The Sierra Nevada range (Granada, SE Spain) (36°55′-37°15^′ N, 2°31′-3°40^′ W) reaches a maximum elevation of 3482 m a.s.l. The summits experience a high-elevation Mediterranean climate, characterized by warm, dry, and ice-free conditions from June to October. The meteorological station at 2507 m a.s.l. reports an annual mean temperature of 4.4 °C and 700 mm of precipitation, 80 % of which falls as snow between October and April. Sierra Nevada lies between European and African biogeographic regions, about 60 km from the coast, and experiences a semi-arid Mediterranean climate (Zamora and Oliva, 2022). Since 1864 CE, the Sierra Nevada region has warmed by 1.56 °C (0.12 °C per decade) mainly due to increase spring and summer temperature (Sigro et al., 2024). Summer precipitation has declined, particularly from 1975 to 2020 CE (-13.4 % per decade) (Sigro et al., 2024). Sierra Nevada hosts approximately 50 small glacial lakes at elevations of 2800–3100 m a.s.l., formed during the glacial retreat after the last glacial cycle (Castillo Martín, 2009). The geologic substrate in lake catchment basins is composed of slow-weathering, metamorphosed siliceous bedrock mainly mica-schist with graphite and mica-schist with feldspar, devoid of carbonated rocks. These lakes are typically oligotrophic or oligo-mesotrophic, characterized by cold, oxygen-rich waters with low alkalinity and mineralization (Medina-Sánchez et al., 2022). Due to its proximity to Africa, Sierra Nevada lakes are ideal for studying Saharan input effects, as southeastern Iberia is frequently exposed to Saharan dust intrusions (Morales-Baquero and Pérez-Martínez, 2016; Pulido-Villena et al., 2006) and Saharan dust fertilizes in phosphorus, calcium, and alkalizing minerals the oligotrophic lakes that receive it (Morales-Baquero et al., 2006b). Saharan dust deposition has significantly increased in recent decades (Moulin and Chiapello, 2006; Mulitza et al., 2010; Prospero and Lamb, 2003), further influencing ecosystem chemistry and dynamics.

This study focuses on Borreguil Lake (Fig. 1), an oligo-mesotrophic lake at 2980 m a.s.l (37°03^′09.53^′′ N, 3°17^′59.03^′′ W) with a maximum depth of 2.8 m and an area of 0.18 ha (Table S1 in the Supplement). The lake is surrounded by approximately 0.56 ha of alpine meadows, the flora of which is mainly composed of Cyperaceae, Poaceae and Fabaceae, with frequent genera including Carex, Nardus, Festuca and Scorzoneroides (Pérez-Luque et al., 2015). The wetter meadow shows bryophyte species such as Drepanocladus fluitans. The lake freezes annually from October to May, with interannual variability. It is fishless and does not thermally stratify in summer. The lake basin shows permanent inlets and outlets supplying water during the ice-free period. Physicochemical data are available in Jiménez et al. (2018) and Pérez-Martínez et al. (2020b).

To determine the composition of the phytoplankton, several water column samples were taken at the deepest zone in Borreguil Lake. Phytoplankton (100 mL) was collected from a homogenized volume of water sampled with a tube sampler. Phytoplankton samples were immediately fixed in acetic Lugol's solution and the algae species were identified and counted using an inverted microscope following the Utermöhl technique. The planktonic algal community of Borreguil Lake currently consists of four classes, dominated by Chlorophyceae and Cyanobacteria, followed by Chrysophyceae and Cryptophyceae (Fig. S1 in the Supplement). Chlorophyceae is the most diverse group, with 17 taxa. Most of the diatoms are periphytic species (Llodrà-Llabrés et al., 2024).

In the lakes of the Sierra Nevada, the planktonic community coexists with three additional algal communities (Cuesta Linares et al., 2003; Sánchez Castillo and Morales Torres, 1980): (1) a bog-associated algal community comprising Chlorophyceae (particularly Zygnemataceae and Desmidiaceae), along with diatoms and cyanobacteria, some of which are nitrogen-fixing (such as Nostoc, Anabaena, and Cylindrospermum); (2) epilithic biofilms primarily made up of cyanobacteria and diatoms and epipelic diatoms; and (3) algal mats predominantly composed of filamentous Zygnemataceae and cyanobacteria growing at the littoral and/or bottom of the lake.

