Introduction

Rivers are the primary source of continental freshwater input to the oceans1. In semi-enclosed coastal systems, rivers are an essential driver of circulation processes2 and the bridge between terrestrial and marine ecosystems, transporting nutrients, organic matter, and sediments that supplement those from the coastal and open ocean3. Recent studies have reported alterations in the volume and timing of freshwater inputs to the coastal ocean4. Analyses of time series and hydrological models show different spatiotemporal patterns5, ranging, for example, from strong increases in freshwater discharges at high latitudes6—due to increased ice melt—to declining trends at mid-latitudes due to increased hydrological drought7. Regardless of the direction and magnitude of recently reported trends, climate change has been identified as a major driver of altered hydrological regimes8, while impacts on coastal waters and the ecosystems they support have yet to be fully assessed and understood9.

Temporal changes in streamflow, influenced by climate change (changing precipitation patterns and rising temperatures) and anthropogenic pressures such as hydroelectric dams or reservoirs10,11, irrigation diversions, and land use changes12, pose a challenge to analysing trends in freshwater inputs13,14. For example, several studies have shown that areas with a high diversity of forests display greater ecosystem resilience during droughts than those dominated by forest plantations or agriculture15,16. At mid-latitudes, understanding future climate change impacts on the coastal ocean via disentangling the effects of climate change and local anthropogenic disturbances requires observations from reference systems with limited alteration so that changes in the hydrological regime can be separated from local human disturbances.

Western Patagonia (> 41°S on the Pacific coast of South America) emerges as one of these well-preserved regions, where ongoing research is essential to monitor and safeguard these ecosystems. In this region, free-flowing rivers interact with complex coastal geomorphology to shape one of the most extensive macro-estuaries in the world17,18. Due to their remoteness, Patagonia’s river basins remain remarkably undisturbed, with a high native forest cover19, low human population density, and in general free of dams, reservoirs, and irrigation infrastructure. The vulnerability of these coastal ecosystems to environmental changes depends mainly, by extension, on their interaction with freshwater sources. Freshwater inputs into the coastal waters of Northwestern Patagonia (NWP; 41–46 ºS), including direct precipitation and river runoff, interact with oceanic water masses, especially the high-salinity Subantarctic Water (SAAW). This, in turn, influences patterns of circulation, water renewal, and ventilation20 through the generation of strong salinity gradients, with a high degree of seasonal variation and spatial heterogeneity in the halocline21. The interaction between these two water sources also influences the composition of planktonic communities22, affects light attenuation23, and leads to a dilution of Nitrogen and Phosphorus concentrations in the SAAW24 and enhanced orthosilicate supply in the surface layer25,26. Moreover, the freshwater input results in a strong vertical gradient of dissolved oxygen (DO) in these estuarine waters, with the upper layer generally showing high DO concentration (6–8 mL L−1)27.

The relatively limited intervention characterising Patagonia’s continental basins contrasts with the intensive use of the coastline, especially by the aquaculture industry, a key economic sector in Chile. Currently ranked among the world’s top aquaculture producers, Chile is the second-largest exporter of salmon and trout, and the largest exporter of mussels28. Salmon and mussel farming, which currently occupies ca. 2300 concession sites with a total area of ca. 24,000 ha in the region, generates significant economic value in southern Chile29. This development is not without environmental impacts, especially in the case of salmon farming, an activity that contributes with exogenous nutrients to these ecosystems30. It has also been shown to be highly vulnerable to climate-related changes31,32.

In recent decades, studies carried out in NWP have reported a decline in freshwater inputs33 and projected a decrease in streamflow during summer and autumn34, as well as prolongation of hydrological droughts35. These changes in freshwater inputs influence immediate environmental parameters and have cascading effects on broader ecosystem dynamics. For example, extreme conditions in coastal surface water during summer-autumn droughts have become more frequent, and, in some cases, have impacted the aquaculture industry through hypoxia36, increased parasitism, and harmful algal blooms32,37,38 (HABs).

Current knowledge on the decline in freshwater inputs across NWP is based on the analysis of hydrological metrics39, which describe the statistical or dynamical properties of hydrologic data series derived from a limited number of long-term records and minimal spatial coverage. As for the corresponding hydrographic conditions in the estuarine waters of NWP or the adjacent coastal ocean, there are no available long-term in-situ records. Given the vulnerability of coastal socio-economic activities to climate change impacts, adopting a comprehensive land-sea interface approach to management and adaptation to climate change is imperative40. To address this challenge, we have integrated hydrological modelling (1990–2020) with satellite-derived and reanalysis data on sea surface conditions (1993–2020). This approach has two main goals: (1) to assess historical trends in hydrological metrics at a regional scale and (2) to examine the implications of these changes for the adjacent coastal system, particularly regarding sea surface salinity, temperature, and turbidity.

Methods

Study area

The marine coastal system of the NWP (Fig. 1), defined here as the region between 41–46°S and 72–76°W, is strongly modulated by a seasonally shifting balance between oceanic and terrestrial influence21. On the oceanic side, changes in prevailing winds along the outer NWP shores determine seasonal changes, from downwelling conditions in autumn–winter to upwelling conditions in spring-summer41. This change is thought to modify the origin and, therefore, the nutrient composition of waters entering through Boca del Guafo (Fig. 1). Hydrographic conditions in the region are determined by the interaction of Subantarctic Water (SAAW), normally found from the surface down to 150 m depth, and occasional intrusions of Equatorial Subsurface Water (ESSW) at depths of 150–300 m, with the low-salinity surface layer generated by freshwater discharges from rivers and runoff on the eastern side of the region42. The inner sea of NWP is a large macro-estuary, that concentrates most of the aquaculture activity in southern Chile31, and that is strongly influenced by tides with amplitudes of 1–2 m43 as well as turbid freshwater plumes throughout the year44. On the terrestrial side, the seasonal change and spatial heterogeneity of freshwater discharges across the study area determine marked spatial and temporal changes in the salinity-driven stratification of the water column23,45 and, consequently, in the occurrence and depth of the halocline21 (i.e., the depth at which salinity changes rapidly).

