Temperature Trends and Seasonality in Neritic and Transitional Waters of the Southern Bay of Biscay from 1998 to 2023

Abstract

Temporal and spatial variations in water temperature were analyzed from 1998 to 2023 across two contrasting southeast Basque coast estuaries: the deeper, stratified estuary of Bilbao and the shallower, mixed estuary of Urdaibai. We assessed long-term trends, seasonality, intra- and inter-estuary differences, and links to hydro-meteorological drivers using time-series decomposition, clustering, cumulative sum, regression, and correlation analyses. The largest differences in interannual and seasonal patterns occurred between outer neritic and shallow transitional waters. Most water masses warmed overall, with increases until 2003–2006, followed by cooling until 2013–2015, and sharp warming in 2020–2023. The strongest trends (0.24–0.25 °C decade⁻¹) occurred in middle-estuary waters, while inner above-halocline waters of the stratified estuary showed no trend or slight cooling. The strongest warming occurred in spring, particularly in the easternmost mixed estuary (0.49 °C decade⁻¹), especially in May (0.88 °C decade⁻¹). Seasonal minima and maxima occurred earlier in surface transitional waters than in neritic and deep transitional waters of the stratified system. Over time, temperature maxima advanced, minima were delayed, shortening the warming phase, and springs became warmer, extending the warm season. Air temperature was the main driver of water temperature trends, while river flow modulated patterns at annual and seasonal scales, with negative correlations with temperature, mainly in spring.

1. Introduction

Temperature is a key driver of biological processes from molecular to ecosystem levels and varies over daily to multidecadal time scales. Much effort focuses on developing theories of biological responses to climate change and predicting these responses across levels of organization and across spatial and temporal scales [1,2,3]. A major focus of attention is the analysis of spatial and temporal temperature patterns in terrestrial and marine ecosystems in the context of the expected climate warming throughout the 21st century [4]. Globally, sea surface temperature (SST) is rising [5], but with exceptions [6]. Warming rates vary strongly [7], both between ocean basins [8] and within regions [9,10,11]. Coastal warming is highly heterogeneous, with temperature trends differing across continental shelves and from coastal to open-ocean locations [12,13,14]. Studies in Australia [15] and South Florida [16] have shown that estuarine waters warm faster than nearby shelf areas. They are also warming at rates higher than those predicted by global ocean and atmospheric models, suggesting these models may underestimate estuarine warming [15]. A recent global assessment of estuarine surface water warming rates has shown that less than half of the studied estuaries have exhibited warming trends over the recent 30-year period (1985–2022). Additionally, estuarine water warming trends have shown significant differences across hemispheres, continents, and between small and large estuaries [17].

Moreover, climate change affects not only mean SST trends but also variability, extreme values, and seasonality, although these aspects remain less explored [6]. Models predict regional variability in SST warming in large marine systems such as the Baltic Sea, with clear latitude-dependent seasonal warming differences [18,19]. Both predicted and observed water temperature changes also show spatial and seasonal differences in estuaries [20,21].

However, aquatic systems are three-dimensional, and surface changes alone do not capture their full dynamics. Studies on long-term water temperature variability with depth are more limited due to the scarcity of data at subsurface and bottom layers, but the existing evidence suggests that warming trends differ across depths due to the influence of different forcing factors at each layer [22]. Similarly, the intensity, duration, and spatial extension of heatwaves vary with water depth [23,24]. These differences are expected to be larger in coastal areas and estuaries, where natural temperature fluctuations occur at smaller spatial and temporal scales. As a result, these systems are more likely to face increased ecological stress [25].

Estuaries are ecologically and economically valuable systems [26], but due to their high warming response to the current climate crisis [15,16], the biodiversity and uniqueness of estuarine biota, as well as the commercial and recreational services of estuaries, are threatened. The rise in temperature is a common driver of water quality impairment in estuaries [27]. It can enhance the severity and extent of hypoxia [28] as well as the level of pathogens and infectious diseases [29] and increase the frequency, severity, and duration of harmful algal blooms [30]. Extreme thermal events can also affect fisheries due to catastrophic fish mortality [31]. However, estuaries vary widely in size, morphology, and hydrodynamics, and therefore encompass a broad range of vulnerabilities to climate warming [15]. Estuarine water temperature is shaped not only by atmospheric drivers but also by external factors such as river discharge, upstream temperature, marine intrusions, and coastal upwelling; internal processes—mixing, dispersion, and residence time—also play a role, e.g., [32,33,34,35]. This complexity in temperature patterns across spatial and temporal scales leads to different vulnerabilities to warming in the context of climate change, with implications for estuarine functions now and in the future [36]. Therefore, disentangling the spatial and temporal patterns and rates of temperature change in different types of estuaries and identifying their main driving factors is essential to improve our understanding of future environmental and biotic changes. Such insights are vital for informing management actions in these dynamic systems.

The aim of this study was to analyze water temperature variability at interannual and seasonal scales, as well as along axial and vertical spatial gradients, in two estuaries of the Basque coast, the estuary of Bilbao and the estuary of Urdaibai (Bay of Biscay), during the last 26-year period (1998–2023). The specific objectives were the following:

(a)

To identify distinct water mass clusters based on temperature variability similarities.

(b)

To describe long-term water temperature trends and assess rates of change.

(c)

To establish seasonal patterns, phenological changes, and seasonal and monthly differences in interannual temperature variations.

(d)

To determine the relationship of air temperature, river flow, and upwelling with water temperature variability at the different spatial and temporal scales studied.

(e)

To assess the role of estuarine morphology in shaping thermal conditions in the estuaries under study.