A slide-hammer gravity corer (Aquatic Research Instruments, Hope, ID, USA) with an inner core-tube diameter of 6.8 cm was used to collect a 26 cm sediment core (referred to as SSBG-21 hereafter) from the lake central area in September 2021. The sediment core was sectioned on-site following Glew et al. (2001) at 0.25 cm intervals for the first 9.25 cm and at 0.5 cm intervals for the remaining sediment with a core extruding apparatus provided by Aquatic research Instruments (http://www.aquaticresearch.com/core_extruding_apparatus.htm, last access: 25 June 2026). For each section, 1 cm³ were immediately taken from the centre of the core, using disposable material, and stored in small plastic polipropilene 5 cm³ vials for pigment analysis avoiding the exposure of the sediment to light. The vials were completely full of sediment to avoid pigment degradation by oxygen and were kept at 4 °C during transportation to the lab. Once in the laboratory, the vials were immediately placed at -80 and at -20 °C after freeze-drying. The remaining sample was extruded into plastic zip-bags, later freeze-dried, and then stored at a desiccator.

Freeze-dried subsamples were analysed for ²¹⁰Pb, ²²⁶Ra and ¹³⁷Cs by direct gamma spectroscopy in the Liverpool University Environmental Radioactivity Laboratory. A total of 18 subsamples between the top of the core and the 14 cm depth sediment were analysed. Sediment ages were estimated from unsupported ²¹⁰Pb activities using the constant rate of supply (CRS) and constant initial concentration (CIC) models (Appleby and Oldfield, 1978). ¹³⁷Cs was used as an additional dating marker, representing fallout emitted during thermonuclear bomb testing (with its peak in 1963) and nuclear accidents such as the Chernobyl accident in 1986.

A total of 51 samples were used for pigment analyses, using freeze-dried sediments stored at -20 °C in the dark. The pigments were extracted using 90 % acetone with a probe sonicator (Sonopuls GM70 Delft, The Netherlands) (50 W, 2 min). The extract was centrifuged (4 min at 3000 rpm, 4 °C), filtered through a Whatman ANODISC 25 (0.1 mm) filter and analyzed with ultrahigh-performance liquid chromatography (UHPLC) following a modification of the method described by Buchaca and Catalan (2007). The UHPLC system used was a Waters Acquity UPLC equipped with a PDA detector (λ 300–800 nm), refrigerated autosampler (4 °C), and upgraded with a Hexa/THF kit and 20 µL loop (Waters, Milford, MA, USA). Analytical separations were performed on an HSS C18 SB column (100 × 2.1 mm, 1.8 µm particle size) at 25 °C. The PDA channel was set at 440 nm for pigment detection and quantification. Peak integration was performed automatically using Empower software (Waters Corporation), which controls the UHPLC system. After sample injection (7.5µL), a constant flow rate of 0.7 mL min⁻¹ was maintained. Pigments were eluted using a linear gradient from 100 % solvent B (51:36:13 methanol : acetonitrile : Milli-Q water, v/v/v, with 0.3 M ammonium acetate) to 75 % B and 25 % A (70:30 ethyl acetate:acetonitrile, v/v) over 3 min, followed by a 0.45 min isocratic hold at 75 % B and a 2 min gradient to 100 % A. Initial conditions (100 % B) were restored in 0.65 min, with a 1 min re-equilibration before the next injection. To address reproducibility, chromatographic performance was evaluated by calculating the relative standard deviation (RSD) for retention times and peak areas across replicate injections. All pigments exhibited RSD values within commonly accepted limits for HPLC analyses (typically ≤ 1 % for retention time and ≤ 2 % for peak area), confirming that the method provides consistent and reliable results under the tested conditions.

Pigments were identified by comparison with a library of pigment spectra obtained from extracts of pure algal cultures from the Culture Collection of Algae and Protozoa (CCAP, Oban, Scotland, UK) and pigment standards (DHI Water and Environment, Hørsholm, Denmark). β-carotene, a ubiquitous pigment present across algal groups, together with a suite of carotenoids with high taxonomic affinity was selected for analysis. These marker pigments included scytonemin (UV-screen pigment in cyanobacteria), alloxanthin (cryptophytes), aphanizophyll (N₂-fixing cyanobacteria), canthaxanthin and echinenone (mainly cyanobacteria and zooplankton), astaxanthin (zooplankton and some chlorophytes), diadinoxanthin (mainly dinoflagellates), diatoxanthin (diatoms), zeaxanthin (chlorophytes, zygnematophytes and cyanobacteria) and lutein (chlorophytes and zygnematophytes). We based our interpretation on carotenoid profiles instead of phorbins because carotenoids are less prone to degradation to colourless compounds (Buchaca and Catalan, 2008). Values for β-carotene were expressed as nmol g⁻¹ DW and used as a proxy for total algal community biomass, whereas concentrations of the ten selected marker pigments were expressed as relative abundances (%) of their summed concentrations (excluding β-carotene). The molecular weights of the different pigments were obtained from Jeffrey et al. (1997).