Fig. 1
figure 1

Study region. The size of the dark gray circles (n = 896) is proportional to the area of each basin that flows into the coastal zone of Northwestern Patagonia. The coloured polygons correspond to the main basins. The crosses and grey polygons indicate the stream gauges and their delimited basins, respectively. Blue shading indicates bathymetry.

Full size image

The inner sea of NWP can be divided into two zones: north and south of Boca del Guafo (Fig. 1). The northern zone spans the Inner Sea of Chiloé and Reloncaví Sound, while the southern zone includes the Guaitecas Archipelago and the Moraleda Channel as its main features (Fig. 1). This division is based on current knowledge of coastal currents in the region, and of the water exchange across Boca del Guafo27,46,47. Along the complex coastline of this region, 896 basins with area > 5 km2 discharge freshwater to the NWP coastal zone, with varying degrees of rain and snow influence (Fig. 1). In terms of size, the six more important basins (> 5000 km2) correspond to Petrohué (~ 200 m3 s−1), Puelo (~ 600 m3 s−1), Yelcho (~ 800 m3 s−1), Palena (~ 900 m3 s−1), Cisnes (~ 300 m3 s−1), and Aysén Rivers (~ 600 m3 s−1) (Fig. 1). These rivers represent a north–south transition in hydrological regime, from pluvial-nival to mostly nival, with the winter peak due to precipitation becoming less important towards the south, as the snowmelt contributions of spring and summer increase33,35.

Data sources and validation

Hydrology of freshwater inputs

As in many remote areas, only a few stream gauges in the NWP have produced long records. Given this limitation, runoff data for each of the 896 basins with surface area > 5 km2 that drain into the coastal zone were obtained from the streamflow series generated by the Variable Infiltration Capacity (VIC) model48 calibrated for the study region49. VIC is a semi-distributed, physically based hydrological model that simulates snow accumulation and melting, evapotranspiration, canopy interception, surface runoff, baseflow, and other hydrological processes at sub-daily time steps. The model was forced with gridded meteorological data from PMET-sim50 and ERA5-Land51 for the period 1985–2020. The VIC hydrological model was specifically calibrated (1985–2004) and validated (2005–2020) for 43 basins at a daily time step, with streamflow as the objective function and using a split-sample approach. The calibration was performed using the Shuffled Complex Evolution algorithm included in the SPOTPY framework52. The modelling approach achieved an adequate performance of hydrological fluxes with Kling-Gupta53 Efficiencies (KGE)53. This goodness-of-fit indicator combines the three components of model errors (i.e., correlation, bias and coefficients of variation), of 0.75 ± 0.12 and 0.64 ± 0.23 in the calibration and validation phases, respectively (with 1 being the optimal value).

In addition to the validation described above, we assessed the ability of the model to reproduce the historical hydrological metrics (i.e., signatures). To this end, we used streamflow time series from the PMET-obs50 dataset, which compiles and quality-controls daily hydrometeorological records from eight institutions in Chile and Argentina. From the 43 stream gauges used to calibrate the VIC model, we selected 19 with more than 75% of the records for the study period (1985–2020). To isolate the effect of record gaps on the hydrological metrics, we used the MissForest algorithm54 to fill in the missing records using nearby stream gauges. This algorithm has performed well in data-poor regions such as south-central Chile. Once the filling process was complete, the validation consisted of 16 hydrological metrics that characterise the intra- and inter-annual variability of streamflow conditions (Table 1).

Table 1 Selected hydrological metrics to characterise the natural streamflow regime of each basin in Northwestern Patagonia (based on Richter et al.55). The metrics related to Magnitude of seasonal/annual water conditions are calculated by season (n = 4) and year (i.e., n = 5 in the details box).
Full size table

Surface layer of NWP coastal system

The spatiotemporal patterns of the surface conditions across the NWP coastal system were characterised using data on sea surface temperature, turbidity, and salinity obtained from two sources. Monthly time series of sea surface temperature (SST) and turbidity were derived from Level-1 daily images obtained by the Moderate Resolution Imaging Spectroradiometer (MODIS) onboard the satellite Aqua. Monthly composites with 1 km spatial resolution were produced from quality-controlled daily swaths distributed by NASA (http://oceancolor.gsfc.nasa.gov). As a proxy for surface turbidity, we used the surface reflectance at 645 nm (R645 product), previously used to track river plumes and low-salinity surface waters typical of river-influenced coastal areas44,56. The time series were obtained for each pixel in the NWP coastal region with > 30% of data in the 20 years from January 2003 to December 2022.