2. Materials and Methods

2.1. Study Area

The nearby estuaries of Bilbao (43°23′ N, 03°07′ W) and Urdaibai (43°22′ N, 02°43′ W) are located on the Basque coast (inner Bay of Biscay), within the eastern North Atlantic temperate region (Figure 1a). They share a temperate oceanic climate characterized by mild winters, warm summers, and frequent rainfall [37]. However, they differ largely in morphology, topography, and watershed characteristics [38]. The estuary of Bilbao is a highly man-modified system currently turned into an artificial channel 25 m wide and 0.5 m deep in the upper fluvial reaches, 270 m wide in the widest middle reaches, and 10 m deep in the deepest lower reaches. This channel meets the Abra harbor, a semiconfined funnel-shaped embayment that increases in depth from around 12 m in the inner Abra to 32 m at the coastline (Figure 1b). Extensive land reclamation has greatly reduced intertidal areas, and dredging has deepened the estuary for navigation [39]. In contrast, the estuary of Urdaibai has remained much more natural, with an average depth of ~3 m at high tide and a greater proportion of intertidal and supratidal zones. However, a straight artificial channel was built alongside the natural meandering one in the upper estuary to facilitate navigation (Figure 1c).

Both estuaries experience a meso-macrotidal, semidiurnal tidal regime, but the estuary of Urdaibai has a much higher tidal prism volume/total volume ratio [40]. The estuary of Bilbao receives greater freshwater inflows, leading to a stratified system with a two-layer circulation [41] and a well-defined estuarine plume drifting offshore [42], whereas the estuary of Urdaibai is mostly a well-mixed system. Under usual river discharge conditions, both estuaries are sea-dominated systems with high salinity (euhaline > 30) water masses reaching the upper estuary below the halocline in the estuary of Bilbao and filling the outer half of the estuary of Urdaibai at high tide. However, the torrential nature of Basque coastal rivers [43] means that extreme flood events can, for brief periods of time, flush freshwater to the estuary mouths.

In the adjacent marine region, downwelling events prevail in autumn and winter due to southwesterly winds, while frequent episodes of weak upwelling occur in spring and summer under easterly and northeasterly winds [44].

2.2. Data Acquisition

Water temperature data were obtained from an ongoing hydrography and plankton monitoring program in the estuaries of Bilbao and Urdaibai. A Lagrangian sampling approach was used to assess changes in water masses of specific salinities instead of samplings at fixed locations. From 1998 to 2023, monthly measurements were taken at sites with salinities of 35, 34, 33, and 30 in the estuary of Bilbao and 35, 33, 30, and 26 in the estuary of Urdaibai, all of them below the halocline. These salinity sites were selected based on the axial distribution of water masses, ranging from the main salinity habitat (~35 salinity) for neritic plankton to brackish plankton habitats at varying salinities in each estuary. At each sampling site, water temperature was measured at 0.5 m intervals from just below the surface (where the sensor became submerged; hereafter referred to as surface (0.0 m)) to the bottom using different multiparameter meters (WTW and YSI) during the study period. To minimize tidal current interference and short-term physical stressor variability, sampling was conducted, whenever possible, in the morning, close to the high tide slack, and during the second neap tide of the month.

To account for local influences, air temperature data for the marine area adjacent to the estuaries were obtained from the Copernicus Climate Data Store (https://cds.climate.copernicus.eu/datasets, accessed on 24 January 2024). Additionally, air temperature over land was sourced from three hydrometeorological stations (Figure 1a): Sondika (https://www.aemet.es, accessed on 23 January 2024), located between the basins of the two estuaries; Abusu (https://www.euskalmet.euskadi.eus/, accessed on 23 January 2024), near the head of the estuary of Bilbao; and Muxika (https://www.euskalmet.euskadi.eus/, accessed on 23 January 2024), near the head of the estuary of Urdaibai. River flow data were also obtained from the latter two gauging stations (https://www.uragentzia.euskadi.eus/hasiera/, accessed on 23 January 2024), representing the combined flow of the Ibaizabal and Nerbioi rivers in the estuary of Bilbao and the Oka River in the estuary of Urdaibai. Upwelling index data were accessed from the Spanish Institute of Oceanography (http://www.indicedeafloramiento.ieo.es/index.html, accessed on 24 January 2024).

2.3. Data Selection

A total of 23 water temperature time series (combinations of sites and depths) were used for the present study. The selected temperature measurements included surface (0.0 m) and 2.5 m depths at all salinity sites in both estuaries and 5.0 m and 7.5 m depths at all salinity sites in the estuary of Bilbao except for 30 salinity at 7.5 m, where this depth was generally unavailable. Due to the shallowness of the estuary of Urdaibai, data below 2.5 m were not consistently available. Similarly, in the estuary of Bilbao, measurements at 5.0 m and 7.5 m were occasionally missing due to shallow water depths or cable length limitation (8 m) of the multiparameter meter used until 2003. To fill these gaps at some depths, missing values were primarily replaced with values from the depth nearest to the missing depth. Likewise, for other occasional missing data, two gap-filling methods were applied: (1) When a predefined salinity was not found, or temperature data were missing due to instrument failure (overall < 4.5%), missing values of temperature were estimated from polynomial regression models of temperature variation as a function of salinity. To this purpose, data from other sampling sites of the same date were used; (2) when a sampling was canceled, missing values were estimated using the average of the preceding and following sampling dates (<2.5%).

Finally, the criterion to establish the months included in each season was that the three coldest and three warmest consecutive months in the highest salinity time series are winter and summer, respectively. As a result, January, February, and March were considered as winter; April, May, and June as spring; July, August, and September as summer; and October, November, and December as winter.

2.4. Data Processing

First, the interannual trend component of the 23 water temperature time series and of the aforementioned hydrometeorological factors were extracted using the multiple seasonal decomposition function (mstl) from the forecast v.8.21 package in R [45].

