Numerical computation of hurricane effects on historic coastal hydrology in Southern Florida
© Swain et al.; licensee Springer. 2015
Received: 1 August 2014
Accepted: 13 December 2014
Published: 12 February 2015
Numerical models are critical for assessing the effects of sea level rise (SLR), hurricanes, and storm surge on vegetation change in the Everglades National Park. The model must be capable of representing short-timescale hydrodynamics, salinity transport, and groundwater interaction. However, there is also a strong need to adapt these numerical models to hindcast past conditions in order to examine long-term effects on the distribution of vegetation that cannot be determined using only the modern record.
Based on parameters developed for a numerical model developed for the recent 1996 to 2004 period, a hindcast model was developed to represent sea level and water management for the period of 1926 to 1932, constrained by the limited hydrology and meteorology data available from the historic past. Realistic hurricane-wind and storm surge representations, required for the hindcast model, are based on information synthesized from modern storm data. A series of simulation scenarios with various hurricane representations inserted into both hindcast and recent numerical models were used to assess the utility of the storm representation in the model and compare the two simulations.
The comparison of the hindcast and recent models showed differences in the hydrology patterns that are consistent with known differences in water delivery systems and sea level rise. A 30× lower-resolution spatially variable wind grid for the hindcast produced similar results to the original high-resolution full wind grid representation of the recent simulation. Storm effects on hydrologic patterns demonstrated with the simulations show hydrologic processes that could have a long-term effect on vegetation change.
The hindcast simulation estimated hydrologic processes for the 1926 to 1932 period. It shows promise as a simulator in long-term ecological studies to test hypotheses based on theoretical or empirical-based studies at larger landscape scales.
KeywordsHindcast Numerical models Hurricanes Wind fields Storm surge Sea level rise Coastal hydrology
Hurricanes and cyclones are major drivers of coastal ecological processes at all levels of biological organization from populations to communities to ecosystems and operate across a hierarchy of spatial and temporal scales (Michener et al. 1997). Hurricane and cyclone effects are receiving greater emphasis and study worldwide with recent high-profile devastating landfall storms (i.e., Hurricane Katrina 2005, Superstorm Sandy 2012, Super Typhoon Haiyan 2013) but also with climate change resulting in rising sea levels and intensification of tropical cyclones (Khairoutdinov and Emanuel 2013) with unpredictable potential effects.
As part of a larger U.S. Geological Survey study focused on ecological interactions in the mangrove estuaries ecosystem fringing the Everglades, an existing numerical model is modified to represent hurricane disturbance. To investigate patterns and processes involved in long-term vegetation change, numerical models are used to simulate hydrologic conditions during the periods of 1926 to 1932 and 1996 to 2004 coincident with vegetation occurrence documented in geo-referenced aerial photos and charts (Smith et al. 2002; Smith et al. 2010). Retrospective analyses of the hindcast output (1926 to 1932) and a time series of past habitat types will be used to identify patterns of ecosystem change and generate and test hypotheses of long-term ecological-hydrological processes. Ultimately, the goal of further research would be to develop simulations of future hydrology to explore scenarios of climate change and adaptive management. However, to realistically simulate future ecological processes and outcomes, it is first necessary to develop a reliable hindcast that accurately reproduces the effects of major storm events.
Hydrodynamic simulation of hurricane storm surge can include complex three-dimensional flow and barotropic effects, as in the application of the CH3D-SSMS-integrated storm surge modeling system to the northeast Florida coast (Sheng and Paramygin, 2010). Circulation patterns and inundation are represented in a simulation of Tropical Storm Fay in 2008. When groundwater interactions are of interest, coupled surface-water/groundwater models of the South Florida area have been developed to examine coastal hydrology and related issues. The primary simulator developed for representing hydrodynamic surface-water flow coupled to groundwater in a highly porous aquifer is called Flow and Transport in a Linked Overland/Aquifer Density-Dependent System (FTLOADDS) (Langevin et al. 2005). FTLOADDS links the two-dimensional hydrodynamic flow and transport code SWIFT2D (Swain, 2005) with the widely used three-dimensional groundwater flow and transport code SEAWAT (Guo and Langevin, 2002). Both regimes incorporate the effects of salinity concentration on water density. The capability to simulate surface-water heat transport and temperature was incorporated into the code with the added benefit of cell-by-cell evapotranspiration computations (Swain and Decker, 2010).
The FTLOADDS numerical simulator is needed to represent three unique aspects of the South Florida hydrology: (1) the low water-level gradients and flat topography, (2) the highly porous aquifer, and (3) salinity transport in both the surface-water and groundwater systems. The unique capability of FTLOADDS to simulate the full hydrodynamic solution, which is especially suited to dynamic events with substantial inertial forcing, allows for the representation of short-term surface-water transient effects such as tidal changes and wind-driven flux while simulating salinity mixing. With the SWIFT2D/SEAWAT linkage simulating the exchange of surface water with the groundwater, all the important controlling processes including tide, wind, precipitation, evapotranspiration, and surface-water management are represented. In comparison, a simpler surface-water/groundwater model does not have the ability to represent short-term hydrodynamic storm events and other transient phenomena while also simulating the impact of these short-lived and spatially explicit events on long-term and landscape-scale hydrodynamic processes, which are important to the resultant short- and long-term ecological response.
