Imprints of the Little Ice Age and the severe earthquake of AD 2001 on the aquatic ecosystem of a tropical maar lake in El Salvador

Lithological description

The sediment core APA 1.2 was transported to the laboratory and stored in cool conditions. The core was then cut in half length-wise and the sediment profile photographed, described (colour, texture). One half of the core was used for dating, whereas the other was sub-sampled at 2 cm-intervals for multiproxy analysis.

Geochronology

The core chronology was established using radionuclide and radiocarbon dating. Dried sediment samples were analysed for ²¹⁰Pb, ²²⁶Ra and ¹³⁷Cs by direct gamma assay in the Environmental Radiometric Facility at University College London, using an ORTEC HPGe GWL series well-type coaxial low-background intrinsic germanium detector. ²¹⁰Pb was determined via its gamma emissions at 46.5 keV, and ²²⁶Ra by the 295 keV and 352 keV gamma rays emitted by its daughter isotope ²¹⁴Pb following 3 weeks storage in sealed containers to allow radioactive equilibration. ¹³⁷Cs was measured by their emissions at 662 keV (Appleby et al., 1986). The absolute efficiencies of the detector were determined using calibrated sources and sediment samples of known activity. Corrections were made for the effect of self-absorption of low energy gamma rays within the sample (Appleby et al., 1992). ²¹⁰Pb activities were calculated by subtracting ²²⁶Ra activity (as supported ²¹⁰Pb) from total ²¹⁰Pb activity. The use of the CIC (constant initial concentration) model was precluded by the non–monotonic variation in unsupported ²¹⁰Pb activities. ²¹⁰Pb chronologies were calculated using the CRS (constant rate of ²¹⁰Pb supply) dating model (Appleby, 2001). Radiocarbon dating of extracted pollen by accelerator mass spectrometry (AMS) was performed for the bottom layer (34–32 cm) in a commercial laboratory (Beta Analytic). The ¹⁴ C date was calibrated using the curve IntCal20 (Reimer et al., 2020). The pollen extract was prepared from fresh sediments (ca. 4 g) according to the standard method (Berglund and Ralska-Jasiewiczowa, 1986).

Elemental composition and geochemical proxies

The bulk sediment elemental composition was measured on the upper 30 cm of the sequence at 2-cm resolution to determine total carbon (TC), total inorganic carbon (TIC), total organic carbon (TOC), and total nitrogen (TN). TC and TN were determined by an elemental analyser (LECO TruSpec CHN Macro). The C/N molar ratios were calculated from TC and TN. For the determination of the TIC content, CO₂ was evolved during hot (70°C) phosphoric acid (H₃PO₄) treatment and quantified conductometrically. The TOC content was calculated by subtraction of TIC from TC. Organic matter and carbonate content as well as the residual (total mass minus CaCO₃ and organic matter) were determined by thermogravimetric loss-on-ignition (LOI) analyses with ignition at 550°C and 950°C, respectively (Heiri et al., 2001).

Concentrations of PO₄, S, Ca, Mg, Fe, Mn and Ti were measured by ICP-OES spectrometer (Perkin Elmer Optima 2100 DV) according to DIN EN 13346 (Anonymous, 2001). The pre-treatment of sediments included sieving (2 mm sieve) and separating into two fractions (<2 mm to ⩾63 and <63 µm). Subsequently, samples were dried, weighed, ground into powder, and measured.

A portable energy-dispersive X-ray fluorescence spectrometer (EDXRF) (Analyticon NITON XL3t) was utilized to analyse concentration of K, Zr, Si, Ti and Al. Each sample (~4 g) was dried at 55°C, homogenized and ground into powder. Thereafter, sediments were transferred into plastic cups and covered with a mylar foil, placed on the EDXRF, and analysed for 120 s. The calibration was conducted using the certified reference material (CRM): GBW07312 (stream sediment), LKSD-2 and LKSD-4 (lake sediments; Lynch, 1990). PO₄ and S were determined to follow changes in lake productivity (Boyle, 2001), while elemental concentrations of Ca, Mg, K, Si, Al, Ti and Zr were measured in order to infer detrital influx (Rothwell and Croudace, 2015). Furthermore, K/Zr and K/Ti ratios were calculated to estimate grain-size variability throughout the core (Chawchai et al., 2016; Kříbek et al., 2017). Results of Fe and Mn measurements, expressed as their molar ratio (Fe/Mn), were used to assess changes in redox conditions (Boyle, 2001).

Spearman correlation analysis was used to identify correlations among geochemical and lithological parameters. Statistical analysis on the z-scored data was run using R software (R Core Team, 2020) and significance of correlation was assessed at p < 0.05.

To identify the mineralogical composition of littoral and profundal sediments, powder X-ray diffraction analysis (PXRD) was used. Sediment samples were ground into powder and then analysed using a Bruker AXS D8 Advance X-Ray Diffractometer. The PXRD results were evaluated using the EVAluation software (EVA) of Bruker.

Biological proxies

For the analyses of biological proxies, the sediment sequence was sampled at 2 cm intervals and further divided into subsamples for each proxy.

Cladocera analysis was carried out following Frey’s (1986) method. Two cm³ of wet sediment were heated in KOH solution (10%) to remove organic matter. The residue on a 38 µm mesh sieve was backwashed and used as the sample with the volume brought up to 5 mL with distilled water. The samples were then dyed with safranine. Three to four slides (0.1 cm³ each) were analysed for each sample at 100×, 200× and 400× magnification. Identification was made following Wojewódka et al. (2020a, 2020b). The most common cladoceran remains, such as head shield, shell, postabdomen were identified, and at least 70 individuals were counted for each sample (Kurek et al., 2010). The cladoceran abundance was expressed as a percentage of the sum of total individuals counted.

