Glacier Retreat and Groundwater Recharge in Central Chile: Analysis to Inform Decision-Making for Sustainable Water Resources Management

Abstract

Glaciers worldwide are in retreat, and their meltwater can modulate mountain aquifers. We examined whether mass loss of the Juncal Norte Glacier (central Chile) has affected groundwater storage in the Juncal River basin between 1990 and 2022. Recession-curve modeling of daily streamflow shows no statistically significant trend in basin-scale groundwater reserves (τ = 0.06, p > 0.05). In contrast, glacier volume declined significantly (−3.8 hm³/yr, p < 0.05), and precipitation at the nearby Riecillos station fell sharply during the 2008–2017 megadrought (p < 0.05) but exhibited no significant change beforehand. Given the simultaneous decreases in meteoric inputs (rain + snow) and glacier mass, one would expect groundwater storage to decline; its observed stability therefore suggests that enhanced glacier-melt recharge may be temporarily offsetting drier conditions. Isotopic evidence from comparable Andean catchments supports such glacio-groundwater coupling, although time lags of months to years complicate detection with recession models alone. Hence, while our results do not yet demonstrate a direct glacier–groundwater link, they are consistent with the hypothesis that ongoing ice loss is buffering aquifer storage. Longer records and tracer studies are required to verify this mechanism and to inform sustainable water resources planning.

1. Introduction

Glaciers are one of the main reserves of fresh water on the planet, storing 75% of it [1]. They play a fundamental role for ecosystems and human communities, given the contribution of water they deliver to basins in the summer period and in times of drought [2]. Worldwide, glaciers reached their latest maximum extension during the Little Ice Age in the mid-19th century [3]. Since then, they have remained in a retreat phase, which increased in the second half of the 20th century [4], due to natural and climate warming.

One of the most important glacier reserves in South America is located in Chile, with 80% of the region’s glaciers [5]. However, the country has lost nearly 10% of its total glacial surface area in the last two decades [6]. This is a concern for authorities due to the impact it has on water resources, climate regulation, and sea level rise. In this country, water availability is not equitable throughout its distribution [7]. This is due to climatic factors such as the fact that annual rainfall geographically decreases south to north [8]. Greater pressure on water resources is also observed in the central-northern portion of the country, exceeding the available flow due to agriculture, mining, and the increased demand for drinking water [9].

In the central Chilean mountain range, a reduction in peak flows has been observed due to melting [10]. An example of this is the Maipo River basin with its notable glacial retreat in recent decades, where no positive trends in glacial flows are observed, and it is believed that the peak water level was reached before the studied period [4]. Moreover, the effects of glacier retreat on groundwater recharge at a watershed level have been investigated around the world. He et al. [11], for example, evaluated the influence of glacier meltwater on aquifer evolution in a Chinese region, concluding that glacier retreat will dramatically affect groundwater resources, resulting in less underground volumes once the glacier disappears. Similar results were found in glaciated watersheds in China [12], Iceland [13], Tibet [14], and India [15], among many other countries. It has been estimated that up to 50% of glacier meltwater can be destined to recharge local aquifers [16], while groundwater in general is showing significant alterations from climate change, one way or the other [17]. However, other studies show little effect of glacier meltwater on local aquifers, e.g., [18].

In this context, the question arises as to whether glaciers and their natural summer melting, possibly accelerated by climate change, have in any way influenced aquifer recharge, either by increasing their volume due to greater circulating processes, or by reducing their stored volume as a result of glacial loss, or whether this process has had generally no greater impact on such recharge. For this reason, the mass balances of a glacier located in the central region of Chile were evaluated, and the recession curves of the hydrographs generated downstream of the glacier were analyzed, thus allowing for the estimation of groundwater reserves and their temporal behavior. The results from this study will inform decision makers about future water resources sustainability in the region.

