Three-hundred-year hydrological changes in a subtropical estuary, Rookery Bay (Florida): Human impact versus natural variability


  • Timme H. Donders,

    1. Palaeoecology, Institute of Environmental Biology, Laboratory of Palaeobotany and Palynology, Science Faculty, Utrecht University, Utrecht, Netherlands
    Search for more papers by this author
  • P. Martijn Gorissen,

    1. Palaeoecology, Institute of Environmental Biology, Laboratory of Palaeobotany and Palynology, Science Faculty, Utrecht University, Utrecht, Netherlands
    2. Netherlands Organization for Applied Scientific Research TNO, National Geological Survey, Utrecht, Netherlands
    Search for more papers by this author
  • Francesca Sangiorgi,

    1. Palaeoecology, Institute of Environmental Biology, Laboratory of Palaeobotany and Palynology, Science Faculty, Utrecht University, Utrecht, Netherlands
    Search for more papers by this author
  • Holger Cremer,

    1. Netherlands Organization for Applied Scientific Research TNO, National Geological Survey, Utrecht, Netherlands
    Search for more papers by this author
  • Friederike Wagner-Cremer,

    1. Palaeoecology, Institute of Environmental Biology, Laboratory of Palaeobotany and Palynology, Science Faculty, Utrecht University, Utrecht, Netherlands
    Search for more papers by this author
  • Vicky McGee

    1. Rookery Bay National Estuarine Research Reserve, Florida Department of Environmental Protection, Naples, Florida, USA
    Search for more papers by this author


The coastal wetland ecosystems in Florida are highly sensitive to changes in freshwater budget, which is driven by regional sea surface temperature, tropical storm activity, and the El Niño–Southern Oscillation (ENSO). Although studying Florida wetlands is pivotal to the understanding of these interacting climate systems, wetland dynamics have been severely altered by recent land use and drainage activities. To gather insights into the natural variability of the coastal ecosystems in Florida versus the effects of anthropogenic impact in the area, we present a 300-year record of changes in the hydrological cycle from a shallow subtropical estuary (Rookery Bay) on the western shelf of Florida, Gulf of Mexico. Palynological (pollen and organic-walled dinoflagellate cysts), diatom, and sedimentological analyses of a sediment core reveal significant changes in past runoff and wetland development. The onset and development of human impact in Florida are evident from high influx of Ambrosia pollen at about A.D. 1900, indicative of land clearance and disturbed conditions. To date, this is the southernmost record of Ambrosia increase related to human impact in the United States. Wetland drainage and deforestation since A.D. 1900 are evident from the reduced freshwater wetland and pine vegetation, and lower abundances of phytoplankton species indicative of lagoonal and brackish conditions. High runoff after A.D. 1900 relates to increased erosion and may correspondingly reflect higher impact from hurricane landfalls in SW Florida. Several phases with high siliciclastic input and greater wetland pollen abundance occur that predate the human impact period. These phases are interpreted as periods with higher runoff and are likely related to regional longer-term climate variability.

1. Introduction

Florida is sensitive to both low-latitude climate systems, such as the El Niño–Southern Oscillation [Vega et al., 1998] and the occurrence of Atlantic hurricanes, and forcing from higher latitudes by surface compensating flow through the Florida Straight in response to North Atlantic Deep Water (NADW) formation [Lund and Curry, 2004]. The Florida peninsula is hence a key location for studying the interaction of high- and low-latitude climate systems and to understand past and present hydrological changes. However, human impact has heavily affected Florida wetland ecosystems, thus impeding the possibility to link wetlands recent development to natural climate variability. It is therefore essential to retrieve sediment records, which predate human impact, not only to study past natural variability and recent human influence, but also to evaluate the sensitivity of a natural and an impacted ecosystem to hurricane occurrences. This information will eventually benefit ecosystem restoration programs, such as the large hydrologic Comprehensive Everglades Restoration Plan (CERP) that was started in A.D. 2000 [Sklar et al., 2005]. The past activity of coupled land-sea climate systems are best studied in sediments from shallow marine sites that provide archives of both marine and terrestrial environments. Natural hydrology and sheet flow across southern Florida have been impacted severely by drainage activities leading to altered water levels in wetlands [e.g., Willard et al., 2001a; Donders et al., 2005b], reduced freshwater runoff into estuaries [e.g., Brewster-Wingard and Ishman, 1999; Swart et al., 1999; Surge et al., 2003; Xu et al., 2007], and land development and forest clearance [e.g., Burns, 1984; Snyder and Davidson, 1994; Shirley and Brandt-Williams, 2003]. Most quantitative ecosystem studies have been conducted since the 1950s [Fourqurean and Robblee, 1999] when human impact on the southern Florida Watershed had already started [Light and Dineen, 1994]. Documentation of pre-anthropogenic conditions and long-term natural changes are therefore dependent on paleoecological and paleoclimatological studies, most of which focused so far on the Everglades and Florida Bay area (Figure 1), while the areas west of the national parks have received little attention [Surge et al., 2003].