A total of 51 samples were used for C / N, δ15N and for visible reflectance spectroscopy (VRS) inferred Chl-a and DOCw analyses. The carbon to nitrogen (C / N) ratio is commonly used to identify organic matter sources in lacustrine sediments and also potential N limitation on primary production. Algal-derived organic matter typically has C / N ratios between 4 and 10, while terrestrial vascular plants show C / N ratios above 20. C / N values between 10 and 20 suggest a mix of both sources (Meyers and Teranes, 2001). For elemental analyses of carbon and nitrogen content of the organic matter, a sediment fraction of ∼ 1 g of wet weight (WW) sediment was freeze-dried. To remove inorganic carbonates, samples were acidified in situ by adding 1 M HCl dropwise until effervescence ceased and overnight at 50 °C. The elemental analysis was performed using a Thermo Scientific Flash 2000 coupled to a mass spectrometer at the Scientific Instrumentation Center of the University of Granada (CIC) determining the content of sediment C / N ratio calculated from the mass data and expressed as atomic ratios.

Nitrogen isotope (δ15N) composition was determined from the same bulk sediment organic matter used for the C / N analyses. It was performed on the remaining carbonate-free sediment samples using the elemental analyser Flash HT Plus coupled to IRMS DELTA V Advantage at the CIC.

Sediment samples were analysed for inferred trends in total lake-water organic carbon following the methodology described in Meyer-Jacob et al. (2017) and sedimentary chlorophyll-a including its derivatives (visible reflectance spectroscopy (VRS) inferred Chl-a) were analysed according to Michelutti et al. (2010) at the Paleoecological Environmental Assessment and Research Laboratory (PEARL), Queen's University (Canada). Briefly, prior to analyses, lyophilized sediment samples were sieved (125 µm mesh) to remove the influence of particle size and water content on the spectral signal. Next, VNIR spectra were recorded with a FOSS XDS Rapid Content Analyzer in diffuse reflectance mode. Each sediment sample spectra represents a mean of 32 scans at 2 nm resolution in the wavelength range from 400 to 2500 nm. For VRS-inferred chl-a, the absorbance peak of chlorophyll-a between 650 and 700 nm was considered in relation to calibration samples covering a range of sedimentary chlorophyll-a concentrations, as measured by high performance liquid chromatography (HPLC). Sedimentary VRS chl-a concentrations were inferred using log transformed data from Michelutti et al. (2010) with the equation: chl a + derivatives = EXP (0.83784 ⋅ LN (peak area 650–700 nm) + (-2.48861)). Our chlorophyll-a inferences, which include all algal and cyanobacterial groups, also include all chlorophyll isomers and its major derivatives (pheophytin and pheophorbide) (Michelutti et al., 2010; Michelutti and Smol, 2016). For inferred water TOC, an orthogonal partial least squares regression model between VNIR spectral information (400–2500 nm) and measured surface-water TOC concentrations in 345 lakes was used to reconstruct lake-water TOC levels from sediment samples (Meyer-Jacob et al, 2017). In our case, lake-water TOC is considered virtually equivalent to DOC in lake water due to the dominant contribution of DOC in the Sierra Nevada lake's organic carbon pool (Mladenov et al., 2008), a pattern also observed in lakes with similar characteristics (Meyer-Jacob et al., 2019). From this point onward, analyzed TOC in water will be referred to as DOCw, as we consider relative changes in TOC to reflect relative changes in DOC in the water.

