The NAH14 hydro-climate record
The apparent co-variability of δ13C values and Mg/Ca, Sr/Ca and Ba/Ca ratios is a widespread observation in speleothem geochemical records, and is commonly interpreted as a dominant hydrological control of these proxies [40, 41]. Here, the strongest relationship exists between δ13C and Mg/Ca, which are correlated throughout the record (r = 0.66). Prior calcite precipitation (PCP) results in covariations of trace elements such as Sr and Mg, and the evolving δ13C composition of the solution, and hence the composition of calcite further down the water flow-line [42, 43]. Thus, higher δ13C and Mg/Ca values occur in response to enhanced PCP during periods with lower recharge and allows to interpret synchronous peaks in Mg/Ca and δ13C values as a sign of aridity [40, 43]. Sr/Ca and Ba/Ca values in speleothems are often masked by additional processes, such as growth rate or minor detrital contamination, which is reflected by the slightly weaker correlation of these elements to Mg/Ca or δ13C values [41, 44]. On the decadal to multi-decadal timescales, frequent excursions towards higher values (i.e. pronounced dry conditions) occur simultaneously not only in δ13C and Mg/Ca, but also in the δ18O values (Fig. 2). Since the oxygen isotopic composition of the drip water is usually less affected by PCP, this argues for a possible additional kinetic control of the proxies. We suggest that the driving process is most probably carbonate deposition in response to decreasing drip rates and/or enhanced degassing of CO2 [42, 45].
Besides the previously mentioned short-term spikes in all proxies, δ18O values appear to be largely uncorrelated with δ13C or Mg/Ca values and out of phase on centennial timescales (Fig. 4). This is, for example, visible at c. 10,900, c. 10,300, and c. 9,600 a BP, when δ13C and Mg/Ca values persist on a relatively high level for roughly 100 years longer than the δ18O values, while the δ18O values switch towards the most negative values of -4 to -6‰ (Fig. 2). Following the traditional interpretation of δ18O values in tropical speleothems, these low values could evidence higher precipitation amount and thus recharge [5, 46], while δ13C and Mg/Ca argue for persistent dry conditions for another century. Thereafter, all three proxies reveal lower values in concert with the interpretation of increasing precipitation and recharge. This apparent decoupling of the proxies’ responses suggests that the changes in the oxygen isotopic composition of the drip water reflect other effects in addition to effective recharge, which is regarded as the dominant driver of δ13C and Mg/Ca values.
In the low latitudes, tropical cyclones can alter the oxygen isotopic composition of the drip water, due to the particularly low δ18O values of tropical cyclone rainfall, and these low δ18O values are then transmitted to the growing stalagmite [46, 47]. Intervals in which the NAH14 δ18O values drop prior to δ13C values may thus indicate periods when enhanced tropical cyclone activity masks the isotope-derived precipitation records [8, 48]. In the NAH14 record, the uncorrelated shifts towards very negative δ18O values occur especially during times with high δ13C or Mg/Ca values, which are thought to indicate persistently dry conditions or even multi-annual droughts. This suggests that the previous described effect is most notable when the water volume of the reservoir is small . Thus, we suppose that the speleothem δ18O values appear to show a bimodal origin related to both precipitation-derived recharge and drought index coupled to the regional tropical cyclone activity. Nevertheless, kinetic fractionation during the carbonate formation cannot be ruled out completely. Consequently, we therefore interpret the speleothem δ13C (and Mg/Ca) values of stalagmite NAH14, which are largely unaffected by tropical cyclones, as the most reliable indicators of the underlying structure of effective recharge, and drought variability.