Monthly salinity data for 1993–2021 were obtained from the Global Ocean Physics Reanalysis GLORYS12V157,58. This product contains global ocean data since January 1993 on a regular 1/12° grid (~ 8 km) and over 50 vertical levels from the surface to bottom (> 5000 m). GLORYS12V1 has been proven useful for coastal zone characterization59,60. In this study, we processed information for each month and node in the grid at 19 depth levels, from the surface (~ 0.5 m) to ~ 55 m depth. Salinity data were validated using a procedure to that used to calculate the uncertainty in satellite products61,62, considering 1833 depth-specific measurements from CTD profiles collected between 2013 and 2019 during oceanographic surveys conducted by the Chilean Fisheries Development Institute (IFOP). Each in-situ CTD-salinity profile was matched with the same date (month and year) salinity profile from the reanalysis product by using the nine pixels closest to the location of the in-situ sample. The mean salinity profile of these pixels was taken as the reanalysis estimate and compared with in-situ CTD-salinity data for the upper ~ 20 m depth. The Pearson linear correlation coefficient (R-value), statistical significance (p value) and root mean square error (RMSE) were calculated as statistical metrics for salinity reanalysis product validation. The validated salinity profiles were used to compute three indices of freshwater influence on the hydrographic conditions across the estuarine region spanned by this study (Fig. 1). First, we tracked the depth of the layer with salinity values in the 32.9–33.2 interval, considering that any salinity value below 32.9 corresponds to the upper layer of the water column that is influenced by freshwater. The salinity threshold of 32.9 was selected considering the previously described relationship between the influence of continental freshwater and oceanic water masses present in the study area. In NWP, the oceanic Subantarctic Water mass (SAAW) is the dominant surface water with a characteristic salinity value of 33.863, seasonally modified by the freshwater coming from intensified spring runoff and summer melting from the fjord region between 40° and 45°S giving rise to Summer Subantarctic Water mass (SSAW) with characteristic salinity value of 33.064. Therefore, values equal or below 32.9 could be associated with a predominant influence of freshwater discharge. Finally, as a proxy for the intensity and location of the halocline, for each monthly profile we computed the vertical salinity gradient (ds/dz) and recorded the maximum value and depth of this maximum (halocline-depth proxy) for each month and node in the GLORYS12V1 grid.

Trend and association analysis

The trend analysis varied according to the variable being analysed, as the hydrological metrics have an annual time step, which masks the seasonal variability shown by the oceanographic monthly time series. For the hydrological metrics, the magnitudes of temporal trends (1990–2020) were assessed using Sen’s slope65. The significance of the potential trends was estimated using the modified Mann–Kendall (MK) test of Yue and Wang66, which is an adaptation of the traditional MK test designed to handle the presence of serial autocorrelation in the time series data. For the monthly composites of satellite data on sea surface conditions (SST and R645, 2003–2022) and the monthly grids of salinity indices derived from GLORYS12V1 (1993–2001), a harmonic analysis based on a least squares procedure67 was used to assess the intensity of seasonal variations as the amplitude of the annual harmonic at each node of the respective grid. This approach is commonly used in oceanography to fit periodic signals in environmental variables. Although it is most often applied to sea-level records to assess the amplitude and phase of tidal constituents68, it has been extended to assessing and mapping seasonal variability in sea surface temperature69. Normality and heteroscedasticity in the different types of data/time series were examined prior to applying this procedure. Based on this evaluation, a log10 transformation was applied to R645 data, whereas no transformation was used for temperature and halocline-depth data. Residuals were calculated for each pixel by subtracting the fitted annual harmonic from the monthly time series, and the fraction of non-seasonal variability in each time series (residual variance) was computed as the ratio between the variance of residuals and the variance of the original time series. Finally, a trend analysis based on the above-mentioned non-parametric tests (Mann–Kendall and Sen’s slope) was performed on each series of residuals to test whether SST, R645, and the salinity indices have significantly changed over time and to determine the sign and magnitude of that change.

To identify temporal changes in the dominant frequencies of variability and to assess the association between the hydrological (1990–2020) and sea surface conditions data, Wavelet and Cross-Wavelet Transforms70,71 were applied to the time series (monthly data between 1993 and 2022) of streamflow and freshwater influence indices calculated from salinity validated data (GLORYS12V1). The Cross-Wavelet Transform (XWT) allows the assessment of temporal changes in the coherence (i.e., correlation) and phase (i.e., sign and time lag of the correlation) of fluctuations in two-time series. The statistical significance of coherence values was determined with a Monte Carlo randomisation technique as in Grinsted et al.70.

Results

Freshwater inputs

Validation of hydrological metrics

The assessment of simulated discharge using hydrological metrics (Table 1) provided insights into the effectiveness of the previously calibrated VIC model in reproducing both intra- and inter-annual streamflow dynamics (Fig. 2). Specifically, the analysis revealed that the VIC model performed proficiently in capturing the magnitude of seasonal and annual water conditions in terms of KGE (Fig. 2a), with correlation values > 0.6–0.7 and a bias close to the optimal value of 1. In this group, summer mean values (DJF) were characterised by a slight overestimation (Fig. 2c). The magnitude and duration of the annual extreme low streamflows also showed good performance for durations greater than 30 days (Fig. 2a). These metrics showed similar biases with different levels of correlation depending on the duration (Fig. 2b). The model performance was more nuanced when assessing the timing of annual extreme water conditions. While the date of maximum discharge showed good correlations (Fig. 2b), other temporal aspects showed varying degrees of accuracy, indicating room for improvement in capturing the timing of such events. The validation also highlighted areas where the model had shortcomings, particularly in relation to the frequency and duration of pulses and the rate of change of water conditions, metrics that showed poor performance, characterised by lower correlations (Fig. 2b), increased biases (Fig. 2c) and a limited ability to capture the full range of hydrological variability (Fig. 2d).