Two major applications developed with FTLOADDS are the TIME model of the Everglades National Park area (Wang et al. 2007) and the BISCAYNE model of the eastern urban area and Biscayne Bay (Lohmann et al. 2012). Together, these applications encompass the natural and urban areas of Miami-Dade County (Figure 1). The initial period of simulation was 1996 to 2002 but was later expanded to 1996 to 2004, incorporating a larger variety of annual hydrologic conditions (Swain and Lohmann, US Geological Survey, March 2014, written communication). Applications of FTLOADDS have been utilized to develop surface-water discharges for salinity targets (Swain and James, 2007), examine the effects of ecosystem restoration (Swain et al. 2008, Obeysekera et al. 2011), provide salinity and temperature estimates for manatee habitats (Stith et al. 2011), determine causes of hypersalinity in Biscayne Bay (Lohmann et al. 2012), represent the effects of hydrological restoration scenarios on American crocodiles (Green et al. 2014), and forecast the impact of future precipitation changes and sea level rise in coastal areas (Swain, Stefanova, and Smith, 2014).
TIME/BISCAYNE covers an area with a maximum extent of 93 km north-south and 129.5 km east-west. Each model cell is 0.5 × 0.5 km and approximately 55,000 grid cells cover the area. The groundwater simulation has a time step of 1 day whereas the surface water is simulated at a 10-min interval. Model inputs are defined in various time intervals: 30-min tidal levels, 6-h average rainfall, and 4-h average wind. However, the option exists to input spatially variable wind fields for specific periods at any desired time intervals, a useful option for representing storm events. Variables output by the model include groundwater and surface-water salinity, hydroperiod, flow, stage, depth, water temperature, and other time series.
The lower sea level during the hindcast period compared to recent times must be accounted for in the simulations. The coastal hydrology and inundation must be simulated correctly to provide useful results.
Water-delivery schemes were quite different 85 years ago and not as well recorded. Most of the hydraulic structures that control canal flows in the recent simulation were not yet or only in the process of being built in the hindcast period.
The recent simulation is parameterized with modern high-resolution observed data and information. However, for the hindcast simulation, there were less observed data with greater uncertainty, requiring the use of modern surrogate data or synthesis of comparable data based on assumed differences between past and present conditions.
Representation of major storms with a spatially uniform wind field has been considered adequate for the recent simulation, as there was not a major onshore hurricane event during that interval. A storm in the hindcast simulation, such as the Great Miami Hurricane of 1926, however, involves a strong wind field with high-spatial variability and requires an accommodation to properly represent the wind forcing.
Due to the proximity of offshore boundaries in both TIME and BISCAYNE, the model cannot represent the long offshore fetch over which high-wind forces on the ocean surface pile up water against shorelines, producing a storm surge. It is therefore necessary to indirectly represent the effects of the surge at the model boundary. The water level or velocity can be specified at the boundary to induce the proper surge height or surge momentum.
The proper temporal scale for the representation of a major storm event is shorter than the 4-h averaged period used for the spatially uniform wind. The center of a hurricane traveling 20 km/h moves 80 km between 4-h time steps, a major portion of the combined TIME/BISCAYNE domain. The model must be capable of resolving finer time steps for the storm event duration and represent the longer time steps for the rest of the simulation.
The spatial resolution at which the hurricane processes are represented must be based on the response of the affected hydrologic system. A model depiction with high-spatial resolution of a forcing function may not yield any more information than a depiction with a lower resolution forcing function, because the dynamics of the hydrologic system have a spatial scale that controls the response.
Purpose and scope
Hindcast water-control structures
The modifications in topography for the 1926 to 1932 simulation are predominantly along the eastern coast (Figure 3) and primarily reflect the resetting of coastal elevations near modern canals to pre-development values. With minimal historic data, no alterations were attempted in the western TIME area, but the elevations near coastal canals were altered to match the adjacent cell elevations. The same leveling process was applied to the site of modern Turkey Point nuclear plant.
Hindcast surface-water inflows and tidal levels
The estimation of flows across the northern model boundary of the Tamiami Canal for the hindcast was accomplished through the construction of an empirical relationship between recent flow and Lake Okeechobee (location in Figure 1) water levels. Although there have been a number of changes to the system over the years, the recent relation was used to estimate historic flow from historic water levels recorded back to the year 1912. Details of the empirical equation's development are discussed in Appendix 1.
Tidal water levels at the remaining boundaries were lowered from the 1996 to 2004 levels to values indicative of the 1926 to 1932 period based on a regression of historical water levels at the Key West Tidal Station, which has continuous record back to the year 1913 and sporadic measurements back to 1846 (Maul and Martin 1993). Tidal data indicate the mean sea level rose at the rate of 2.4 mm/year from 1933 to 2004, a total of 172.8 mm. The mean tidal levels computed for 1926 and 1932 are −0.390- and −0.373-m NAVD88, respectively, based on a −0.20-m mean level for 2004.
Historic groundwater conditions are poorly documented and insufficient information exists to define time-variant groundwater levels. The groundwater boundaries along the coastal areas are controlled by the overlying tidal levels and therefore change with the defined sea level. The Miami River was uncontrolled during the hindcast period and probably drained the area around it more than during the recent period. However, overall drainage through the region increased over time, so it is difficult to generalize the overall change to groundwater levels.