According to similarities in interannual trends, the 23 water temperature time series were grouped into six distinct water mass clusters using Ward’s hierarchical clustering method, based on Spearman’s rank correlation (r = 0.8), with clusters defined at a 0.3 height threshold. For each sampling date, mean temperatures were calculated for each cluster of time series. To identify breakpoints in the variations in these mean temperatures throughout the study period, a cumulative sum (CuSum) analysis was conducted for each water mass cluster.

Additionally, the monthly mean for the entire study period was calculated for each water temperature time series to establish the standard seasonal pattern.

Ordinary least squares (OLS) regression analysis was performed to estimate temperature change rates from the trend component of each water temperature time series, both overall and by season. Because the seasonal and high-frequency variability had been removed during decomposition, the residuals were less affected by seasonal persistence or cyclic autocorrelation, which are known to bias uncertainty estimates in time series regression [46]. OLS provides unbiased estimates of linear trends under these conditions, and its assumptions are therefore more appropriate for the filtered trend component analyzed here. For each regression model, we report the intercept, slope decade⁻¹, 95% confidence intervals (confint function in R), and associated parameters (p-values and R²) from the model summary.

To examine relationships between water temperature and environmental drivers, Spearman’s rank correlation analyses were conducted (cor.test function, method = “spearman” in R). For each correlation, we report the correlation coefficient (r), indicating significance.

For all analyses, statistical significance was considered at p < 0.05. All statistical analyses were performed by means of R 4.5.0 software [47].

3. Results

3.1. Water Temperature Time-Series Clustering

Hierarchical clustering analysis of the 23 water temperature time series identified six water mass clusters based on similarities in interannual trends (Figure 2), herein termed as follows:

(1)

BTD (Bilbao-Transitional-Deep). It includes time series from the intermediate and inner (transitional) waters of the estuary of Bilbao, specifically from below the halocline (deep). This water mass cluster comprises data from 2.5, 5.0, and 7.5 m depths at the 34 salinity site; 5.0 and 7.5 m at the 33 salinity site; and 5.0 m at the 30 salinity site.

(2)

UN (Urdaibai-Neritic). It includes time series from the neritic waters of the 35 salinity site (both surface and 2.5 m depths) of the estuary of Urdaibai.

(3)

BND (Bilbao-Neritic-Deep). It includes time series from the outer zone (neritic) of the estuary of Bilbao, specifically from below the halocline (deep). This group comprises data from 2.5, 5.0, and 7.5 m at the 35 salinity site.

(4)

UT (Urdaibai-Transitional). It includes time series from the transitional waters of the estuary of Urdaibai, comprising data from surface and 2.5 m depths at the 33, 30, and 26 salinity sites.

(5)

BTS (Bilbao-Transitional-Surface). It includes time series from the intermediate and inner (transitional) surface waters of the estuary of Bilbao. The group comprises data from 0.0 m depth at the 34, 33, and 30 salinity sites.

(6)

BTG (Bilbao-Transitional-Gradient). It is a mixed group from the estuary of Bilbao, which includes the time series from the 35 salinity site at the surface and the 33 and 30 salinity sites at 2.5 m depth.

Figure 2. Similarity dendrogram of water temperatures from the 23 time series, constructed using Ward’s hierarchical clustering method based on Spearman’s rho correlations. Water masses (see text for definition of BTD, UN, BND, UT, BTS, and BTG waters) are defined by a 0.3 height threshold (dotted line), with water clusters outlined by boxes.

Figure 2. Similarity dendrogram of water temperatures from the 23 time series, constructed using Ward’s hierarchical clustering method based on Spearman’s rho correlations. Water masses (see text for definition of BTD, UN, BND, UT, BTS, and BTG waters) are defined by a 0.3 height threshold (dotted line), with water clusters outlined by boxes.

Neritic waters comprise outer estuarine waters influenced primarily by oceanic conditions (salinity ≥ 35 and closest to the mouth of the estuaries), whereas transitional waters correspond to inner and intermediate estuarine areas subject to stronger freshwater influence (salinity ≤ 34, shallower depths, and increasing distance from the mouth). As shown in Figure 2, BND and UN were grouped into a single cluster at the set threshold, thereby combining data from both estuaries. However, this was an exception, as all other groups were associated with a single estuary. Given their distinct estuarine origins and different hydrographic regimes, we considered it ecologically and physically relevant to treat BND and UN as separate clusters. Their dissimilarity was close to the statistical threshold, and separating them allowed us to examine differences in the thermal behavior of neritic waters associated with both estuarine location along the Bay of Biscay thermal gradient (BND to the west of UN) and depth (subsurface waters > 2.5 m in BND vs. surface waters ≤ 2.5 m in UN). For these reasons, we retained BND and UN as distinct clusters. At a lower similarity level, BTD merged with the neritic assemblage (BND and UN), while the other transitional waters (UT, BTS, and BTG) formed a separate group.

3.2. Interannual Trends in Water Temperature Across Water Masses

Interannual trends of water temperature at each depth, salinity site, and estuary showed an overall significant increase in most cases during the study period (Figure 3 and

Table 1. Intercept, rate of change decade⁻¹, 95% confidence intervals (minimum and maximum; rate of change decade⁻¹), and parameters (R² and p-value) of the ordinary least squares linear regression model describing the variation in the trend component of each water temperature time series (point) in the identified water clusters over the study period (see Figure 3). Significant p-values in bold.