2. Materials and Methods

2.1. Study Site

The Juncal River sub-basin, formerly known as Estero Juncalillo (ID: 05400), is located in the upper part of the Aconcagua River basin, in the Valparaíso Region of central Chile (Figure 1). In the highlands, the Aconcagua River has a markedly snowy regime, so its flow increases in the spring and summer months due to the melting of snow from the mountain range. The Juncal River drains 359 km² upstream of the Río Juncal gauging station (05401003-6). Channel width ranges from 15 to 30 m in austral summer, with thalweg depths of 0.5–2 m. Mean annual discharge for the 1990–2022 period was 6.2 m³/s (Dirección General de Aguas [DGA] records). Summer flows (December–March) are dominated by snow/glacier melt [19,20]. Two types of climates are distinguished according to the Köppen–Geiger classification: the winter rain tundra climate (ET(s)) and the cold Mediterranean winter rain climate (Csc). Mean annual precipitation at the Riecillos (ID 300450) and Juncalito (ID 300470) stations is 650–720 mm/yr, 80% of which falls as winter snowfall (May–August). Summer precipitation seldom exceeds 40 mm/mo. Mean annual air temperature at 2900 masl is 4 °C, with a July mean of −2 °C and a January mean of 9 °C (DMC, 1990–2022) [21]. Juncal Norte is a temperate valley glacier that covered 25.3 km² in 1987 and 17.7 km² in 2022. It extends from ≈5300 m a.s.l. at the accumulation zone to ≈2900 masl at the terminus, with a present-day median elevation of ≈4040 m [22]. Average ice thickness in 2022 was estimated at 90 ± 15 m from ground-penetrating-radar transects.

2.2. Data Analysis

The total amounts of data used for flow, precipitation, and glacier area were 43,643, 348, and 7, respectively. To carry out the study, R 4.4.1 [23] and Qgis 3.34.13 [24] were used, in order to establish the temporary value of the groundwater reserves in the sub-sub-basin from 1990 to 2022 and also to determine if there have been significant variations in the mass of the Juncal Norte Glacier from 1990 to 2022. Likewise, the data obtained from the DGA were not complete for all years, so it was not possible to calculate the volume of groundwater for the year 2008. However, this was not a problem in carrying out the study.

2.3. Procedure for the Estimation of Groundwater Volumes

For the analysis, measured data corresponding to the last flood recorded before March 31 of the year under study were used. These data were adjusted using the Remenieras model (see further down for details), which allowed for an estimate of the base volume. The recession flow calculation was performed in the sub-basin using daily average flow data provided by the DGA [25], which can be downloaded from the DGA website by searching for station 05401003-6 (Río Juncal in Juncal).

For the 1990–2022 analysis period, using the aforementioned information, flood hydrographs were reconstructed for the interval that defines the beginning of the hydrological year and the end of the previous one, that is, between March 30 and April 2 for each year.

2.3.1. Hydrograph Separation

Flood hydrographs were used to estimate the recessionary flows. In the descent curve of this, the beginning of the recessionary curve was studied, which theoretically, and according to Linsley et al. [26], begins at the second break point of the descent curve of the hydrograph, as observed in Balocchi et al. [27]. From that point on, the circulating flow represents the flow that comes exclusively from groundwater. To model the recessionary flow, the mathematical models of Remenieras [27] (Equation (1)), Potencial [28] (Equation (2)), and Balocchi et al. [27] (Equation (3)) were used, where Q(t) is the recessive flow at time t, in m³/s, Q₀ is the initial recessive flow at time t₀, in m³/s, e is the Neper constant, and α is the recession constant. To estimate the volume stored in the basin, each model had to be adjusted with the daily flow data available for the basin during the defined period.

$$Q\left(t\right)=Q_{0 }\times e^{-\alpha t}$$

(1)

$$Q \left(t\right)=Q_0{\left(1+\alpha \times t\right)}^{-2}$$

(2)

$$Q\left(t\right)=Q_0\times e^{(-2\alpha \sqrt{t} )}$$

(3)

From the annual hydrographs, the last major flood and the peak flow that defined said flood were selected, and this was done for dates prior to 31 March and for each year. It should be noted that the peak flow was not always recorded in the months close to the study date, since it could be found at the end of the previous year, that is, at the end of December. From the selected peak flow, the logarithms of the flows were graphed as a function of the time of the hydrograph’s decline curve, leaving the semilogarithmic relationship lnQ v/s t, as indicated by the hydrograph separation method [27].

The second breaking point was then graphically established, and the coordinates were obtained, which corresponded to the time at which exclusive groundwater contributions began, t₀, and the flow rate recorded at time t₀, i.e., Q₀. Then, a flow rate value was chosen that corresponded to days after t₀, implying a time t₀ + Δt, where a Q_t value is found. From this, the recession constant α was calculated according to Equations (1)–(3). It should be noted that the relationship from time t₀ onwards is defined by a negative slope, i.e., dQ/dt ≤ 0, which is explained because it is expected that the basin will begin to empty when there are no contributions and therefore the recessive flow will always decrease over time.

