Skip to main content
NCBI home page
Proceedings of the National Academy of Sciences of the United States of America logoLink to Proceedings of the National Academy of Sciences of the United States of America
. 2016 Mar 7;113(12):3169–3174. doi: 10.1073/pnas.1519566113

Shipwreck rates reveal Caribbean tropical cyclone response to past radiative forcing

Valerie Trouet a,1, Grant L Harley b, Marta Domínguez-Delmás c,d
PMCID: PMC4812713  PMID: 26951648

Significance

Twenty-first-century North Atlantic tropical cyclone (TC) projections are crucial for the development of adaptation and mitigation strategies, but they are subject to large uncertainties, particularly with respect to TC response to radiative forcing. We used a combination of tree-ring data and historical shipwreck data to show that TC activity in the Caribbean was distinctly suppressed during the Maunder Minimum (1645–1715 CE), a period when solar irradiance was severely reduced. This solar fingerprint on decadal-scale Caribbean TC variability implies modulation by a combination of basin-wide climatic phenomena. Our findings highlight the need to enhance our understanding of the response of atmospheric circulation patterns to radiative forcing and climate change to improve the skill of future TC projections.

Keywords: Caribbean, tropical cyclone, Maunder Minimum, dendrochronology, documentary data

Abstract

Assessing the impact of future climate change on North Atlantic tropical cyclone (TC) activity is of crucial societal importance, but the limited quantity and quality of observational records interferes with the skill of future TC projections. In particular, North Atlantic TC response to radiative forcing is poorly understood and creates the dominant source of uncertainty for twenty-first-century projections. Here, we study TC variability in the Caribbean during the Maunder Minimum (MM; 1645–1715 CE), a period defined by the most severe reduction in solar irradiance in documented history (1610–present). For this purpose, we combine a documentary time series of Spanish shipwrecks in the Caribbean (1495–1825 CE) with a tree-growth suppression chronology from the Florida Keys (1707–2009 CE). We find a 75% reduction in decadal-scale Caribbean TC activity during the MM, which suggests modulation of the influence of reduced solar irradiance by the cumulative effect of cool North Atlantic sea surface temperatures, El Niño–like conditions, and a negative phase of the North Atlantic Oscillation. Our results emphasize the need to enhance our understanding of the response of these oceanic and atmospheric circulation patterns to radiative forcing and climate change to improve the skill of future TC projections.

Landfalling tropical cyclones (TCs) bring devastation to natural and human landscapes with floods, winds, and storm surges. In recent decades, TC-related human mortality and economic losses have risen in step with increasing populations in high-risk coastal communities (1). TC damage is expected to further increase in the near future with rising exposure and projected anthropogenic climate change (2). This is particularly the case for the North Atlantic Basin, which is one of the most TC-active basins globally. The development of successful adaptation and mitigation strategies relies on skillful projections of North Atlantic TC activity, as well as an improved understanding of the drivers of its variability.

Modeling studies of twenty-first-century global TC activity generally converge in their projections of increased TC intensity and decreased frequency, but the magnitude range of projected North Atlantic TC variability is wide (3). Uncertainties in twenty-first-century North Atlantic TC projections are largely driven by the chaotic nature of the climate system and by our limited understanding of TC response to radiative forcing, including anthropogenic greenhouse gases and aerosols, as well as natural variability in volcanic and solar activity (4). Response uncertainty is the dominant source of uncertainty toward the end of the twenty-first century (4), with different model runs resulting in TC responses of opposing sign to projected radiative forcing (3). Our understanding of TC response to radiative forcing—and thus the skill of future TC projections—is restricted by limitations in the time-series length and quality of observational records (5) that hinder trend detection and attribution (3).

To attribute significant TC changes to specific climate forcings, recent TC activity needs to be placed in a longer-term context (6). The Maunder Minimum (MM; 1645–1715 CE), the most severe change in solar irradiance in documented history (7, 8), is of particular interest in this context, but TC proxy records that cover this period are scarce, often present a conservative estimate of the total number of storm events (9), and largely have insufficient time resolution to distinguish the MM (6, 10–12). Documentary data sets are the main source of paleotempestology information of appropriate temporal resolution, but most document-based TC studies have primarily focused on long-term TC climatology (e.g., seasonality, recurrence intervals) rather than interannual or decadal-scale variability (13, 14). Here, we combine two annual resolution proxy records—a documentary time series of Spanish shipwrecks in the Caribbean (TCship) and a tree-growth suppression chronology from the Florida Keys (TCsupp)—to extend the observational Caribbean TC (CTC) record back over the last 500 y and thus to cover the MM.

Over the past centuries, TCs have caused destruction of human settlements and wreaked havoc at sea. In the Caribbean, TCs were the primary documented cause of shipwrecks in the sixteenth through eighteenth centuries (15) and they left their mark on regional history. For instance, Spanish hegemony over Cuba was secured in 1640 after a hurricane decimated a Dutch fleet poised to attack Havana (15), leading to an additional century of Spanish monopoly over trade between the Caribbean and Europe. We make use of the well-documented maritime TC legacy in the Caribbean region to reconstruct CTC variability. Our reconstruction (TCship) is based on a comprehensive documentary compilation (16) of 657 ships of Spanish origin that wrecked in the Caribbean Basin (Fig. 1A and Tables S1 and S2) over the period 1495–1825 CE due to storms or unspecified factors.

Fig. 1.

Fig. 1.

Open in a new tab

Geographical location, storm-induced tree growth suppression, and its climatic signal at Big Pine Key, Florida. (A) Geographical location of the BPK tree-ring site (red dot), HURDAT-derived (1851–2010 CE) (17) category 1–5 TCs (white lines; n = 44) that tracked within 160 km (red buffer ring) of the site, and countries/states for which shipwrecks were recorded (black shading) (16); (B) Slash pine (P. elliottii var. densa) section (dated over 1707–1829 CE) with tree-growth suppressions (multiple consecutive narrower-than-average annual rings) resulting from TC events indicated by date of recorded TC event; (C) SEA (1895–2009 CE) of TCsupp event series (defined as years when >75% of samples showed growth suppression; n = 12) with July–November Palmer Drought Severity Index (PDSI) for Florida climate division 7.