Data for daily parameters with no historical measures
As hydrologic information from this period is limited, the hindcast simulation utilized data from the 1996 to 2004 simulation as a surrogate for the unknown quantities, under the assumption that these time series were at least similar to values in the 1926 to 1932 period and no other assumed values were considered better. This modern time series includes parameters used to compute heat transport and evapotranspiration (ET) (solar radiation, relative humidity, air temperature), the wind field (with a modification discussed below), groundwater-level boundaries, offshore salinities, and tidal water-level fluctuations. A lower sea level in 1926 might mean that groundwater levels are lower inland as well, but the lack of drainage could mean a higher level. Groundwater levels at the coast are more controlled by sea level.
Hindcast long-term rainfall
Compared to the 107 rainfall gages available for the 1996 to 2004 recent simulation, only a few gages had been constructed by the 1926 to 1932 hindcast period. Rainfall values were averaged from the Coconut Grove 7S Station (Station # 06168) and the Homestead IFAS Station (Station #HB872) shown in Figure 3, referenced through the South Florida Water Management District DBHydro database (http://www.sfwmd.gov/dbhydroplsql/show_dbkey_info.main_menu).
Comparison of daily rainfall between the gages showed data were recorded throughout the 1926 to 1932 period. At least one station was in operation at all times, with the exception of the Great Miami Hurricane aftermath, when both stations were destroyed. In general, these stations provided a consistent measure of precipitation for the area. However, on several occasions, the gages provided substantially different measurements, likely related to the distribution and path of the storms across the model domain. For example, a storm on 7 September 1927 showed rainfall values of 4.1 and 0.39 in., for the Coconut Grove and Homestead gages, respectively. This type of observational difference, known as a correlation distance effect (Székely et al. 2007), is inherent to the limited spatial sampling of historical data. In general, the longer the averaging timescales, the more uniform the spatial distribution of rainfall, so monthly and yearly average tend to be more consistent between rainfall gages.
To assess the adequacy of using only two gages to represent past conditions, the data were input into the initial hindcast model and inundation output was mapped and examined on a daily time step to identify sudden inundation signatures indicative of a major rain event. The timing of each event was compared to newspaper and publication reports of past tropical storms. Inundation signatures corresponded with dates of the known major hurricanes in HURDAT, the official National Oceanic and Atmospheric Administration hurricane database (Landsea et al. 2008). Furthermore, the verification process identified major rain events not associated with tropical storms. The most notable was described as the ‘Florida Disturbance’ 27 May to 19 June 1930 (Fish, 1930). It is still one of the largest rainfall events ever recorded in the Miami area when 33.16 in. fell over the 24-day period. Although not associated with major winds and storm surge, such precipitation events can impact terrestrial and perhaps some marine species.
Missing rainfall during the 1926 Great Miami Hurricane
A major break in the rainfall data occurred when the rain gages were inoperable for 59 days following landfall of the 1926 Great Miami Hurricane. Two different methods to estimate the missing rainfall are discussed below: one based on historic estimates of barometric pressure extrapolated to rainfall, the other based on the forward speed of the storm. Mitchell (1926) reported that from midnight to 6:45 am on 18 September 1926, when the eye came ashore, rainfall was recorded at 0.28 in./h. However, a substantial portion of any hurricane's rainfall occurs in its tail, during which there were no measurements. Barometric pressure recorded from Miami at midnight was 29.5 mb but did not rise back to this level until about 6:00 pm (Mitchell 1926). If 18 h represents the period of rainfall, and the rate is assumed to be 0.28 in./h, then a total rainfall of 5.04 in. was estimated.
The alternative method developed by R. H. Kraft (Pfost 2000) is an empirical relationship, 100 divided by the forward speed of the hurricane (knots), which, for modern storms, provides a reasonable estimate of the resulting rainfall amount in inches. One example is Hurricane Andrew, which traveled at about 16 knots and dropped about 7 in. of rain; the Kraft method yielded an estimate of 6.25 in. The Great Miami Hurricane was estimated to travel at 18.75 mile/h (16.3 knots) as it approached Miami and 11.5 mile/h (10.0 knots) after hitting land (Mitchell 1926). This method yields estimates of 6.1 and 10.0 in. of rain from the storm, respectively, with probably more weighting to the lower value, which corresponds to landfall. Considering the previous estimate extrapolated from Mitchell (1926) of 5 in., a 6-in. total rainfall (0.152 m) was input into the simulation for 18 September 1926 as estimated rainfall from the Great Miami Hurricane.
Modeling storm surge
The method used to approximate storm surge requires modifying the boundary sea level conditions to represent water accumulated from hurricane winds offshore outside the model domain. When approximating surge for a specific hurricane in the recent simulation, the boundary stage conditions were iteratively modified until the simulated stage at coastal river outlets matched measured values.
Storm surge estimates for the Great Miami Hurricane were based on information from Mitchell (1926). Tidal water levels were set to values indicative of the hindcast period as discussed previously. The maximum storm surge height was recorded as approximately 8 ft (2.4 m) on the Miami side of Biscayne Bay and also at Miami Beach (Mitchell, 1926). The datum was not specifically defined, so it is assumed to be approximately defined relative to sea level at the time of the storm. These storm surge data were incorporated into the simulated tidal level with linear rises and falls occurring over a 6-h period consistent with the standard time step of weather prediction models (Landsea et al. 2012).
Storm wind field representations
Spatially uniform wind fields
The peak velocity and storm duration recorded by Mitchell (1926) were used to make a spatially uniform storm wind field for the Great Miami Hurricane. The documented wind velocity increased to a peak of 56 m/s (125 mile/h) over 5-h period and dropped back to nominal velocities over the next 15 h. This time series was delineated as close as possible within the 4-h intervals of the uniform wind field developed for the recent simulation.