The procedure for diatom analysis was adapted from a standard method (Battarbee et al., 2001): 0.4 g wet sediment was treated with 37% HCl and 30% H₂O₂ and heated at 70°C to remove carbonates and to oxidize organic matter, respectively. Samples were washed with distilled water during preparation to dilute remnants of peroxide and acids. Naphrax^® was used as a mounting medium to make permanent slides for light microscope (LM) analyses. Slides were analysed using a Leica DM 5000 B LM with Differential Interference Contrast under oil immersion at 1000 × magnification. A minimum of 400 valves was counted in each sample. Diatom concentration was determined according to Battarbee and Kneen (1982) and identification was based on standard identification literature together with relevant taxonomic publications (e.g. Hofmann et al., 2011; Krahn et al., 2018, 2021). Species relative abundances were expressed as percentages of the total diatom individuals counted.

For chironomid analysis, 3 g of sediment were deflocculated in warm KOH solution (10%) for 20 min and then backwashed on a 90 mm sieve (Walker and Paterson, 1985). Chironomid head capsules were hand-picked and permanently mounted in Berlese mounting medium. All head capsules were picked from each sample, ranging from 21 to 103 individuals. Taxonomic identification was performed under a compound microscope at up to 400 × magnification, with reference to Hamerlík and Silva (2020). Species abundances were expressed as percentages of the sum of total remains counted.

For testate amoeba analysis, 1 cm³ of sediments was backwashed on a sieve with 63 µm mesh. The resulting fraction was then analysed under a ZEISS Stemi 508 stereomicroscope, and tests were extracted using a fine brush (Ellison and Ogden, 1987). Taxonomic identification was made using an optical microscope and a Scanning Electron Microscope, following relevant literature such as Kumar and Dalby (1998), and Sigala et al. (2016). Results were expressed as the number of individual tests per 1 cm³.

Three grams of wet sediments were used for ostracod analysis. Sediments were initially disaggregated with a solution of 3% H₂O₂ for 12 h and then flushed with distilled water and sieved through 63 µm mesh sieves. Ostracods were picked using a stereomicroscope and adult and juvenile valves were counted. Valves were stored in micropalaeontological slides. Species identification was conducted in valves of adult specimens only and following available literature (Cohuo et al., 2017; Pérez et al., 2013). Results were expressed as the number of specimens per 1 g of dry sediment.

Due to a high variation in sedimentation rate throughout the core, the total abundances of aquatic biota were expressed as fluxed, that is, as number of individuals per cm²yr⁻¹. Data were plotted with C2 version 1.7.7. (Juggins, 2007).

Lake Apastepeque limnological conditions

In 2013, the water of Lake Apastepeque was alkaline (pH = 8.6) with a low conductivity (ca. 100 µS cm⁻¹). HCO₃^– was the dominant anion, followed by Cl^–. The dominant cations were Ca²⁺ and Mg²⁺. Moreover, a high concentration of SiO₂ (ca. 28 mg L⁻¹) (Table S1) was observed in the lake water. Water transparency was ca. 6 m, and the lake had a low content of PO₄³⁻ (<0.01 mg L⁻¹) and NO₃⁻ (0.4 mg L⁻¹).

Core chronology

Total ²¹⁰Pb activities reach equilibrium with the supported ²¹⁰Pb at a depth of ca. 15 cm in the core. There is a decline in unsupported ²¹⁰Pb activities in the top 6 cm, and little net decline from 6 to 14 cm (Supplemental Figure S2, available online), suggesting changes in sedimentation rates. ¹³⁷Cs was detected from 2 to 15.5 cm. The ¹³⁷Cs activity versus depth shows a peak at 6.3 cm (Supplemental Figure S2, available online), but as the activities were low, confidence of using the peak for dating is not high. The low caesium activities put into question the real presence of the ¹³⁷Cs peaks, and cause a mismatch between the peak in ¹³⁷Cs and the ²¹⁰Pb record. Therefore, ¹³⁷Cs dating was excluded from further consideration.

²¹⁰Pb chronologies were calculated using the CRS (constant rate of ²¹⁰Pb supply) dating model (Appleby, 2001). Radiometric chronologies and accumulation rates of the sediments are given in Supplemental Table S2 and shown in Supplemental Figure S2a, available online. Sediment core dating shows that the top 13.3 cm of the core covers approximately the last 100 years. Sedimentation rates show a slight increase from 13.3 to 8.3 cm (from 1910s until the 1950s) and peak at 0.1 g cm⁻² yr⁻¹ (0.272 cm yr⁻¹) at 8.3 cm. Sedimentation rate at 13–13.3 cm was lowest with 0.0119 g cm⁻² yr⁻¹ (0.028 cm yr⁻¹).

The lowermost sample (34–32 cm depth) was solely used for dating and yielded a conventional radiocarbon age of 520 ± 30 years BP based on extracted pollen. Calibration yielded an age of AD 1327–1349 (7.9%) or AD 1395–1444 (87.5%) with 95% probability (Supplemental Table S3, available online). The age of the sequence between 34 and 13.3 cm was interpolated by linear regression and using the date calculated by weighted average (AD 1413) for the bottom layer (34 cm). Results from both ²¹⁰Pb and ¹⁴C were used to create the age-depth model (Supplemental Figure S3b, available online). It should be stressed that interpolated age data are affected by errors resulting from assumptions about linear changes of sedimentation rate. Therefore, ages given for sediment layers between the bottom of the core and 13.3 cm below lake floor are approximated.