The climate data that was used for statistical analysis to explore relationships with the pigment matrix was obtained from Sigro et al. (2024), who developed a high quality daily climate data base specifically for the Sierra Nevada Mountains (http://www.c3.urv.cat/climadata.php, last access: 25 June 2026). We used the mean annual temperature anomaly (MATA) and annual precipitation anomaly (APA). Seasonal mean temperature anomaly (MTA)- springMTA, summerMTA, fallMTA, and winterMTA-, and seasonal precipitation anomaly (PA)- springPA, summerPA, fallPA, and winterPA-, were used to correlate with the annual climate data through Pearson's correlation analysis. Data were tested for normality before the analysis. This database has been quality-controlled and homogenized using CLIMATOL (https://climatol.eu/, last access: 25 June 2026). Climate data was available for the period ∼ 1894–2020 years CE.

The intensity of Saharan dust emission and transport has been linked to the Sahel drought (Chiapello et al., 2005; Moulin and Chiapello, 2004). Furthermore, Jiménez et al. (2018) demonstrated that the Sahel Precipitation Index (SPI) can be utilised as a predictor of the transport and intensity of Saharan dust events in Sierra Nevada, reflecting atmospheric P and Ca deposition trends in this region. These indices exhibited strong correlations with the zirconium-to-aluminium (Zr / Al) ratio – a proxy for Saharan dust deposition—measured in a sediment core from one of the Sierra Nevada lakes, as well as with Saharan calcium concentrations in an ice core from the French Alps, which are indicative of Saharan dust events (Preunkert and Legrand, 2013). Consequently, SPI was employed for this purpose in the present study. The Sahel precipitation index (SPI), extending back to 1900 years CE, was accessed from the NOAA Physical Sciences laboratory (https://psl.noaa.gov/data/timeseries/month/SAHELRAIN/, last access: 25 June 2026). It provides a standardized rainfall index data for the Sahelian zone of northern Africa. Negative SPI values indicate severe drought conditions, significantly enhancing dust emission and atmospheric loading.

The annually resolved climate and dust metrics were averaged over the period of accumulation for each dated interval, i.e. the data were binned in exactly the same intervals as those obtained from the sediment core, thereby integrating the instrumental data with the paleolimnological data. Final weighted data are only available for the period 1905–2020 years CE (corresponding to samples 1–22).

Before statistical analyses were applied, data were transformed to reduce the asymmetry of the distributions and to standardise them, i.e. remove incomparabilities in the total variance. The pigment variables (expressed as relative abundance) were square root transformed (Hellinger transformation) to minimize the impact of dominant species and handle the sparse nature of the dataset (presence of many zeros). The β-carotene pigment was excluded from the pigment matrix because it was considered to be a proxy for total algal biomass (Buchaca et al., 2019; Zhang et al., 2019). The explanatory biological and paleo-environmental variables ((C / N ratio, DOCw, chlorophyll-a and β-carotene) were logarithmically transformed followed by a relative scale transformation (y′=yi/ymax) to remove asymmetry and incomparabilities in the total variance. The δ15N values were used untransformed as these met the conditions of normality and homoscedasticity. Finally, the climate and dust metric variables (MATA, APA, and SPI) were relative scale transformed. All transformed variables met the assumptions of normality and homoscedasticity for the application of parametric statistics

The number of significant pigment zones (CONISS; constrained hierarchical clustering; Grimm, 2011) was determined using the broken stick model (Bennet, 1996). Principal component analysis (PCA) was performed on the matrix of individual pigments to summarize the major patterns of variability in pigment assemblages into a few axes. Previously, detrended correspondence analysis (DCA) results (gradient length < 2 SD) indicated that linear ordination was most appropriate (Legendre and Birks, 2012).

Redundancy analysis (RDA) was conducted to identify the climate and atmospheric explanatory variables of algal assemblage changes. Generalized linear models (GLM) analysis was conducted to explore climate and atmospheric drivers of the main variables reflecting primary producer abundance (VRS-inferred Chl-a and β-carotene). Both analyses were performed for the period during which atmospheric and climate data were available (∼ 1905–2020 years CE; 22 samples). The explanatory variables used in both analyses included annual temperature (MATA), precipitation data (APA) and SPI. Seasonal temperature and precipitation data were not included in these analyses to reduce the number of explanatory variables. Pearson correlations between annual and seasonal climate variables were performed.

The RDA (with forward/backward selection) was performed to identify the statistical independence and relative strength of each of the explanatory variables (MATA, APA and SPI) of the pigment assemblage changes, using Monte Carlo permutation tests (999 unrestricted permutations) with a significance level of p<0.05. Akaike's information criterion (AIC) (Burnham and Anderson, 2002) was used to extract the environmental variables that significantly increased the amount of explained variation of pigment data, thereby optimising the global AIC. AIC was selected because it helps identify the most parsimonious model for explaining the variation in algal community assemblage. Significant environmental variables were incorporated into the final model and subjected to RDA once again.