Table S1.

Annual number of Spanish shipwrecks in the Caribbean (TCship) as reported by Marx (16)

Year No. wrecks Year No. wrecks Year No. wrecks Year No. wrecks Year No. wrecks Year No. wrecks Year No. wrecks
1495 6 1543 0 1591 29 1639 3 1687 1 1735 0 1783 2
1496 0 1544 1 1592 1 1640 0 1688 1 1736 0 1784 0
1497 0 1545 26 1593 2 1641 9 1689 3 1737 2 1785 0
1498 0 1546 0 1594 0 1642 0 1690 0 1738 2 1786 1
1499 0 1547 1 1595 1 1643 2 1691 0 1739 1 1787 0
1500 4 1548 0 1596 1 1644 0 1692 0 1740 1 1788 2
1501 2 1549 0 1597 2 1645 0 1693 0 1741 4 1789 0
1502 26 1550 3 1598 0 1646 0 1694 0 1742 2 1790 1
1503 0 1551 5 1599 0 1647 6 1695 3 1743 0 1791 0
1504 0 1552 2 1600 15 1648 0 1696 1 1744 0 1792 3
1505 1 1553 38 1601 0 1649 0 1697 0 1745 0 1793 0
1506 0 1554 5 1602 1 1650 0 1698 0 1746 0 1794 13
1507 0 1555 6 1603 5 1651 1 1699 1 1747 0 1795 0
1508 20 1556 5 1604 1 1652 0 1700 0 1748 1 1796 0
1509 21 1557 0 1605 3 1653 0 1701 0 1749 0 1797 1
1510 0 1558 0 1606 0 1654 0 1702 0 1750 1 1798 0
1511 1 1559 7 1607 0 1655 0 1703 0 1751 1 1799 0
1512 0 1560 1 1608 0 1656 0 1704 0 1752 7 1800 0
1513 0 1561 0 1609 2 1657 3 1705 4 1753 0 1801 0
1514 0 1562 1 1610 7 1658 0 1706 1 1754 0 1802 1
1515 1 1563 1 1611 0 1659 4 1707 0 1755 1 1803 2
1516 0 1564 5 1612 0 1660 0 1708 0 1756 0 1804 0
1517 0 1565 5 1613 0 1661 0 1709 0 1757 1 1805 1
1518 0 1566 2 1614 12 1662 0 1710 0 1758 2 1806 0
1519 2 1567 13 1615 1 1663 0 1711 1 1759 0 1807 0
1520 1 1568 0 1616 1 1664 0 1712 0 1760 1 1808 3
1521 1 1569 0 1617 1 1665 0 1713 0 1761 0 1809 1
1522 0 1570 0 1618 2 1666 0 1714 1 1762 1 1810 2
1523 2 1571 4 1619 0 1667 0 1715 5 1763 2 1811 2
1524 0 1572 7 1620 0 1668 0 1716 0 1764 0 1812 0
1525 2 1573 0 1621 3 1669 0 1717 0 1765 0 1813 2
1526 2 1574 4 1622 12 1670 0 1718 0 1766 12 1814 1
1527 4 1575 0 1623 6 1671 0 1719 3 1767 0 1815 0
1528 0 1576 1 1624 0 1672 4 1720 2 1768 19 1816 1
1529 0 1577 1 1625 0 1673 0 1721 2 1769 0 1817 0
1530 0 1578 0 1626 0 1674 0 1722 0 1770 2 1818 3
1531 0 1579 2 1627 0 1675 0 1723 0 1771 1 1819 0
1532 0 1580 0 1628 10 1676 0 1724 5 1772 4 1820 0
1533 0 1581 0 1629 0 1677 0 1725 0 1773 0 1821 6
1534 0 1582 2 1630 4 1678 0 1726 0 1774 1 1822 3
1535 0 1583 0 1631 20 1679 0 1727 0 1775 0 1823 3
1536 0 1584 4 1632 0 1680 5 1728 0 1776 0 1824 2
1537 2 1585 0 1633 0 1681 0 1729 0 1777 1 1825 0
1538 0 1586 4 1634 2 1682 0 1730 2 1778 0
1539 0 1587 1 1635 2 1683 0 1731 2 1779 5
1540 0 1588 1 1636 1 1684 0 1732 1 1780 1
1541 0 1589 4 1637 0 1685 0 1733 23 1781 0
1542 1 1590 16 1638 0 1686 0 1734 2 1782 0
Open in a new tab

Table S2.

Decadal number (TCship) and percentage (TCship%) of Spanish shipwrecks in the Caribbean

Decade No. wrecks Decade No. wrecks
1495 38 1590 0.11
1505 43 1600 0.07
1515 7 1610 0.14
1525 8 1620 0.17
1535 4 1630 0.30
1545 80 1640 0.06
1555 26 1650 0.04
1565 35 1660 0.00
1575 10 1670 0.02
1585 58 1680 0.06
1595 26 1690 0.03
1605 24 1700 0.04
1615 26 1710 0.10
1625 36 1720 0.06
1635 17 1730 0.32
1645 7 1740 0.04
1655 7 1750 0.06
1665 4 1760 0.15
1675 5 1770 0.06
1685 5
1695 5
1705 7
1715 17
1725 30
1735 12
1745 10
1755 7
1765 39
1775 9
1785 20
1795 4
1805 12
1815 18
Open in a new tab

TCship does not overlap in time with the North Atlantic Hurricane Database (HURDAT; 1851–2010 CE) (17) and therefore we used TCsupp (1707–2009 CE) to assess its validity as a CTC proxy. TCsupp is based on 38 south Florida slash pine (Pinus elliottii var. densa) trees from Big Pine Key that show common patterns of suppressed growth (Fig. 1B). The primary causes of tree-growth suppressions in the Florida Keys are high-energy winds and storm-surge-induced saltwater intrusion and TCsupp thus reflects interannual variability in landfalling CTCs. Owing to the geography and regional positioning of the Florida Keys relative to the Caribbean Sea and Atlantic Ocean (Fig. 1A) and to the decadal-scale relationship between landfalling and basin-wide TC dynamics (18, 19), TCsupp can also be interpreted to reflect Caribbean basin-wide TC dynamics.