Spatially variable wind fields
As hurricanes have high-spatial variability in wind speed and direction as they cross land, a spatially variable dataset should improve the ability to simulate these storms events. A historic reconstruction for the Great Miami Hurricane (Landsea et al. 2012) was considered, but not used, as only the wind at the location of the hurricane's eye is reported every 6 h. Instead, an approximation was developed for a modern-era hurricane utilizing data from the 2005 Hurricane Wilma H*Wind Gridded Surface Wind Analysis (Powell et al. 1998), which is a data compilation based primarily on actual observations at sea, on land, from satellites, and from Hurricane Hunter aircraft. Past numerical simulations of storm surge have used wind fields created on the basis of fitting analytical cyclone models (Holland, 1980) or on surface wind field observations from H*Wind analyses of real storms (Zhang et al. 2008, Zhang et al. 2012). Real wind analyses provided more realistic wind velocity input for models, and storm surge output driven by these wind fields was compared with field measurements (Zhang et al. 2008, Zhang et al. 2012) for validation. No modern hurricane faithfully tracked the Great Miami Hurricane; however, Hurricane Wilma in 2005 was similar in spatial scale, and its track across South Florida from southwest to northeast can be compared to that of the Great Miami Hurricane from southeast to northwest when location and wind direction are switched from east to west (Figure 2). Using a real storm wind analysis from the study area provided the opportunity to compare different wind representations in the recent simulation to actual hydrology measurements in the field. Comparing output from the same storm run in both the recent and hindcast simulations allowed assessment of the efficacy of modifications to the recent model to produce a hindcast. Furthermore, using a modern storm provided experience with developing a diversity of known-storm scenarios that could be useful simulations for research and management questions.
The wind variable presented in the H*Wind data set is the maximum sustained wind speed over a 1-min average, the standard for hurricane wind measurements (Powell et al. 1996). The wind data are initially gathered at a 10-min average measurement but then multiplied by a gust factor to approximate the 1-min standard. The factor increases the value of the 10-min average velocity by approximately 11%. Because the minimum interval for the TIME/BISCAYNE simulations is also a 10-min average, this gust factor was removed by division of the output data values by the standardized gust factor of 1.11 (Powell et al. 1996).
In addition to the full resolution wind field described above, a reduced resolution representation was also developed to provide smaller data volume and computational effort for the hindcast. Whereas the full resolution grid bilinearly interpolates the wind data to the 0.5-km TIME/BISCAYNE grid, the reduced resolution grid divided the simulation area equally into a 4 × 4 grid (Figure 4) for a grid cell spacing of approximately 32 km in the east-west direction and 24 km in the north-south direction. The 16 cells formed a 4 × 4 matrix of spatially variable wind values that changed with each of the 7 hourly measurement from the Hurricane Wilma H*Wind analysis during the passage of the storm. Thus, the increase in area and reduction in resolution of each grid cell is approximately 30×.
In order to synthesize a spatially variable wind field surrogate for the Great Miami Hurricane, it was noted that the track of Hurricane Wilma from southwest to northeast is reflective of the Great Miami Hurricane from southeast to northwest (Figure 2) when location and wind direction are switched from east to west. The 4 × 4 variable wind grid from Wilma (Figure 4) was transformed in a similar manner, and the resulting wind field represented a storm traveling from southeast to northwest like the Great Miami Hurricane. The transformed time series of wind data from Wilma was scaled in wind speed magnitude and used as a spatially variable surrogate for the Great Miami Hurricane wind field. The peak wind speeds for the Wilma data are approximately 45 m/s (100 mile/h) whereas the peak winds reported for the Great Miami Hurricane were on the order of 56 m/s (125 mile/h) (Mitchell, 1926), so the wind speed data were all multiplied by a factor of 1.25 to approximate the Great Miami Hurricane wind field.
Short-term storm representation in long-term simulation
The spatially variable wind is defined on a shorter time step (1 h) than the default longer-term spatially uniform wind (4 h). In order to accommodate transitions between these two timescales, the FTLOADDS model code was modified to accept a separate gridded data set for the duration of the storm and use the standard time series of uniform wind for the rest of the time. Negative time step values indicate use of gridded data for storms. The ratio of the wind grid size to the model grid size is specified in the main surface-water input data set (Swain, 2005). Not only can this input format be used to represent a single storm, but the wind data set can be flagged at any time during the simulation and the subsequent array of wind in the gridded wind set will be used.
Verification and validation of the new components models
TIME/BISCAYNE has been tested thoroughly and calibrated for the recent 1996 to 2004 period by comparison to measured field data (Swain and Lohmann, US Geological Survey, March 2014, written communication). Much of the hindcast 1926 to 1932 simulation input is based, by necessity, on spatial information used in the calibrated recent model. However, the available historic field data for sea level, tidal level, rainfall, and surface-water inflows, although limited, are major drivers of key hydrologic processes. With reasonable depictions of past water control features and more realistic representations of wind fields and storm surge for landfall hurricanes, differences in hydrologic parameters that affect ecological response, such as salinity and percent time inundated, should be apparent in comparisons of output from the hindcast and recent models and should be consistent with expectations from known hydrologic principles and historic observations. Verification and validation of the efficiency of the hindcast modifications can be assessed based on differences between models and not between model output and observations. Furthermore, the time periods and storm events depicted in the simulations were also designed to provide insight into how the variability of hydrologic conditions can affect flow and salinity regimes in coastal surface waters and groundwater. Comparisons of these simulations yield information not possible without numerical modeling.