In the GLM analyses, the optimum model was selected using Akaike's information criterion adjusted for sample size (AICc; Burnham and Anderson, 2004). Models with an AICc difference of less than 2 compared to the lowest AICc were considered the best models and statistically equivalent. The contribution of each variable to the final model was determined by assessing its significance and percentage of variance explained. Residuals of the final models were examined to check for normality of data and absence of over-dispersion.

Ordination and model selection analyses were performed using the vegan (Oksanen et al., 2022) and the MuMIn (Multi-Model Inference; Bartón, 2025) packages for the R software environment, respectively. In all the analyses the variance inflation factor (VIF) was used to eliminate highly correlated variables before applying analyses. All the explanatory variables that yielded VIFs > 5 were eliminated in the analysis due to the high degree of collinearity.

The Borreguil lake record consists of 26 cm of peaty (primarily bryophytes) and silty clays. The total activity of ²¹⁰Pb (Fig. 2A) reached levels close to equilibrium with the supporting ²²⁶Ra at a depth of 6–7 cm in the core SSBG-21. According to the CRS and CIC models (Appleby and Oldfield, 1978), the concentrations of unsupported ²¹⁰Pb below 1 cm decrease exponentially with depth (unshown data). This indicates a relatively uniform accumulation of sediments over much of the time period spanned by the core. The relatively uniform concentrations observed in the top 1 cm may be attributed to a recent slight increase in sedimentation rates. Excluding the top 1 cm, the average sedimentation rate is calculated to be 0.0098 ± 0.008 g cm⁻² yr⁻¹ (or 0.041 cm yr⁻¹) (Fig. 2B). Additionally, ¹³⁷Cs concentrations showed a distinct peak at a depth of 1.5–1.75 cm (Fig. 2A), likely reflecting fallout from the Chernobyl accident in 1986. The reliable dating period for the core extends from 6.63 cm to the present, covering approximately ∼ 1850 to 2020 years CE, with an estimated mean temporal resolution of 5 years per sampling interval. The ²¹⁰Pb chronology only constrains the upper part of the sediment sequence, whereas ages assigned to deeper intervals rely on extrapolation of sedimentation rates beyond the dated section. This approach implicitly assumes constant sedimentation through time, negligible compaction, and the absence of hiatuses or sediment mixing, assumptions that we acknowledge explicitly given the lack of independent chronological control below the ²¹⁰Pb-supported interval. Therefore, the inferred basal age (∼ 400 years) and the temporal interpretation of deeper sediments should be taken cautiously.

Throughout the whole core, the pigments associated with Chlorophyceae and Zygnematophyceae (lutein and zeaxanthin) and those associated with diatoms (diatoxanthin) were predominant (Figs. 3, S2). Both lutein and zeaxanthin were attributed to an origin in green algae, given that their concentration profiles change in parallel (Fig. S2). Other significant pigments are those associated with cyanobacteria (aphanizophyll, echinenone and scytonemin) (Table S2). The least abundant pigments were diadinoxanthin (a marker pigment for Dinophyceae) and alloxanthin (a marker pigment of cryptophytes), which remained below 5 % abundance throughout the profile (Fig. 3). In addition to the aforementioned pigments, astaxanthin and canthaxanthin, typically associated with zooplankton, were also detected (Fig. 3, Table S2). Figure S2 and Table S2 show the pigment profiles expressed as concentrations (nmol PGM/g DW) and all the pigments identified together with their taxonomic affinities, respectively. In shallow high-mountain lakes, pigment records are often affected by intense photo-oxidative processes in the water column. To evaluate whether such processes influence downcore variability, we assessed pigment preservation using the Chl-a/a-phorbins index (Fig. S2), which remains remarkably constant throughout the sediment sequence. This indicates that, while overall pigment degradation may be high, preservation conditions have not changed substantially through time, and therefore relative stratigraphic trends are unlikely to result from variable degradation.

In these systems, sedimentary pigment assemblages reflect an inherent integration of primary producers from both the water column and benthic biofilms developing at or near the sediment surface. Previous studies in Pyrenean lakes have demonstrated that lake depth largely controls the relative importance of planktonic versus benthic pigment signals, with shallow lakes showing a stronger benthic imprint and short transport distances between production and deposition (Buchaca and Catalan, 2007, 2008). Thus, pigment-based reconstructions in Borreguil Lake should be interpreted as recording relative changes in integrated photoautotrophic communities.

