Tropical cyclones (TCs) are among the most catastrophic atmospheric phenomena in terms of their potential to cause loss of both life and property1. In light of this, much effort has been put into studying how TC regimes might change under anthropogenic increase in atmospheric greenhouse gas (GHG) concentrations and related global warming - often with conflicting results2,3. For the most recent decades, we possess comprehensive instrumental records and research has documented relationships among TC track, frequency and intensity and key aspects of basin-scale climate changes over seasonal to decadal timescales4,5,6,7. In order to better understand the mechanisms driving long-term variations in TCs we require records of TC strikes that span periods exceeding the instrumental record, ideally at near-annual temporal resolution; unfortunately, such records are rare. Here we report on a highly detailed sedimentary proxy record of paleo-hurricane activity from the Blue Hole of Lighthouse Reef in Belize, one of the only records from the Caribbean basin that possesses near-annual temporal resolution of TC strikes (Fig. 1). This proxy record provides a unique contribution to the paleo-storm record, filling in a much needed gap in our understanding of Western Caribbean TC regimes; filling this knowledge gap may help answer important questions regarding local versus basin-wide climate controls on Caribbean and Gulf of Mexico TC frequency. This work builds and improves on a previous study that established the sub-annual temporal resolution of normal background sedimentation in the Blue Hole and the exceptional preservation potential of event-layers deposited by the passage of TC strikes8.

Figure 1
figure 1

Geographical location of the Blue Hole on Lighthouse Reef, an isolated platform surrounded by deep water, 100 km offshore the Belize coastline.

Inset details tropical cyclone tracks (1996–1945) passing within 100 km of the Blue Hole11. Bathymetric contours 0 m, 100 m and 500-m intervals below 500 m below sea level. This figure was produced by the corresponding author by the digitizing a nautical chart of the coast of Belize in Adobe Illustrator. The hurricane tracks were downloaded from NOAA HURDAT and superimposed over the digitized map. The figure was produced using US government data (HURDAT) and US Defence Mapping Agency chart 28167. Under section 101 of the US Copyright Act, these documents and data sets are not entitled to domestic copyright protection and the US DMA does not assert copyright protection over its nautical chart products.

A recent comparison of continuous, high-resolution (e.g., decadal or better resolution) paleo-hurricane records for the Western Atlantic, eastern Caribbean and northeast GoM coasts has suggested that basin-scale climatic patterns have controlled Atlantic Basin TC track, frequency and intensity over century timescales and longer9. However, other studies have suggested that TC strike frequencies may be influenced more by local climate phenomena than hemispheric-scale forcing10. This uncertainty and our inability to understand long-term hemispheric patterns in TC track, frequency and intensity cannot be resolved without additional high-resolution TC records from the Caribbean - especially from the western Caribbean that, at present, lacks good representation. Sediment cores, collected from the Blue Hole of Lighthouse Reef, Belize, an anoxic sinkhole located on an offshore isolated carbonate platform in the Mesoamerican Barrier Reef Complex (Fig. 1), contain an undisturbed record of sub-annually resolved sedimentation produced by a combination of fair-weather infilling and tropical cyclone strikes (tropical storms, major and minor hurricanes)8. This newly acquired archive of tropical-cyclone-driven event sedimentation records the impacts of tropical cyclones over the last 1200 y, providing the most continuous and thoroughly dated, sub-annually resolved record of TC activity for the Caribbean to date. This record will allow us to more reliably evaluate the mechanisms driving long-term variations in TC regimes in the Caribbean Basin and consequently better understand the mechanisms driving long-term variations in TCs within the North Atlantic Basin. We hypothesize that, if the latest Holocene record of TC strikes preserved in Blue Hole sediments is largely controlled by local climatic and oceanographic conditions, then the Blue Hole TC record should be generally decoupled from elsewhere in the Gulf of Mexico and Caribbean Sea. Conversely, if Gulf of Mexico and Caribbean TC activity is largely driven by basin-wide climate and oceanographic conditions, then the timing of hurricane regime shifts should correspond to major documented shifts in basin-scale climate and oceanographic forcing mechanisms such as sea surface temperatures (SSTs) from the main development region (MDR) for Atlantic hurricanes and persistent shifts in the phase of the North Atlantic Oscillation (NAO) index.