Water Cluster	Point	Intercept	Rate of Change Decade−1	95% CI (Min–Max)	R2	p-Value
UN	U35_0.0	−21.23	0.19	0.13–0.25	0.11	<0.001
	U35_2.5	−20.91	0.18	0.12–0.25	0.10	<0.001
BND	B35_2.5	−3.21	0.10	0.03–0.16	0.03	0.002
	B35_5.0	−9.75	0.13	0.07–0.19	0.06	<0.001
	B35_7.5	−13.98	0.15	0.09–0.21	0.07	<0.001
BTD	B34_2.5	−20.81	0.18	0.12–0.24	0.10	<0.001
	B34_5.0	−20.83	0.18	0.12–0.24	0.11	<0.001
	B34_7.5	−22.39	0.19	0.13–0.25	0.11	<0.001
	B33_5.0	−32.13	0.24	0.19–0.29	0.20	<0.001
	B33_7.5	−32.55	0.24	0.19–0.29	0.22	<0.001
	B30_5.0	−24.89	0.20	0.14–0.27	0.12	<0.001
UT	U33_0.0	−30.52	0.23	0.17–0.30	0.14	<0.001
	U33_2.5	−25.29	0.21	0.14–0.28	0.10	<0.001
	U30_0.0	−12.66	0.14	0.07–0.22	0.05	<0.001
	U30_2.5	−19.72	0.18	0.11–0.25	0.07	<0.001
	U26_0.0	−33.82	0.25	0.18–0.32	0.12	<0.001
	U26_2.5	−33.87	0.25	0.18–0.32	0.13	<0.001
BTG	B35_0.0	11.60	0.02	−0.05–0.10	0.00	0.570
	B33_2.5	−23.49	0.20	0.13–0.26	0.11	<0.001
	B30_2.5	−1.06	0.08	0.01–0.16	0.02	0.022
BTS	B34_0.0	−13.95	0.15	0.07–0.23	0.04	<0.001
	B33_0.0	18.95	-0.02	−0.11–0.08	0.00	0.740
	B30_0.0	75.66	-0.30	−0.40–−0.20	0.10	<0.001

).

Exceptions were observed in some surface waters of the estuary of Bilbao, where a clear downward trend (−0.3 °C decade⁻¹ at the 30 salinity site) or no trend (at the 33 and 35 salinity sites) was found. The strongest increases were observed at the 26 salinity site in the UT water masses and at the 33 salinity site in the BTD ones, with warming rates of 0.25 °C and 0.24 °C decade⁻¹, respectively. Additionally, UN waters showed a slightly higher increase (0.18–0.19 °C decade⁻¹) than BND ones (0.10–0.15 °C decade⁻¹).

However, the increase was not gradual. Water temperature trends showed a decrease across all time series from 2003–2006 to 2013–2015, followed by an increase in recent years of the study period. In addition, the warmest and coldest years varied between clusters. Annual mean values for each water mass (cluster) showed that 2023 was the warmest year in neritic waters (UN and BND) and 2022 in transitional waters (UT, BTG, BTS). In the first half of the study period, temperature peaked in 2006 in neritic waters and in 2003 in transitional waters. The coldest years were 2015 in neritic waters and 2013 in transitional waters. Temperature in BTD waters (the nearest ones to the neritic waters of the estuary of Bilbao) was an exception because it aligned better with the neritic pattern (Figure 3).

The CuSum analysis of water temperature data calculated for the six water mass clusters (Figure 4) revealed five distinct periods: (1) 1998–2002, characterized by values below the mean in the main water masses (UN, BND, BTD, and UT); (2) 2003–2006, with values above the mean, especially in the earliest years in UT, BTG, and BTS, and in later years in BND, UN, and BTD; (3) 2007–2012, with values below the mean but closer to it in UT, BTG, and BTS waters; (4) 2013–2019, characterized by the lowest values below the mean across all waters; and (5) 2020–2023, with values significantly above the mean.

The linear interannual trends and the periods of increase and decrease identified by the CuSum analysis primarily reflected the interannual patterns of spring (April–June) water temperature, which showed stronger variations and higher rates of increase than other seasons’ temperatures (Figure 5 and

Table 2. Rate in temperature change decade⁻¹ for each water mass cluster in each season over the study period (see Figure 5).

	Rate of Change Decade−1
Water Cluster	Winter	Spring	Summer	Autumn
UN	0.14	0.42	−0.10	0.28
BND	−0.01	0.44	−0.05	0.13
BTD	−0.02	0.43	0.23	0.22
UT	0.07	0.49	0.05	0.27
BTG	−0.02	0.31	0.14	0.01
BTS	−0.18	0.17	0.10	0.08

). Indeed, spring showed the strongest warming across all water masses, with UT showing the largest spring increase (0.49 °C decade⁻¹) and BTS waters the smallest one (0.17 °C decade⁻¹). Autumn also showed considerable warming (0.28 to 0.22 °C decade⁻¹ in UN, UT, and BTD clusters).

In contrast, summer and winter displayed greater variability, with summer showing both warming (e.g., in BTD waters, 0.23 °C decade⁻¹) and cooling (e.g., minimum rise in UN waters: −0.10 °C decade⁻¹), while winter had the most inconsistent patterns, with predominantly negative trends. However, none of these relationships was statistically significant. Similarly, the interannual rates of change in water temperature at the monthly scale (Figure 6) were positive for all spring months, with May showing the strongest rates across most water masses (0.88 °C decade⁻¹ in UT and between 0.5 and 0.6 °C decade⁻¹ in BND, BTD, and UN). In autumn months, the highest rates were observed in November or December (e.g., 0.63 °C decade⁻¹ in December in UT). Negative change rates were primarily observed in summer and winter months, especially in August for UN (−0.37 °C decade⁻¹) and in March for BTS (−0.60 °C decade⁻¹).