Results are described using a stratigraphically constrained cluster analysis based on pigment composition (excluding chl-a and β-carotene) (Fig. 3). According to the CONISS analysis the sediment pigment sequence can be divided in five significant zones of change (Fig. 3) with changes at ∼ 1970, ∼ 1843, ∼ 1804, and ∼ 1690 years CE.

Zone 1 (between 1600 and 1690): This period is characterised by the highest levels of lutein, as well as relatively high levels of diatoxanthin and zeaxanthin. This indicates a period of dominance by diatoms and green algae. Other pigments present in this zone are astaxanthin and canthaxanthin, which show particularly highly abundances during this interval. In contrast, VRS-inferred Chl-a and β-carotene, both indicators of overall algal biomass, show the lowest values recorded in the entire sequence.

Zone 2 (between 1690 and 1804): A decrease in lutein accompanied by an increase in diatoxanthin indicates a shift toward greater dominance of diatoms over green algae. The diversification of the algal community is notable during this period. Pigments associated with cyanobacteria, such as echinenone and aphanizophyll, appeared for the first time, although they were present in low concentrations. Notably, echinenone production began to increase around 1750 CE and remained relatively stable at around 5 % until the top of the core, whereas aphanizophyll showed low abundances between 1750 and 1800 CE before disappearing. Scytonemin, a pigment associated with the UV protection in cyanobacteria, also appeared at this time, albeit at very low relative abundances. Diadinoxanthin, a marker pigment for Dinophyceae, was first recorded around 1750 CE. Astaxanthin and canthaxanthin exhibited their highest values during this period. These pigments were most abundant from the bottom of the core to around 1750 CE, after which there was a progressive decline to the top of the core. Indicators of algal biomass (VRS-inferred Chl-a and β-carotene) and the C / N ratio show the lowest values on record during the first half of the period (until around 1750 CE), but increase during the second half, exhibiting significant fluctuations with several peaks and troughs.

Zone 3 (between 1804 and 1843): In the algal assemblage, an increase in green algae (indicated by lutein) is observed at the expense of diatoms (represented by diatoxanthin). All pigments detected in the previous period remain in this one, with the exception of aphanizophyll, which disappears. This interval is also characterised by high peaks in VRS-inferred Chl-a, β-carotene, C / N ratios and DOCw.

Zone 4 (between 1843 and 1970): There are few changes observed in the algal assemblage, which largely appears to be a continuation of the previous period, except for the reappearance of aphanizophyll. The concentration of diadinoxanthin (a marker pigment for Dinophyceae) declined from 1870 CE onwards, while alloxanthin first appeared around 1900 CE, maintaining low concentrations until the top of the core. In contrast, more pronounced changes are evident in other paleoenvironmental variables (Fig. 3). C / N and DOCw show fluctuating values, whereas ∂15N decreases slightly during this period. VRS-inferred Chl-a and β-carotene show low values until approximately 1900 CE, after which they show a marked increase.

Zone 5 (between 1970 and 2020): This period exhibits the most pronounced changes in the entire record. The highest values recorded for VRS-inferred Chl-a and β-carotene are observed during this interval. The abundance of diatoxanthin declined from 1970 towards the top of the core, whereas the abundance of lutein and zeaxanthin increased. This indicates a shift in dominance from diatoms to green algae. These post-1970 changes coincide with a significant increase in temperature (MATA) and Saharan dust deposition (SPI) in the Sierra Nevada (Fig. 4).

Echinenone production maintains a stable abundance of around 5 % throughout the core (Fig. 3). This suggests that either the primary producer (cyanobacteria) or the transformation pathway (the degradation of other pigments into echinenone) is constantly operating, regardless of climate and environmental factors. Aphanizophyll, which is produced by nitrogen-fixing cyanobacteria, shows the highest values (10 %) towards the top of the core (Fig. 3). This increase coincides with a steep decline in δ15N values, suggesting NOx transformation and fixation by N₂-fixing cyanobacteria. Aphanizophyll exhibits a strong negative Pearson correlation with δ15N (r=-0.84), and the two variables display opposite profiles throughout the entire core (Fig. 3). A steep decline in C / N ratio values are also recorded at this period. Scytonemin, a pigment associated to protection against UV light, persisted at 2 % until the top (Fig. 3). DOCw values, a good estimator of UV transparency (Laurion et al., 2000), show an increase since ∼ 1975. The least abundant pigments were diadinoxanthin (a marker pigment for dinoflagellates) and alloxanthin (a marker pigment for cryptophytes), with an abundance of less than 5 % throughout the record (Fig. 3).