Figure 1.

Geographical location of the Rookery Bay National Estuarine Research Reserve in southern Florida, USA. WCA, water conservation area. Location of core ROB03 (inset) is indicated by the black star. The simplified map of dominant vegetation types in the Rookery Bay area is based on data from Shirley and Brandt-Williams [2003].

Here we present a multiproxy record of organic (pollen and organic-walled dinoflagellate cysts, hereafter dinocysts) and siliceous (diatoms) microfossils, and sedimentological properties (organic carbon and carbonate content) that signals relative runoff, salinity and wetland vegetation changes during the past ∼300 years in a shallow estuary (Rookery Bay) in SW Florida. Palynological indicators are useful to document the beginning of widespread human impact [e.g., Willard et al., 2003; Donders et al., 2005b], and therefore permit to distinguish between periods of natural variability and human-induced vegetation, runoff and salinity changes. Here, we also examine how historically known hurricanes since A.D. 1850 in the direct vicinity of the site are preserved as signals in the sediment composition, and compare the local hurricane occurrence with a basin-wide record of Atlantic hurricane frequencies [Nyberg et al., 2007]. We further discuss possible relations between pre-human impact variations in the proxy data and the strength of regional climate factors. This study presents the initial results of the Utrecht University “Hurricanes and Global Change” project that is scheduled to run until 2012.

2. Geological and Environmental Setting

The study site, Rookery Bay, lies on the west central Florida shelf at the transition between a barrier-island chain to the north and the sediment-starved Ten Thousand Islands mangrove coast to the south. The west central Florida shelf is a broad Cenozoic carbonate platform with a quartz dominated coastal zone. The shelf has a low to moderate wave climate and tidal range owing to the low slope gradient [Cremer et al., 2007; Hine et al., 2003] (Figure 1). The subtropical climate of southeastern Florida has a mean annual temperature of ∼23°C with low seasonality. Annual rainfall is high (∼1350 mm/a) and shows strong seasonality with relatively dry winters and wet summers [Cahoon and Lynch, 1997] (KNMI climate explorer, The climate in Florida is significantly influenced by the El Niño-Southern Oscillation (ENSO) [Vega et al., 1998; Cronin et al., 2002], leading to significantly higher winter precipitation during El Niño years. The late Holocene intensification of ENSO has created wetter conditions in Florida [Donders et al., 2005a, 2008] since winter precipitation largely controls the length of the inundation period in Florida swamps [Donders et al., 2005b].

The Quaternary geological units in the immediate surrounding of Rookery Bay are quartz rich and poor in clays and organic matter [Shirley and Brandt-Williams, 2003]. The Rookery Bay National Estuarine Research Reserve (RB NERR, hereafter RB) is a relatively pristine area bordering the freshwater mixed-cypress swamp forests of the Fakahatchee Strand Preserve State Park (FSPSP) and Big Cypress National Preserve (BCNP) (Figure 1) toward the east, and the city of Naples toward the north. The bay is a shallow (∼1 m average depth), non-stratified, mesohaline estuary with an annual mean tidal range of 0.6 m. The main freshwater tributary, Henderson Creek, has an average annual discharge of 0.68 m3 s−1 into RB, which provides a relatively low amount of terrigenous sediment [Cahoon and Lynch, 1997].