Simulation runs, time periods, and storm configurations
Wind field grid used
none (TIME/BISCAYNE daily wind only)
Spatially variable, low resolution
Spatially variable, high resolution
Spatially variable, low resolution
none (TIME/BISCAYNE daily wind only)
Spatially variable, low resolution rotated
Simulation sets for the recent period
Simulations of the recent period were needed to examine the effects of a representative modern storm and the efficiency of a reduced wind field representation to approximate a known storm. The original 1996 to 2004 simulation provides a reasonable time series, but does not contain an appropriate modern storm with sufficient hydrologic measurements. Instead, wind, rain, and storm surge representations for Hurricane Wilma, which struck the Everglades on 24 October 2005, were synthetically inserted on 18 September 1996. This is 262 days into the recent simulation, the same point as the 1926 Great Miami Hurricane strike in the hindcast simulation, with a sufficient period following the storm to examine long-term processes such as groundwater and surface-water salinity effects. Rainfall data collected at the modern stations when Hurricane Wilma struck the study area were used to represent storm precipitation. The storm-surge-height time series was obtained from data collected at the Harney River and Shark River sites (Figure 3) during the actual storm.
with the 4 × 4 reduced wind data for Hurricane Wilma (Simulation RWVL);
the full wind data grid for Wilma, which is on a 5.1-km spacing (Simulation RWVH); and
the ‘base’ simulation, which included no spatially variable wind, only the uniform wind from the original daily time step (Simulation RN).
Comparison of the three simulations provides information on how much spatial resolution is needed to represent the wind effects on hydrology. The simulations also provide a validation of the wind field and storm surge used in the hydrodynamic models when salinity- and stage- simulated output is compared to empirical data collected during and after Hurricane Wilma.
Simulation sets for the hindcast period
with no simulated wind field for the storm (Simulation HN),
the surrogate Great Miami Hurricane with spatially uniform wind field (Simulation HGU),
the surrogate Great Miami Hurricane with low-resolution spatially variable wind field (Simulation HGVL), and
the modern Wilma low-resolution spatially variable wind field (Simulation HWVL).
The comparison of the first three simulations indicates effects based on three representations of the wind field. Including a modern storm in the hindcast period (Simulation HWVL) allows a comparison of the same storm in the two time periods, yielding some insight into the importance of various ambient hydrologic conditions on the response to a particular storm event. Furthermore, the insertion of these major storms at the beginning of the recent and hindcast simulations provides a 7-year period to examine long-term effects to surface-water and groundwater hydrology that is not yet available for any recent, naturally occurring major hurricane in the region.
Results and discussion
Simulations in the recent period (1996 to 2004)
Calibration of Wilma storm surge values to empirical measurements. [Runs RWVL and RWVH]
Effects of different wind field grid resolution [Runs RWVL and RWVH]
Verification of efficiency of the Wilma storm representation [Runs RWVL and RWVH]
Efficacy of surrogate variable wind field for the Great Miami Hurricane [Run HGVL]
Results described above demonstrate the validity of the storm simulation in the recent environment (1996 to 2004), setting the stage for representing the Great Miami Hurricane in 1926 through the insertion of a surrogate hurricane wind field. The hypothesis was that a transformed time series of wind data from Wilma could be scaled in wind magnitude and used as a spatially variable surrogate for the Great Miami Hurricane wind field, capturing storm processes salient to known hydrological response.
Effect of different wind field grid resolutions on hydrological response [Runs HGU, HGVL, and HN]
The hindcast period was simulated with three different representations of wind for the 16-h period on 18 September 1926 during the Great Miami Hurricane, described as HGU, HGVL, and HN above. All input parameters for the first two simulations are identical, including storm surge and rainfall, with the exception of the wind field on the day of the hurricane strike. The third simulation does not have the simulated storm surge and storm rainfall, so all hurricane-related phenomena are absent from HN.
As expected, simulations without any hurricane wind field (HN) show almost no change in salinity and inundation during the days immediately around landfall. One month later on October 18, substantial differences still exist due to the hurricane. Even after another month, on November 18, the differences in inundation and salinity are still noticeable, although the storm and no-storm simulations are starting to show similarity. An important consideration is that, over the 30 days following the Great Miami Hurricane of 1926, 6.2 in. (15.7 cm) of rain were measured. The equivalent 30-day period after the Wilma-type surrogate storm in 1996 had 9.8 in. (24.9 cm) of rain. The drier conditions in 1926 allow the inundation and salinity affects to remain longer than if there had been additional rainfall.