Fig. 2
figure 2

Validation of model-based runoff using the selected hydrological metrics (Table 1). The streamflow records were obtained from 19 stream gauges in the PMET-obs50 dataset. The Kling–Gupta Efficiency (KGE; a) is a metric that evaluates model performance by combining three error components: correlation (Pearson correlation; b) for capturing the linear relationship between observed and simulated data, bias (c) for assessing the systematic deviation between their means, and the coefficient of variation (d) for comparing the relative variability between observed and simulated data. For all metrics, the optimal value is one.

Full size image

Spatio-temporal patterns in freshwater inputs

In the NWP region, model-based runoff estimates show a significant prevalence of high specific runoff along the coast, with 18% of the continental study area having annual specific runoff greater than 4,000 mm yr−1 (Fig. 3a), with no clear latitudinal differences. This corresponds to an area-weighted mean runoff of 2650 mm yr−1 (Fig. 3a). Over the historical period (1990–2020), the coastal area received an average freshwater flux of 7346 m3 s−1 (Fig. 3b), with the six macro-basins spanning 61% of the total area and contributing 48% of this input (Fig. 3b).

Fig. 3
figure 3

Long-term average freshwater input (1990–2020) in terms of (a) specific runoff (mm yr−1) and (b) streamflow (m3 s−1). The values in parentheses represent the long-term average freshwater input for the main basins.

Full size image

The analysis of temporal trends revealed significant changes in the hydrological regime or streamflow patterns of the NWP rivers (Fig. 4 and S1). Within the macro-basins (Table S1), there was an increasing trend in prolonged dry periods (90-day min). Four of the six macro-basins showed a significant trend (p < 0.05), indicating a significant increase in the duration and intensity of dry periods over time. Specifically, these trends showed a decrease in water availability ranging from − 2 to − 4% per decade (Fig. 4). These changes coincide with reduced seasonal freshwater inputs from the rivers stretching from Petrohué to Cisnes, especially during the dry season, which includes summer and, in recent years, autumn (Fig. 4; Table S1). In contrast to summer, which shows a generalised regional trend, autumn follows a latitudinal pattern, with larger decreases observed in the northern area (Fig. S1). Winter is the only season with significant increases due to an early melt season, suggesting a shift in hydrological dynamics. The timing of peak discharge events also differed between northern and southern rivers, and between rivers draining large and medium-sized basins versus those originating in small basins (Fig. 4 and S1). For example, the Petrohué River tends to produce peak streamflows during the spring period, while the Palena and Cisnes Rivers peak streamflows are during the winter months. Consequently, these changes have significant implications for coastal ecosystems such as Reloncaví Fjord and Reloncaví Sound, as well as the northernmost segment of the Inner Sea of Chiloé.

Fig. 4
figure 4

Trends in aggregated regional (n = 896 basins) hydrological metrics based on Sen’s slope65 (period 1990–2020). The nine selected metrics correspond to the hydrological metrics that achieved a good performance in the validation stage (Fig. 2). The different symbols represent the six target macro-basins. More details on the macro-basins can be found in Table S1. The spatial patterns shown by the 896 basins can be found in Fig. S1.

Full size image

Surface layer of NWP coastal system

Spatial and temporal patterns in sea surface temperature and turbidity (R645)

The mean spatial pattern of sea surface temperature (SST) and the surface reflectance at 645 nm (R645) revealed a high level of spatial heterogeneity in these conditions over the NWP region (Fig. 5a, e). Warmer waters occur in the inner section of the most prominent fjords and inside the Reloncaví Sound, and there is a clear difference in long-term mean SST between the northern and southern sections of the inner sea of Chiloe (Fig. 5a). Nearshore waters appear to be much cooler in the southern section, south of Boca del Guafo, especially around the many islands west of the Moraleda Channel (~ 44 \(^\circ\)S). In general, areas that are warmer on average exhibit a more pronounced seasonal cycle, with amplitudes of up to 5 \(^\circ\) C inside the fjords Puyuhuapi and Comau and around 4 °C in areas like the Reloncaví Sound (Fig. 5b). The smallest seasonal changes in SST (amplitudes of 1 °C) were observed on the southeast shore of Chiloe Island and at the northwest tip of the island, outside of the Chacao channel (Fig. 5b). Despite these spatial differences in mean SST and in the amplitude of the seasonal fluctuations, in 93.5% of the region analysed the annual harmonic explained \(\ge\) 80% of total SST variability (i.e., residual variance \(\le\) 0.2), with residual variance reaching maximum values of 70% on the exposed shore north of Chiloe island (Fig. 5c).

Fig. 5
figure 5

Summary of spatial and temporal patterns of surface and hydrographic conditions in the Northwestern Patagonia coastal system. Each row of maps shows a different variable: sea surface temperature (a–d), the turbidity proxy R645 (e–h), and the halocline depth (i–d) corresponding to the depth of maximum salinity gradient. Columns show the long-term means (a, e, i), the amplitude of the seasonal signal (b, f, j), the non-seasonal variability or residual variance (c, g, k), and statistically significant interannual trends (d, h, l). Data from MODIS-Aqua monthly composites (SST and R645; 2003–2022) and the Copernicus GLORYS12V1 reanalysis (salinity profiles 1993–2021). Panels d, h, and l show only the pixels with significant (α = 0.05) negative (blue) and positive (red) trends.

Full size image

As for the turbidity proxy (R645), the highest long-term mean values were observed at the mouth of rivers Yelcho and Palena and inside the Aysén and Reloncaví fjords (Fig. 5e). Also, high levels of turbidity were found on the southwest shore of Chiloe Island, despite the absence of major rivers in the area. The seasonal fluctuations of turbidity were more apparent near river mouths and in nearshore pixels throughout the region (Fig. 5f), and the annual harmonic explained at best ca. 20% of total variability along the central sections of the inner sea and along the outer shores of Chiloe island (Fig. 5g).