3.3. Seasonal Patterns in Water Temperature Across Water Masses

Overall, as depth and salinity increased, winter temperatures were higher, summer temperatures were lower, and the annual temperature range narrowed (Figure 7). An exception was observed in the estuary of Bilbao, where summer maxima were higher at the surface of the 33 salinity site compared to the 30 salinity site. Also, the estuary of Urdaibai showed higher temperatures at the same depths and salinity sites than the estuary of Bilbao. The ranges of average water temperatures of each water mass cluster (Figure 7) were the following: from 12.2–12.4 to 20.6–21.1 °C in BND waters, from 12.0–12.2 to 21.7–21.8 °C in UN ones, from 11.9–12.4 to 21.0–21.9 °C in BTD ones, from 10.1–11.6 to 22.5–23.4 °C in UT ones, and from 8.4–10.9 to 21.5–22.6 °C for the joint BTG plus BTS waters.

The timing of the annual water temperature minima and maxima, as well as the coldest and warmest periods, varied between the different water masses. The coldest periods occurred from January to March, with annual minima in February for the BND, UN, and BTD waters, while for BTS and UT, the coldest periods were from December to February, with annual minima in January. Annual maxima were skewed toward July in UN, UT, and BTS waters (Figure 7). Spring water temperatures were always higher than autumn temperatures, with the smallest differences observed in deep water masses of the estuary of Bilbao (BND and BTD) and the highest in inner surface ones (BTS) (Figure 7). In the estuary of Urdaibai, such seasonal differences were also found from UN to UT waters, with an increase in the difference between spring and autumn temperatures and decreases in the temperature difference between spring and summer and between autumn and winter.

In all water masses, temperature seasonality (Figure 8) was similar in the five time periods identified by the CuSum analysis. The most significant variation over time of the annual cycle was the delay of the coldest winter values towards March and the advance of the warmest summer values towards July, which led to a shortening of the seasonal warming period (although this process did not occur in a totally gradual manner). The delay of the winter minima was more evident in the neritic waters of both systems and the transitional waters of the estuary of Urdaibai, where a faster and greater warming from April to May was also observed in the last period (2020–2023). In contrast, the first period (1998–2002) showed the lowest May temperatures and the coldest springs in all water masses, except in BTS. The 2003–2006 period was characterized by the coldest winters (at least in January and February or February and March) in all water masses and the warmest summers (except in BTD waters), and therefore, the highest temperature ranges. The 2007–2012 time span showed the coldest autumns in all water masses. The 2013–2019 period was characterized by water temperatures below the time series mean for all the seasons in all water masses and by the lowest range of temperature variation between the warmest and coldest periods. In contrast, the 2020–2023 period showed water temperatures above the time series mean for all the seasons, with the warmest springs and autumns in all water masses, the warmest winters in UN, BND, and BTD, and the warmest summers in BTD.

3.4. Relationship of Water Temperature with Environmental Factors

Interannual trends of environmental factors showed a significant increase over the study period (Figure 9 and

Table 3. Intercept, rate of change decade⁻¹, 95% confidence intervals (minimum and maximum; rate of change decade⁻¹), and parameters (R² and p-value) of the ordinary least squares linear regression model describing the variation in the trend component of air temperature (from various sources), river flow, and upwelling index over the study period (see Figure 9). Significant p-values in bold.

Variable	Type	Intercept	Rate of Change Decade−1	95% CI (Min–Max)	R2	p-Value
Air temperature	Copernicus	−41.03	0.27	0.21–0.33	0.22	<0.001
	Sondika	−73.63	0.44	0.37–0.51	0.34	<0.001
	Abusu	−26.17	0.20	0.13–0.27	0.10	<0.001
	Muxika	−63.59	0.39	0.32–0.45	0.28	<0.001
River flow	Abusu	−396.56	0.21	0.91–3.26	0.04	<0.001
	Muxika	−5.61	0.03	0.00–0.06	0.02	0.023
Upwelling index		−7985.5	38.05	12.31–63.78	0.03	0.004

). Regarding air temperatures, the rate of change varied considerably across meteorological station sources (Table 3).

At the monthly scale (Figure 10), warming rates were quite similar between over-land and over-ocean air temperature data. The assessment of the month-specific rate of warming/cooling showed an overall increasing trend of warming rates from winter (with decreasing rates in January) to autumn, with the highest warming rates in November-December. Similarly, at the monthly scale, river flow trends showed an almost coincident seasonal pattern between basins, with increasing rates in late autumn and early winter (from November to February) and clear decreasing rates in early spring (March–April) and early autumn (October). The upwelling index showed a somewhat opposite pattern to that of river flow, with a marked increase in April and October but a decline in January.

Regarding the relationship between environmental factors and water temperature (

Table 4. Spearman’s rank correlation coefficients between water temperature and hydrometeorological factors (AT: air temperature, RF: river flow, UPW: upwelling index) analyzed by season at each water mass cluster. Only statistically significant correlations (p < 0.05) are shown. Correlations with p < 0.001 are highlighted in bold. Shaded cells indicate the highest correlation for each identified water mass cluster and season. Data sources of local factors include Copernicus (c), Sondika (s), Abusu (a), and Muxika (m).