RB and the surrounding rivers and tidal creeks are fringed by mangrove vegetation (Figure 1), which form tall and complex communities that grow on thick sediment deposits. Mangrove zonation is tied to salinity gradients and edaphic conditions that determine the distribution relative to the coast. Shoreline species are red and black mangrove (Rhizophora mangle and Avicennia germinans), while white mangrove (Laguncularia racemosa), buttonwood (Conocarpus erecta), and saltwort (Batis maritima) are found on higher elevations and backswamps. Behind the mangroves, saltmarsh taxa like leather fern (Acrostichum sp.), saltwort, and glasswort (Salicornia perennis) occur together with grass and sedge communities [Tomlinson, 1986; Willard et al., 2001b]. The mangrove stands and salt marshes grade into freshwater marshes and pine flatwoods further inland, which consist of Pinus elliottii var. densa and an understory of saw palmetto (Serenoa repens), gallberry (Ilex glabra) and wax myrtle (Myrica cerifera) (Figure 1). The pine flatwoods are susceptible to invasion of non-native species, particularly melaleuca (Melaleuca quinquenervia) and Brazilian pepper (Schinus terebinthifolius) [Shirley and Brandt-Williams, 2003]. The inland pine flatwoods alternate with wetter Taxodium distichum-dominated forested wetlands. Drier inland sites are hardwood hammocks with cabbage palm (Sabal palmetto) and live oak (Quercus virginiana), and areas with dry scrub vegetation with scrub oaks (Quercus geminata, Q. myrtifolia, Q. chapmanii) and rosemary bushes (Ceratiola ericoides) [Myers and Ewel, 1990; Shirley and Brandt-Williams, 2003; Donders et al., 2005b]. Toward the north and east agricultural and built-up disturbed areas now occur (Figure 1).

3. Material and Methods

3.1. Core Description and Chronology

A sediment core was obtained using a modified Livingstone piston corer in February 2003. Water depth at the coring site (81°44′42″W, 26°01′40″N) (Figure 1) was 120 cm. Core ROB03 consists of 4 partly overlapping sections spliced to reach a total depth of (after correction) 290 cm below the seafloor (Figure 2a). The upper 130 cm consists of sandy calcareous mud rich in shell fragments. Lithology changes into organic-rich mud at 130 cm and below 148 cm into peaty sediment intercalated with sandy layers. Within the peat section a second organic-rich muddy interval occurs between 228 and 238 cm.

Figure 2.

(a) Core sections and lithology. Total core depth was corrected on the basis of the lithological correlations between the overlapping sections. (b) Radiocarbon (closed squares) and biostratigraphic ages (open squares) used in the age-depth model (see Table 1).

The core has been dated by five AMS 14C measurements on both peat and shell (gastropod) fragments (Table 1). 14C ages were calibrated with CALIB 5.0.2 [Stuiver and Reimer, 1993; Stuiver et al., 2006] and the IntCal04 data set of Reimer et al. [2004a] for the peat samples. The gastropod samples were calibrated with the marine04 calibration curve [Hughen et al., 2004]. Following Stuiver and Braziunas [1993], we applied a marine reservoir correction (ΔR) of 33 ± 16 years based on a marine reservoir age estimate from Florida Bay (The Rocks, offshore the Florida Keys; The uppermost calibrated 14C age points to a modern age and has been converted to calendar years with the 20th century bomb carbon anomaly [see, e.g., Donders et al., 2004] using the CALIBomb program of Reimer et al. [2004b]. In addition, two pollen dates that mark the first occurrences of exotic plant species are used as age calibration points. Schinus terebinthifolius (Brazilian peppertree), and Casuarina equisetifolia (Australian pine) were introduced to Florida in ∼A.D. 1845 and ∼A.D. 1900, respectively [Alexander and Crook, 1974] (IFAS Centre for Aquatic and Invasive Plants).

Table 1. Radiocarbon Age Calibration
Depth (cm)MaterialUtCδ13CMass (mg)Radiocarbon AgesCalibration CurveCalibrated Range
14C Age ± 1σΔR ± 1σF14CaCal B.P. ± 1σ
  • a

    F14C = EXP-(14Cage / 8033).


3.2. Sample Processing

In this study the top 60 cm of section ROB03 was continuously sampled at 2-cm intervals for palynological and loss-on-ignition (LOI) [Dean, 1974] analyses. In addition, samples for quantitative diatom analyses were taken at 10-cm intervals. LOI samples were freeze-dried overnight, weighed, and weight loss after combustion at 550°C and 950°C was measured to determine the relative amounts of organic carbon and carbonate, respectively. Samples were heated for 4 h each time and were allowed to cool in an exicator before weighing to prevent moisture to condensate. Measurement errors for this method do generally not exceed 1% of the total (dry) sample weight [Heiri et al., 2001].

Samples for palynology were freeze-dried, weighed and spiked with a known amount of Lycopodium clavatum spores to calculate concentrations prior to further processing according to standard methods [Wood et al., 1996]. Carbonates were removed with HCl (10%), and silicates with HF (40%). A second HCl treatment was used to remove fluoride gels before samples were treated in an ultrasonic bath and sieved with a 250 μ and 10 μ mesh for removal of coarse and fine fractions, respectively. Residues were mounted on glass slides in glycerin for light microscopic examination. The pollen and dinocyst contents of the slides were analyzed quantitatively under a Leitz light microscope at 400× magnification. Pollen identification follows Willard et al. [2004] (online at and an unpublished Florida Pollen database by T. H. Donders (Utrecht University). Dinocyst identification follows Rochon et al. [1999], Marret and Zonneveld [2003], Fensome and Williams [2004], and Cremer et al. [2007]. Diatom slides were prepared and analyzed according to the procedure described by Cremer et al. [2007]. Diatom identification was carried out at 1000 × magnification using an Olympus BX51 microscope equipped with differential interference contrast.