The results of this study are anchored on a robust age/depth framework, based upon an AMS 14C age model that shows a very strong correlation with an independent age model determined by counting annual couplets. The layer-counting approach was validated previously for this location8, although our radiometric age model shows a constant age offset of −339 y, compared to the layer-counting model (Fig. 2). Excess 210Pb is found throughout the upper portions of the Blue Hole sedimentary sequence, to at least a depth of 63 cm (Fig. 2) and suggests that the uppermost seabed materials were deposited during a time interval equal to or less than the last 110 y (or ~five times the 210Pb half-life of 22.3 y). The consistency of the age offset between the layer-counting and 14C age models in combination with the presence of 210PbXS indicate that this constant temporal offset represents a reservoir effect linked to an older source of 14C within the system that has yet to be documented at this particular site and does not represent a hiatus at the top of or within the sedimentary record.

Figure 2
figure 2

(A) Combined plot of sediment age (Years AD) vs. depth in core (cm) as determined by both calibrated 14C AMS geochronology and by varve counting and plot of calibrated 14C AMS age vs. varve-count age (Years AD). (B) 210PbXS and mean grain-size profile for sediment gravity core BZE-BH-GC2. The lack of a log-linear decrease of 210Pb with depth and the evident variability of 210Pb activity with depth suggests that sediment properties and/or sediment accumulation rates have varied sharply over time. This is consistent with event sedimentation such as that caused by tropical cyclones. This variability however, also prevents calculation of sediment accumulation rates using common methods that require (near) steady-state depositional processes30,34.

Core stratigraphy includes two classes of sediment layers: fine-grained couplets (mean grain size 10–30 μm) <2.5 mm thick each comprised of one buff-coloured and one green-coloured lamina and thicker (>2.5 mm), coarser-grained layers (mean grain size > 30 μm) (Fig. 3). The previous Blue Hole study identified the couplets as annual sediment layers and the thicker, coarser-grained layers as being TC-generated sedimentary event layers8. In order to evaluate probable mechanisms for formation of these thick event layers, the known chronology of historical TCs passing within 100 km of Lighthouse Reef was compared with the distribution of event layers observed in sediment cores during the same time period (Fig. 1 inset; Table 1)11. Of the 13 documented tropical systems (intensity ≥ tropical storm) recorded in the vicinity of the Blue Hole, nine were matched to event layers near the anticipated core depth, with an age offset from layer counting of −0.4 ± 2 y with respect to the known age of the event. Four historically documented tropical systems were not recorded in the core near the anticipated depth in the sediment, including three tropical storms and one Category 1 hurricane (Table 1). One event layer preserved in the sediment did not match the age of any known TCs (Table 1). For TCs not recorded in the sediment core, it is likely that these storms were not detectable because they lacked the necessary energy to induce storm sedimentation (overwash) at the study location due to a combination of distance from the Blue Hole and/or local wave intensity. Because the comparison of the proxy record based on the sedimentary character of TC-generated event layers to the record of known storms does not generate a perfect match, missing ~23% of known events, we must consider this to be the minimum uncertainty associated with the record.

Table 1 A comparison of historical tropical systems (1945–1996) passing within 100 km of the study location11 (Intensity ≥ Tropical Storm). The upper ~3 cm of core material is disturbed and no correlations were made for this interval
Figure 3
figure 3

Representative photo of the Blue Hole sediment core (204 – 226 cm interval of vibracore BZE-BH-SVC4).

The normal background sedimentation consists of fine-grained, buff/green laminated calcium carbonate couplets (~2.5 mm in thickness). The laminations reflect variations is organic matter content of the sediment resulting from seasonal variations in the organic productivity of Lighthouse Reef lagoon surface waters8. TC-event layers can be distinguished from this normal background sedimentation on the basis of grain-size, layer thickness and colour8.