2.3.2. Goodness of Fit

The model that presented the best fit was evaluated with the Kling–Gupta Efficiency (KGE) test and the coefficient of determination (R²). To carry out this calculation, the R 4.4.1 program was used, using the “Metric” package [29], which contains the KGE equation defined by Gupta et al. [30] and expressed in Equation (4), where r is the linear correlation coefficient between the observed and modeled series, α is the measure that compares the standard deviation of the simulated data with the observed data, and β is the ratio between the mean of the simulated and observed data.

$$KGE=1-\sqrt{{\left(r-1\right)}^2+{(\alpha -1)}^2+{(\beta -1)}^2}$$

(4)

The coefficient of determination (R²), expressed in Equation (5), was also calculated, where RSS is the sum of squares of the residuals and TSS is the total sum of squares.

$$R^2=1-{\displaystyle \frac{RSS}{TSS}}$$

(5)

2.3.3. Estimating Stored Volume

Once the recession flow model was adjusted, the annual volumes of groundwater stored (m³) were estimated. Since the stored volume was to be obtained by 31 March of each year (Equation (6)), it was assumed that there was no further water input to the channel, either from precipitation or snow or glacier melt.

$$Volume\left({Hm}^3\right)={\displaystyle \frac{24\left(h\right)\times 3600\left(s\right)}{\mathrm{1,000,000}}}{\int }_{31march}^{\mathrm{\infty }}Q\left(t\right)dt$$

(6)

2.4. Glacier Mass Variation

Using the vector layers of the glacier provided by the DGA, where the polygon of the ice body is delimited, the area of the ice body was calculated for multiple years in the period 1987–2023.

Glaciers have different shapes, adapted to the geography and environmental conditions of their location. Multiple authors have found relationships between the area and volume of glaciers in different areas of the planet [31]. Some of these empirical equations were used to estimate the volumetric variation in the Juncal Norte Glacier (

Table 1. Equations for estimating glacier volume.

Autor	Equation
Chen & Ohmura [32]	$V=28.5\times S^{1.357}$
Bahr et al. [33]	$V=27.6\times S^{1.36}$
Radic & Hock [34]	$V=36.5\times S^{1.375}$
Huss and Farinotti [35]	$V=28.4028\times S^{1.327}$
Grinsted [36]	$V=28.4028\times S^{1.327}$
Huenante [31]	$V=2.45\times S^{1.27}$

), and are based on Equation (8), where V is the volume of the glacier based on its surface area (Hm³), S is the glacier surface (km²), and C and Y are empirical constants.

$$V=C\times S^Y$$

(7)

2.5. Precipitation

The calculation of precipitation was carried out with the daily accumulated precipitation data recorded by the DGA and stored in a CR2 database [37] from 1990 to 2018, belonging to a meteorological station close to the fluviometric station. The selected station was Riecillos (ID: 5403006), since there were no meteorological stations within the sub-sub-basin for more than 10 years. The analysis was carried out annually in order to determine the behavior of precipitation and its possible impact on the water reserves of the basin.

2.6. Trend Analysis

The nonparametric Mann–Kendall (MK) test (Equation (8)) was applied to verify the existence of trends and whether they were significant according to the confidence level provided (p < 0.05). The formulas for calculating the test statistics are shown in Equations (8)–(11), where n is the sample size, t_p is the frequency of ties in a group, and q is the number of groups with ties (p).

$$S=\sum _{k=1}^{n-1}\sum _{j=k+1}^{n}sgn(X_j-X_k)$$

(8)

$$Sgn\left(X_j-X_k\right)=\left\{\begin{array}{c}1 if \left(X_j-X_k\right)>0\\ 0 if \left(X_j-X_k\right)=0\\ -1 if \left(X_j-X_k\right)<0\end{array}\right.$$

(9)

$$VAR\left(S\right)={\displaystyle \frac{1}{18}}\left[n\left(n-1\right)\left(2n+5\right)-\sum _{p=1}^q{t}_p(t_p-1)({2t}_p+5)\right]$$

(10)

$$Z=\left\{\begin{array}{c}{\displaystyle \frac{S-1}{\sqrt{VAR\left(S\right)}}};si S>0\\ 0;si S=0\\ {\displaystyle \frac{S+1}{\sqrt{VAR\left(S\right)}}};si S<0\end{array}\right.$$

(11)

The Z statistic allows for inferring whether the trend is significant for the confidence value used. For a confidence level of 95%, the trend is significant if −1.96 > Z > 1.96. The Mann–Kendall test was performed using the R 4.4.1 program, using the “rkt” package [38].