The analysis of temporal trends in SST showed a warming trend of up to 0.35 °C per decade in the northern section of the Chiloe inner sea, across Boca del Guafo, and on the open waters west and north of Chiloe island (Fig. 5d). The rate of surface warming appeared slightly higher in offshore waters than in Boca del Guafo and the northern section of the Chiloe inner sea (Fig. 5d). As for significant trends in the turbidity proxy R645, there was a predominance of decreasing trends across the southern section of the inner sea, in coastal waters northwest of Chiloe island, and in offshore waters (Fig. 5h). The most negative trends in R645 (i.e., waters becoming less turbid over the past two decades) were observed along the central and SE sections of the Inner Sea of Chiloe. This pattern was particularly evident in the coastal area influenced by the Yelcho River (Figs. 1 and 5h, dark blue symbols). On the other hand, significant increases in turbidity were detected on the west and southern shores of Chiloe Island and on the exposed shores of the many islands located between 45 and 46 °S (Fig. 5h, red symbols).

In addition to the trends just described, the occurrence of a significant increase in surface turbidity was detected at the inner section of the Reloncaví Fjord during the spring–summer of 2014–2015. This significant shift in the residuals (Pettitt test72, p < 0.05), which were obtained by subtracting a periodic, stationary seasonal signal from the monthly time series of R645, reveals a sudden increase in the mean levels or surface turbidity in the area where the rivers Puelo and Petrohué drain. Furthermore, an inspection of the residuals grouped by season (Fig. S2) suggested that the upward shift in turbidity was more pronounced in fall and winter (March through August) than in spring and summer (September through February).

Spatio-temporal patterns in salinity

The validation of the GLORYS12V1 product showed different levels of reliability between the northern and southern zones of the study area (Table S2 and Fig. S3). The correlation analyses applied to the reanalysis and observed data revealed a better agreement along the northern zone (i.e., the inner sea of Chiloé and Reloncaví Sound). Within the top 10 m of the water column, a correlation coefficient (r) of 0.62 (p =  < 0.01) and a low error (RMSE = 1.75) were found for the northern zone (more details in Table S2 and Fig. S3). In contrast, the GLORYS12V1 product did not perform as well in the southern zone (r = 0.25; p =  < 0.01; RMSE = 4.08) and tended to overestimate surface salinity (Table S2 and Fig. S3).

There was a clear seasonal pattern in the salinity structure across the region, with two low-salinity poles separated by the main connection to the Pacific Ocean at Boca del Guafo (Fig. 6). The first of these low-salinity poles was in the northern section of the study region where large rivers (Petrohué, Cochamó, Puelo) flow into the Reloncaví Fjord and adjacent coastal areas. In the south, the rivers Aysén, Cisnes, and Palena appear to have a strong influence on the thickness of the low-salinity surface layer along the Moraleda Channel (Fig. 6), i.e., areas with high depth values imply that the water column above that depth would be mainly dominated by freshwater influence. On average, the ocean signal (i.e., salinity values > 32.9) shoals during the summer months and reaches its shallowest depth in autumn (MAM, Fig. 6). The saline layer 32.9–33.2 then deepens significantly during the winter months (JJA), especially at the northern end. In spring (SON), the freshwater-influenced layer expands southwards (Fig. 6).

Fig. 6
figure 6

Seasonal variability across the NWP coastal system in the depth of salinity values in the range [32.9–33.2], used as a proxy for freshwater vs. ocean influence. Maps correspond to mean values for 1993–2021. Darker areas indicate stronger freshwater influence (i.e., thicker low-salinity surface layer). The lighter shade corresponds to areas where salinity values were higher than 33.2 (source: GLORYS12V1).

Full size image

The spatial pattern for the halocline depth (i.e., depth of maximum vertical gradient) derived from the salinity profiles revealed differences among coastal zones in the region, with greater freshwater influence and a shallower halocline towards the NE section and at the southern tip of the study region (Fig. 5i). In particular, the Reloncaví Sound and the outlet of the Chacao Channel were the most dynamic zones, where the vertical salinity change was most pronounced in the top 10 m of the water column (Fig. 5i). The amplitude of seasonal fluctuations in halocline depth with largest along the shore north of Chiloe island, and secondarily on the southwest shore of the island and near the Palena river mouth (Fig. 5j). Seasonal fluctuations in the halocline depth were small in areas most strongly influenced by river plumes, which is consistent with the year-round signal detected in previous studies44. Even in the areas where the seasonal amplitude was largest, the residual variance was > 50% (Fig. 5k), thus highlighting the importance of non-seasonal variability for salinity stratification in this region. The residuals for the halocline-depth proxy revealed a shoaling trend (i.e., thinning of the low-salinity layer) in the continental coastal zones where the rivers of interest in this study are tributaries (Fig. 5l, blue shading), at a rate of ca. 0.5 m  year−1. The changes in the halocline depth have been particularly evident in the Reloncaví Sound, possibly due to the reduced freshwater discharge from the Puelo and Petrohué rivers (Fig. 1), as well as other important rivers such as the Vodudahue (into the Comau Fjord) and the rivers Negro and Blanco in Hornopirén (Figs. 1 and 5l).

Discussion

The risks associated with the recurrence and magnitude of drought events are increasing. Our results highlight the need to design and implement climate change adaptation measures to limit the impacts of changing seasonal runoff patterns32,35,37 on the NWP land-sea interface and the socio-economic activities it supports.