4. Results

4.1. Chronology

Radiocarbon ages indicate a basal age of 4600 years cal B.P. at 290 cm depth. The transition from peat to marine sediments occurs approximately around 2650 years cal B.P. based on linear interpolation of the ages. We suspect the lowermost radiocarbon age to be anomalously young. If the age was correct, the sedimentation rate between 290 cm and 220 cm would be much higher than the overlying units. Unfortunately, we have no further age control to verify this. The CALIBomb calibration of the uppermost 14C age at 10 cm results in several calibration intervals (Table 1). The record of S. terebinthifolius is sparse but its first occurrences down core clearly predate Casuarina (Figure 4), and just follow the increase in Ambrosia, which is characteristic for disturbed environments (see section 5.1). The first occurrence of S. terebinthifolius likely reflects the initial ∼A.D. 1845 introduction into Florida (IFAS Centre for Aquatic and Invasive Plants), while it only shows a significant peak in the second half of the 20th century, in line with documentary evidence of broad expansion in that period [Morton, 1978].

The stratigraphic pollen ages indicate that the radiocarbon age at 10 cm must post-date A.D. 1900. Therefore, the most recent of the multiple calibration intercepts can be selected, resulting in an age of A.D. 1952 at the initial rise of the bomb carbon anomaly at 10 cm depth. The uppermost 80 cm contain more age control points and sample ages in between 0–80 cm depth are interpolated using a third-order polynomial fit with a set intercept at A.D. 2003 (−53 B.P.) (Figure 2).

4.2. Loss-On-Ignition Data

The LOI results (Figure 3) reveal significant variations in the amount of carbonate and organic carbon, which are closely correlated (r = 0.83). Each fraction varies between 6 and 11% of the total dry weight. This points to a varying input of the residue sediment fraction, which mostly contains siliciclastic material. These siliciclastic sediments are ultimately derived from southern Appalachian river drainage during sea level low stands, which have been deposited mainly on the eastern shelf. A lack of effective transport across the western shelf has favored quartz deposition in the coastal zone [Hine et al., 2003]. The geographical distribution and down core variations in the siliciclastic fraction relative to a stable organic carbon/carbonate ratio are best explained by invoking a variable runoff signal.

Figure 3.

Loss-on-ignition (LOI) results for the upper 60 cm of core ROB03 showing relative weight loss after 550 and 950°C, representative of the organic carbon and carbonate dry sediment fractions, respectively [Dean, 1974].

4.3. Microfossil Data

The palynological analyses from the upper 60 cm revealed well-preserved pollen and spore assemblages that are represented as percent abundances in Figure 4. On the basis of stratigraphically constrained numerical clustering (CONISS) [Grimm, 1987], the pollen diagram is divided in three zones RB A-C. The highly abundant taxa Pinus and Ambrosia are large pollen producers and typically over-represented in pollen diagrams. In a marine setting, abundance changes of these two taxa can represent both changes in transport (mainly for Pinus that is optimally adapted to wind dispersal) [Hooghiemstra, 1988] and land clearance [Bazzaz, 1974; Willard et al., 2003], next to mere vegetation changes within the natural communities. Pinus and disturbance indicators (primarily Ambrosia) are therefore excluded from the pollen percentage sum since we want the percent abundances to primarily reflect changes between natural communities. The percent abundances of Pinus and Ambrosia were then calculated relative to the total pollen sum so they can be compared to the abundances of other taxa, without biasing the results. Individual taxa have been grouped into different plant communities following the classification scheme of Willard et al. [2001a] (Figure 4). Individual taxa often represent more than one community but are here grouped according to their highest abundance within different vegetation types [Myers et al., 1990] and modern pollen samples in Florida [Willard et al., 2001a, 2001b; Donders et al., 2005b]. Counts of microscopic charred particles are expressed relative to the total pollen sum, and are indicative of past fire frequency.

Figure 4.