The study site is located ~200 km north of the active strike-slip zone along the Caribbean/North American plate boundary, so we considered the potential of seismic events as mechanisms for the deposition of overwash sedimentary event layers. Unfortunately there are no widely applicable diagnostic criteria to distinguish storm versus tsunami-generated event beds in such a depositional setting. However, the far-field tsunami potential for the Belize coast is low12 and estimated earthquake return period for the region is relatively long (>150 y), especially when compared to anticipated TC return periods; so that seismically assciated event beds, if present at all, could only account for a small fraction of the many events recorded in the Blue Hole.

Based on the strong correlation between the age and depth of observed event layers and the known timing of historical TCs and the long return period of potential tsunami-generated event layers, thick event layers are interpreted to be formed by TCs, with a detection uncertainty of 23%. Although TC event layers are distinguishable from annual couplets by grain size, colour and layer thickness, our evaluation of the historical portion of the core indicates that layer thickness is a sufficient recognition criterion to make this distinction.

As proxy for TC frequency, we present detectable event return period averaged over 100 y (T100), with T100 mean and standard deviation over the entire record of 11 ± 6 y (Fig. 4). From ca. 800–1350 AD, an interval corresponding to the Medieval Climate Anomaly (MCA) that is characterized as the most recent pre-industrial warm interval13, the Blue Hole sediments record the most active period of TC activity in our proxy record. During this interval we observe T100 of 4 – 7 y, reaching the shortest T100 during this interval ca. 900–1000 AD y BP (i.e., highest TC frequency; Fig. 4). In contrast there is an abrupt decrease in the frequency of recorded TC events beginning ca. 1400 AD, near the onset of the Little Ice Age14 during which we observe a sustained lull in TC activity that lasts until ~1900 AD (Fig. 4). This is followed by an upswing in TC activity that extends until present day, with the T100 approaching 7 y - nearing activity levels observed during the MCA. We interpret these results to mean that the interval corresponding to the MCA was a time of high TC frequency, as have been the past 100 y and that the interval corresponding to the LIA was a time of reduced TC frequency.

Figure 4
figure 4

(Lower Panel) Total of all detected event layers per 20 y period (gray histogram), 100 y averaged event-return period (red dash, T100, y).

(Upper Panel) Select, high-resolution sedimentary records of land falling hurricanes are presented for the Atlantic Basin including New England (Boston, MA)23, the Florida Gulf Coast (Mullet Pond, FL)25, the Eastern Caribbean (Vieques, PR)24 and the Bahamas26.

Previous studies have relied upon grain size for estimates of relative intensity. However, in the present case, event-layer thickness is likely to be a more sensitive indicator of hurricane combined intensity and duration than grain size. This is because the coarsest sediment on the reef top that is found in our cores is readily resuspended by waves generated from the winds of even the weakest TCs. From linear wave theory15, a weak Category 1 hurricane with wind blowing 32 m/s over a 10 km fetch and water 3 m deep can generate local waves of 0.85 m height and 3.3 s period. Bed shear stresses from these waves reach ~3.7 Pa, sufficient to resuspend calcite spheres ~5 mm in diameter. In contrast, the coarsest grains in our cores are lower-density flakes from the calcareous green alga Halimeda, up to ~4 mm in maximum dimension (Corey shape factor ~0.2), with theoretical critical shear stress for resuspension on the order of 2 Pa16,17. These calculations do not take into account currents or propagation of swell across the reef surface, which would be likely to result in even greater bed shear stresses produced by minimal TCs. It is likely that stronger TCs would produce stronger bed shear stresses, but the peak grain size recording the event could be limited by the size of sediment available for transport. In contrast, bedload sediment transport rate is generally considered to be a power function of the amount by which local bed shear stress exceeds critical shear stress for grain transport15,17. As a result, the bedload transport rate can continue to increase due to increasing wave-current intensity (even if sediment grain size does not increase appreciably during a storm) and can persist as long as shear stresses remain high and sediment is available for resuspension.