Hydrological changes

The model-based runoff simulations showed a significant prevalence of high specific runoff along the coast, with 18% of the continental study area experiencing annual specific runoff greater than 4000 mm (Fig. 3a). The coastal area received an average freshwater flux almost 7300 m3 s−1 over the historical period (1990–2020), with macro-basins contributing 48% of this input (Fig. 3b). This spatial variability in runoff distribution highlights the complex interplay of natural climate variability, climate change and local geography in shaping hydrological patterns in the region. The spatial variability in the region can be attributed to its precipitation patterns, which are largely driven by disturbances within the westerly flow73. These disturbances create pronounced orographic gradients, with windward uplift promoting hyper-humid conditions along the western Andean slopes, where precipitation exceeds 5000 mm. Since previous studies considered only a limited number of long-term records in the macro-basins24,32,35, this study is the first to provide a complete coverage of the spatial hydrological patterns. Nevertheless, while the model-based data provide valuable insights, acknowledging uncertainties in the modelling workflow and data sources (Fig. 2) is crucial to accurately interpret these results and guide future research efforts.

The significant trends detected in many catchments (Fig. 4 and S1; Table S1) are consistent with previously observed climate trends in the NWP region35, characterised by a decline in precipitation (− 8 ± 8% per decade) alongside a slight increase in air temperature, particularly in summer (0.3 ± 0.1 °C in December), and a significant reduction in snow cover, notably affecting basins towards the south such as Yelcho, Palena, and Aysén. Our analysis indicates that the previously observed changes in the Puelo River streamflow24,32 may be representative of many tributaries to the NWP coastal system (Fig. 4 and S1), with significant trends in four of the six target macro-basins, resulting in reduced dry season freshwater input to semi-enclosed coastal systems. These changes are more pronounced during the summer months, resulting in decreasing trends in minimum streamflow metrics on both a monthly and seasonal basis, with all macro-basins showing a decrease in 90-day minimum streamflows ranging from − 2 to − 8% per decade (Fig. 4). Consequently, these changes have significant implications for coastal ecosystems such as Reloncaví Fjord and Reloncaví Sound (Fig. 1), as well as the northernmost segment of the Inner Sea of Chiloé. These findings highlight the intricate interplay between climate variability, hydrological processes, and the resilience of coastal systems, emphasizing the need for integrated management approaches to mitigate potential impacts on freshwater resources and coastal ecosystems in the NWP region.

River-to-sea interactions

In our study, the analysis of satellite and reanalysis data on sea surface temperature, turbidity, and vertical salinity structure showed clear signs of a reduced influence of freshwater input in the upper water column (Fig. 5). The increase in SST across the northern section of the Chiloe inner sea could indicate a weakening of river influence and a stronger influence of southerly winds, which would increase the residence time of surface waters. The lack of significant trends in SST in the southern half of the inner sea points to a stronger oceanic influence associated with the weakening of freshwater inputs from rivers such as Yelcho and Palena (Fig. 5d and S1) and because these areas are strongly influenced by energetic tidal currents74, with velocities often > 1 m s−1, and high levels of vertical mixing, which probably attenuate or counteract the surface warming.

Regarding surface turbidity, previous studies conducted in this region44 have reported a good correlation (r ~ 0.6) between time series of river discharge and R645 from satellite imagery, which is particularly evident during periods of high streamflow and low salinity in the coastal surface layer. Our analysis revealed that in some areas, there is a significant trend to a reduction in the suspended sediment load, especially in the area influenced by the Yelcho River (Fig. 5h). Though several mechanisms could contribute to this trend24,75, the most likely explanation is a reduction in peak streamflows from Yelcho’s main tributary, the Futaleufú River, as a consequence of hydropeaking by a hydroelectric complex on the Argentine side of this basin. Second, the natural trapping of sediment (Total Suspended Solids, TSS) in Lago Yelcho76 and the distinctly U-shaped or threshold TSS:Q curves for the Yelcho discharge imply that streamflows > 500–600 m3 s−1 are required for TSS load to exceed lake effects (Fig. S4). Unlike the other macro-basins considered here, Yelcho has a high-order lakes with a significantly longer estimated water residence time, of approximately 1.5 years77. The problem with this explanation is that the hydroelectric dams have been in operation since 1978, while the turbidity time series analysed here begins in 2003. Thus, a third factor is necessary to avoid discarding this hypothesized mechanism, namely decreased overall precipitation and runoff due to climate change. While the climate change driver is general, affecting all large and small basins considered here, the combination of global change effects with the existing intervention (hydropeaking and sediment trapping in the upper basin) may result in a much more pronounced reduction in the streamflow exceedance needed for TSS transport in the Yelcho basin over recent decades. A potential reduction in the proportion of glacial meltwater, which may contribute to higher TSS load during low-streamflow periods (Fig. S4) may be a complementary factor (noting that estimated glacial peak water for the Yelcho basin corresponds with the recent remote sensing time period). Changes in hydrograph signatures, i.e., rise/fall rates and frequency of pulses (Fig. 2) may also potentially alter TSS dynamics, with corresponding differential trends between northern and southern basins.