Pollen percentage diagram of the upper 60 cm of core ROB03. The pollen percentage sum contains all tree and herb pollen except for Pinus and invasive species. The CONISS output [Grimm, 1987] was used to define the pollen assemblage zones. The pollen taxa are ordered into vegetation communities for which they are most representative (see section 3.3).

Dominant pollen throughout the record are Pinus, Quercus, Chenopodiaceae/Amaranthaceae (largely Salicornia), and Rhizophora mangle (red mangrove). Even though R. mangle is known to produce abundant pollen [Van Campo and Bengo, 2004], and it is the dominant vegetation type surrounding the site (Figure 1), it does generally not exceed 20% abundance in the RB samples (9% when both Pinus and Ambrosia are included in the pollen sum). Pollen assemblage zone RB-A (A.D. 1680–1830) contains relatively high levels of wet prairie (Myrica cerifera, Cyperaceae, Chenopodiaceae/Amaranthaceae) and mixed cypress swamp vegetation (Taxodium distichum and Iva-type), particularly after A.D. 1730. In zone RB-B (A.D. 1830–1935) herbs typical of saltmarsh and mangrove vegetation (Salicornia and Batis maritima) increase in abundance, while Chenopodiaceae/Amaranthaceae, grasses and sedges decrease. Clear disturbance indicators increasingly occur (Ambrosia and fern spores). The uppermost zone RB-C (A.D. 1935–1990) shows increased mangrove and (wet) prairie vegetation, and decreases in saltmarsh and wetland taxa. The amount of charred particles are relatively stable except for the base of RB-A and the transition between RB-B and base of RB-C. Pinus shows a high and varying abundance in zones RB-A and B, which is equivalent to 80% of all counted pollen, and decreases sharply in zone RB-C down to 60% of the total pollen count.

Dinocysts are at least one order of magnitude less abundant than pollen. Owing to the low count sum, we preferred to use this proxy only for qualitative interpretations and refrain from interpreting small abundance fluctuations and minor groups. We refer to the Cremer et al. [2007] study for a detailed description of the fossil dinocyst taxa encountered at Rookery Bay. The down core distribution of typical lagoonal dinocyst species (Operculodinium israelianum, Polysphaeridium zoharyi, and Tuberculodinium vancampoe) [Marret and Zonneveld, 2003, and references therein] mirrors the altered conditions in zone RB-C that is seen in the vegetation and runoff signals (LOI residue) (Figures 3 and 4). The dinocyst data point toward more marine higher salinity conditions during the 20th century. These observations are supported by the quantitative diatom data that indicate strong increases in the abundance of fully marine taxa (Cymatosira belgica and Paralia sulcata) and a decrease of typically brackish water indicators during the 20th century (Figure 5). The diatom turnover occurs at the same time as the mangrove increases described above, pointing to a change toward more saline conditions in RB after A.D. 1930. Diatom preservation is limited to corrosion-resistant taxa in the upper 40 cm. Presence further down core is restricted to single valves most likely owing to strong silica dissolution (Figure 5).

Figure 5.

Absolute diatom concentration and species abundance in core ROB03 based on a set of samples taken at 10-cm intervals. For species descriptions and additional information, see Cremer et al. [2007].