Lithology and geochemical characteristics

The sediments are almost homogeneous with a fine-grained texture (clay) and dark brown colour. Sediments have a low carbonate content (<2%), while organic matter content varies from ca. 8% to 18%. The vertical profiles of organic matter and carbonates display similar trends. The residual mineral matter content ranges from ca. 80% to 92% (Supplemental Figure S4, available online) and mainly consists of fine-grained siliciclastic material. The alkaline metals (K, Ca, Mg) show inverse trends to TOC, PO₄, organic matter and carbonate content. TOC varies slightly between 4.8% and 7%, while C/N ratio ranges between 13 and 15 throughout the core.

Correlation analysis displays two main groups of parameters positively correlated with each other at the p-value of 0.05 (Supplemental Figure S5, available online). The first group encompasses alkaline metals (Ca, Mg, K), K/Ti and K/Zr. Of these, Ca and Mg are strongly positively correlated (>0.75) with residual mineral matter and negatively with carbonates (<−0.75). The second group is composed of organic matter, carbonates, TOC, TN, PO₄, S and Fe. Elements from this group are moderately or strongly correlated (>0.5). Particular elements between these two groups, such as Ca, Mg, K (group 1) and TOC, TN, PO₄, S (group 2), are negatively correlated (Supplemental Figure S5, available online).

The highest sediment moisture (up to 80%) is detected at the bottom of the core (30–24 cm), and corresponds to high contents of organic matter (up to 16%), highest TOC and low carbonate content (1.4–1.5%, Supplemental Figure S4, available online). Between 30 and 24 cm, slightly lower ratios of C/N (13.3–13.6), higher TOC (6.3–7%) and TN (0.5–0.6%) are observed in comparison to the overlying sediment layer (24–14 cm). Furthermore, lower K/Zr (42‒50) and K/Ti values (0.8–0.9) as well as high PO₄ value are documented. Lowest organic matter values (7.4–10.5) are recorded between 24 and 14 cm, coinciding with low values of carbonate (<1.4%), PO₄ (up to 0.19%), and TOC (up to 6.2%). Concentrations of alkaline metals such as K, Ca, Mg increase up to 0.6%, 1.6% and 0.8%, respectively (Supplemental Figure S4, available online). At the same time, the concentration of residual mineral matter in the sediments also rises. Zr and Si have their lowest contents in the layers between 24 and 18 cm. Sediments at the depths of 14–2 cm are characterized by gradually increasing concentrations of S, PO₄, TOC and carbonates along with gradually decreasing Ca, Mg and K content. The uppermost part of the sediment core (2–0 cm) is marked by the highest content of organic matter (>18%), S (0.52%), high Fe/Mn (158) and lower values of K/Zr (50.7) and K/Ti ratio (0.9).

XRD analysis of littoral sediment samples includes the following mineral phases: plagioclase (likely andesine), montmorillonite, quartz, and amphibole (magnesiohornblende), diopside, forsterite (Supplemental Figure S6, available online). Profundal sediments are composed of plagioclase (likely andesine), montmorillonite (bentonite), halloysite, kaolinite, forsterite, quartz, diopside, microcline (Supplemental Figure S6, available online). All of these phases are silicate minerals that include Ca and/or Mg, Na, Al, sometimes Fe, and K (microcline and halloysite).

Changes in aquatic fauna and flora assemblages

Lake ecosystem responses to the Little Ice Age (zones I and II, 30–14 cm, 16^th–19^th centuries)

The Little Ice Age (LIA) has been identified as a phase of fluctuating climatic conditions expressed by temperature decrease and changes in humidity due to an alteration of precipitation and evaporation (Cuna et al., 2014; Hodell et al., 2005; Lozano-García et al., 2010; Winter et al., 2000). The onset of the LIA in Mexico and Central America has mostly been dated from the beginning of the 15^th century in the lowlands and middle elevation areas (Hodell et al., 2005; Lozano-García et al., 2010; Pérez et al., 2010; Rodríguez-Ramírez et al., 2015; Wu et al., 2017), although by 1360 AD, the effects of the LIA were already observed in high mountain areas of central Mexico (Cuna et al., 2014).