In addition to the interannual trends in surface turbidity detected by our analysis, it is noteworthy that specific areas like the inner section of Reloncaví Fjord exhibited a sudden and significant upward shift in turbidity during the spring–summer season of 2014–2015. This shift could be deemed counterintuitive given the decreasing streamflows observed in the Petrohué and Puelo rivers (Fig. 4). However, the increase in turbidity may be related to the change in the timing of peak streamflows observed in these rivers, from rainfall-driven winter floods to short spring floods mainly driven by snowmelt24. Another plausible mechanism is that, in recent decades, the first major precipitation events, which typically occur in late autumn and early winter, are interacting with basins exposed to increasingly longer droughts, thus increasing the potential for allochthonous material transport75 into the fjords during floods78,79. The effects of reduced freshwater input were evident in temperature and turbidity, but it was challenging for our study to capture this effect on salinity, which is most influential for hydrographic conditions and the circulation in NWP estuarine waters. In this regard, the performance of the GLORYS12V1 product in the northern zone (r = 0.62; p < 0.01; RMSE = 1.75) allowed us to observe how the influence of freshwater determines clear spatial and temporal patterns of salinity in this area. However, it’s important to recognize that in the southern zone, these observations are subject to high uncertainties (r = 0.25; p < 0.01; RMSE = 4.08), highlighting the need for further research to fill these gaps and improve the understanding of coastal hydrological processes in the NWP region.

On a regional scale, the spatial configuration of freshwater inflows from the tributary network to the NWP inland coastal system (Fig. 3) determines the presence of two low-salinity poles separated by an open connection to the Pacific Ocean at Boca del Guafo (Figs. 5i and 6). Temporally, the layer with salinity values in the 32.9–33.2 range appeared to shoal in summer and especially in autumn, coinciding with the lowest influence of river freshwater (Figs. 4 and 6). During winter, precipitation increased, and river peaks occurred, causing the mixed layer to deepen, especially in the northern region (Fig. 6). In spring, the influence of freshwater expanded southward, coupling the mixed and nival hydrological regimes of the northern and southern rivers (Fig. 6). In this context, the analysis of temporal trends in halocline depth identified coastal areas with different levels of freshwater influence in their local hydrography (Fig. 5l). Reloncaví Sound stood out as one of the most dynamic zones, exhibiting a shoaling trend toward shallower haloclines at a rate of about half a meter per year (Fig. 5l). In other semi-enclosed coastal zones such as Hornopiren and Comau, where the Negro, Blanco and Vodudahue Rivers streamflow out, the trend analysis also showed a remarkable thinning of the halocline indicators (Fig. 5l).

Based on the good performance of GLORYS12V1 in the northern zone, under the influence of Puelo River, which shows marked runoff decline trends (Fig. 4), we used salinity stratification proxies for this section of NWP to examine the temporal relationships between the monthly time series of Puelo streamflow and the maximum vertical salinity gradient in the Reloncaví Sound. The Wavelet and Cross-Wavelet Transform analysis performed on these time series revealed an overall dominance of the annual cycle (i.e., seasonal signal), with discrete periods where the spectral peaks around this period were statistically significant: 1998–2000, 2003–2005 and 2013–2017 for the Puelo streamflow (Fig. 7, top) and 2003–2009, 2011–2014 for the vertical salinity gradient (Fig. 7, middle). Notably, the Puelo streamflow wavelet showed a departure in the dominant periodicity, towards shorter periods (8–10 months) in 2003–2008, and from annual to biannual, during 2014–2017, with even longer periods concentrating most of the variability from 2017 onwards (Fig. 7, top), though not significantly. Fluctuations with periods of 6–8 months in the vertical salinity gradient appeared as significant in 2015–2016 (Fig. 7, middle). Notably, the Puelo streamflow wavelet showed a departure in the dominant periodicity towards shorter periods (8–10 months) in 2003–2008 and from annual to biannual during 2014–2017, with even longer periods concentrating most of the variability from 2017 onwards (Fig. 7, top), though not significantly. The timing of these changes in the periodicity of freshwater discharge from the Puelo river is consistent with the sudden increase in surface turbidity detected across the inner section of Reloncavi fjord, and likely reflects the changes in climate forcing associated with concurrent positive phases of the Southern Annular Mode (SAM), Pacific Decadal Oscillation (PDO) and El Niño conditions44. As for the vertical salinity gradient, fluctuations with periods of 6–8 months (i.e., bi-annual variability) appeared significant in 2015–2016 (Fig. 7, middle). The cross-wavelet coherence analysis revealed an interannually-variable association between the Puelo streamflow and maximum salinity gradient time series, with the strongest coherence (i.e., correlation) between the freshwater input and estuarine stratification at the annual scale (12-month period) prior to 2001 and during the 2005–2015 decade. During both periods, there was a phase shift of 45–60° (i.e., 1.5–2 months) between both time series, thus indicating that seasonal changes in estuarine stratification tracked seasonal changes in freshwater discharges by 1.5–2 months. From 2014 onwards, however, the correlation between these time series became negative and shifted from annual to biannual periodicity (Fig. 7, bottom). Similar multiscale patterns and synchronies between satellite-derived R645 and Chl-a time series for this region were found in a recent study encompassing this region80. These synchronies were particularly evident in the northern sections of the Inner Sea of Chiloé, where they identified changes in synchronies from one (2008–2015) to two-year periods during the last decades (> 2014). Comparing our results with those from previous studies that identified synchrony between oceanographic and fluvial variables, it becomes increasingly clear that the change in the hydrological signature of NWP rivers has reached a magnitude that is already influential on the NWP coastal system.

Fig. 7
figure 7

Continuous wavelet transforms of the monthly time series for the Puelo River streamflow (top) and the maximum vertical salinity gradient in Reloncaví Sound (middle) between 1993 and 2021. Black contours indicate time–frequency ranges with statistically significant power, and the dashed horizontal lines indicate a 12-month period (i.e., annual cycle). Bottom panel: Cross-wavelet coherence analysis for both monthly time series. The thick black contours indicate significant coherence (colour scale). Arrows indicate the phase of the coherence between the variables (right = in-phase, left = anti-phase).