5. Discussion

5.1. Recent Changes: Human Impact

Drainage of the Everglades commenced around ∼A.D. 1880, and from ∼A.D. 1950 complex water control structures were built to diminish the risk of flooding from the larger lakes (e.g., the Hoover Dike south of Lake Okeechobee) and gain agricultural land. The RB and Big Cypress watershed was first affected by the dredging and canalization of the Caloosahatchee River to the north (Figure 1), beginning in A.D. 1884 ( [Light and Dineen, 1994]. Locally, the main impact on the watershed came from the construction of State Road 41 (Tamiami trail; A.D. 1915–1928) and Interstate 75 (Alligator Alley; A.D. 1967) that connect the eastern and western Florida coasts, but block much of the natural sheet flow through the wetlands despite the presence of numerous culverts [Duever et al., 1986; Light and Dineen, 1994; Donders et al., 2005b]. The Golden Gate Watershed/Henderson Creek Canal was made in the 1960s for housing development and drains directly into RB through a manually operated weir (Figure 1), causing unnaturally high runoff peak [Surge and Lohmann, 2002]. The most active logging phase in the Big Cypress Swamp region occurred between 1947 and 1952 [Burns, 1984], but this area east of Naples has not been developed extensively since it has proved difficult to drain. The reconstructed vegetation communities clearly reflect drainage activities and changing land use during the 20th century A.D. Pinus elliottii (slash pine) is a durable wood and has been used extensively for timber. The first significant Pinus decrease starts ∼A.D. 1910. It shortly recovers and decreases again sharply at ∼A.D. 1950. Ambrosia (ragweed) is an early successional plant that rapidly (<1 year) occupies cleared sites [Bazzaz, 1974]. Ambrosia is continually present from ∼A.D. 1850, increases abundance in ∼A.D. 1910, and again in ∼A.D. 1950 (Figure 4). When Pinus and Ambrosia are included within the percentage sum, Ambrosia abundance reaches only 3% owing to the overrepresentation of Pinus, which reaches 70% of total pollen in zone RB-C. A sharp increase in Ambrosia pollen is widely used in North America to mark the beginning of deforestation and urban development [e.g., Willard et al., 2003]. The RB record shows a continuous significant presence of Ambrosia from ∼A.D. 1900, indicative of widespread disturbance of natural vegetation. The abundance increase of Ambrosia observed here, relative to the large amount of Pinus pollen, is striking since a large increase related to human impact had never been observed previously in Southern Florida (D. A. Willard, U.S. Geological Survey, personal communication, 2007). Around 3% Ambrosia (of total pollen) has been observed in the upper levels at Lake Tulane on the Lake Wales Ridge [Grimm et al., 1993]. But there Pinus reached values of only 30% during the Ambrosia increase, and the area is drier and more suitable for Ambrosia growth. Most other studies have focused on the Everglades marsh ecosystem [Willard et al., 2001a, 2001b], while the southwestern cypress and pine ecosystems are relatively more elevated [Duever et al., 1986] and contain less surface water. The southwestern area thus potentially provides a better habitat for Ambrosia, which likely evolved in subarid warm conditions [Bazzaz, 1974]. Ambrosia indeed was abundant during the last Glacial in south central Florida when conditions were significantly drier and slightly cooler than present, although the vegetation remained typically Floridian [Grimm et al., 2006]. Ambrosia was still present during the warm but relatively dry early Holocene, but essentially disappeared when conditions became increasingly moist during the late Holocene [Grimm et al., 2006].

The lowest Pinus abundance (Figure 4) is reached around A.D. 1950 when regional logging activities were maximal [Burns, 1984]. Since Pinus pollen are easily transported by wind or floating, the Pinus fluctuation may also partly relate to changing water discharge conditions. The Ambrosia and Pinus disturbance signal is likely related to the development of the city of Naples nearby Rookery Bay, which was founded in A.D. 1880 and caused rapid urban development and forest clearance in the area in the second half of the 20th century [Shirley and Brandt-Williams, 2003].

A small but sustained increase in mangrove pollen occurs from ∼A.D. 1910, while swamp wetland communities typical for long hydroperiods decrease (Figure 4). The total mangrove abundance increases sharply at A.D. 1930, and shows significant variability afterward, while never reaching the low values of the period prior to A.D. 1900 (Figure 6e). These changes point to increased saltwater intrusion caused by higher sea level and/or lower freshwater input. Documented regional sea level rise of the past 70 years is around 3–4 mm a−1, which is nearly 10 times that of the past 3200 years [Wanless et al., 1994]. However, sediment accretion and shallow subsidence studies of mangrove soils suggest that the recent sea level increase is mostly balanced by the elevation change of the mangrove soils in Rookery Bay [Cahoon and Lynch, 1997]. Increase in the mangrove species Laguncularia (white mangrove) and associated Batis, Poaceae and Cyperaceae point to more salt intrusion into the inland marshes since these species represent are typical of the transition zone between mangroves and (salt)marshes [Tomlinson, 1986]. Subsequent increases of more salt-tolerant Avicennia and Rhizophora in the uppermost zone complete the plant succession due to increased salt intrusion into the wetlands. These trends confirm other paleoecological studies of freshwater wetlands that related 20th century vegetation changes to the impact of wetland drainage [Willard et al., 2001a; Donders et al., 2005b].

Figure 6.