Regarding climate variables, Mean Annual Temperature Anomalies (MATA) increased steadily from 1905 to 1970, while Annual Precipitation Anomalies (APA) decreased slightly until 1920 and then increased slightly (Fig. 4). The Sahel Precipitation Index (SPI) shows relatively high values (less dust in the atmosphere) with no significant changes. The SPI record (Becker et al., 2013) indicated a predominantly wet period (positive values) in the Sahel from 1900 years CE to approximately 1970 years CE (Fig. 4).

A significant increase in temperature (MATA) and Saharan dust deposition (SPI) in the Sierra Nevada (Fig. 4) was observed post-1970. The warming in MATA was mainly driven by summer and spring temperatures (Sigro et al., 2024). This trend was also observed in the correlation analysis, which yielded a correlation factor of 0.97 and 0.94 between MATA, and summer and spring mean temperatures, respectively (Fig. S3). Mean annual (Fig. 4) and seasonal mean temperatures (Fig. S4) show negative anomalies prior to the 1940s and there were negative values around the 1970s. Regarding the precipitation anomalies, a downward trend was observed during the period 1975–2020 years CE, which exhibited a significant negative trend of -12.9 % per decade in summer precipitation (Fig. 4; Sigro et al., 2024). The SPI record indicated a relatively stable and dry period (negative values) from around 1970 years CE to the present, with the lowest values observed during the 1980s–1990s years CE (Fig. 4). Substantial negative SPI values indicate severe drought conditions, significantly enhancing dust emission and atmospheric loading (Jiménez et al., 2018).

PCA of the 51 core samples explained 73.91 % of the total pigment variation (45.84 % for axis 1 and 28.07 % for axis 2) (Fig. 5). Axis 1 showed two periods: a decrease until ∼ 1670 years CE and an upward trend after that, with some fluctuations (Fig. 3). Axis 2 tracked an increasing trend around ∼ 1705 to ∼ 1780 years CE and ∼ 1850 to ∼ 1900 years CE, followed by a sharp decline after ∼ 1970 years CE (Fig. 3). Axis 1 was linked to echinenone, aphanizophyll, diadinoxanthin, and alloxanthin in the positive region, and Crustacea-related pigments (astaxanthin, canthaxanthin) in the negative region (Fig. 5). Axis 2 was associated with zeaxanthin, lutein, and diatoxanthin, with scytonemin showing no clear relationship to either axis.

Prior to all the analyses, VIF > 5 analyses were performed on the explanatory variables and no variables were discarded. The resulting RDA model produced an R2 adjusted value of 0.468, with MATA and SPI identified as the primary drivers of pigment composition across the period ∼ 1905–2020 years CE (Fig. 10; Table 1).

To explore the relationships between VRS-inferred Chl-a and β-carotene and annual climate and atmospheric variables, GLM analyses were undertaken (Table 2). MATA was selected as the main explanatory variable for VRS-inferred Chl-a and β-carotene, being APA secondary explanatory variable for β-carotene.

The long-term changes in primary production and/or algal biomass throughout the core was assessed by the combined analysis of VRS-inferred Chl-a and β-carotene, indicators of algal biomass. Both variables show similar trends throughout the core. From the bottom of the core to around 1750 years CE, both remained stable and low, reaching their lowest levels. This low biomass period may be linked to low temperatures during the Little Ice Age (LIA), as shown by Jiménez-Moreno et al. (2023) for Río Seco Lake in Sierra Nevada and from Oliva et al (2018) for the Iberian Mountains. Between ∼ 1750 and 1800 CE, concentrations of VRS-inferred Chl-a and β-carotene fluctuated, reaching peak levels, despite this period coinciding with the lowest temperatures of the Little Ice Age (LIA) in the Sierra Nevada, according to Jiménez-Moreno et al. (2023). Although we do not have a clear explanation for this increase, we hypothesize that this phenomenon may be attributed to reduced herbivory pressure (reduction of astaxanthin and canthaxanthin), as the development of herbivore populations is constrained by the brief duration of ice-free periods. However, Oliva et al. (2018) indicate that the period 1760–1800 CE was a time of climate extremes in the Iberian Mountains, which could explain the fluctuations observed in VRS-inferred Chl-a and β-carotene in the Sierra Nevada during that period.