Full size image

Mussel farming is an important sector of Chilean aquaculture with Chile being the second-largest producer of farmed mussels (Mytilus chilensis) and the top exporter of mussels in the world. A unique aspect of mussel farming in the NWP region is that “seeds” (i.e., planktonic larvae) are obtained from wild mussel beds. In fact, about 99% of the “seeds” for mussel farming in the region are collected using long-line systems deployed near natural mussel beds, which to date are mostly located in the Reloncaví Fjord and the Hornopirén area, both of which are strongly influenced by high freshwater inputs81,82 (Figs. 1, 5 and 6). Lower salinity of the upper water column seems to favor the dominance of M. chilensis over other species, thus facilitating the production of larvae and the collection of seeds mostly of this species. Therefore, these ecosystems function as natural hatcheries. Later, the collected seeds are transported to several areas, mostly located on the eastern coast of Chiloé Island, where they undergo the fattening processes.

Management needs for adaptation to climate change

The climate projections for NWP towards the middle of the century are not very encouraging, with recent studies projecting more intense and prolonged hydrological droughts34,35. This is important from an oceanographic and socio-economic perspective, as hydrological changes such as prolonged droughts will increase the risk of oceanographic anomalies. On a local scale, studies carried out in the Reloncaví Fjord have shown anomalous events in the upper water column in recent decades under low summer-autumn streamflow of the Puelo River. In such scenarios, the influence of the freshwater inflow on the surface layer of the fjord tends to weaken33,36,83, leading to a more frequent recurrence of autumn phytoplankton blooms (> 2 mg Chl-a m−3, > 500 cell mL−1) and an increase in the synchronicity between wind events, higher surface salinity and low DO saturation (< 50%) and concentration (< 5 mg L−1). The results of our study show that the changing trend of the NWP rivers will continue to generate a lower freshwater inflow, which promotes a shallower halocline and higher surface temperatures (Fig. 5), thus creating favourable scenarios for the onset of harmful algal blooms and hypoxia events, increasing their frequency and duration, as currently observed in the NWP22,84.

Recent risk assessments and climate-change risk maps developed for various economic sectors including salmon and mussel farming use freshwater dependency as a relevant sensitivity factor under climate change (ARClim project85). Among many other ecosystem services, coastal areas influenced by high freshwater inflows used for salmon farming (around 50% of salmon farming in Chile) have shown lower impacts of sea lice (Caligus rogercresseyi) one of the main sanitary problems in Chilean salmon farming86. This parasite has an optimal development in high-salinity waters, but its larval stages are sensitive to low salinity87,88, a feature that affects its settlement rates and impact on salmon production. On the other hand, the weakening of stratification by decreasing freshwater inputs can enhance the occurrence of HABs. One of the most extreme examples occurred in the summer of 2016, under the synergy of the SAM positive phase and El Niño events, when the lowest freshwater inputs in 70 years created conditions for the occurrence of a record HAB in Northwestern Patagonia32. At that time, the HAB caused in a few days levels of salmon mortality equivalent to what the entire salmon farming industry typically experiences in two years, amounting to US$800 million in losses. Extreme events such as the one in 2016 were again observed in April 2021, following another very dry summer37, when a HAB led to significant mortality of farmed salmon in the Comau Fjord with an estimated loss of US$ 25 million37. It should be recalled that in these coastal zones the analysis of interannual trends in surface conditions showed a remarkable thinning of the halocline proxy (Fig. 5l).

Consequently, expected changes in the upper water column due to reduced freshwater input in the main basins should be considered as relevant threats to salmon and mussel farming and must be included in the future management of the sector31. In recent years, the Chilean government has followed United Nations recommendations on adapting to climate change: developing sectoral adaptation plans, implementing training programs, introducing new legislation, and decentralizing decision-making. However, there is a need to accelerate the implementation of these initiatives in remote and well-preserved areas where the signs of climate change will be increasingly evident and its effects could also have strong socio-economic and environmental impacts, as in the case of the NWP. Climate change adaptation of aquaculture, fisheries and marine biodiversity also depends on the way we manage the land and basins. This implies not only the design and implementation of mitigation and adaptation actions and strategies at the affected marine farming sites, but also the consideration of ecosystem-scale nature-based solutions, including appropriate management and protection of basins that feed the coastal system in the face of climate change.

Our results suggest that the influence of rivers on adjacent estuarine and coastal waters of NWP has changed (lower total volume, especially in the summer, and different timing of peak streamflows) and could continue to change due to the hydrological projections predicted for this region. These findings could be seen as an invitation to prevent additional disturbances to river basins in the region that could amplify or propagate their effects. To this end, it seems appropriate to advance the protection and adequate management of NWP basins to limit the joint impacts of climate change and land-use change. Currently, there is growing concern about the increase of anthropic pressure on the landscape in this area of Chile, where an increase in road connectivity could be a gateway to more extreme events with high disturbance potential (e.g., wildfires89). Similarly, the rivers of NWP have always been of interest for their hydroelectric potential, but thus far no legal instruments exist that limit or regulate the realization of hydroelectric projects. Among their many impacts, the potential increase in this type of anthropogenic pressure on landscapes and riverscapes could enhance the change in hydrological regime of NWP rivers due to climate change, and lead to as yet unanticipated consequences for the structure, productivity, and fate of estuarine and marine ecosystems in the region. There is an urgent need to continue and improve the monitoring of main rivers and their basins and continue studies to better understand changes in the marine system functioning as well as the response of aquaculture, fisheries and biodiversity to reduced freshwater inputs.