With sub-annual resolution recording a mean storm recurrence of ~12 TCs per century, the Blue Hole is a unique and faithful archive of paleo-hurricane and tropical storm strikes for the Western Caribbean. Although uncertainty of detection is ±23% per event, mostly for TCs of tropical storm intensity, this proxy record appears to have the finest temporal resolution (the only near-annual record) and continuity (a single record with no apparent hiatuses) of all TC proxy records from the Western Caribbean. Our results show that this archive records strikes of TCs from tropical storm to major hurricane intensity, not only major hurricane (H4 and H5) activity as has been previously reported18 and show evidence for three distinct Caribbean TC (frequency) regimes over the past 1200 y. Although TC activity is observed throughout the entirety of the time period represented in our sediment cores, the results of this study unequivocally demonstrate that the development of hurricanes affecting the western Caribbean has varied over centennial timescales (Fig. 4). The hurricane-strike frequency distribution we present is broadly consistent with the findings of a previous study from the Blue Hole8. However, we offer improved resolution and detection of relatively weak TC events, quantitative time-series analysis and an age model with reduced uncertainty as compared to this previous study. We note that our TC frequency distribution is inconsistent with a recent study of TC strikes from the coastal mainland of Belize that attributed variations in TC strike frequency to local climate phenomena10. Recording only five likely TC events layers in sediments younger than 256 y b1950, the mainland study is not calibrated to the modern instrumental record and has significantly lower temporal resolution and site sensitivity than our Blue Hole sedimentary proxy record.

Our proxy record demonstrates that the timing of observed major shifts in hurricane frequency and event layer thicknesses in the Western Caribbean occur contemporaneously with documented shifts in basin-scale climate phenomena (Fig. 5). As such we interpret the centennial-scale variations in TC frequency in terms of shifts in documented basin-scale climatic patterns that are understood to drive TC frequency variability in the Caribbean including; (1) the NAO that influences the tracking of TCs through shifting North Atlantic Basin atmospheric pressure gradients19 and by association the favourability of the thermodynamic environment that they encounter20; and (2) SSTs from the MDR for Atlantic hurricanes (10–14° N, 20–70°W) that affect the favourability of the thermodynamic environment and the vertical wind shear and tropospheric stability of the environment of TC development21.

Figure 5
figure 5

TC frequency per 20 y and TC layer thickness with decadal-scale surface temperature and paleoclimate reconstructions.

(A) Winter NAO reconstruction22; (B) MDR SST9; and (C) hurricane frequency per 20 y, with tropical cyclone event layer thicknesses and average background sediment varve thickness (2.5 mm).

We observe increased TC activity between ~800 and ~1350 AD, corresponding to the MCA, a time period during which widespread indices of climatic shift have been observed22. Among these indices are a persistently positive NAO index and changes (warming) of SSTs (Fig. 5) both of which are conducive to the development and intensification of TCs. Conversely during the LIA, we see evidence of persistently negative NAO indices and cooler MDR SSTs, indicative of climatic shift towards unfavourable TC development conditions (Fig. 4).

The Blue Hole proxy TC record fills a spatial gap in regional coverage of TC records, with the finest temporal resolution of TC proxy records for the western Caribbean Sea. Figure 4 compares the Blue Hole proxy record to other records of comparable temporal resolution23,24,25,26 and demonstrates that active periods for TCs for the western Caribbean are consistent not only across the greater Caribbean/Gulf of Mexico region, but also across much of the Atlantic Basin supporting the idea that centennial-scale paleo-TC activity has indeed varied at the basin scale9. Timing of shifts in hurricane activity levels recorded in Blue Hole sediments appear to be contemporaneous with variations observed in other paleo-storm archives and also with documented centennial-scale shifts in climatic conditions (including the mean NAO index and MDR SSTs) known to affect the development and intensity of TCs (Fig. 5). Collectively, these results indicate that basin-scale climatic and oceanographic variables, rather than local SST conditions may be the primary drivers for variations in the observed shifts in centennial-scale TC frequency regimes over the last 1200 y.