TCsupp captures 40 of the 44 storms—occurring during 31 out of a total of 34 storm years—that tracked within 160 km of Big Pine Key over the instrumental period (Figs. 1A, 2A, and S1; Tables S3 and S4). Widespread suppressions in tree growth (high TCsupp values) corresponded to storm years (χ2 = 255, P < 0.001) and typically occurred the same year (t0; P < 0.001) or 1 y after (t+1; P < 0.01) a TC event (Figs. 1C and 2A). We interpret the double peak in the superposed epoch analysis (SEA) results to represent the immediate or delayed effect of wind and storm surge damage on tree growth. The timing of a TC event during the hurricane season (August to October) relative to the growing season of trees (February to November) (20) likely dictates whether suppression in growth occurs at t0 or t+1, but a significant pattern was not found.

Fig. 2.

Fig. 2.

Open in a new tab

Instrumental and reconstructed CTC activity. SEA of the TCsupp time series with instrumental [HURDAT, ref. 17; 1851–2010 CE] (A); and reconstructed (TCship; 1495–1825 CE) (B) CTC events. SEA was performed for TCsupp contemporaneous to (year 0) and lagging (1 y before, 5 y following) the TC event years. HURDAT TC event years were selected based on category 1–5 storms that tracked within 160 km radius from Big Pine Key (A). TCship event years were defined as years with five or more shipwrecks (16) (B). The combined event list (C) consisted of TCship, HURDAT, and documented historical storms (14) (1826–1849 CE; asterisks above TCsupp).

Fig. S1.

Fig. S1.

Open in a new tab

SEA of TCsupp with varying HURDAT TC tracking radii. (A) Geographical location of the BPK tree-ring site (red dot), HURDAT-derived (1851–2010 CE) (17) category 1–5 TCs (white lines) that tracked within 100 km (red buffer ring; n = 23) and 300 km (dashed red ring; n = 60) of the site; (B and C) SEA of the TCsupp time series with HURDAT TC event years based on category 1–5 storms that tracked within 100 km (B) and 300 km (C) radii from Big Pine Key. SEA was performed for TCsupp contemporaneous to (year 0) and lagging (1 y before, 5 y following) the HURDAT TC event years.

Table S3.

Annual percentage of suppressed BPK trees (TCsupp)

Year Supp (%) No. Year Supp (%) No. Year Supp (%) No. Year Supp (%) No. Year Supp (%) No.
1707 0 1 1781 63 8 1855 0 12 1929 52 23 2003 0 24
1708 0 1 1782 38 8 1856 0 12 1930 96 23 2004 75 24
1709 0 1 1783 13 8 1857 8 12 1931 61 23 2005 75 24
1710 0 1 1784 0 8 1858 8 12 1932 79 24 2006 61 23
1711 100 1 1785 50 8 1859 17 12 1933 13 23 2007 91 23
1712 0 1 1786 100 8 1860 75 12 1934 13 23 2008 57 23
1713 0 1 1787 38 8 1861 58 12 1935 46 24 2009 75 12
1714 0 1 1788 0 8 1862 25 12 1936 75 24
1715 100 1 1789 11 9 1863 0 12 1937 29 24
1716 100 1 1790 89 9 1864 0 12 1938 0 24
1717 100 1 1791 0 9 1865 75 12 1939 0 24
1718 100 1 1792 0 9 1866 25 12 1940 0 24
1719 100 1 1793 0 9 1867 15 13 1941 92 24
1720 0 1 1794 22 9 1868 8 13 1942 96 24
1721 0 2 1795 56 9 1869 0 13 1943 40 25
1722 0 2 1796 0 9 1870 64 14 1944 36 25
1723 0 2 1797 0 9 1871 29 14 1945 8 25
1724 0 2 1798 0 9 1872 0 14 1946 4 26
1725 50 2 1799 80 10 1873 0 15 1947 65 26
1726 0 2 1800 40 10 1874 0 15 1948 38 26
1727 0 2 1801 50 10 1875 80 15 1949 65 26
1728 0 2 1802 50 10 1876 40 15 1950 8 26
1729 0 2 1803 0 10 1877 27 15 1951 8 26
1730 0 2 1804 0 10 1878 27 15 1952 4 26
1731 0 2 1805 60 10 1879 0 15 1953 4 27
1732 0 2 1806 80 10 1880 0 16 1954 26 27
1733 100 2 1807 0 10 1881 0 16 1955 7 27
1734 100 2 1808 0 10 1882 63 16 1956 4 26
1735 0 2 1809 0 10 1883 88 16 1957 27 26
1736 0 2 1810 0 10 1884 0 16 1958 8 26
1737 0 2 1811 45 11 1885 0 16 1959 4 26
1738 0 2 1812 0 11 1886 81 16 1960 81 26
1739 0 2 1813 0 11 1887 61 18 1961 54 26
1740 0 2 1814 42 12 1888 0 18 1962 38 26
1741 0 2 1815 0 13 1889 0 19 1963 35 26
1742 0 2 1816 54 13 1890 0 19 1964 77 26
1743 0 2 1817 92 13 1891 5 19 1965 27 26
1744 100 2 1818 54 13 1892 89 19 1966 36 25
1745 100 2 1819 38 13 1893 32 19 1967 36 25
1746 50 2 1820 0 13 1894 0 19 1968 0 25
1747 0 3 1821 85 13 1895 26 19 1969 0 26
1748 0 3 1822 92 13 1896 37 19 1970 0 26
1749 0 3 1823 77 13 1897 0 19 1971 12 26
1750 0 4 1824 31 13 1898 0 19 1972 4 26
1751 0 4 1825 42 12 1899 47 19 1973 0 26
1752 100 4 1826 0 12 1900 74 19 1974 4 26
1753 50 4 1827 0 12 1901 26 19 1975 8 26
1754 25 4 1828 75 12 1902 0 19 1976 35 26
1755 0 4 1829 0 12 1903 0 19 1977 4 26
1756 0 5 1830 0 11 1904 100 19 1978 8 26
1757 0 5 1831 55 11 1905 6 18 1979 4 26
1758 0 5 1832 100 11 1906 6 18 1980 8 26
1759 0 5 1833 45 11 1907 28 18 1981 35 26
1760 0 5 1834 27 11 1908 26 19 1982 0 26
1761 0 5 1835 0 11 1909 53 19 1983 0 26
1762 0 5 1836 0 11 1910 25 20 1984 4 26
1763 0 6 1837 0 11 1911 0 18 1985 0 26
1764 0 6 1838 91 11 1912 39 18 1986 0 26
1765 0 6 1839 0 12 1913 0 18 1987 38 26
1766 100 6 1840 83 12 1914 39 18 1988 65 26
1767 29 7 1841 42 12 1915 6 18 1989 65 26
1768 57 7 1842 0 12 1916 0 19 1990 46 26
1769 0 7 1843 25 12 1917 5 19 1991 54 26
1770 0 7 1844 92 12 1918 0 19 1992 96 25
1771 14 7 1845 83 12 1919 81 21 1993 56 25
1772 0 7 1846 33 12 1920 48 21 1994 0 24
1773 0 7 1847 0 12 1921 27 22 1995 8 24
1774 0 7 1848 0 13 1922 0 22 1996 0 24
1775 0 7 1849 23 13 1923 0 22 1997 0 24
1776 100 7 1850 0 13 1924 0 22 1998 25 24
1777 71 7 1851 15 13 1925 0 22 1999 50 24
1778 29 7 1852 83 12 1926 0 22 2000 38 24
1779 0 8 1853 92 12 1927 0 22 2001 50 24
1780 0 8 1854 17 12 1928 0 22 2002 0 24
Open in a new tab