Overview of the main palynological and sedimentological trends in the RB record in relation to different forcing factors mentioned in the main text. (a) ENSO (NINO3) variability reconstructed by a tree ring network from southwestern USA and Mexico [D'Arrigo et al., 2005]. A LOESS smoother (red curve) has been applied to highlight decadal trends. (b) Total solar irradiance estimates based on sunspot data from Lean [2000], including the background changes, and a reconstruction from Bard et al. [2000] using linear scaling of Δ14C variability and a conservative estimate (−0.25%) for the long-term TSI decrease in the Maunder minimum. (c) Weight percent sediment fraction after removal of organic carbon and carbonate comnponents by LOI treatment. The residue fraction represents siliciclastic material. (d) Relative lagoonal dinocysts abundance (percent sum of Operculodinium israelianum, Polysphaeridium zoharyi, and Tuberculodinium vancampoe) in core ROB03. (e–g) Relative pollen abundances in core ROB03 representing salt (mangroves), freshwater (cypress swamp), and total disturbance-indicating vegetation communities (primarily Ambrosia). (h) Influx rates of total disturbance-indicating pollen. (i) Basin-wide reconstruction of major Atlantic hurricane frequencies based on coral δ18O and luminescence data [Nyberg et al., 2007]. (j) Historical hurricane events in the RB region with the recorded Saffir-Simpson category for southeastern (SE) and southwestern (SW) Florida (see section 4.2 for data source).

Water flow into Henderson Creek and RB was impacted by water control structures, canalization and land use changes in the second half of the 20th century, which greatly increased maximum drainage levels [Surge and Lohmann, 2002] ( The dinocyst and diatom data (Figures 5 and 6d) provide evidence for a long-term shift to more saline conditions in the 20th century. These salinity changes are coeval with increased siliciclastic input as a result of erosion during peak runoff (Figure 6c), which was likely caused by land clearance and canalization such as the construction of the A.D. 1930 Tamiami Canal that resulted in less buffer capacity. Hence, the wetland drainage has reduced freshwater retention in the coastal wetlands, while increased the salt water intrusion and mean salinities in RB, and increased erosion due to deforestation and canalization.

5.2. Recent Changes: Hurricane Impacts

Several large hurricanes have impacted on SW Florida during the past century. The most significant historically documented impacts that directly affected the study site from A.D. 1850 onward are listed in Figure 6j (all hurricane data from and; the HURDAT re-analysis data adapted from Jarrell et al. [1992]). The events in Figure 6j represent regional category 3–5 hurricanes on the Saffir-Simpson scale, and hurricanes of lower category that potentially had a local effect on the study area owing to their proximity. Several very large storms impacted the area between A.D. 1926 and A.D. 1960, while periods before and after experienced less severe local storm impacts. Especially the A.D. 1926 “Great Miami,” A.D. 1947 and A.D. 1960 Hurricane Donna were major storms in the southwestern sector. The A.D. 1926 and A.D. 1947 storms are known to have caused widespread flooding in the Everglades, which accelerated the construction of water control systems that are currently in place across southern Florida [Light and Dineen, 1994]. A recent basin wide proxy-based reconstruction of Atlantic hurricanes [Nyberg et al., 2007] indeed shows a minimum in hurricane frequencies between A.D. 1910 and A.D. 1930, and after A.D. 1960 (Figure 6i). Hence, regional documented storms around RB correspond with basin-wide maxima in the number of large hurricanes, indicating that the local hurricane events are at least partly indicative for large-scale trends.

The maximum influx of disturbance-indicating Ambrosia pollen is between ∼A.D. 1925 and A.D. 1950, which corresponds to the regional maximum in hurricane impacts described above (Figures 6h and 6j). Later, Ambrosia influx declined and became more constant, while human impact in the region was still active (see section 5.1). In contrast, recorded category 3 storms in the period prior to human impact seem to have had little or no influence on Ambrosia influx rates as large peaks are absent in the A.D. 1870–1890 period. The peaks in Ambrosia influx imply that the recorded human impact by deforestation and canal construction from ∼A.D. 1900 onward [e.g., Light and Dineen, 1994] are more pronounced during a phase with severe hurricanes. Disturbance by strong wind and storm-related precipitation in deforested areas possibly amplifies the disturbance signal as less natural vegetation is present to buffer the impact.

Studies on the impact of the category 4/5 Hurricanes Donna (1960) and Andrew (1992) show that the main impact on the natural communities occurs as loss of interior wetland, rather than coastal erosion [Wanless et al., 1994; Doyle et al., 1995]. The resolution of our analyses precludes an exact one-to-one correlation between single hurricane events and precise levels in the sedimentary sequence. However, storm impacts on mangrove communities are very significant for the forest structure and function, and the effect lasts up to several decades [Doyle et al., 1995]. The best documented and most recent large storm that directly impacted the Ten Thousand Island coast is the category 4 Hurricane Donna, which directly passed over the Rookery Bay area in A.D. 1960. The hurricane caused some coastal erosion but is known to have mainly impacted on interior mangrove forest [Wanless et al., 1994]. Figure 6e shows that the trend toward more mangrove vegetation in the second half the 20th century, caused by reduced freshwater flow into RB, seems interrupted by the effects of Hurricane Donna. Further down core correlation is only tentative owing to the dating uncertainties, but following our present age model all category 3–5 hurricanes in the SW sector correspond to multisample minima in the total mangrove abundance (Figure 6e). The mangrove minimum around A.D. 1905 is potentially related to the 1910 cat. Two storm that passed parallel to the southwestern Florida coast and made landfall just north of RB, but this needs confirmation.