At the end of the LIA (1800–1840 years CE), a temperature peak at the Sierra Nevada summit was recorded (García-Alix et al., 2020; Jiménez-Moreno et al., 2023). The observed algal biomass increase at the 8–10 cm depth of our record was likely related to this temperature peak (although the uncertainty of our dating prior to 1850 should be noted) and is consistent with trends in other mountain lakes (Hu et al., 2014; Huo et al., 2022; Lami et al., 2010; Oleksy et al., 2020). Moreover, the C / N peak ∼ 1840 years CE suggests an influx of external organic matter, potentially driven by thaw-induced transport. The main changes in VRS-inferred Chl-a and β-carotene occurred from ∼ 1920 years CE onwards, when a striking increase in both variables was observed. This biomass increase has no precedent in the last 400 years in the lake and according to GLM analysis (Table 2), climate factors – particularly rising temperatures – were identified as the main drivers of these changes. In alpine lakes, higher temperatures and lower precipitation may extend the ice-free period (Anderson et al., 1996; Rogora et al., 2018), increasing the growing season and resulting in a higher accumulation of annual algal biomass. Moreover, longer ice-free seasons in alpine lakes boost light availability, water temperature, and solute inputs through snowmelt and weathering (Preston et al., 2016; Sommaruga-Wögrath et al., 1997) all enhancing biological production (Douglas and Smol, 2010; López-Merino et al., 2011). These facts were observed across the Sierra Nevada lakes (Jiménez et al., 2018) and across arctic and alpine lakes (Adrian et al., 2009; Rühland et al., 2008, 2015). We therefore suggest that temperature controls how much biomass is produced in the lake. This means that the big rise in temperature and fall in rain in summer in Sierra Nevada (Fig. S4; Sigro et al., 2024) could be the reason for the large and unprecedented increase in the amount of algae since the 1960s years CE to the present.

Additionally, Saharan dust may increase nutrients in Borreguil Lake, boosting algal production. Sierra Nevada has experienced an increase in the frequency of dust events particularly since the 1970s years CE. Longer ice-free periods may have further amplified dust exposure, boosting nutrient availability and promoting algal growth (Korbee et al., 2012; Saros et al., 2005; Yang et al., 1996). Saharan dust inputs influence the structure and composition of biological communities in Sierra Nevada lakes. For example, phosphorus inputs enhance algal biomass, leading to higher chlorophyll-a concentrations (Carrillo et al., 1990; Morales-Baquero et al., 2006b). Similar Saharan dust fertilization effects have been observed in the Pyrenees (Camarero and Catalan, 2012). Recent results from Borreguil Lake show that, in 2022 years CE, following a massive influx of Saharan dust in Sierra Nevada, there was a significant increase in phytoplankton water column density and a shift in the lake algal composition, with a marked increase in the percentage of cyanobacteria (Fig. S1). Thus, the increase in algal biomass in Borreguil Lake since the 1970s can be attributed primarily to temperature and secondarily to Saharan P input.

Since the early 1900s years CE, anthropogenic nitrogen deposition has become increasingly evident in lake sediment cores from high-latitude regions of the American Northern Hemisphere (Holtgrieve et al., 2011), particularly during the latter half of the 20th century (around the 1970s), coinciding with the “Great Acceleration” of global environmental change (Steffen et al., 2007). It is conceivable that nitrogen deposition has influenced the increase in algal biomass observed in Borreguil Lake and the sharp decline in the C / N ratio post-1970 CE could reflect this N enrichment in the lake. However, nitrogen deposition in Sierra Nevada is lower than in other Mediterranean regions (Morales-Baquero et al., 2006b, 2013) and heavily industrialized areas of Central Europe (Holland et al., 2005), with a mean total nitrogen deposition (dry and wet) of 115.2 mg m⁻² d⁻¹ between 2000–2002 years CE. Moreover, the parallel dynamics between the C / N ratio and the δ15N ratio, as well as the increase in Aphanizophyll, suggest that N enrichment in the lake is associated with an increase in N₂-fixing cyanobacteria. Therefore, it is unlikely that the increase in algal biomass in Borreguil Lake was mainly influenced by atmospheric N deposition.