Water Cluster	Season	ATc	ATs	ATa/m	RFa/m	UPW
UN	Winter	0.589	0.602	0.494	−0.306
	Spring	0.907	0.881	0.868	−0.537	0.430
	Summer	0.593	0.514	0.521
	Autumn	0.871	0.861	0.780	−0.483
BND	Winter	0.496	0.473	0.410	−0.418
	Spring	0.893	0.874	0.882	−0.442	0.398
	Summer	0.559	0.503	0.551
	Autumn	0.844	0.835	0.805	−0.466
BTD	Winter	0.599	0.537	0.536	−0.434
	Spring	0.918	0.893	0.897	−0.525	0.411
	Summer	0.645	0.609	0.647
	Autumn	0.868	0.858	0.833	−0.509
UT	Winter	0.681	0.700	0.632
	Spring	0.896	0.886	0.864	−0.607	0.476
	Summer	0.608	0.525	0.624		0.313
	Autumn	0.835	0.839	0.811	−0.526
BTG	Winter	0.667	0.650	0.624	−0.486
	Spring	0.925	0.904	0.907	−0.504	0.448
	Summer	0.679	0.624	0.701	−0.273
	Autumn	0.836	0.812	0.812	−0.669	0.241
BTS	Winter	0.737	0.737	0.722	−0.267
	Spring	0.855	0.843	0.854	−0.491	0.481
	Summer	0.633	0.552	0.678	−0.319	0.298
	Autumn	0.878	0.853	0.866	−0.624

), air temperature showed the highest positive correlations across all cases, with over-ocean air temperature data (ATc) being the most consistent. However, over-land air temperatures showed stronger correlations during winter and/or summer in some water masses, the air temperature measured at the head of the estuaries (ATa/m) being the most strongly correlated factor with water temperature in all transitional water masses (BTD, UT, BTG, and BTS). River flow showed significant negative correlations, except during the summer in the estuary of Urdaibai (UN and UT) and below the halocline in the estuary of Bilbao (BND and BTD), where no correlations were observed. The upwelling index was positively correlated with water temperature in spring in all water masses.

4. Discussion

4.1. Differences in Thermal Environments

Water masses differed between and within estuaries in their interannual variability, seasonal pattern, and range of temperature. Neritic waters of both estuaries showed a very similar pattern, likely influenced by the connectivity of waters along the adjacent coastal area. In contrast, the thermal features of transitional waters diverged by estuary due to differences in water circulation and mixing patterns, which are, in turn, shaped by the estuarine morphology and river discharges. In the estuary of Bilbao, higher river discharges and deeper water channels produce vertical stratification, leading to differences in thermal features between surface and deeper transitional waters. In this estuary, deep transitional waters derive from high-salinity marine intrusions [41]. The present study showed that their limited exposure to air temperature and river flow enables them to maintain a neritic thermal behavior. By contrast, surface transitional waters exhibited a different thermal environment due to the extent and persistence of stratification, which allows a greater influence of air temperature. In contrast, in the estuary of Urdaibai, being shallower with lower river discharges, mixing dominates, yielding a similar thermal behavior over the entire water column. Deeper water temperatures were, therefore, also influenced by air temperatures in the estuary of Urdaibai. These thermal differences agree with a characteristic estuarine spatial gradient, where upper reaches are consistently warmer than the lower estuary [33]. In our estuaries, the delimitation between the thermal environments of neritic and transitional waters is well reflected in the changes from zooplankton communities dominated by neritic to brackish species [38,48]. Nevertheless, results also revealed we also found that the main waters of the estuary of Bilbao were, on average, colder and more thermally stable than those of the estuary of Urdaibai. This implies differences in the thermal habitats of brackish communities that may partly explain the observed between-estuary differences in the dominant taxa and phenology of brackish zooplankton [49].

4.2. Long-Term Temperature Trends and Change Rates

Over 1998–2023, all water masses warmed except the lowest-salinity, above-halocline waters of the inner estuary of Bilbao, which cooled slightly. Our data further suggests that warming trends may reverse if poorly mixed freshwater masses expand towards lower reaches in the event of increased river flows. In estuaries, freshwater discharge primarily affects upper layers more than bottom layers, influencing heating and cooling gradients and rates, particularly in frontal areas [50]. This agrees with observations in the stratified estuary of Bilbao in the present study, where the correlation between water temperature and river flow was higher in transitional gradient waters than in bottom or inner surface transitional waters during the rainiest seasons. As observed in our study, sites within estuaries that deviate from the prevailing warming trend have been reported from many other estuaries [51,52].

The comparison of warming and cooling rates between data series of the same area or similar types of waters has to be made with caution when they differ in the time window analyzed, due to the non-steady pace of global warming. Both of our systems showed the highest warming in transitional waters, with rates of 0.25–0.24 °C decade⁻¹ (considering data from all months in the series). The average warming rate from 1998 to 2023 in the main transitional waters of these two estuaries was 0.21 °C decade⁻¹, lower than the average increase of 0.37 °C decade⁻¹ reported for all modeled English estuaries between 1990 and 2022 [52]. However, our absence of colder early 1990s data, as shown by SST data from the nearby Aquarium of San Sebastian [53], may bias estimates downward. Still, inter-regional differences between our south-eastern Bay of Biscay estuaries and English estuaries are smaller than the intraregional variability (0.70–0.05 °C decade⁻¹) reported for the English systems [52]. Neritic waters warmed at an average of 0.15 °C decade⁻¹ in our two systems, confirming that estuaries are more sensitive to global warming than open sea waters [15,16]. Warming rates in neritic waters decreased substantially from surface layers in the estuary of Urdaibai (0.19 °C decade⁻¹) to subsurface layers in the estuary of Bilbao (0.13 °C decade⁻¹). This pattern is very likely due to the combined effect of depth and the west-to-east gradient of temperature increase that develops in the inner Bay of Biscay during the warm period [11]. In fact, in the Bay of Biscay the southeastern corner has shown the strongest warming trend [54], and our data have also shown higher warming rates in the neritic waters of the eastern site (estuary of Urdaibai) than in those of the western site (estuary of Bilbao). Moreover, further to the east along the Basque coastal waters, higher rates of SST increase (0.23 °C decade⁻¹) were observed from 1980 to 2019 [55], but as mentioned above, comparison must consider data gaps in our time series in the 1980s-1990s. In the neritic waters of the estuary of Bilbao, surface and deeper waters showed different thermal behaviors, in agreement with findings in shelf waters [22].