Early Holocene precipitation variability on the north-eastern Yucatán peninsula
Due to its unprecedented and unique time resolution, the NAH14 record provides the opportunity to investigate decadal to centennial early Holocene climate variability on the north-eastern Yucatán peninsula. In particular, the speleothem covers a 1,500 a-long continuous interval between c. 11,040 and 9,520 a BP, during a time when previous evidence provides only a fragmentary picture of past precipitation patterns in the tropical Americas as illustrated in Fig. 5 and Figure S 1 [18–20]. Paleo-precipitation records spanning the transition of the last glacial to the mid Holocene in the Caribbean and Central American realm describe a dry-to-wet transition between c. 11,000 and 9,000 a BP following the increasing boreal summer insolation, and a relatively stable hydro-climate thereafter  (Fig. 5). In the NAH14 record, however, the above described trend is not identified. In addition, low water tables and terrestrial environments persisted after the deglaciation until c. 8,000–9,000 a BP in the Lake Chichancanab basin on the northern Yucatán peninsula [4, 11], and SST reconstructions from the Gulf of Mexico also lack a pronounced orbital trend [17, 31]. This indicates that the progressive northward movement of the ITCZ, as suggested by other proxy records located further to the south [2, 21], is not (yet) noticeably reflected on the north-eastern part of the Yucatán peninsula (Fig. 5).
δ18O, δ13C and Mg/Ca values of NAH14 demonstrate that droughts on the decadal to multi-decadal scale were frequently occurring events, as evidenced by repeated peaks towards positive values concurrently in all proxies (Fig. 2 and Fig. 3). A higher drought frequency with c. 5–10 drought phases per 100a is identified between 10,950–10,800 a, 10,680–10,480 a, 10,330–10,200 a, 10,080–10,000 a, and 9,950–9,520 a BP. In contrast, during the intervals 10,800–10,680, and 10,200–10,090 a BP, relatively humid and stable conditions prevailed on the north-eastern Yucatán peninsula. In the century between 10,420–10,320 a BP, climate was also generally wetter, but humid conditions were interrupted by a drought phase between c. 10,370 and 10,350 a BP. Superimposed on this high-frequency signal, strong swings on the centennial scale indicate pronounced changes in regional rainfall patterns. The NAH14 record suggests that during the more humid phases, decadal-scale droughts were exceedingly rare and climatic conditions were relatively stable. On the other hand, during the already drier intervals, additional, severe droughts occurred much more frequently, with the documented pronounced decadal periodicities of 11-17a in the NAH14 record translating in 5–10 droughts of a duration of up to several years.
Major droughts occurring on a decadal to centennial scale have been widely reported for Mesoamerica during the last millennia [4–6, 24], thus reflecting a pattern similar to that preserved in NAH14 for the early Holocene. Increased drought frequency during the late Holocene, however, came along with generally drier conditions, resulting from a progressive southward displacement of the ITCZ after 7,000 to 3,000 a BP [11, 21, 30]. In contrast, the mid- Holocene northernmost position of the ITCZ was accompanied by reduced precipitation variability. Similarly, the here observed early Holocene decrease in precipitation variability occurs during presumably more humid phases indicative of a northward movement of the ITCZ. This displacement of the convective centers positions the Yucatán peninsula within the seasonal latitudinal bounds of the ITCZ for a longer period each year [30, 53]. In contrast, during the drier phases of the record, the study site experienced prolonged dry seasons, which leads to a higher seasonal contrast, and thus higher precipitation variability.
The centennial- to millennial scale swings between wet and dry conditions in the north-eastern Yucatán peninsula appear to be largely concurrent with meridional shifts of the ITCZ (Fig. 5), as indicated from the high-resolution reflectance record from the Cariaco basin . Before c. 10,800 a, the potential link is less obvious, which may be related to the superimposed larger variability of the NAH14 record. It may nevertheless also hint towards a more complex behavior of the ITCZ and its northernmost extent over Mesoamerica, as compared to the position of the Cariaco basin closer to the convective center of the ITCZ over South America (Fig. 1, ).