Long-term effects of surface-water and groundwater salinities under three wind field scenarios [Runs HGVL and HN]
Comparison of hindcast and recent simulations
Comparison of the overall hindcast and recent simulations [Runs HWVL and RN]
Effect of existing conditions and surface-water control features in hydrologic response to the same storm [Runs HWVL, RWVL, and RN]
Comparison of measured and simulated river salinity values for Wilma-type storm inserted into the recent and hindcast
Comparisons of model simulations and conclusions
Purpose of comparison
RWVH and RWVL
Effects of wind field grid resolution
Low resolution wind grid is sufficient to represent hurricanes, saving data and computational effort
HGU and HGVL
Effects of spatially uniform wind field grid and spatially variable wind field grid
The depiction of hurricane winds as uniform does not provide necessary storm wind geometry
HN, HGU, and HGVL
Effects of major storms on hydrology
Storms can have substantial long-term effects on groundwater salinity with consequences for vegetation
RWVL and HWVL
Effects of same storm in different time periods
Storm effects are quite similar for the same storm in different time periods
RN and HGVL
Hindcast and recent simulation general comparison
Historical changes in hydrology are primarily traceable to sea level variability and water-management changes
The hindcast model shows promise for use in ecological studies to test hypotheses based on theoretical- or empirical-based studies at larger landscapes and timescales. Processes on a wide range of timescales are represented by the simulations: surface-water flow at 10 min, wind at 1 to 4 h, rain at 6 h to 1 day, and groundwater flow at 1 day, with a multiyear simulation period amenable to the timescale of ecologic changes. With the strong dependence of coastal vegetation composition on salinity, results of the model comparison could provide useful input for a variety of ecological models. The information can be input to any relevant study or application as a difference in surface-water and groundwater salinity between the spatially variable wind hurricane scenario and the no-hurricane scenario. By identifying the salinity difference between model runs, errors in model mean values largely cancel out. This difference can be combined with field-derived information and applied to habitat models, vegetation succession models, and individually-based models to provide realistic storm-induced salinity related evaluations.
While the effects of hurricane wind fields on surface-water and groundwater salinity can have an inherent time lag of multiple months or years (Wilson et al. 2011), the effect on coastal vegetation is possibly much slower and of longer duration. Saha et al. (2011) and Jiang et al. (2012) have shown that the population of mangroves and tropical hardwood hammocks can be substantially affected for decades after a salinity event. The balance between the two plant populations heavily depends on soil salinity. Clearly, individual, large storm events must be considered when determining the long-term change in coastal ecology. The results of this study demonstrate that a storm event may have effects on groundwater salinity extending for years, depending on location and in situ hydrology.
If hindcast studies integrating this numerical model with ecological models reliably show process and effects in the past, then these methods can be applied with more confidence in forecasting in areas of further research, which have even more data and resolution limitations. Any representation of rare events like hurricanes in a forecast will require generation of random events to represent a storm. Rather than be restricted to random presentations, we believe that users of the model would be better served if they can generate and modify various actual storm scenarios and apply it to a particular area of interest in a series of what-if questions. This is the technique we applied for representation of the missing wind field for the 1926 Great Miami Hurricane. While clearly, no two hurricane wind fields are the same, the simulation of a known storm provides some basis for comparison, given the approximations needed to make a hindcast, which are amplified when making a forecast. The insertion of a synthetic Hurricane Wilma shows the technique works with modern simulations with high-data resolution, giving us more confidence in the hindcast and forecast projections. Scenario planning (Peterson et al. 2003) has been embraced by the National Park Service as a tool to develop management plans under the uncertainty of future climate change (Weeks et al. 2011). Forecasting a range of plausible future conditions and disturbances allows managers to explore multiple strategies for water resource management in response to climate change.
hindcast time period with Great Miami Hurricane, spatially Uniform wind field
hindcast time period with Great Miami Hurricane, spatially variable wind field at low resolution
hindcast time period with no storm
hindcast time period with Wilma storm, spatially variable wind field at low resolution
recent time period with no storm
recent time period with Wilma storm, spatially variable wind field at high resolution
recent time period with Wilma storm, spatially variable wind field at low resolution
flow and transport in a linked overland/aquifer density-dependent system
surface-water integrated flow and transport in two dimensions
tides and inflows in mangroves of the everglades
Hurricane Wind Analysis System developed by the National Oceanic and Atmospheric Administration's (NOAA) Hurricane Research Division (HRD) (Powell et al. 1998)
This research is part of the U.S. Geological Survey (USGS) interdisciplinary research project, Future Impacts of Sea Level Rise on Coastal Habitats and Species (FISCHS), funded by the USGS Ecosystems Mapping, the USGS Greater Everglades Priority Ecosystems Science, and the USGS Climate and Land Use Change - Research and Development Program. FISCHS' team members, who were integrally involved in discussions on the development of the hindcast models, included Don DeAngelis, Ann Foster, Jiang Jiang, Melinda Lohmann, Thomas Smith, Brad Stith, and Zuzanna Zajac. Thomas Smith provided the surface-water salinity and stage data during Hurricane Wilma in 2005, which were used to validate the hydrological model.