4.3. Temperature Trends and Change Rates at Intra-Annual Scales

Seasonal analysis showed the strongest warming in spring (April to June), weaker warming in autumn (November and December), and the weakest or even cooling trends in summer (August) and winter (February and March) in our estuaries. These intra-annual trends are consistent with observations in shelf waters of the Bay of Biscay from 1982 to 2014 [11], yet rates in our systems were higher. This suggests that intra-annual variations in water temperature at our local scale are in agreement with the regional pattern, but warming has been more intense in nearshore zones.

Our results also confirm that warming was mainly driven by an extended warm season rather than milder winters or warmer summers [11]. This pattern appears to characterize the first quarter of the 21st century in the Bay of Biscay, but it does not match observations during the last decades of the 20th century. For example, during the 1965–2004 period, [56] reported a clear seasonal dependence of warming rates, with summer temperature trends being twice as strong as those of winter in surface waters. In the southeastern Bay of Biscay, from 1972 to 1993, both winter and summer showed warming trends, but the winter increase was slightly higher [54]. Likewise, along the Cantabrian Sea coastal waters, spring and summer showed similar warming trends from 1985 to 2005 [57]. Results from other marine (e.g., NW Mediterranean [58]; northeastern continental shelf of North America [13]) and estuarine areas (e.g., Yangtze River Estuary [21]; Hudson River Estuary [59]; South Florida estuaries [16]) further confirm that seasonal variations in water temperature trends and rates of change can vary greatly as a function of the time window and sites/areas analyzed. Overall, these differences highlight the need for joint analysis of water temperature series from different regions and time periods to better understand the spatial and temporal effects of climate change in coastal and transitional waters.

4.4. Interannual Patterns

Our estuaries showed alternating warming and cooling periods with inflection points in 2003–2006 and 2013–2015, paralleling the large-scale subpolar North Atlantic SST trend reversal from warming (1994–2004) to cooling (2005–2015) [60]. Earlier Bay of Biscay data (1965–2004) showed non-uniform warming with early cooling until the 1970s [56]. A six-decade SST analysis from a site near our area also revealed 8-, 11-, and 18-year cycles likely tied to broader climate oscillations [53]. These alternating cooling-warming trends in the Bay of Biscay were more patent during the first decade of the 2000s (2002–2014) under atmospheric conditions favorable for plume enhancement [61]. Notably, in our study, cooling phases corresponded with higher river discharges.

Regarding the timing of extreme temperatures, in the neritic waters of our systems, the warmest summer values of the whole time series were registered in 2020 and 2023, and the coldest winter values in 2015. In the first half of the data series, the warmest values were in 2003 and 2006, and the coldest values were in 2005, consistent with SST records further east in Cantabrian coastal waters [53].

4.5. Seasonal Patterns

The largest seasonal differences were between the neritic subsurface waters and the inner area surface waters of the estuary of Bilbao, reflecting dominant oceanic vs. continental influences. Shallower waters in mid to inner areas showed larger temperature variations, a pattern observed in other estuaries too, e.g., [62]. Temperature minima were lowest and occurred earliest (January) in the above-halocline waters of the estuary of Bilbao and the transitional shallow waters of the estuary of Urdaibai. In contrast, in neritic and below-halocline waters of the estuary of Bilbao, the coldest temperatures were milder and occurred later (March), matching patterns in adjacent Cantabrian Sea coastal/shelf waters [57]. However, no such differences were observed for the annual warmest values and their timing. A similar pattern has been reported for other European marine areas, including the Baltic Sea, where summer SST peaked in August, but winter minima occurred in February in shallow waters and in March in deeper areas [63].

Seasonal patterns of temperature also showed interannual differences. Throughout the study period, the seasonal warming phase shortened due to a delay in winter minima, an advance in summer maxima, and an earlier start of an extended warm period. In agreement with this pattern, broader spatial scale observations across the North Atlantic have shown that areas like the Bay of Biscay have had earlier spring warming transitions [64]. Similarly, SST trends in the North American northeastern continental shelf showed earlier onset and longer duration of summers between 1982 and 2014 [13].

4.6. Relationships with Local Climate and Hydrographic Factors

In our two estuaries, at the interannual scale, air temperature was the main driver of water temperature variations. In addition, warming rates in transitional waters (0.25–0.24 °C decade⁻¹) were comparable to trends in offshore air temperatures from the Copernicus database. However, the waters (neritic and transitional) of both estuaries warmed more slowly than the air overall. In contrast, some estuaries (e.g., Chesapeake Bay) warmed faster than the air [51], while others (e.g., Hudson estuary) showed no significant difference between air and water warming rates [59]. However, the seasonal patterns of air and water temperature from our two systems showed differences, which suggests that local factors modulate estuarine water warming. Unlike air temperature, spring (April–June) water temperature showed strong warming trends. A similar pattern was reported for the Hudson River estuary, where water warming maximum occurred in April–August, but air temperature highest warming rates were registered in December and February–March [59].

Despite differences in river flow between estuaries, the overall trend of increase in freshwater discharge had a cooling effect on water temperature in both systems, which was reflected in a temperature decline when river discharge increased until 2013–2015, followed by a temperature rise as river discharge decreased. A similar cooling influence of freshwater discharge has been reported for the Bay of Biscay during 1982–2014, where river-influenced areas cooled while offshore waters warmed [61]. In the present study, the highest cooling effect was observed in the above-halocline water masses of the inner estuary of Bilbao. However, responses of water temperature to river discharge and air temperature vary across estuaries. For example, in the Hudson River estuary, increased freshwater input did not counteract water warming related to the increase in air temperature [59], while in the San Francisco Bay-Delta system, the combined effects of atmospheric and riverine influences differed spatially [34]. In our study, water temperature correlated strongly with river flow in spring and autumn, but no relationship was found in summer. Similarly, in the upper San Francisco Estuary, the negative relationship between water temperature and river inflow was most common, although positive relationships were found in winter [35].