Forcing mechanisms of precipitation variability
On seasonal and inter-annual timescales, the distribution of precipitation over the Yucatán peninsula is modulated by the so-called Pacific-Atlantic ‘inter-basin mode and is thus influenced by several forcing mechanisms, such as ENSO activity, radiative forcing, or variations in Atlantic SSTs [4, 6–8, 22, 23]. In particular, some of these studies suggest an influence of volcanic eruptions and aerosol emission on the regional evaporation/precipitation balance in Mesoamerica, and that these exacerbated or prolonged predominant drought intervals during the past centuries in Mesoamerica [8, 54]. To assess the influence on our record, we compare our data with a reconstruction of volcanic forcing generated from sulphate concentrations of the Greenland GISP2 ice core , which documents several events during the time covered by the NAH Mg/Ca and δ13C record (Fig. 5). However, the number of recorded droughts in the NAH14 proxies exceeds by far the number of known strong volcanic eruptions. Coeval to two major volcanic events at c. 10,400 and 10,250 a BP apparently strong dry events are recorded in the NAH14 proxies. In contrast, the most pronounced dry phases, such as for example after c. 10,930 a, 10,650 a, or 10,050 a, rather occur in absence of major detected volcanic eruptions in the ice core sulphate loads. Even though, we cannot exclude the possibility that volcanic forcing might have punctually contributed to or exacerbated droughts occurring during the early Holocene; we argue that other mechanisms were probably more dominant.
Spectral analyses revealed dominant periodicities with frequencies of 11–17 a, 25–35 a and c. 250 a (Fig. 3). The pronounced spectral power of these periods reflect reported periodicities in records of modern Mesoamerican climate variability, and are most strikingly similar to solar activity cycles (11 a Schwabe solar (or sunspot) cycle and reoccurrence of the c. 210–220 a de Vries/Suess solar cycle [55, 56]. This is strongly supported by the agreement of the NAH14 proxies with a reconstruction of sunspot numbers , as well as solar activity reflected in the residual cosmogenic 14C , as shown in Fig. 5. A relation of early Holocene speleothem δ18O values with solar forcing has been previously suggested by Stinnesbeck et al.  to better constrain the chronology of a speleothem record from Chan Hol cave on the north-eastern Yucatán peninsula between 9,000 and 13,000 a BP. We here confirm this link, but based on a precisely dated record of much higher temporal resolution.
The influence of solar variability on tropical hydro-climate has been predicted by model studies [57, 58], and especially the centennial-scale Suess-cycle is documented in various early to late Holocene records throughout the tropical and subtropical Americas [4, 12, 24, 34, 53, 59]. For example, Pollock et al.  or Bernal et al.  report similar periodicities of 200–250 years in stalagmites from Belize and south-western Mexico, and hypothesized that solar variability is a significant force causing precipitation variability in North Central America during the mid-Holocene. Times of reduced solar output are associated with cooler conditions in the North Atlantic, as well as a development of more El Niño-like conditions in the eastern equatorial Pacific (EEP), resulting in a more arid circum‐Caribbean climate [16, 28, 60].
In particular, these solar cycles modulate ENSO variability via the so-called “ocean dynamical thermostat” response, because negative (positive) radiative forcing results in a dynamic warming (cooling) of the eastern tropical Pacific [28, 61]. This model prediction has been widely supported by paleo-environmental reconstructions mainly from the last millennium [26, 28, 57]. ENSO itself influences the precipitation pattern over the Yucatán peninsula, where a warm El Niño event is associated with anomalous subsidence and dry conditions, whereas the delayed warming of SSTs may increase convective activity in the following year [6, 23]. As a consequence, the question of the role of solar activity in climate variability of the tropical Americas has been frequently debated [7, 24, 59]. We here present a record that is sufficiently well resolved and clearly provides evidence for an empirical similarity between solar activity and early Holocene Mesoamerican precipitation. The origin of this coupling, however, may be suspected in ENSO activity, which evolved over the course of the Holocene [27, 62] . During the early to mid-Holocene, a dampening of ENSO activity has been reported , while other studies suggest that ENSO variance was close to modern levels . For example, an early Holocene (c. 9,500 a) model experiment suggests that freshwater flux and remnant ice sheet forcing factors led to significant remote responses in the tropical Pacific strengthening the amplitude of eastern Pacific El Niño events . The visual inverse relationship of the NAH14 record to Pacific SSTs  provides additional support for a strong influence of tropical Pacific dynamics on Mesoamerican precipitation (Fig. 5). The late Holocene onset of major dry events after c. 3,000 to 4,000 a BP coincided with ENSO and ITCZ changes, with an increasing dominance of the Pacific Ocean [18, 53]. Consequently, the persistent decadal to centennial solar cycle forcing of the regional hydro-climate, is modulated by an ENSO activity which is similar as the one in late Holocene leading to pronounced decadal droughts in northern Yucatán and Mesoamerica.