Table S4.

Annual number of category 1–5 storms in HURDAT that tracked within 160 km of BPK

Year No. storms Year No. storms Year No. storms Year No. storms
1851 0 1891 1 1931 0 1971 0
1852 1 1892 0 1932 0 1972 0
1853 0 1893 0 1933 1 1973 0
1854 0 1894 1 1934 0 1974 0
1855 0 1895 0 1935 3 1975 0
1856 0 1896 0 1936 0 1976 0
1857 0 1897 0 1937 0 1977 0
1858 0 1898 0 1938 0 1978 0
1859 0 1899 0 1939 0 1979 0
1860 0 1900 0 1940 0 1980 0
1861 1 1901 0 1941 1 1981 0
1862 0 1902 0 1942 0 1982 0
1863 0 1903 0 1943 0 1983 0
1864 0 1904 2 1944 0 1984 0
1865 1 1905 0 1945 1 1985 0
1866 0 1906 2 1946 0 1986 0
1867 0 1907 0 1947 1 1987 1
1868 0 1908 0 1948 2 1988 0
1869 0 1909 1 1949 0 1989 0
1870 2 1910 1 1950 1 1990 0
1871 0 1911 0 1951 0 1991 0
1872 0 1912 0 1952 0 1992 1
1873 0 1913 0 1953 0 1993 0
1874 0 1914 0 1954 0 1994 0
1875 0 1915 0 1955 0 1995 0
1876 1 1916 0 1956 0 1996 0
1877 0 1917 0 1957 0 1997 0
1878 1 1918 0 1958 0 1998 1
1879 0 1919 1 1959 0 1999 1
1880 0 1920 0 1960 1 2000 0
1881 0 1921 0 1961 0 2001 0
1882 0 1922 0 1962 0 2002 0
1883 0 1923 0 1963 0 2003 0
1884 0 1924 1 1964 1 2004 1
1885 0 1925 0 1965 1 2005 3
1886 1 1926 2 1966 2 2006 0
1887 0 1927 0 1967 0 2007 0
1888 0 1928 0 1968 0 2008 0
1889 0 1929 1 1969 0 2009 0
1890 0 1930 0 1970 0
Open in a new tab

A similar pattern occurs when comparing TCsupp to TCship events: anomalously high TCsupp values occurred in the year of (t0; P < 0.001) or the year after (t+1; P < 0.05) severe TCship events (χ2 = 470, P < 0.0001; Fig. 2B), suggesting a high synchronicity between years of damage at sea (shipwrecks) and on land (pines). Comparable patterns were found when selecting shorter but better replicated TCsupp time series (Fig. S2) and when using a TCship time series based only on shipwrecks that were documented to be caused by storms (Fig. S3). Furthermore, this “double peak” SEA pattern also occurs when using a combined storm event list (1707–2010 CE) including TCship, documented historical storms (14), and HURDAT data (Fig. 2C).

Fig. S2.

Fig. S2.

Open in a new tab

SEA of TCsupp with varying minimum sample depth. SEA of the TCsupp time series with a minimum sample depth of two (1721–1825 CE) (A) and four (1750–1825 CE) (B) suppression samples and TCship event years. SEA was performed for TCsupp contemporaneous to (year 0) and lagging (1 y before, 5 y following) the TCship event years. TCship event years (n = 10 for A and n = 7 for B) were defined as years with five (four) or more shipwrecks in A (B).

Fig. S3.

Fig. S3.

Open in a new tab

SEA and decadal time series of storm caused only TCship. SEA of the TCsupp time series and TCship event years (A) and decadal-scale TCship% (B) based only on shipwrecks documented to be caused by storms. SEA was performed for TCsupp contemporaneous to (year 0) and lagging (1 y before, 5 y following) the TCship event years. TCship event years (n = 8) were defined as years with four or more storm-caused shipwrecks. Gray shaded area in B indicates the MM (1645–1715 CE).