The LIA in the northern Neotropics is recognized as an environmental disruptor of ecological stability of aquatic systems, for example being associated with species turnovers recorded in lake sediments. This period was characterized by the presence of cold-water zoo- and phytoplankton species (Cuna et al., 2014). In Lake Apastepeque, LIA effects on the biological community resulted in a relatively low diversity and frequent species turnovers (Figure 2). Biota and geochemistry revealed three main hydroecological zones over the last 500 years and one abrupt, geological event, likely related to the earthquake occurring on 13^th February 2001 (GE I; Figure 2). Our reconstruction reveals the heterogeneous character of the LIA, manifested mainly in water level changes which were reflected in aquatic biota assemblages. The basal sediments of the core (30–26 cm) represent the older part of the LIA (probably 16^th century; Figure 2). Based on biological and geochemical evidence, we infer that Lake Apastepeque was characterized by deep waters and moderate biological productivity during this time. Cladoceran and diatom assemblages were dominated by planktonic taxa, suggesting relatively deep-water conditions together with a rather weakly developed littoral zone. High water levels are also indicated by the presence of only few centropyxids in the testate amoebae assemblage, and the ostracod Keysercypria sp., which is well adapted (large swimming setae) to inhabit the water column. The low diversity and abundance of chironomids reflects limited habitat availability. The limited littoral zone is likely a result of the typical conical morphometry of maar type lakes. The presence of the cladoceran Bosmina sp. and the diatom A. granulata s.l. together with the prevalence of the chironomid genus Labrundinia, may indicate mesotrophic waters (Figure 2) (Hamerlík et al., 2018). The relatively low share of A. neotropicum, a species that dominates through almost the entire sediment sequence, may suggest a different mixing behaviour of the lake or changes in water depth. A. neotropicum is a new species described from the sediment core of Lake Apastepeque, therefore, its ecological requirements are not well known yet. Although all species of Achnanthidium, except for A. catenatum, are considered benthic (Marquardt et al., 2017), Vázquez and Caballero (2013) observed high abundances of the species A. minutissimum in plankton samples of eastern Mexico. The authors suggest that this taxon possibly had an epiphytic habitat attached to floating algae, which flourished during times of increased water column stratification. However, lake level changes with expansion of the littoral zone favouring benthic habitats might also be a possibility. Therefore, we assume that the domination of Fragilaria species and low share of A. neotropicum in the oldest part of the sequence may be the result of deep water conditions with limited benthic habitat and/or more frequent/pronounced turnover of epilimnion under drier conditions.

OPEN IN VIEWER

Figure 2.

Summary diagram of aquatic proxies and main zones and event of lake evolution during the last ca. 530 yr. Selected biotic and geochemical features of the record were grouped by environmental, explanatory factors, where ‘+’ means more positive relationship of a given palaeoindicator with environmental factor, and ‘−’ means more negative relationship. The uppermost layer of Zone III encompasses surface sediments collected by Ekman grab. Fe, Mn, Mg, Ca, S and PO₄ were determined by ICP-OES spectrometer. TOC was determined by an elementary analyser.

High concentration of TOC, TN and PO₄ in the sediment also support relatively high trophic status. At the same time, the lower Fe/Mn ratio indicates more oxygenated conditions in the epilimnion, probably due to increase volumetric ratio of epilimnion to hypolimnion (Townsend, 1999) and/or weaker stratification. On the other hand, the low number of ostracods and testate amoebae indicate unfavourable conditions for their development or their poor preservation (likely associated with anoxic conditions due to decomposition of organic matter in the deeper part of the water column). The species assemblage of testate amoebae suggests persistent stressful environmental conditions, related to deep water conditions, oxygen availability, limited vegetation and low organic matter content throughout the whole time spanning the core (Charqueño-Celis et al., 2022).

Stable C/N ratios of 13–14 indicate a mixture of aquatic and terrestrial biomass (Meyers and Ishiwatari, 1993). The slightly lower C/N ratios recorded in this zone could have been caused by a lower input of terrigenous organic matter. Lower terrigenous fluxes are further supported by lower content of residual mineral matter. At the same time, low K/Zr ratios suggest coarser sediments (Kříbek et al., 2017) than in the following period. In this case, however, low K/Zr ratios were likely caused by a lower input of clay sediments from the catchment, not by higher input of coarse particles (sand fraction). This is further supported by low variability in the content of both Ti and Zr throughout the record.

In the younger part of the LIA (probably between the beginning of 17^th and the end of 19^th century, zone II), a marked decrease in total fauna and flora abundance in the sediment can be observed. The development of the littoral zone significantly changed the habitats of Lake Apastepeque during this period. Based on the responses of individual proxies, two different climatic scenarios and related changes of the littoral zone are proposed (Figure 3): (1) relative expansion of the littoral zone by lake level drop due to precipitation decrease (Scenario 1) or, alternatively, (2) increasing lake level in a wetter environment due to lower evaporation and transport of biological remains from the littoral to profundal zone near to the coring site (Scenario 2). Increasing contribution of littoral cladocerans and benthic diatoms may have resulted from environmental changes caused by both Scenario 1 and 2 (Figure 3). Concurrently, the chironomid record is marked by a relative increase of the subfamily Chironominae, especially of the genus Goeldichironomus, which mostly inhabits fine sediments in the profundal zone (supporting Scenario 2). However, this genus is ecologically versatile with some species also living on macrophytes (Scenario 1 and 2). Predominance of Goeldichironomus along with the presence of the testate amoeba Arcella discoides (Charqueño-Celis et al., 2020; Sigala Regalado et al., 2018) suggests hypoxic conditions, at least temporarily (Figure 2). Furthermore, high water levels with more diluted waters (low carbonate content) could have created unfavourable conditions for ostracod development. The absence of ostracods in the sediments could have been also caused by poor preservation, or some combination of both high-water and low oxygen levels as well as poor preservation.

OPEN IN VIEWER

Figure 3.

Possible scenario of water level fluctuation in Lake Apastepeque during the Little Ice Age: Scenario 1 – lake level drop, Scenario 2 – lake level rise.

Considering that Lake Apastepeque is a maar lake, characterized by steep and almost vertical walls, surrounded by berms and low hills that rise sharply from the lakeshore, we assume that the presence of both benthic and littoral species resulted from a broadening of the littoral zone and an expansion of benthic habitats related to the increase of lake level from a wet environment (Figure 3, Scenario 2). We assume that apart from the mixing and turbulence of the water column, the steep morphology of the maar lake led to efficient transportation of benthic biota remains from shallower areas to deeper sites.