The spectral analyses also suggest the presence of inter-decadal periods of c. 25–35 years in the NAH14 record (Fig. 3). Long-term ENSO variability may modulate the Pacific Decadal oscillation (PDO), which is not a single physical mode of ocean variability, but rather the sum of several processes with different dynamic origins . Enhanced rainfall over Mesoamerica on the multi-decadal scales may be also related to a warmer North Atlantic Ocean related to a pattern similar to the late Holocene Atlantic multidecadal oscillation (AMO) [6, 23]. However, the wavelet analyses of the NAH14 record show that the longer multi-decadal periods are not a persistent feature in this record and only significant during a few, short intervals. This is similar to the results of Lachniet et al., , who reported no clear link to the PDO or AMO regarding the strength of wet season rainfall in the Mesoamerican monsoon based on late Holocene stalagmite records from south-western Mexico. Similarly, recent studies suggest that especially the AMO is no internal oscillation of Earth’s climate, but rather an externally forced quasiperiodic variation .
Deglacial and early Holocene records from the tropical Atlantic realm suggest a link of local precipitation to sub-centennial to millennial freshwater input into the North Atlantic and the Gulf of Mexico [19, 20]. The comparison of NAH14 speleothem data with marine records from the Gulf of Mexico (Fig. 5), shows that meltwater flux into the Gulf of Mexico resulted in decreasing SSTs and drier conditions on the north-eastern Yucatán peninsula [17, 31, 52]. The centuries-long dry-wet cycles also roughly agree with previous paleoclimatic evidence deduced from lacustrine sediments of Lake Petén Itza, southern Yucatán, suggesting a series of wet–dry cycles of about 250 a duration with severe dry events occurring at c. 11,200, 10,900, 10,700 and 10,400 cal. a BP, concurrent to pre-boreal meltwater flooding events in the Gulf of Mexico [15, 19, 52].
A large-scale comparison of the NAH14 record with other speleothem-based reconstructions from North and South America highlights the associated link of both low and mid-latitude precipitation patterns during the early Holocene. The centennial-scale dry/wet cycles on the north-eastern YP are largely anti-phased to precipitation recorded in speleothems from Brazil  (Fig. 5) and Peru  (Figure S 1). Both records reflect the strength of the South American Monsoon System (SAMS), thus the visual anti-correlation is interpreted as an expression of a hemispheric anti-phasing in the low-latitudinal Americas, similar to late Holocene observations . In addition, an inverse relationship of precipitation in the northern American low- and mid-latitudes may be suggested by comparing the NAH14 record with the hydro-climate reconstructions from the Great Basin  and New Mexico  (Figure S 1). Despite the lower temporal resolution of these records compared to NAH14, periods of higher precipitation in Central America appear to coincide with drier conditions in New Mexico, and more northerly moisture sources in the Great Basin. This comparison thus further supports previous evidence that the meridional displacement of the ITCZ shifts the Hadley cell and drives circulation processes in the higher latitudes [9, 34].
Southward shifts of the ITCZ and the Hadley Cells reduce the meridional pressure gradient in the northern hemisphere, inducing expansion and southward displacement of the polar and mid-latitude pressure cells . Consequently, the polar jet and westerlies (Fig. 1) shifted southward and weakened, and monsoonal systems propagated less far northward . Similar to the past 2,000 a, the observed low-latitude hemispheric anti-phasing over decadal-centennial timescales in response to meridional shifts of the ITCZ during the early Holocene, is most likely a result of a combination of external forcing. These encompass in particular solar activity and maybe also volcanic eruptions, but also internal feedback mechanisms, including meltwater fluxes into the Gulf of Mexico and the North Atlantic Ocean.