TCsupp and TCship combined provide an annual resolution record of CTC variability that covers the Little Ice Age (ca. 1500–1850 CE), a period characterized by unusually cool global temperatures driven largely by a combination of strong volcanic activity and fluctuations in solar irradiance (21). Over this period, TCship shows distinct decadal-scale variability (Fig. 3A) that corresponds to decadal-scale variations in an independent shipwreck compilation (Fig. S4). By calculating decadal-scale percentages of shipwrecks per number of active North Atlantic Spanish ships (TCship%), we verified that decadal-scale fluctuations in TCship were not a function of fluctuations in Spanish transatlantic shipping activity (Figs. 3B and S5). Both TCship and TCship% reveal a distinct reduction of CTC activity during the MM (Fig. 3 A–D), which is also expressed as a reduction in the number of years with shipwrecks per decade (Fig. 3C). This finding is supported by historical evidence from ship logbooks (14) suggesting increased CTC activity during the 1550s–1640s followed by a marked reduction until the 1720s (Fig. 3B) and by a lake sediment geochemical record from Jamaica (22) (Fig. 3E). Reduced seventeenth-century CTC activity has also been recorded in an array of Caribbean and Gulf of Mexico lower-resolution TC proxies (10–12, 23–25) and is concurrent with increased ocean stratification—possibly linked to reduced TC-induced mixing—near Great Bahama Bank (26) and increased aridity in various proxy records from the circum-Caribbean region (27, 28).

Fig. 3.

Fig. 3.

Open in a new tab

Reduced CTC activity during the Maunder Minimum. (A) Decadal-scale TCship and instrumental (HURDAT) time series sums and TCsupp averages; (B) TCship% and decadal documentary Caribbean storm counts (14); (C) number of years per decade with more than one or four shipwrecks in TCship; (D) spectral solar irradiance (8); (E) tropical Atlantic SST, relative SST proxies (calculated as the residual between tropical Atlantic SST anomalies and averaged Indian, west, and east Pacific Oceans SST anomalies), and storm proxies from Cariaco Basin foraminifera (34), tropical Atlantic corals (35), a Jamaica lake sediment geochemical record (22), mean SST anomalies over the Main Development Region (36), and tropical west Atlantic, west and east Pacific, and Indian Oceans corals (37). Gray-shaded area indicates the MM (1645–1715 CE). Smoothing techniques in E include loess (34) and 30-y spline (36, 37). The extended hurricane activity index in E is calculated over the region 15–25°N and 70–80°W (22).

Fig. S4.

Fig. S4.

Open in a new tab

Comparison of shipwreck numbers from two sources. (A) Comparison of the decadal number of shipwrecks of Spanish origin reported in Marx (16) with the decadal number of Florida shipwrecks (1515–1815 CE) reported in Singer (42). Only Florida wrecks in Marx were used for comparison. (B) Decadal-scale sums for the Singer Florida shipwrecks. Gray shaded area in B indicates the MM (1645–1715 CE).

Fig. S5.

Fig. S5.

Open in a new tab

Variability in transatlantic Spanish shipping activity. Decadal sums of the number of active transatlantic Spanish ships (1590–1780 CE) as compiled from the AGI.

Modern CTC variability is influenced by four key large-scale climate state variables (3, 29–31): (i) absolute tropical Atlantic sea surface temperatures (SSTs) (32); (ii) tropical Atlantic SSTs relative to tropical mean SSTs (relative SSTs) (33); (iii) the El Niño–Southern Oscillation (ENSO); and (iv) the North Atlantic Oscillation (NAO). A combination of proxy records suggests that the states of these climate factors converged during the Little Ice Age, and the MM in particular, to create a drastically unfavorable environment for CTC development. The distinct reduction in CTC activity during the MM can be attributed to cool absolute SSTs in the Caribbean (34, 35) and in the main development region (36): warmer ocean waters provide more thermal and kinetic energy and thus increase the potential for a greater number of severe storms (31) (Fig. 3E). Moreover, west Atlantic SST anomalies during the MM were on average −0.21 °C cooler than SST anomalies in the other (Indian, west Pacific, and east Pacific) tropical ocean basins (based on data from ref. 37; Fig. 3E), resulting in cool relative SSTs. Anomalously cool absolute and relative SSTs resulting from generally low sunspot activity during the MM (8) could therefore have impacted the thermodynamic environment, vertical wind shear, and tropospheric stability, thus creating conditions adverse to CTC development.

The MM was likely also a period of broad-scale climate shifts. Model simulations suggest a southerly position of the intertropical convergence zone, an El Niño–like mean state, and negative NAO conditions (7) during the late seventeenth century, that have been confirmed by proxy evidence (35, 36, 38, 39). Modern CTC climatology—with generally suppressed TC activity during El Niño years and negative NAO phases—alludes to a likely contribution of the state of these major circulation patterns to TC suppression in the Caribbean during the MM. The majority of sedimentary TC proxies on the northeastern US coast, however, did not record a TC suppression during the MM (9, 36), suggesting an important influence of regional oceanic conditions and shifting genesis locations as well as a potential modulation—and northeastward redirection—of the prevailing storm tracks related to the above-described changes in atmospheric circulation patterns (30).

Our high-resolution CTC proxy records suggest that a combination of cold Caribbean SSTs, El Niño–like conditions, and a negative NAO during the MM might have provided a pathway for an unfavorable effect of reduced solar activity on decadal-scale CTC activity. However, on interannual and even daily time scales, the solar–TC relationship is complex, spatially explicit, and modulated by oceanic heat content (40). In regions with high oceanic heat content (e.g., the twentieth-century Caribbean), TC activity is typically reduced during years of strong solar activity, when tropopause temperatures are warmed and the vertical temperature differential responsible for tropospheric convection is weakened (41). In contrast, in regions where ocean heat content is limited in summer (e.g., the twentieth-century east Atlantic), strong solar activity can provide a marginal SST increase necessary for cyclogenesis, resulting in more TCs. This modulation of the solar–TC relationship by ocean heat content can also be applied to decadal temporal scales rather than spatial scales: tropical western Atlantic SST reconstructions suggest that MM Caribbean SSTs were up to 1 °C cooler than present (34, 37; Fig. 3E) and as such represent a potential MM shift in the ocean heat budget and the solar–TC relationship.

Our findings suggest an abrupt CTC suppression in response to a severe reduction in solar irradiance and thus provide a benchmark for modeling studies of North Atlantic TC response to radiative forcing. The unequivocal strength and sign of the CTC response to reduced solar activity imply modulation by synchronized oceanic and atmospheric circulation patterns. To further improve our understanding and to constrain estimates of future North Atlantic TC activity, additional high-resolution proxy records and targeted climate modeling experiments are needed that can support a process-based investigation of the responses of North Atlantic TC activity, SSTs, ENSO, and NAO to radiative forcing and future climate change.