3. Results

3.1. Estimation of Groundwater Volumes

The fit of the Remenieras [27], Potencial [28], and Balocchi et al. [27] models is shown in

Table 2. Values of the adjustment of the mathematical models.

	R2			KGE
Year	Remenieras [27]	Potencial [28]	Balocchi et al. [27]	Remenieras [27]	Potencial [28]	Balocchi et al. [27]
1990	0.837	0.805	0.704	0.730	0.686	0.641
1991	0.145	0.154	0.201	0.287	0.289	0.405
1992	0.862	0.840	0.708	0.803	0.798	0.655
1993	0.229	0.234	0.343	0.420	0.424	0.576
1994	0.375	0.361	0.258	0.427	0.419	0.290
1995	0.808	0.807	0.757	0.874	0.860	0.740
1996	0.254	0.242	0.196	0.499	0.486	0.419
1997	0.546	0.542	0.441	0.303	0.303	0.224
1998	0.928	0.924	0.889	0.792	0.817	0.745
1999	0.707	0.693	0.590	0.606	0.600	0.472
2000	0.704	0.708	0.717	0.830	0.840	0.788
2001	0.902	0.885	0.785	0.867	0.862	0.713
2002	0.704	0.718	0.715	0.808	0.819	0.742
2003	0.825	0.820	0.792	0.734	0.722	0.779
2004	0.339	0.326	0.223	0.558	0.547	0.425
2005	0.416	0.363	0.285	0.594	0.535	0.428
2006	0.923	0.929	0.872	0.823	0.829	0.703
2007	0.479	0.488	0.562	0.557	0.548	0.666
2009	0.664	0.670	0.668	0.742	0.748	0.660
2010	0.336	0.334	0.331	0.515	0.514	0.447
2011	0.466	0.467	0.464	0.556	0.560	0.483
2012	0.462	0.465	0.456	0.333	0.336	0.280
2013	0.670	0.668	0.574	0.568	0.567	0.446
2014	0.823	0.832	0.779	0.862	0.847	0.708
2015	0.546	0.526	0.434	0.664	0.642	0.522
2016	0.660	0.650	0.680	0.567	0.549	0.691
2017	0.689	0.688	0.619	0.609	0.611	0.506
2018	0.743	0.764	0.811	0.681	0.709	0.706
2019	0.687	0.683	0.644	0.787	0.795	0.700
2020	0.638	0.626	0.446	0.327	0.325	0.227
2021	0.363	0.358	0.346	0.485	0.483	0.410
2022	0.495	0.489	0.383	0.232	0.230	0.136
Mean *	0.601	0.596	0.552	0.608	0.603	0.542

(*) A value closer to one indicates a more accurate fit of the model to the observed data.

. The averages of R² and KGE indicate that the model that explains the largest proportion of the observed data and is therefore considered to have a good fit is the Remenieras model, which is why it was chosen over the other two. The volumes stored as of March 31 for each year were estimated based on the Remenieras model and its integral, as shown in Figure 2.

First, the period 1990–2009 was considered to examine the trend in groundwater volumes without the influence of the megadrought; this is in contrast to the period 1990–2022, where the total period is analyzed. Then, the periods 1990–2000, 2000–2010, and 2010–2022 were analyzed to assess trends over shorter periods. As seen in

Table 3. Mann–Kendall trend for groundwater volumes.

Period	Theil–Sen Slope (Hm3)	Tau	Score	Var	p-Value	Zmk
1990–2009	0.433	0.135	23	817	0.441	0.77
1990–2022	0.012	0.060	30	3803	0.638	0.47
1990–2000	0.328	0.091	5	165	0.755	0.31
1990–2005	−1.196	−0.117	−14	493	0.558	−0.59
2000–2010	9.638	0.556	25	125	0.032 *	2.15
2010–2022	−1.211	−0.308	−24	269	0.161	−1.40

(*) Statistically significant value.

, the only significant trend is in the period 2000–2010, which indicates that groundwater storage volumes increased during this period. In contrast, a non-significant negative trend is observed for the years 2010–2022. While this indicates insufficient evidence to confirm a clear trend in groundwater volumes, the trend reverses for years with megadroughts. Thus, the 1990–2000 period shows a non-significant positive trend. When compared to the 1990–2005 period, it can be seen that adding just five years changes the trend; it is now negative but remains non-significant.