We present sedimentary proxy data for paleo-hurricane activity in two sediment cores from the Blue Hole of Lighthouse Reef, Belize: a short gravity core (10 cm diameter, 63 cm long), BZE-BH-GC2 (17° 18.97N, 87° 32.1W, 120 m water depth), collected with undisturbed sediment surface and a longer vibracore (7.5 cm diameter, 532 cm long) BZE-BH-SVC4 (17° 18.97N, 87° 32.1W, 120 m water depth) that was collected using a Rossfelder® P-3 system. Whole cores were returned to Memorial University of Newfoundland for analysis, where they were first measured for bulk density using a Geotek Multi Sensor Core Logger gamma densitometer at 0.5 cm intervals. The cores were then split and imaged using Geotek Geoscan line-scanning system at 0.05 mm resolution, with images stored in RGP BMP format. Following Geoscan imaging, cores were subsampled for X-radiography, grain size measurement and 210Pb/137Cs geochronological analysis. For X-radiography, axial slabs (1 cm thick) were prepared and imaged using a Thales Flashscan35 digital X-ray detector illuminated by a Lorad industrial X-ray generator. Of the density, X-radiograph and colour-imaging data sets, the Geoscan BMP images possessed the best combination of accurate and precise spatial reference and high spatial resolution, so those images were used for quantitative analysis of bedding thickness and depth described below.

Following the RGB image capture, Geotek ImageTools image analysis software was used to extract a horizontally averaged (20 pixels wide) RGB colour data profile along the stratigraphic axis of the core image at 0.1 mm (1 pixel) depth intervals27. The aperture-calibrated RGB time-series data was then converted to a greyscale/luminance (L*) time-series using the following equations:

where: IA is the aperture setting used for core image capture and CA is the aperture setting used for light and dark field calibration

where: L* is calibrated greyscale/luminance28.

In the case of laminated core material, layer boundaries correspond to the inflection points of the greyscale luminance curves. Our method, like other established laminae detection methods uses a differential filter for edge detection29. In terms of core material, the filter determines the depth-referenced contact between light and dark laminae. To implement the filter, a simple Matlab program was developed to produce a laminae count, core depth of contacts and individual bed/lamina thicknesses. Images and data generated by the automated analysis were then compared manually to correct for errors in automated analysis associated with thick, coarse-grained beds and artefacts from core collection and processing.

For gravity core BZE-BH-GC2 (0–62.5 cm), sub-samples for grain-size analysis and 210Pb/137Cs radioisotope geochronology were taken at relatively coarse (0.5 cm) intervals to ensure that there was sufficient material for all necessary analyses along the entire length of the gravity core (63 cm). For granulometric analysis samples were dispersed in 0.05% sodium metaphosphate solution, subsequently disaggregated in an ultrasonic bath and then analysed using a Horiba LA-950 PARTICA particle size analyser. Grain size was estimated visually along the entire length of the vibracore BZE-BH-SVC4 (to a maximum depth in core of 532 cm). Activities of excess 210Pb and 137Cs (Becquerels/g or Bq/g) were determined by γ-spectroscopy analysis of dried sediment (46.5 KeV peak for 210Pb and 661 KeV peak for 137Cs)30. Cesium-137 was not detectable at activities above 3 Bq/g. Additionally, ten Accelerator Mass Spectrometry (AMS) 14C ages from organic residue samples were obtained (sample preparation performed by the Université Laval and analysis at University of California, Irvine) from the core BZE-BH-SVC4 in order to constrain the age model. Organic residue was separated from the bulk sediment material by dissolving the carbonate with HCl, washing the sample with NaOH and repeating until no more carbonate material remained. These ages were calibrated using the Intcal09 and Marine09 curves31 and the OxCal program32,33. As no site-specific ΔR value is available for the Belize Atoll system, a global marine reservoir age of 405 y was used to calibrate these 14C ages. A further calibration step was undertaken by a regression between known varve age and depth with calibrated radiocarbon age, allowing development of the age model referenced to year of collection (2009) in Fig. 2.