Data

TCship.

We compiled a total of 656 Spanish shipwreck events (1495–1825 CE) that occurred in the Caribbean realm and were included in Marx (16). This data set is not complete, as some shipwrecks were never documented and some documents likely lost due to deterioration with age, war, or fire (15), but it is a comprehensive collection of shipwrecks in the Americas that includes ca. 4,000 entries cataloged by year and location. Ships of British, French, Dutch, and other nationalities also crossed the Atlantic and wrecked in the Caribbean from the seventeenth century onward, but to avoid trends in TCship due to increasing numbers of ships, we compiled only shipwrecks of Spanish origin. We compiled all Spanish shipwreck events that were recorded to have occurred (i) due to storm activity (66%) or for unspecified causes (34%), (ii) during the hurricane season (July–November) or with unknown seasonality, and (c) in Florida (21%), on the Atlantic Coast of Mexico (25%), in Hispaniola (22%), Cuba (13%), the Lesser Antilles (5%), and Puerto Rico, Texas, South Carolina, Louisiana, Mississippi, Jamaica, the Cayman Islands, Bermuda, and the Bahamas (less than 5% each; Fig. 1A). Quantities were recorded as “many,” “several,” or in similar terms on six occasions in the record and we counted these as five shipwrecks, which is the median number of ships reported to have wrecked during storms when more than one ship wrecked (range: 2–29 shipwrecks; n = 39). The TCship time series thus consists of the number of Spanish ships that wrecked per year in the Caribbean region due to storm activity or for unspecified causes (Fig. 2).

We calculated decadal sums based on TCship (Fig. 3A) and verified that decadal-scale fluctuations were not a function of fluctuations in the number of active transatlantic Spanish ships by calculating decadal-scale percentages of shipwrecks per number of active ships in the Caribbean (TCship%; Fig. 3B). For this purpose, we compiled a time series of the total number of ships departing from Seville to the Caribbean and Gulf of Mexico during the period 1511–1784 CE based on the General Archive of the Indies (AGI) (pares.mcu.es). Within the AGI, we tallied each ship departure among the 643 registers in the Registros de Ida. Because of discrepancies in the AGI related to low ship log entries during early and late decades, we calculated annual ship occurrences for the period 1590–1780 CE only (n = 3,881; Fig. S5). The data set was quality checked for potential duplicate records by considering the ship name, type, and captain name, and duplicate records were excluded from the final time series. The record of annual ship counts was binned into decadal sums (Fig. S5) and used to calculate TCship% (Fig. 3B).

We further compared the decadal-scale TCship time series to a shipwreck time series (1515–1815 CE; number of shipwrecks = 232) for Florida based on the data compiled in ref. 42. This decadal-scale Florida time series correlated strongly (r = 0.77, n = 31, P < 0.001) with the Florida component of the TCship time series (Fig. S4).

TCsupp.

We aimed our sampling strategy at using radial growth anomalies from Pinus elliottii var. densa (hereafter slash pine) trees to detect TCs in the Florida Keys/Caribbean region. For this purpose, we collected partial (live trees) or full (dead trees) cross-sections from 38 mature old-growth pines that showed no signs of visible injury to the tree (i.e., fire scar, beetle gallery, lightning scar) and were located at the highest elevations (≥ 2 m a.s.l.) within the island interior of Big Pine Key (BPK; Fig. 1A). Trees located at higher BPK elevations are able to survive multiple storm surge events because salt water has a short residence time at the surface (43). Moreover, BPK slash pine trees exhibit a different growth form compared with individuals on mainland Florida and windthrow-induced mortality associated with TCs is uncommon. We used standard dendrochronological techniques (44) for sample preparation, measurement, and cross-dating of the tree-ring data.

Slash pine ring-width data can contain a weak late summer (September) moisture availability signal (45), but this was not the case at the BPK site (Fig. S6) and we focused our analysis on tree-growth suppressions (Fig. 1B) in the slash pine data set. We expected a TC signal due to loss of tree biomass (leaves, branches) to be manifested within the tree-ring record as a period of tree-growth suppression (46) and therefore defined suppression events as periods in which raw ring width was ≥25% below the 10-y running mean of all growth rings (n = 1,228 for 38 tree samples over the period 1707–2009 CE). We than calculated an annual time series TCsupp as the percentage of samples in a given year t that contained a suppression. Widespread tree growth suppressions (defined as suppression in ≥75% of samples) occurred during wet [positive Palmer Drought Severity Index (PDSI)] summers, most likely caused by TC-induced rainfall (Fig. 1C).

Fig. S6.

Fig. S6.

Open in a new tab

BPK climate–growth relationships. Pearson correlation coefficients (1895–2009 CE) between the BPK tree-ring width chronology and monthly precipitation, temperature, and PDSI values for Florida climate division 7 (Florida Keys). Correlation coefficients were calculated for previous year’s (uppercase) February through current year (lowercase) December.

Methods

The relationship between TCsupp and instrumental and reconstructed TC events was evaluated in a contingency analysis and in an SEA. We compared the frequency distribution of growth suppressions in TCsupp to storms in the HURDAT (17) and shipwrecks in the TCship time series in a contingency analysis and calculated statistical significance levels using a Pearson’s χ2 (χ2) test. The HURDAT time series (1851–2010 CE) consisted of the number of category 1–5 storms that tracked within 160 km of BPK per year (Fig. 2A). Large storms within this radius are documented to cause major storm surges on BPK (47). For instance, Hurricane Wilma passed at 145 km from BPK and caused growth suppression in 75% of the TCsupp samples in 2005 (Table S3).

We compared lagged relationships in an SEA by superposing a window of contemporaneous and lagged annual growth suppressions over each HURDAT (1851–2010 CE) or TCship-based (1707–1825 CE) TC event year (48). Event years were defined as years in which category 1–5 TCs tracked within 160 km of BPK (n = 34) for HURDAT and as years with five or more shipwrecks events (n = 9; Fig. 2B) for TCship. Results are consistent when using smaller (100 km) and larger (300 km) tracking radii for HURDAT TCs (Fig. S1) or shorter but better replicated TCsupp time series (Fig. S2). In addition to this, we used a combined event list of TCship, documented TCs/storms from ref. 14 (n = 2; 1826–1849 CE), and HURDAT (Fig. 2C). Monte Carlo simulations were used (n = 1,000) to develop bootstrapped confidence intervals to determine whether a significantly higher than average number of trees showed growth suppression in a 7-y window (t–1 through t+5) centered around the TC event year (t0).