The Mann–Kendall test was applied to the precipitation data to assess their behavior during the study period.

Table 4. Mann–Kendall trend for precipitation.

Period	Theil–Sen Slope (Hm3)	Tau	Score	Var	p-Value	Zmk
1990–2009	1.080	0.042	8	950	0.820	0.23
1990–2018	−2.696	−0.281	−114	2842	0.034 *	−2.16

(*) Statistically significant value.

shows a positive trend for the period 1990–2009, but this value is not significant, so it cannot be stated with complete certainty that precipitation increased during this period. However, when considering the total data (1990–2018), the trend reverses and is significant, indicating a decrease in precipitation during the megadrought period.

3.2. Variation in Glacial Mass

As can be seen in Figure 3, there is a decrease in the area of the glacier between the years 1987 and 2023. It can also be seen that the right side of the glacier fragments into three different parts by the year 2023.

Glacier volume was calculated from its area using formulas developed by six different authors, as shown in

Table 5. Volumes (Hm³) of the Juncal Norte Glacier according to various formulas.

Year	Area (km2)	Chen & Ohmura [32]	Bahr et al. [33]	Radic & Hock [34]	Huss and Farinotti [35]	Grinsted [36]	Huenante [31]
1987	9.6	614.39	599.04	819.56	572.11	703.52	397.48
1997	8.3	501.58	488.83	667.28	469.16	585.35	328.74
2006	9.0	561.09	546.96	747.56	523.52	647.97	365.12
2008	8.7	532.82	519.34	709.41	497.72	618.31	347.87
2010	9.2	574.79	560.35	766.06	536.02	662.30	373.45
2021	7.4	434.12	422.95	576.44	407.36	513.53	287.18
2022	7.4	429.69	418.63	570.47	403.30	508.78	284.44
2023	6.5	363.57	354.08	481.62	342.50	437.27	243.26

. Glacier volumes generally showed a downward trend from 1987 to 2023, regardless of the formula used. While both area and volume increased from 2006 to 2010, a decrease in volume was observed again starting in 2010, which continued until 2023.

Table 6. Application of Mann–Kendall for glacier volumes calculated using Huenante’s formula.

Period	Theil–Sen Slope (Hm3)	Tau	Score	Var (S)	p-Value	Zmk
1987–2023	−3.786	−0.643	−18	65.3	0.035	−2.35

shows a negative and significant volume trend for the Juncal Norte Glacier, suggesting a decrease in the volume of this ice body for the 1987–2023 period. This could imply an increase in the meltwater contribution to the study basin.

3.3. Analysis of the Relationship Between Glacier Mass Variation, Groundwater Volumes, and Precipitation

Figure 4 shows a comparison of the three variables analyzed in this study. Basically, no direct relationship is observed between the variables analyzed; however, an inverse relationship is observed between glacier volume and the modeled recession volume.

4. Discussion

The results indicate that the estimated volumes of water stored in the sub-basin as of March 31 do not show a clear trend for the period 1990–2022, as the Mann–Kendall test shows a non-significant positive trend for the period. This is the exception for the period 2000–2010, where the trend is significant and indicates an increase in the stored underground volume.

Regarding precipitation, the Mann–Kendall test was applied to see its behavior during the study period. Table 4 shows that for the period 1990–2009, it showed a positive trend, but this value is not significant, so it cannot be stated with total certainty that rainfall increased for this period. However, when considering all the data (1990–2018), the trend is reversed and is significant, which would indicate a decrease in rainfall if the megadrought period is considered. This situation is consistent with what is reported in the literature, with a decrease of up to 30% in rainfall between the Coquimbo and Araucanía regions [10]. The analysis is carried out at the beginning of the hydrological year, when the channel, in theory, does not receive new water inputs; that is, precipitation should not be contributing water to the basin at that time [39,40]. This is added to the fact that the basin is located in the upper part of the Aconcagua River, so it presents a snow regime [19]. Given the above, the volumes of stored water should not be affected by precipitation. Figure 4 shows the decrease in precipitation throughout the period studied, a situation that is confirmed by the Mann–Kendall test in Table 4. This behavior is contrary to the non-significant positive trend of stored water reserves for the period 1990–2022, which would indicate an increase in showing a decrease in volume, which would be greater from 2010 onwards.