Scenario 2 is further supported by geochemical indices which suggest increasing humidity in the area. As shown by the XRD results, modern Apastepeque sediments consist of minerals rich in alkaline metals, silica, and aluminium. Thus, a gradual increase of alkaline metals (Ca, Mg, K) and the K/Zr ratio suggest enhanced chemical weathering of soils and/or rocks in the catchment of Lake Apastepeque (basaltic ash and andesite are rich in Si, Al, K, Ca, Ti, Fe, Zr, Mg) under more humid conditions (Colman and Dethier, 1986), leading to higher input of clay minerals (weathering products). On the other hand, the lack of significant changes in Al and Si content may be explained by their lower rates of migration in dissolution processes, while alkaline metals are at the beginning of the element migration ranking (Lo et al., 2017). Consequently, strongly weathered rocks consist mainly of Si, Al and Fe due to the previous leaching of bases (e.g. K, Ca, Mg and Na) from the crystalline structure. Subsequently, rather stable concentrations of Ti and Zr exclude significant changes in detrital input from mechanical weathering (Haug et al., 2001). Therefore, the higher input of alkaline metals suggests increased effective moisture, which in turn supports Scenario 2. This period, in all likelihood, was characterized by lower evaporation compared to the older part of the LIA (probably the period between the end of 15^th to end of 16^th centuries, zone I) and thus reduced water losses.

Subsequently, in the younger part of the LIA (probably the period between the beginning of 17^th and the end of 19^th century), the trophic state of Lake Apastepeque was lower than during the older part of the LIA. This period overlaps with the decimation in the indigenous population associated with the arrival of Spanish conquistadors, and, in turn, decreasing human pressure. Forest regeneration, which occurred in El Salvador at that time, could have reduced watershed erosion, causing lower nutrient influx to surface freshwaters. Therefore, environmental change at that time in Lake Apastepeque presumably was induced by both a reduction in human activities and a climatic shift. In our record, the decrease in nutrient flux and trophic state is indicated by geochemical indices and the reduced abundance of biota. The same trends shown by carbonates, organic matter, TOC, sulphur, PO₄ and their positive correlation, underline the relationship of CaCO₃ precipitation with lake productivity. Increased productivity may cause a rise in biological consumption of CO₂, and, in turn, precipitation of CaCO₃ (Boyle, 2001). The CaCO₃ record (determined by LOI), however, does not follow the TIC trend (determined by the elementary analyser) but this may result from the limitations of the LOI method (Santisteban et al., 2004) and therefore should be interpreted with caution. In spite of that, decreased trophic levels are indicated by lower TOC and PO₄, and likely caused a reduction in flora abundance and lower biogenic removal of CO₂, and, in turn, a decrease of CaCO₃ production (Boyle, 2001).

Our finding of higher lake levels during the LIA corresponds with Dull’s (2004a) finding in the Laguna Llano del Espino (western El Salvador). He observed an increased contribution of Potamogeton sp. during the LIA which was interpreted as lake deepening. However, Dull (2004a) hypothesized that this was driven by subsidence, because he did not observe similar hydrological changes in two other lake records from El Salvador (Lake Verde and Lake Cuzcachapa; Dull, 2004a). It is possible that differences between Salvadorian lakes are due to their topographic settings and/or other abiotic factors at the local level. Laguna Verde is located ca. 1600 m a.s.l. within the Cordillera de Apaneca, while Lake Cuzcachapa and Laguna Llano del Espino are situated at mid-elevations (ca. 700 m a.s.l.). However, Lake Cuzcachapa is located at the lee side of the Cordillera de Apaneca, that, in turn, could have caused a rain shadow effect. Therefore, we hypothesize that mid-elevation lakes (500–1000 m a.s.l.; Echeverría Galindo et al., 2019) located in a more open landscape, such as Lake Llano del Espino and Lake Apastepeque, were under the influence of moist air masses. On the other hand, the region of Lake Cuzcachapa and Lake Verde today is characterized by higher precipitation and lower evaporation than the areas of Lakes Llano del Espino and Apastepeque (UNESCO, 2006). These differences emphasize the need to extend multi-proxy palaeolimnological research in El Salvador to improve our understanding of local climate dynamics.

It is also important to mention that during the younger part of the LIA, (ca. AD 1700−1750, 22−20 cm depth) a marked shift in diatom assemblages from A. neotropicum to Fragilaria species occurred. This change likely indicates conditions similar to those recorded in the older part of the LIA, that is lower lake levels with limited littoral and benthic habitats and/or weaker stratification. Interestingly, the environmental turnover at that time was only reflected by the diatom record and may result from diatom sensitivity to changes in water depth and water column mixing (Hofmann et al., 2020; Vázquez and Caballero, 2013). As a result, diatoms may record these smaller-scale changes more accurately than other biological proxies, which respond less sensitively to small fluctuations of water level and water mixing.

The Little Ice Age in Central America – a regional perspective

A comparison of our results with previous studies in the northern Neotropical region (Mexico, Central America and the circum-Caribbean region) provides a wider context for climate variability during the LIA (Supplemental Table S4, available online, Figure 4). The magnitude of cooling across the region varied spatially from ~1°C (Florida coast; Figure 4 – no. 30; Lund and Curry, 2006) to ~2−3°C (Puerto Rico coast; Figure 4 – no. 27; Nyberg et al., 2002; Winter et al., 2000). To date, quantitative temperature reconstructions from lake sediments, however, are scarce in Central America (Supplemental Table S4, available online, Figure 4).

OPEN IN VIEWER

Figure 4.