- Davis SE III, Cable JE, Childers DL, Coronado-Molina C, Day JW Jr, Hittle CD, Madden CJ, Reyes E, Rudnick D, Sklar F (2004) Importance of storm events in controlling ecosystem structure and function in a Florida Gulf coast estuary. J Coastal Res 20(4):1198–1208View ArticleGoogle Scholar
- Davis SM, Childers DL, Lorenze JJ, Wanless HR, Hopkins TE (2005) A conceptual model of ecological interations in the mangrove estuaries of the Florida Everglades. Wetlands 25(4):832View ArticleGoogle Scholar
- Fish GV (1930) Record rainfall for Miami, Florida. Mon Weather Rev 58(6):251–252View ArticleGoogle Scholar
- Green TW, Slone DH, Swain ED, Cherkiss MS, Lohmann M, Mazzotti FJ, Rice KG (2014) Evaluating effects of Everglades restoration on American crocodile populations in South Florida using a spatially-explicit, stage-based population model. Wetlands 34(Suppl 1):S213–S224View ArticleGoogle Scholar
- Guo W, Langevin CD (2002) User's guide to SEAWAT: a computer program for simulation of three-dimensional variable-density ground-water flow. In: Techniques of Water-Resources Investigations Book 6, Chapter A7Google Scholar
- Heffernan JB, Soranno PA, Angilletta MJ, Buckley LB, Gruner DS, Keitt TH, Kellner JR, Kominoski JS, Rocha AV, Xiao J, Harms TK, Goring SJ, Koenig LE, McDowell WH, Powell H, Richardson AD, Stow CA, Vargas R, Weathers KC (2014) Macrosystems ecology: understanding ecological patterns and processes at continental scales. Front Ecol Env 12(1):5–14View ArticleGoogle Scholar
- Holland GJ (1980) An analytic model of the wind and pressure profiles in hurricanes. Mon Weather Rev 108(8):1212–1218View ArticleGoogle Scholar
- Hopkinson CS, Lugo AE, Alber M, Covich AP, Van Bloem SJ (2008) Forecasting effects of sea-level rise and wind storms on coastal and inland ecosystems. Front Ecol Env 6(5):255–263View ArticleGoogle Scholar
- Jiang J, DeAngelis DL (2013) Strong species-environment feedback shapes plant community assembly along environmental gradients. Ecol Evol 3(12):4119–4128, doi:10.1002/ece3.784View ArticleGoogle Scholar
- Jiang J, DeAngelis DL, Smith TJ, Teh SY, Koh HL (2012a) Spatial pattern formation of coastal vegetation in response to external gradients and positive feedbacks affecting soil porewater salinity: a model study. Landscape Ecol 27:109–119View ArticleGoogle Scholar
- Jiang J, Gao D, DeAngelis DL (2012b) Towards a theory of ecotone resilience: coastal vegetation on a salinity gradient. Theoretical Pop Ecol 82:27–37Google Scholar
- Jiang J, DeAngelis DL, Anderson GH, Smith TJ III (2014) Analysis and simulation of propagule dispersal and salinity intrusion from storm surge on the movement of a marsh-mangrove ecotone in South Florida. Estuaries and Coasts 37:24–35, doi:10.1007/s12237-013-9666-4View ArticleGoogle Scholar
- Khairoutdinov MF, Emanuel KA (2013) Rotating radiative-convective equilibrium simulated by a cloud-resolving model. J Adv Model Earth Sys 5(4):816–825, doi:10.1002/2013ms000253View ArticleGoogle Scholar
- Landsea CW, Glenn DA, Bredemeyer W, Chenoweth M, Ellis R, Gamache J, Hufstetler L, Mock C, Perez R, Prieto R, Sánchez-Sesma J, Thomas D, Woolcock L (2008) A reanalysis of the 1911–20 Atlantic hurricane database. J Climate 21(10):2138–2168View ArticleGoogle Scholar
- Landsea CW, Feuer S, Hagen A, Glenn DA, Sims J, Perez R, Chenoweth M, Anderson N (2012) A reanalysis of the 1921-30 Atlantic hurricane database. J Climate 25(3):865–885View ArticleGoogle Scholar
- Langevin CD, Swain ED, Wolfert MA (2005) Simulation of integrated surface-water/groundwater flow and salinity for a coastal wetland and adjacent estuary. J Hydrol 314:212–234View ArticleGoogle Scholar
- Lohmann MA, Swain ED, Wang JD, Dixon J (2012) Evaluation of effects of changes in canal management and precipitation patterns on salinity in Biscayne Bay, Florida, using an integrated surface-water/groundwater model. US Geol Surv Sci Invest Rep 5009:94Google Scholar
- Maul GA, Martin DM (1993) Sea level rise at Key West, Florida, 1846-1992: America's longest instrument record? Geophys Res Lett 20(18):1955–1958, doi:10.1029/93gl02371View ArticleGoogle Scholar
- Michener WK, Blood E, Bildstein K, Brinson M, Gardner L (1997) Climate change, hurricanes, and tropical storms, and rising sea level in coastal wetlands. Ecol App 7(3):770–801View ArticleGoogle Scholar
- Mitchell CL (1926) The West Indian hurricane of September 14-22, 1926. Mon Weather Rev 54(10):409–414View ArticleGoogle Scholar
- Obeysekera J, Browder J, Hornung L, Harwell MA (1999) The natural South Florida system I: climate, geology, and hydrology. Urban Ecosys 3:223–244View ArticleGoogle Scholar
- Obeysekera J, Kubler L, Ahmed S, Chang ML, Engel V, Langevin CD, Swain ED, Wan Y (2011) Use of hydrologic and hydrodynamic modeling for ecosystem restoration. Crit Rev Environ Sci Technol 41(1):447–488View ArticleGoogle Scholar
- Ogden JC, Davis SM, Jacobs KJ, Barnes T, Fling HE (2005) The use of conceptual ecological models to guide ecosystem restoration in South Florida. Wetlands 25 (4):795–809
- Peters DPC, Bestelmeyer BT, Turner MG (2007) Cross-scale interactions and changing pattern-process relationships: consequences for system dynamics. Ecosys 10:790–796View ArticleGoogle Scholar