Significant positive correlations between water temperature and upwelling index were found across all water masses in the estuaries of Bilbao and Urdaibai, but only in spring. Spatial differences in SST show that regions in the SW Bay of Biscay experiencing intense upwelling show lower SST values [54], and coastal upwelling mitigates warming in nearshore waters in comparison to adjacent ocean waters along the western Iberian Peninsula [12,65]. Upwelling and downwelling events also shape the thermohaline properties of the northwestern Iberian estuaries [32,66]. In our study area, downwelling was the dominant process, and it showed a weakening trend (upwelling index increased), which was most intense in April. Accordingly, we only found a significant positive correlation between the upwelling index and temperature in spring. This interannual pattern is opposite to what we would expect if upwelling/downwelling had a dominant effect on SST, e.g., [67,68,69]. Coastal upwelling and downwelling processes are driven by local atmospheric forces, and on the Basque coast, the seasonal pattern shows that spring downwelling relaxation is related to reduced wind stress and changes in wind direction, mainly of the prevailing north-westerly winds [40]. These results suggest that local wind variability influences estuarine water temperatures more strongly than shelf-water advection. Indeed, surface water warming can be influenced by both atmospheric surface heat flux and oceanic horizontal thermal advection, with winds playing a significant role in air-water heat flux [70]. However, studies have reported positive, negative, and uncertain relationships between SST and surface wind speed in coastal waters, e.g., [70,71,72,73]. Therefore, the effect of winds on water temperature in our study area needs to be assessed more thoroughly in future studies.

4.7. Ecological Effects

Ocean warming trends are causing changes in the structure and function of marine ecosystems because of the impact on living beings, from plankton to large marine vertebrates [74]. Warming can enhance growth rates of living beings, but excessive warming can be detrimental. In addition, other physical changes can be induced with warming, such as water column stratification/mixing, that have significant effects, particularly on phytoplankton [75]. Predictions on global marine plankton changes driven by global warming are an overall decline of phytoplankton and zooplankton biomass and the food value for fish, but with clear differences between marine ecosystems and regions [76]. An additional generalized pattern driven by global warming is the poleward migration of species [77]. This is having different effects depending on latitude, because excessive warming is already reducing species diversity in the equatorial region, i.e., the equatorial dip [78], but in temperate marine regions it is causing an increase in species richness [79]. Coastal and estuarine systems are among the most vulnerable marine ecosystems to climate change [80]. In our study area, variations in chlorophyll biomass and in the abundance and composition of neritic and brackish zooplankton provide indicators of potential responses to future temperature change. The observed warming trend coincided with declining phytoplankton biomass and increasing zooplankton abundance in both neritic and transitional waters. Moreover, interannual temperature variability was associated with contrasting contributions of vernal and summer–autumn species to total zooplankton abundance [49,81]. Species richness also increased due to the introduction of non-native thermophilic taxa, including Acartia tonsa, Oithona davisae, and Pseudodiaptomus marinus, which substantially modified the structure and diversity of brackish zooplankton in transitional waters [48,49]. These taxa are expected to be further favored under the projected warming scenarios. However, apart from climate change, local anthropogenic stressors are also known to severely impact coastal and estuarine biota [82,83], so our results may have also been affected to some extent by water quality variations [84].

5. Conclusions

Our results reveal common patterns in water temperature variability across both estuaries, which agree with those observed over wider areas in the Cantabrian Sea, the Bay of Biscay, and the North Atlantic. However, differences between and within estuaries highlight the sensitivity of estuaries to climate warming in a global context, as well as to the specific features that characterize the Basque coast. Based on temperature ranges and seasonal and interannual variability, two main thermal behaviors were identified. One is associated with water masses from the estuary mouths (neritic) and below the halocline in the stratified estuary of Bilbao, which was characterized by lower temperature variability, later annual minima, and an interannual delay in the alternation of warmest and coldest events influenced by the shelf waters’ thermal properties and dynamics. The other one is linked to water masses in the intermediate-inner reaches of the estuary of Urdaibai and surface waters of the estuary of Bilbao, which showed a higher variability and a stronger influence of local land-atmosphere conditions.

Temperature trends showed similar maximum warming rates in transitional water masses of both estuaries, but in different water masses, because these maxima were observed in below-halocline higher salinity waters (euhaline) in the estuary of Bilbao and in shallow lower salinity waters (polyhaline) in the estuary of Urdaibai. Warming rates were higher in the neritic waters of the estuary of Urdaibai than in the estuary of Bilbao, likely due to between-estuary differences in depth and geographic location. Overall, decreasing trends in above-halocline inner waters of the estuary of Bilbao were linked to river discharge and over-land climate factors.

Highest warming rates occurred in spring, particularly in May, except for the above-halocline waters of the estuary of Bilbao, likely due to the higher influence of river discharge in these waters. The strongest cooling in August sharpened summer negative trends in the neritic waters of both estuaries. A similar pattern was observed in winter in the above-halocline waters of the estuary of Bilbao. These cooling trends were attributed to external factors of marine and continental origin, respectively. A delay in the coldest annual temperatures and an advance in the warmest ones, which shortened the annual warming period, were the main water temperature phenological changes. Additionally, a sharp increase in May extended the warm period, reflecting a transition from cooler to warmer conditions over the study period.

Water temperature was primarily driven by air temperature but modulated by river flow, which had a cooling impact, particularly in spring and autumn. The relationship between upwelling index and temperature was complex, suggesting the influence of additional factors.