Map showing paleorecords including information about climatic conditions during the Little Ice Age (LIA). Different colour of dots indicates reconstructed climate conditions. Red: drier conditions, blue: wetter/higher moisture availability and/or lower evapotranspiration, yellow: wetter/higher moisture availability and/or lower evapotranspiration with marked drier beginning of the LIA, red-blue: not clear signal. Black dots: mainly temperature reconstructions. Grey arrows: direction of wind adapted from Kilbourne et al. (2008). Numbers corresponding to discussed records: 1. Lake Santa María del Oro (Rodríguez-Ramírez et al., 2015); 2. Lakes La Luna and El Sol (Cuna et al., 2014); 3. Lake Metztitlán (Olivares-Casillas et al., 2021); 4. Lago Verde (Lozano-García et al., 2007); 5. Lago Aljojuca (Bhattacharya and Byrne, 2016); 6. Laguna San Lorenzo (Franco-Gaviria et al., 2018b); 7. Laguna Esmeralda (Franco-Gaviria et al., 2018b); 8. Lake Lacandón (Vázquez-Molina et al., 2016); 9. Aguada X’caamal (Hodell et al., 2005); 10. Tecoh cave-stalagmite (Medina-Elizalde et al., 2010); 11. Mangrove sediments record (Aragón-Moreno et al., 2012); 12. Lake Punta Laguna (Curtis et al., 1996); 13. Lake Salpetén (Rosenmeier et al., 2002); 14. Lake Petén Itzá (Pérez et al., 2010); 15. Cave Macal Chasm (Webster et al., 2007); 16. Blue Hole (Gischler et al., 2008); 17. Lake Izabal (Obrist-Farner et al., 2022); 18. Lake Amatitlán (Velez et al., 2011); 19. Laguna Llano del Espino (Dull, 2004a); 20. Lake Apastepeque (this study); 21. Lake El Gancho (Harvey et al., 2019; Stansell et al., 2013; 22. Lake Nicaragua (Slate et al., 2013); 23. Laguna Zoncho (Clement and Horn, 2001; Haberyan and Horn, 2005; Lane et al., 2004; Wu et al., 2017; 24. Lake San Carlos (Correa-Metrio et al., 2016); 25. Lake Valencia (Curtis et al., 1999); 26. marine sediment cores, Cariaco Basin (Black et al., 2007; Haug et al., 2001; 27. marine sediment core/corals reef (Nyberg et al., 2002; Winter et al., 2000; 28. Lake Miragoane (Hodell et al., 1991); 29. unnamed lagoon (Peros et al., 2015); 30. marine sediment core (Lund and Curry, 2006); 31. marine sediment core, Fisk Basin (Richey et al., 2009); 32. marine sediment core, Garrison Basin (Richey et al., 2009). Detailed information about the gathered records may be find in the Supplemental Table S4, available online.

Two intervals (Spörer and Maunder minima) of colder conditions during the LIA (AD 1300–1850) are usually recorded in the northern Neotropics. However, there are inconsistencies between different regional records regarding their intensity. Some records from Mexico, Puerto Rico and Panama indicate that the Spörer minimum was the coldest and/or driest period during the LIA (Correa-Metrio et al., 2016; Nyberg et al., 2002; Rodríguez-Ramírez et al., 2015) while others indicate the most severe conditions during the Maunder minimum (Black et al., 2007; Cuna et al., 2014). In Belize, according to stalagmite reconstruction, the LIA climate was relatively wet, with one short dry period at the end of the 15^th century (Webster et al., 2007). An similar scenario for the LIA has been determined using data from western Mexico (Rodríguez-Ramírez et al., 2015) that indicate relatively wet conditions persisting throughout most of the LIA with a short dry interval at the beginning and a slightly pronounced Maunder minimum. Such patterns seem to be similar to those suggested by the proxy data from Lake Apastepeque.

The most distinctive changes of the LIA in the northern Neotropics were associated with water balance (evaporation/precipitation ratio). However, the climate pattern during the LIA is less well recognized and seems to be inconsistent across Central America. Several studies signal a coherent response to climate during the LIA in Central America and the circum-Caribbean region, especially in the north, and display a shift towards drier conditions (Supplemental Table S4, available online, Figure 4 – sites: 1, 2, 9; 13, 21, 22, 25 and 26). By contrast, there are several records where wet/or relatively wet conditions have been reported (Supplemental Table S4, available online, Figure 4 – sites: 3, 4, 5, 6, 7, 8, 10, 15, 16, 18, 19, 23, 24 and 28). In palaeoclimate records, especially from sites located in the middle of Central America, the LIA is documented as a period of increased moisture and lower evapotranspiration (Supplemental Table S4, available online, Figure 4 – sites: 3, 4, 5, 6, 7, 10, 18, 23 and 24). Results of δ¹⁸O in spelaeothems located in the northwest Yucatán Peninsula also suggest relatively wet conditions (Figure 4 – no. 10; Medina-Elizalde et al., 2010), especially between 1490–1580 AD and 1760–1828 AD. The observations mentioned above therefore seem to oppose other interpretations from the Yucatán Peninsula, where data suggest that the LIA was a period of dry conditions (Hodell et al., 2005). Alternatively, a reconstruction from mid elevation in Guatemala (Lake Amatitlán) demonstrates lower lake water levels prior to the LIA (875−1375 AD), followed by increasing lake levels between 1375 and 1875 AD, and finally reduced water level again from 1875 AD to the present (Figure 4 – no. 18; Velez et al., 2011). These changes are opposite to the trend recorded in lowland Lake Petén Itzá, where relatively wet conditions were recorded, although with a trend towards drier conditions (Figure 4 – no. 14; Pérez et al., 2010). These observations could suggest quite high moisture availability during the LIA. On the other hand, the isotopic data from the record of Lake Salpetén (Guatemala, near Lake Petén Itzá) point to rather dry conditions (Figure 4 – no. 13; Rosenmeier et al., 2002). However, Pérez et al. (2010) interpreted a slight decrease in δ¹⁸O values in Lake Salpetén (Rosenmeier et al., 2002) as a higher water level, similar to Lake Petén Itzá. Moreover, Pérez et al. (2010) indicated that wet intervals (especially ca. ~1580 and ~1650 AD) could be caused by increased winter precipitation. A similar conclusion was drawn by Lozano-García et al. (2007) who assumed that a critical factor explaining LIA climates may be winter precipitation. It is noteworthy that records of wet conditions and/or lower evaporation come mainly from middle altitude (500‒1000 m a.s.l.) or highland lakes located <2500 m a.s.l., for example from Lake Santa Maria del Oro (Mexico), which is considered a suitable site for palaeoclimatic study due to its strong dependence on regional variables (Sigala et al., 2017).