- Peterson GD, Cumming GS, Carpenter SR (2003) Scenario planning: a tool for conservation in an uncertain world. Con Biol 17(2):358–366View ArticleGoogle Scholar
- Pfost RL (2000) Operational tropical cyclone quantitative precipitation forecasting. Natl Weather Digest 24:61–66Google Scholar
- Powell MD, Houston SH, Reinhold TA (1996) Hurricane Andrew's landfall in South Florida. Part 1: standardizing measurements for documentation of surface wind fields. Weather Forecast 11:304–328View ArticleGoogle Scholar
- Powell MD, Houston SH, Amat LR, Morisseau-Leroy N (1998) The HRD real-time hurricane wind analysis system. J Wind Eng Ind Aerodynamics 77&78:53–64View ArticleGoogle Scholar
- Renken RA, Dixon J, Koehmstedt J, Ishman S, Lietz AC, Marella RL, Telis P, Rogers J, Memberg S (2005) Impact of anthropogenic development on coastal ground-water hydrology in southeastern Florida, 1900-2000. US Geol Surv Circ 1275:77Google Scholar
- Saha AK, Saha S, Sadle J, Jiang J, Ross MS, Price RM, Sternberg LSLO, Wendelberger KS (2011) Sea level rise and South Florida coastal forests. Climatic Change 107(1):81–108View ArticleGoogle Scholar
- Sheng YP, Paramygin VA (2010) Forecasting storm surge, inundation, and 3D circulation along the Florida coast. Estuarine Coastal Model 2009:744–761, doi:10.1061/41121(388)43View ArticleGoogle Scholar
- Smith TJ III, Tiling-Range G, Jones J, Nelson P, Foster A, Balentine K (2010) The use of historical charts and photographs in ecosystem restoration: examples from the Everglades Historical Air Photo Project. In: Cowley DC, Standring RA, Abicht MJ (eds) Landscapes through the lens: aerial photographs and the historic environment. Occ Pub Aerial Archaeology Res Group, vol No. 2. Oxbow Books, Oxford, UK. pp 179-191
- Smith TJ III, Foster AM, Briere PR, Jones JW, Van Arsdall C (2002) Conversion of historical topographic sheets (T-sheets) from paper to digital form. U S Geological Survey Open-File Report 02-204, Florida Everglades and vicinityGoogle Scholar
- Smoak JM, Breithaupt JL, Smith TJ, Sanders CJ (2013) Sediment accretion and organic carbon burial relative to sea-level rise and storm events in two mangrove forests in Everglades National Park. Catena 104:58–66, doi:10.1016/j.catena.2012.10.009View ArticleGoogle Scholar
- Stith BM, Reid JP, Langtimm CA, Swain ED, Doyle TJ, Slone DH, Decker JD, Soderqvist LE (2011) Temperature inverted haloclines provide winter warm-water refugia for manatees in southwest Florida. Estuaries Coasts 34:106–119View ArticleGoogle Scholar
- Swain ED (2005) A model for simulation of surface-water integrated flow and transport in two dimensions: user's guide for application to coastal wetlands. US Geol Surv Open-File Rep 1033:88Google Scholar
- Swain ED, Decker JD (2010) Measurement-derived heat-budget approaches for simulating coastal wetland temperature with a hydrodynamic model. Wetlands 30(3):635–648, doi:10.1007/s13157-010-0053-7View ArticleGoogle Scholar
- Swain ED, James DE (2007) Inverse modeling of surface-water discharge to achieve restoration salinity performance measures in Florida Bay, Florida. J Hydrol 351:188–202View ArticleGoogle Scholar
- Swain ED, Lohmann MA, Decker JD (2008) Hydrologic simulations of water-management scenarios in support of the Comprehensive Everglades Restoration Plan. In: The role of hydrology in water resources management. IAHS Red Book Series, UNESCO/IAHS Symposium, October 2008, Isle of Capri, Napoli, ItalyGoogle Scholar
- Swain ED, Stefanova L, Smith TJ III (2014) Applying downscaled global climate model data to a hydrodynamic surface-water and groundwater model. Am J Clim Change 3(1):33–49, doi:10.4236/ajcc.2014.31004View ArticleGoogle Scholar
- Székely GJ, Rizzo ML, Bakirov NK (2007) Measuring and testing dependence by correlation of distances. Ann Stat 35(6):2769–2794View ArticleGoogle Scholar
- Teh SY, DeAngelis DL, Sternberg LSL, Miralles-Wilhelm F,R, Smith TJ, Koh H-L (2008) A simulation model for projecting changes in salinity concentrations and species dominance in the coastal margin habitats of the Everglades. Ecol Model 213:245–256View ArticleGoogle Scholar
- Wang JD, Swain ED, Wolfert MA, Langevin CD, James DE, Telis PA (2007) Application of FTLOADDS to simulate flow, salinity, and surface-water stage in the southern Everglades, Florida. US Geol Surv Sci Rep 5010:114Google Scholar
- Weeks D, Malone P, Welling L (2011) Climate change scenario planning: a tool for managing parks into uncertain futures. Park Sci 28(1):26–33Google Scholar
- Wilson AM, Moore WS, Boye SB, Anderson JL, Schutte CA (2011) Storm-driven groundwater flow in a salt marsh. Water Resources Res 47(W02535):1–11, doi:10.1029/2010WR009496Google Scholar
- Zhang K, Xiao C, Shen J (2008) Comparison of the CEST and SLOSH models for storm surge flooding. J of Coastal Res 242:489–499View ArticleGoogle Scholar
- Zhang K, Liu H, Li Y, Xu H, Shen J, Rhome J, Smith TJ III (2012) The role of mangroves in attenuating storm surges. Estuar Coastal Shelf Sci 102–103:11–23View ArticleGoogle Scholar
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.