Further higher resolution studies on new sediment cores will be carried out within the Utrecht University “Hurricane and Climate change” project to assess whether these patterns are consistent, and can be used to interpret, e.g., the mangrove minima in A.D. 1750 and 1730 in the this core.

5.3. Decadal to Centennial Variability

Decadal to centennial variability in the RB record is most prominent in the LOI record. Prior to human impact at ∼A.D. 1900, minima in the residue fraction (Figure 6c) correspond to low amounts of wetland vegetation (Figure 6f). A 10–20 year time lag is apparent between the LOI and wetland records, which points to a delayed response of the vegetation to precipitation changes. This lag can be explained by the fact that plant abundance is limited by growth and reproductive rates. On the basis of the correspondence between LOI and vegetation, and the source of siliciclastic material described in section 2.1, we interpret minima in LOI residue as periods with reduced runoff.

The ultimate cause of these fluctuations is yet unclear. Regionally, marine and terrestrial paleoclimate studies have correlated low humidity, low sea surface temperature (SST), and high-salinity conditions with long-term minima in total solar irradiance (TSI) [Lund and Curry, 2004, 2006; Richey et al., 2007; C. González et al., Late Holocene mangrove dynamics in the Colombian Caribbean: A history of human and natural disturbances, submitted to The Holocene, 2008] (C. González, personal communication, Universität Bremen, 2007). However, the limited length of our RB record and the lack of other records in the area do not yet permit unequivocal determination of the phase relation with TSI estimates (Figure 6). ENSO significantly influences the hydrological cycle in Florida, where a more intense ENSO cycle would result in higher winter precipitation extremes [e.g., Vega et al., 1998; Donders et al., 2005a, 2005b]. Although ENSO-tied winter rainfall in Florida is important for plant growth in early spring, it contributes little to the total annual rainfall which accounts for the runoff patterns. A 600-year reconstruction of ENSO variability based on North American tree ring data [D'Arrigo et al., 2005] shows some multidecadal phases of enhanced and reduced ENSO activity (Figure 6a), but the relation to runoff and humidity patterns are also not conclusive.

6. Concluding Remarks

Our RB multiproxy study shows that runoff patterns and wetland composition in SE Florida have substantially varied over the past ∼300 years. A clear sign of the influence of human impact through land clearance and canalization is seen on the increased terrigenous input into the RB estuary during the 20th century. Human impact is evident from a significant increase in Ambrosia pollen abundance at about 1900 A.D. The Ambrosia influx rates suggest that in human impacted ecosystems the erosional effects of hurricanes are more pronounced. We further show (Figures 5 and 6) that increased wetland drainage has lowered freshwater retention and increased salinities in RB in the second half of the 20th century. These changes are evident from decreases in lagoonal dinocysts and diatom taxa indicative of brackish conditions, and increased abundance of salt-tolerant mangrove vegetation.

The overall palynological data seem to confirm documentary evidence that describe hurricane landfalls as significantly damaging the mangrove wetlands. These impacts can provide a potential tracer for detecting hurricane events prior to historical documentation and will be subject of further high-resolution studies. Any further increase in drainage will most likely lead to significant salt water intrusion and coastal erosion during tropical storms and hurricanes in a coastal area already severely impacted by urban development, water control structures and deforestation.

The decadal- to centennial-scale variability in the RB record is characterized by phases of increased siliciclastic input, here interpreted as high runoff, and higher wetland pollen abundance. Possible explanations for these variations include regional forcing by regional SST variations and ENSO-related precipitation variability, but more data and higher resolution studies are needed to address these questions.


We are very grateful to the RB NERR staff for the fieldwork logistics and general research support. The manuscript benefited greatly from suggestions of Michael Mann and two anonymous reviewers. Thanks to Ton van Druten, Wolfram Kürschner, Oliver Heiri, and Andy Lotter for the fieldwork assistance. Klaas van der Borg and Arie de Jong from the Van de Graaff AMS facility at Utrecht University performed the radiocarbon dating and gave valuable suggestions for the age calibration. The research was carried out in the framework of the Utrecht University High Potential program. This paper is Netherlands School of Sedimentary Geology publication 20080505.