Drier conditions during the LIA have been explained by a more southward displacement of the ITCZ (Haug et al., 2001; Hodell et al., 2005), probably caused by a rise in the meridional temperature gradient of the sea surface (Hodell et al., 2005). Cooler SSTs of the Atlantic correspond to a more southerly position of the ITCZ (Haug et al., 2001) which persisted until the onset of the 19th century (Lechleitner et al., 2017). Additionally, a more southern position of the ITCZ results in an enhancement of the north-easterly trade winds which, in turn, leads to a lower moisture content in the atmosphere and a reduction of precipitation (Nyberg et al., 2002). Additional factors controlling moisture availability in the region are the activity of the North Atlantic monsoon and El Niño or Pacific tropical storms (Metcalfe et al., 2000; Rodríguez-Ramírez et al., 2015; Stansell et al., 2013). The LIA has been linked to a more negative phase of the NAO (Stansell et al., 2013), associated with the more southerly position of the ITCZ (Lechleitner et al., 2017) and high activity of El Niño (Mann et al., 2009). In some regions, orography and fractional convergence may have caused increased rainfall despite a more southern position of the ITCZ (Hastenrath, 1976). However, any of the aforementioned mechanisms do not explain the discrepancies among Central American paleorecords.

The disparities between these independent records suggests that the mechanisms of climate forcing are much more complex, and archives of climatic change are strongly influenced by local hydrological balance and catchments. This region is characterized by high geographic and geomorphic variability which results in diverse ecological, tectonic, lithological, and climatic zones (Pérez et al., 2011). The complex and diverse topography results in considerable gradients in humidity, precipitation and wind (Karmalkar et al., 2011). The climate is additionally complicated by the collision of two air masses from the Atlantic and Pacific Oceans. Geographical configurations, therefore, result in the region being characterized by a broad microclimate spectrum. The diversification of abiotic factors may be a reason for the heterogeneous response of ecosystems to global climate changes related to the LIA. Similar conclusions were drawn in a recently published paper by Obrist-Farner et al. (2022). They suggest that any single climatic factor (e.g. ITCZ, NAO, ENSO) cannot explain the full regional patterns of hydroclimatic changes in Central America and, therefore, the disparate signals derived from paleorecords are a result of a combination of several driving factors.

We would like to also highlight that most of the available palaeorecords are from lakes, which are very diverse in Central America with respect to their origin and other abiotic parameters (Pérez et al., 2011; Sigala et al., 2017). These lakes encompass shallow to deep waterbodies of different origin (volcanic, karstic, and tectonic) and morphometry, that are located from lowlands to highlands. The implication of such diversity could be their different sensitivities to local and regional variables (Sigala et al., 2017). For instance, karstic lowland waterbodies, like Lake Petén Itzá and Lake Salpetén, may be more sensitive to changes in evaporation and precipitation ratio (E/P), while shallow lowland lakes (e.g. Lake El Gancho) and shallower high mountain lakes (e.g. Lake La Luna and Lake El Sol) may display more pronounced drier period signals than deep lakes (e.g. Lake Aljojuca, Lake Petén Itzá). However, Obrist-Farner et al. (2022), further suggest that system diversity (including lakes), may not be the main source of discrepancy as they found disparities even when comparing similar systems or spelaeothem records from the same cave.

Considering the above, further work is required to resolve the character of the LIA in the studied region and to assess the role of local conditions on climate records in Central America.

Post-Little Ice Age environmental conditions in Lake Apastepeque (zone III, 14–2 cm, 20^th century)

Modern environmental conditions were established with the end of the LIA. Historically, the 19^th and 20^th centuries coincide with recovery of the indigenous population that declined over 90% in the 16^th and 17^th centuries due to epidemics associated with the arrival of Spanish conquistadors (Dull, 2007; Dull et al., 2010). This period of low human population was reflected in a decrease in anthropogenic pressure on the environment and, as a consequence, forest expansion (Dull, 2004b). Later, a gradual recovery of the El Salvador population and, in turn, an increase in related anthropogenic disturbance caused environmental transformation in the 19^th and 20^th centuries (Dull, 2007). Demographic pressure and extensive agriculture have led to deforestation, increasing erosion and eutrophication in lakes (Dull, 2007).