Acknowledgments

We thank Justin L. Hart and Mary Glueck for assistance with data collection and development. We are grateful to P. J. Baker, L. Bakkensen, H. Diaz, and J. Esper for providing suggestions. G.L.H. was supported by a grant from the Office of Research from the University of Southern Mississippi, by the National Science Foundation under Grants 1002479 and 0538420, and by the United States Fish and Wildlife Service. M.D.-D. was supported by a University of Arizona Agnese N. Haury Visiting Fellowship and by a Marie Sklodowska Curie Innovative Training Networks Fellowship (ForSEAdiscovery project, PITN-2013-GA-607545). The idea for this collaborative project originated during the Second American Dendrochronology Conference.

Footnotes

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1519566113/-/DCSupplemental.

References

  • 1.Pielke RA, et al. Normalized hurricane damage in the United States: 1900-2005. Nat Hazards Rev. 2008;9:29–42. [Google Scholar]
  • 2.Mendelsohn R, Emanuel K, Chonabayashi S, Bakkensen L. The impact of climate change on global tropical cyclone damage. Nat Clim Chang. 2012;2(3):205–209. [Google Scholar]
  • 3.Knutson TR, et al. Tropical cyclones and climate change. Nat Geosci. 2010;3(3):157–163. [Google Scholar]
  • 4.Villarini G, Vecchi GA. Twenty-first-century projections of North Atlantic tropical storms from CMIP5 models. Nat Clim Chang. 2012;2(8):604–607. [Google Scholar]
  • 5.Landsea CW, Harper BA, Hoarau K, Knaff JA. Climate change. Can we detect trends in extreme tropical cyclones? Science. 2006;313(5786):452–454. doi: 10.1126/science.1128448. [DOI] [PubMed] [Google Scholar]
  • 6.Mann ME, Woodruff JD, Donnelly JP, Zhang Z. Atlantic hurricanes and climate over the past 1,500 years. Nature. 2009;460(7257):880–883. doi: 10.1038/nature08219. [DOI] [PubMed] [Google Scholar]
  • 7.Shindell DT, Schmidt GA, Mann ME, Rind D, Waple A. Solar forcing of regional climate change during the Maunder Minimum. Science. 2001;294(5549):2149–2152. doi: 10.1126/science.1064363. [DOI] [PubMed] [Google Scholar]
  • 8.Lean J. Evolution of the sun’s spectral irradiance since the Maunder Minimum. Geophys Res Lett. 2000;27(16):2425–2428. [Google Scholar]
  • 9.Donnelly JP, et al. Climate forcing of unprecedented intense‐hurricane activity in the last 2000 years. Earths Futur. 2015;3(2):49–65. [Google Scholar]
  • 10.Denommee KC, Bentley SJ, Droxler AW. Climatic controls on hurricane patterns: A 1200-y near-annual record from Lighthouse Reef, Belize. Sci Rep. 2014;4:3876. doi: 10.1038/srep03876. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Lane P, Donnelly JP, Woodruff JD, Hawkes AD. A decadally-resolved paleohurricane record archived in the late Holocene sediments of a Florida sinkhole. Mar Geol. 2011;287(1-4):14–30. [Google Scholar]
  • 12.Frappier AB, et al. Two millennia of tropical cyclone-induced mud layers in a northern Yucatan stalagmite: Multiple overlapping climatic hazards during the Maya Terminal Classic “megadroughts”. Geophys Res Lett. 2014;41(14):5148–5157. [Google Scholar]
  • 13.Chenoweth M, Divine D. A document-based 318-year record of tropical cyclones in the Lesser Antilles, 1690-2007. Geochem Geophys Geosyst. 2008;9(8):Q08013. [Google Scholar]
  • 14.Garcia-Herrera R, Gimeno L, Ribera P, Hernandez E. New records of Atlantic hurricanes from Spanish documentary sources. J Geophys Res. 2005;110(D3):D03109. [Google Scholar]
  • 15.Rappaport EN, Fernandez-Partagas JJ. History of the deadliest Atlantic tropical cyclones since the discovery of the New World. In: Diaz HF, Pulwarty RS, editors. Hurricanes: Climate and Socioeconomic Impacts. Springer-Verlag; Berlin: 1997. pp. 93–108. [Google Scholar]
  • 16.Marx RF. Shipwrecks in the Americas. Crown Publishers; New York: 1983. p. 482. [Google Scholar]
  • 17.Landsea CW, Franklin JL. Atlantic hurricane database uncertainty and presentation of a new database format. Mon Weather Rev. 2013;141(10):3576–3592. [Google Scholar]
  • 18.Kozar ME, Mann ME, Emanuel KA, Evans JL. Long-term variations of North Atlantic tropical cyclone activity downscaled from a coupled model simulation of the last millennium. J Geophys Res Atmos. 2013;118(24):13383–13392. [Google Scholar]
  • 19.Reed AJ, Mann ME, Emanuel KA, Titley DW. An analysis of long-term relationships among count statistics and metrics of synthetic tropical cyclones downscaled from CMIP5 models. J Geophys Res Atmos. 2015;120(15):7506–7519. [Google Scholar]
  • 20.Harley GL, Grissino-Mayer HD, Franklin JA, Anderson C, Kose N. Cambial activity of Pinus elliottii var. densa reveals influence of seasonal insolation on growth dynamics in the Florida Keys. Trees Struct Funct. 2012;26(5):1449–1459. [Google Scholar]
  • 21.Crowley TJ. Causes of climate change over the past 1000 years. Science. 2000;289(5477):270–277. doi: 10.1126/science.289.5477.270. [DOI] [PubMed] [Google Scholar]
