period 1980–1981 to 1998–1999; green line, period 1999–2000 to 2019–2020.
Change in runoff trends during drought
). The lowest observed runoff during the 1998–1999 drought was 57.4 mm/year, which was 55.5 mm lower than the observed runoff (112.9 mm/year) during the 1990–1991 drought, and 60.5 mm lower than the observed runoff during the mega-drought observed from 2010 to the present day (annual values from 1980 to 2020 are shown in Supplementary information, Table S2). Years with higher runoff standard deviation showed a greater shift toward more negative values during years of precipitation deficits (Figure 5). 
Precipitation histograms for all evaluated catchments for drought years. (Annual precipitation values available in Supplementary information, Table S2).
Precipitation histograms for all evaluated catchments for drought years. (Annual precipitation values available in Supplementary information, Table S2).
Runoff histograms for all evaluated catchments for drought years. (Annual runoff values available in Supplementary information, Table S2).
Runoff histograms for all evaluated catchments for drought years. (Annual runoff values available in Supplementary information, Table S2).
Relationship between runoff and indicators during drought
) (<1,190 mm) and higher 
(>1,145 mm) led to a higher AI in the final year of drought, and a subsequently higher mean runoff as runoff (
) is closely related to the AI (Supplementary information, Table S1). In contrast, after a long drought and in normal hydrological years (Supplementary information, Table S2), the mean value of runoff (
) is higher than the mean value of precipitation and PET, and the mean value of PET does not exceed precipitation. For an extreme aridity value above 6.5 (such as during the hydrological years of 1998–1999 and 2019–2020), PET far exceeds the average annual runoff and precipitation (Supplementary information, Table S2). 
(a) Runoff and AI relationship during 13 droughts years. (b) Relationship between the evaporation index (
) and the 
during droughts years. The orange line represents the ‘Water limit’ for which long-term evapotranspiration cannot exceed precipitation. The red line represents the ‘Energy limit’ for which long-term evapotranspiration cannot exceed its potential.
(a) Runoff and AI relationship during 13 droughts years. (b) Relationship between the evaporation index () and the during droughts years. The orange line represents the ‘Water limit’ for which long-term evapotranspiration cannot exceed precipitation. The red line represents the ‘Energy limit’ for which long-term evapotranspiration cannot exceed its potential.
vs. 
relationship during 13 drought years.
DISCUSSION
Although droughts are particularly common in Mediterranean and temperate ecosystems, as they are mainly controlled by precipitation (Ragab & Prudhomme 2002; Rockström et al. 2010) and are synonymous with lower runoffs (Fowler et al. 2022), over the last 40 years, the Mediterranean and temperate ecosystems in Chile have experienced lower than expected runoff during intense and prolonged droughts (Quintana & Aceituno 2012; Masiokas et al. 2016; Garreaud et al. 2017; Alvarez-Garreton et al. 2021). Our results suggest that the observed decrease in precipitation has a larger influence in catchments further north where fewer precipitation events contribute to annual runoff. The southern catchments have been less affected by persistent precipitation deficits over the past 40 years, leading to a significant contrast in aridity in response to decreased precipitation. Our results complement the findings of Alvarez-Garreton et al. (2021) which include examples from California, USA (Avanzi et al. 2020), China (Tian et al. 2020), and Australia (Hughes et al. 2012; Saft et al. 2016b). More importantly, they evidence the vulnerability of Mediterranean and temperate ecosystems to persistent precipitation deficits. The role of drought intensification in runoff response and water supply highlights the need to better understand drought severity. Extreme or severe dry spells have been extremely unusual in terms of runoff, tending to show a greater shift toward negative runoff values. This is extremely important in Mediterranean and temperate ecosystems, where changes in the intensification of droughts and runoff deficit have already been detected due to climate change (Cortés et al. 2011; Boisier et al. 2018; Bozkurt et al. 2018).
During years of severe drought, our results show that there are differences in runoff response, which are significantly correlated with the AI. Runoff is sensitive to the degree of severity of the AI. In years and catchments with less precipitation and high runoff coefficient, the AI is high; therefore, the sensitivity of runoff to climate change is greater in years where AI is low. These results complement the findings of Wang et al. (2021), Ni et al. (2022) and Dai et al. (2023), who claim that with continued increase in warming and aridity, runoff will gradually decrease. This means that runoff in northern catchments will gradually become less sensitive to increases in precipitation, while runoff in southern catchments will become more sensitive to decreases in precipitation.