In the sediment record of Lake Apastepeque, an enhanced nutrient flux is documented by abruptly increasing abundance of cladocerans, a higher share of A. granulata that is an indicator of eutrophication (Kilham et al., 1986), along with higher deposition ratio of lake sediments and slightly higher TOC, TN, and PO₄ values since the beginning of the 20^th century. Extension of croplands (Arino et al., 2012, Supplemental Figure S1, available online) and agricultural activities in the catchment, especially developing sugar cane plantations (Climapesca, 2017; Díaz Ayala et al., 2004) could have induced the trophic state increase. This may have been followed by an expansion of hypoxic waters, as recorded by simultaneous decreases of Mn and an increase of Fe/Mn and S together with the presence of Arcella discoides.

Environmental and ecological conditions in Lake Apastepeque after the catastrophic 13^th February 2001 earthquake (geological event I, 2–0 cm, beginning of the 21 century)

Geological event I (GE I, Figure 2), is characterized by the most significant shift within the biota records, indicating an abrupt alteration of the aquatic ecosystem. The most prominent change was observed in the cladoceran assemblage. Large bodied species characterize the system during GE I (after the earthquake) with Bosminidae species (size 0.2–1 mm) replaced by Daphnia laevis (size 1–2.5 mm). We also recorded the presence of Moina sp. ephippia which indicates high water turbidity and limited presence of fish as species of this genus are large and avoid fish presence. Furthermore, abundances of the diatom A. granulata suddenly increases. This species has often been associated with eutrophic, turbid and turbulent water bodies as well as high physical alterations, for example as a consequence of erosion events (Kilham et al., 1986). Subsequently, highest sulphur concentrations and Fe/Mn values were noted indicating the depletion of oxygen (Boyle, 2001).

We suggest that these alterations were associated with a dramatic earthquake at the beginning of 2001 that was unprecedented in the study area over the investigation period. On 13^th January 2001 and 13^th February 2001, two earthquakes of different origin with magnitudes of M_w = 7.7 and M_w = 6.6, respectively, occurred and caused severe damage to ecosystems and human settlements in El Salvador (Bommer et al., 2002). The most significant for Lake Apastepeque, however, was likely the second, slightly weaker earthquake (Mw = 6.6). The first one was caused by the subduction of the Cocos plate with its epicentre near to the continental shore, deep below the surface (ca. 60 km) and relatively distant to Lake Apastepeque (ca. 90 km). The second earthquake, on the other hand, was an upper-crust earthquake with a shallow epicentre (ca. 10 km), caused by rupture of a part of the fault-zone (Corti et al., 2005) which runs by the Apastepeque Volcanic Field. It was situated only about 30 km away from the study site (Figure 1).

Besides the widely reported loss of buildings and human population (ReliefWeb, 2001), the earthquake on 13^th February 2001 had a series of environmental consequences. The shaking caused localized, shallow landslides (Baum et al., 2001) with more than 25 landslides reported on the flanks of the San Vincente volcano alone (Global Volcanism Program, 2009). Furthermore, it generated liquefaction and, in turn, lateral displacement of the ground on lake shores, for example, Lago de Ilopango (Baum et al., 2001). In Lake Apastepeque, mixing of the water column and likely liquefaction on the shores may have resulted in the release of toxic substances such as hydrogen sulphide, an alteration in water colour and transparency (from transparent to green, reduced light availability), increased extent of the anoxic zone, as well as a massive fish kill observed after earthquake on 13^th February 2001 (BLOG, 2021 ‒ access no longer available; GoogleSite, 2015 ‒ data provided by Turicentro Laguna de Apastepeque). These local reports along with our results allow us to deduce that GE I likely coincided with the environmental transformation caused by the earthquake of 13^th February 2001. Importantly, the earthquake-induced water mixing could have been intensified by strong winds during the dry season that sometimes leads to complete turnover even of deep, crater lakes, for example, Lake Rio Cuarto (Umaña et al., 1999).

The region of Lake Apastepeque is a highly seismic zone, and 68 upper-crust earthquakes with magnitudes between M 5.7 and 6.93 have taken place since AD 1528 in El Salvador and neighbouring countries (Salazar, 2021). However, only the second earthquake seems to have affected Lake Apastepeque. This assumption is supported by the noted fish kill event after the earthquake on 13^th February 2001. We hypothesize that the strong impact of this event on Lake Apastepeque is a result of three combined factors: the upper-crust origin, the high magnitude and its close proximity to the lake. This hypothesis is supported by the fact that shallow crustal earthquakes cause more damage than deep, subduction earthquakes (Bent and Evans, 2004).

Lake recovery after earthquake disturbance (zone III, surface sediments collected using Ekman grab, ca. AD 2013)

Within the cladoceran assemblage, Liederobosmina sp. and Bosmina sp. have reappeared along with some new taxa such as Chydorus cf. sphaericus, Graptoleberis testudinaria and Leydigia louisi louisi. At the same time, Daphnia laevis and Moina sp. almost disappeared, probably because of fish reintroduction and predation (Figure 2). The diatom assemblage indicates a shift towards reduced turbulence and increased light availability by the almost total disappearance of Aulacoseira granulata var. granulata. Simultaneously, the benthic Achnanthidium minutissimum, which is often reported as a pioneer diatom taxon in disturbed environments (Peterson and Stevenson, 1992), occurred more abundantly together with several new taxa (e.g. Achnanthidium cf. saprophilum, Achnanthidium straubianum) (Supplemental Figure S8, available online). Cladoceran and diatom analyses from surface sediments, accumulated since 2013, therefore suggest a partial recovery of the lake ecosystem after the earthquake. Nonetheless, the newly introduced taxa document a persistent effect of this catastrophic event on the aquatic ecosystem, even after several years.