  • 22.Burn MJ, Palmer SE. Atlantic hurricane activity during the last millennium. Sci Rep. 2015;5:12838. doi: 10.1038/srep12838. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Brandon CM, Woodruff JD, Lane DP, Donnelly JP. Tropical cyclone wind speed constraints from resultant storm surge deposition: A 2500 year reconstruction of hurricane activity from St. Marks, FL. Geochem Geophys. 2013;14(8):2993–3008. [Google Scholar]
  • 24.Ercolani C, Muller J, Collins J, Savarese M, Squiccimara L. Intense southwest Florida hurricane landfalls over the past 1000 years. Quat Sci Rev. 2015;126:17–25. [Google Scholar]
  • 25.Donnelly JP, Woodruff JD. Intense hurricane activity over the past 5,000 years controlled by El Niño and the West African monsoon. Nature. 2007;447(7143):465–468. doi: 10.1038/nature05834. [DOI] [PubMed] [Google Scholar]
  • 26.Lund DC, Lynch-Stieglitz J, Curry WB. Gulf Stream density structure and transport during the past millennium. Nature. 2006;444(7119):601–604. doi: 10.1038/nature05277. [DOI] [PubMed] [Google Scholar]
  • 27.Burn MJ, Palmer SE. Solar forcing of Caribbean drought events during the last millennium. J Quat Sci. 2013;29(8):827–836. [Google Scholar]
  • 28.Lane CS, Horn SP, Orvis KH, Thomason JM. Oxygen isotope evidence of Little Ice Age aridity on the Caribbean slope of the Cordillera Central, Dominican Republic. Quat Res. 2011;75(3):461–470. [Google Scholar]
  • 29.Sabbatelli TA, Mann ME. The influence of climate state variables on Atlantic Tropical Cyclone occurrence rates. J Geophys Res Atmos. 2007;112(D17) doi: 10.1029/2007jd008385. [DOI] [Google Scholar]
  • 30.Kozar ME, Mann ME, Camargo SJ, Kossin JP, Evans JL. Stratified statistical models of North Atlantic basin-wide and regional tropical cyclone counts. J Geophys Res Atmos. 2012;117:D18103. [Google Scholar]
  • 31.Goldenberg SB, Landsea CW, Mestas-Nunez AM, Gray WM. The recent increase in Atlantic hurricane activity: Causes and implications. Science. 2001;293(5529):474–479. doi: 10.1126/science.1060040. [DOI] [PubMed] [Google Scholar]
  • 32.Webster PJ, Holland GJ, Curry JA, Chang HR. Changes in tropical cyclone number, duration, and intensity in a warming environment. Science. 2005;309(5742):1844–1846. doi: 10.1126/science.1116448. [DOI] [PubMed] [Google Scholar]
  • 33.Vecchi GA, Swanson KL, Soden BJ. Climate change. Whither hurricane activity? Science. 2008;322(5902):687–689. doi: 10.1126/science.1164396. [DOI] [PubMed] [Google Scholar]
  • 34.Black DE, et al. An 8-century tropical Atlantic SST record from the Cariaco Basin: baseline variability, twentieth-century warming, and Atlantic hurricane frequency. Paleoceanography. 2007;22(4):PA4204. [Google Scholar]
  • 35.Kilbourne KH, Alexander MA, Nye JA. A low latitude paleoclimate perspective on Atlantic multidecadal variability. J Mar Syst. 2014;133:4–13. [Google Scholar]
  • 36.Mann ME, et al. Global signatures and dynamical origins of the Little Ice Age and Medieval Climate Anomaly. Science. 2009;326(5957):1256–1260. doi: 10.1126/science.1177303. [DOI] [PubMed] [Google Scholar]
  • 37.Tierney JE, et al. Tropical sea surface temperatures for the past four centuries reconstructed from coral archives. Paleoceanography. 2015;30(3):226–252. [Google Scholar]
  • 38.Trouet V, et al. Persistent positive North Atlantic oscillation mode dominated the Medieval Climate Anomaly. Science. 2009;324(5923):78–80. doi: 10.1126/science.1166349. [DOI] [PubMed] [Google Scholar]
  • 39.Haug GH, Hughen KA, Sigman DM, Peterson LC, Röhl U. Southward migration of the intertropical convergence zone through the Holocene. Science. 2001;293(5533):1304–1308. doi: 10.1126/science.1059725. [DOI] [PubMed] [Google Scholar]
  • 40.Hodges RE, Elsner JB. The spatial pattern of the sun-hurricane connection across the North Atlantic. ISRN Meteorol. 2012;2012:1–9. [Google Scholar]
  • 41.Hodges RE, Elsner JB. Evidence linking solar variability with US hurricanes. Int J Climatol. 2011;31(13):1897–1907. [Google Scholar]
  • 42.Singer SD. Shipwrecks of Florida. Pineapple Press; Sarasota, FL: 1992. [Google Scholar]
  • 43.Ross MS, Obrien JJ, Sternberg LDL. Sea-level rise and the reduction in pine forests in the Florida Keys. Ecol Appl. 1994;4(1):144–156. [Google Scholar]
  • 44.Stokes MA, Smiley TL. An Introduction to Tree-Ring Dating. Univ of Chicago Press; Chicago: 1968. [Google Scholar]
  • 45.Harley GL, Grissino-Mayer HD, Horn SP. The dendrochronology of Pinus elliottii in the lower Florida Keys: Chronology development and climate response. Tree Ring Res. 2011;67(1):39–50. [Google Scholar]
  • 46.Hart JL, Grissino-Mayer HD. Vegetation patterns and dendroecology of a mixed hardwood forest on the Cumberland Plateau: Implications for stand development. For Ecol Manage. 2008;255(5-6):1960–1975. [Google Scholar]
  • 47.Saha AK, et al. Sea level rise and South Florida coastal forests. Clim Change. 2011;107(1-2):81–108. [Google Scholar]
  • 48.Haurwitz MW, Brier GW. A critique of the Superposed Epoch Analysis method - Its application to solar weather relations. Mon Weather Rev. 1981;109(10):2074–2079. [Google Scholar]

Articles from Proceedings of the National Academy of Sciences of the United States of America are provided here courtesy of National Academy of Sciences

RESOURCES