In Table 3, the Mann–Kendall test for the period 1990–2022 indicated a positive trend, but it is not significant for the stored volume of groundwater in the Juncal River basin. Figure 4 shows the behavior of these volumes compared to the volume of the Juncal Norte Glacier for a similar period, and as previously confirmed by the Mann–Kendall test, the volume of the glacier decreases in the same period. The above could indicate that the decrease in the volume of the glacier given the summer melting positively affected the recharge of the basin’s aquifer, considering that the study by Rodríguez [41] indicates that the greatest influence on the base flow of the basin would be glaciers or snow, but this relationship cannot be assured since there is no significant trend for water reserves.

Regarding the significant positive trend in underground reserves in March for the period 2000–2010 in Table 3, this suggests an increase in underground volumes before the onset of the megadrought. This phenomenon cannot be fully explained by the loss of glacial mass, as there are insufficient data to analyze the glacier trend for the same period. The influence of precipitation on this fact cannot be confirmed or ruled out, given that the positive trend is not significant.

Based on all of the above, it is recommended to apply other models or software to estimate recession flows, allowing comparison with the obtained values. In addition, other variables (ice, snow, temperature), zones, and timings should be analyzed to identify patterns of behavior or relationships with underground volumes.

Global Perspective

The results from this investigation contrast with findings in other glacierized regions where meltwater has been shown to play a major role in sustaining groundwater systems. For instance, in Iceland, a study [16] found that glacier meltwater contributed significantly to proglacial aquifer recharge, comprising up to 50% of groundwater during the melt season and up to 22% of annual river flow. In subarctic Alaska, it was observed that glacier-fed headwaters were significant recharge corridors, with glacier melt contributing between 15 and 28% of annual streamflow and feeding into aquifers that sustained baseflows year round [42]. Similarly, on the Tibetan Plateau, groundwater recharge in some basins was found to be overwhelmingly sourced from glacial melt in the upper Indus Basin [43].

The findings from this investigation underscore that while glacier melt may contribute to aquifer recharge, especially in dry periods, its impact is not always dominant or linear, particularly in regions where other variables like snowpack, evapotranspiration, and geological heterogeneity modulate groundwater dynamics. The contrast with studies in Iceland, Alaska, and Asia highlights the relevance of localized hydrogeological conditions and long-term monitoring in understanding glacier–aquifer interactions. This study concludes that more years of data, incorporation of additional variables such as snow cover and temperature, and refined modeling approaches are necessary to clarify the long-term effects of glacier retreat on groundwater in Chile, a conclusion aligned with the global call for better integrated cryosphere–hydrology research under climate change scenarios.

5. Conclusions and Recommendations

This study offers a novel, long-term analysis of how glacier retreat may influence groundwater recharge in central Chile, an under-researched yet water-stressed region. By combining glacier mass balance, groundwater modeling using recession curves, and precipitation trends from 1990 to 2022, it provides a holistic view of basin hydrology. The use of multiple recession models to estimate groundwater storage is methodologically innovative for this context.

The analysis showed no clear or statistically significant trend in groundwater storage volumes in the Juncal River sub-basin over the 1990–2022 period, although a significant increase in groundwater reserves was observed specifically during the 2000–2010 window. This increase occurred prior to the onset of Chile’s ongoing megadrought, suggesting a possible influence of earlier climatic or glacial conditions.

Moreover, a significant decline in precipitation was observed over the full study period, especially during the megadrought years, which aligns with broader regional trends reported in the literature. Despite this decline, modeled groundwater reserves did not show a corresponding decrease, indicating that precipitation alone does not fully explain groundwater behavior in the basin.

The volume of the Juncal Norte Glacier showed a significant and steady decline from 1987 to 2023. This suggests increased glacial meltwater input over time, although the study could not confirm a statistically significant relationship between glacier loss and groundwater recharge. However, the observed inverse relationship between glacier volume and groundwater storage hints that glacial meltwater may contribute to aquifer recharge, at least temporarily, in this high-altitude Andean basin.

Given these mixed signals and the complexity of hydrological interactions, this study concludes that more years of data, additional variables (such as snowpack and temperature), and alternative modeling approaches are needed to confirm whether glacier retreat is significantly influencing groundwater resources. Nonetheless, the findings raise important considerations for water resources management under climate change, especially in regions where glacier-fed systems are vital for sustaining flows during dry seasons. Similarly, it is important to develop hydrological models able to predict what will happen when glaciers are gone, considering also climate change in general and future water consumption, as done elsewhere, e.g., [44].