The model of changes in PET /annual precipitation ratios represents the runoff trend under the degree of aridity during the period from 1980 to 2020, which is the exchange of energy and water between the land surface and a higher quality atmosphere (Arora 2002; Dai 2013; Feng & Fu 2013; Nastos et al. 2013; Fu & Feng 2014; Sherwood & Fu 2014; Scheff & Frierson 2015; Fu et al. 2016; Park et al. 2018; Greve et al. 2019; Marvel et al. 2019; Zomer et al. 2022). However, the AI does not fully represent the complexity underlying runoff processes (Roderick et al. 2015; Scheff et al. 2017; Greve et al. 2019). Runoff is a complex process that requires a comprehensive assessment of climatic, ecological, geographic, geological, and anthropogenic variables effects (van Dijk et al. 2013; Schewe et al. 2014). For example, anthropic activities such as agriculture and water extraction have undoubtedly contributed to the different responses of catchment runoff to the precipitation deficit since 1980 (Arroyo et al. 2020; Muñoz et al. 2020; Villablanca et al. 2022). The regulation of reservoirs alters hydrological regimes, leading to a decrease in average annual runoff downstream of the upstream reservoirs (Villablanca et al. 2022). These cause a decrease in runoff, which are not directly related to precipitation deficits, but are probably a combined result with hydro-climatic conditions that need to be understood as hydrological and anthropic processes (Alvarez-Garreton et al. 2019, 2021, 2023).
Our study has shown that the relationship between runoff and AI provides a reasonable indicator of the hydrological vulnerability of Mediterranean and temperate ecosystems to unprecedented long-term precipitation deficits due to changes in climate in a single dimensionless number and may provide a helpful warning indicator for monitoring catchment ecosystem water resource availability.
Based on our results, and given that runoff is a complex process that requires a comprehensive assessment of multiple variables and factors, it is important to consider indicators on water resource planning challenges related to changes in precipitation and the response of runoff to global climate change (Fowler et al. 2022). A late start with drier conditions, such as those reported by Garreaud et al. (2017) and Alvarez-Garreton et al. (2021), can induce and amplify future effects related to the supply of water for human consumption, as well as in the biological support and aquifer recharge within a catchment (Eltahir & Yeh 1999; Van De Griend et al. 2002; Lehner et al. 2017; Fowler et al. 2020; Muñoz et al. 2020). However, currently these are not accounted for in runoff response projections.
CONCLUSION
40-year changes in the runoff behavior of 44 catchments located between 35° and 40°S provide evidence of the vulnerability of Mediterranean and temperate ecosystems to uninterrupted precipitation deficits. Our results reveal that the runoff response of these catchments is closely related to their AI and that this response can vary during normal hydrological years, after a long drought, during consecutive droughts, and for years with extreme droughts. The drought and reduced runoff trend is greater in the north than in the south, where precipitation events prevail, and cause a significant contrast in the aridity response to the decrease in precipitation. Consecutive droughts of more than three hydrological years lead to a moderate runoff deficit and a high AI in the final year of drought. Finally, severe single-year droughts, as occurred in the 1998–1999 and 2019–2020 hydrological years, induce larger runoff deficits. We have shown that, for any type of drought, precipitation and evapotranspiration are key factors modulating catchments runoff response. We adhere to calls of previous studies suggesting that policy-makers should consider scientific research for environmental planning challenges related to expected changes in runoff. This will allow us somehow to reduce the uncertainty of the effects of climate change on water availability needs affecting social and productive demands and planning to reduce potential ecological impacts.
ACKNOWLEDGEMENTS
This work was funded and supported by the government of Chile through the ANID IDeA I + D ID23I10011 and ANID BASAL FB210015. We also want to recognize the contribution, support, and infrastructure of Departamento de Obras Civiles of the Universidad Católica del Maule, Centro Nacional de Excelencia para la Industria de la Madera (CENAMAD) of the Pontificia Universidad Católica de Chile, Laboratorio de Investigación de Suelos, Aguas y Bosques (LISAB) and Cooperativa de Productividad Forestal at the Universidad de Concepción, Chile. We also want to recognize the contribution of DGA-Chile, who provided data. We would also like to recognize the associate editor and anonymous reviewers for their feedback during the peer-review process.
AUTHOR CONTRIBUTIONS
G.B. was involved in data analyses, conceptualization, and was the main responsible author for manuscript writing. R.R. wrote and revised the manuscript, and provided funding for postdoctoral research. E.D. and A.P. provided analyses support and revised the manuscript. All the authors have read and approved the final version of the manuscript.
DATA AVAILABILITY STATEMENT
All relevant data are included in the paper or its Supplementary Information.
CONFLICT OF INTEREST
The authors declare there is no conflict.
