Hurricanes Eta and Iota were the most intense events during the 2020 Atlantic hurricane season, and their passage caused serious infrastructure affectations and even human losses in the Archipelago of San Andrés, Providencia, and Santa Catalina due to the extreme winds, storm surge flooding, and rainfall flooding. Numerical modeling and field measurements were used to reconstruct the effects of these events on the archipelago. The simulations were conducted with WAVEWATCHIII, SWAN, XBeach, Storm Water Management Model (SWMM), and a parametric model for hurricane winds. A differentiated contribution of each hazard on physical infrastructure, coastal ecosystems, and population is represented through: winds up to 50 m/s, significant wave heights (Hs) between 1 and 6 m in intermediate waters (around 10 m deep) associated with flood levels in the order of 2 m on the coast, and flood distances varying between 12 and 904 m. A spatial distribution of Hs and the contribution of wave run-up and storm surge in some areas of the archipelago showed the importance of mangrove and coral reef ecosystems to mitigate the intensity of Eta and Iota on the coast. This study encourages science-based decision-making and provides information for policymakers to consolidate risk assessments in vulnerable zones like the archipelago.
1 Introduction
Tropical cyclones (TC) are atmosphere–ocean coupled systems that can generate high-energy winds and waves, coastal flooding from storm surge, and heavy rainfall. These phenomena cause significant human and economic losses, being among the most destructive meteorological events on Earth (Aon 2017). In the Caribbean, hurricanes are by far the most hazardous phenomena, leaving devastating effects on the ecosystems and coastal communities (Spencer and Urquhart 2018; Tanner et al. 1991; Walcker et al. 2019; Wiley and Wunderle 1993), as well as serious political, social, and economic consequences across the region (Ishizawa et al. 2019; Johnson 2015; Pielke et al. 2003; Watson and Johnson 2004).
Specifically, in Colombia, the Archipelago of San Andrés, Providencia, and Santa Catalina (hereafter, the archipelago) (Fig. 1) is the most susceptible region to the occurrence of hurricanes (Montoya et al. 2018; Ortiz-Royero 2012). Recently, Hurricanes Eta and Iota caused in Providencia the death of 3 people, and the complete or partial losses of approximately 98% of building infrastructure and 90% of the tropical dry forest. During Hurricane Iota, one death and the associated damage of around 90% of the houses were reported for San Andrés (UNGRD 2020). The effects of these hurricanes have cast doubt on the current ability of the population to face the possible damages generated by the passage of a hurricane and suggest that the coastal hazards during these events are poorly managed and understood.
A map illustrates the Caribbean archipelago, highlighting the spectral wave conditions around Providencia and Santa Catalina. It marks the locations of wave and current sensors deployed during the field campaign, as well as selected profiles for assessing coastal flooding.
Fig. 1
a Location of the archipelago in the Caribbean Sea, including the location of the spectral wave conditions (red dots) obtained from WWIII and intermediate nested domains with a spatial resolution of 200 m (green rectangles) 50 m (red rectangles) implemented for wave propagation using SWAN model. b Providencia and Santa Catalina, indicating the location of wave/current sensors during the field campaign and selected profiles for coastal flooding, and c San Andrés, indicating the location of the coastal flooding profiles (dashed red lines)
×
Anzeige
Moreover, an increase in the occurrence of extreme wave events has been registered in the Caribbean Sea (Montoya et al. 2018) and a greater intensity of the most extreme TC is expected (Knutson et al. 2020; Seneviratne et al. 2021), so that hurricanes will have greater potential to affect the coast. As a result, all countries dealing with these extreme events each year must endeavor to develop an effective response to possible mass-casualty incidents. Due to the critical role of long- and short-term warnings on risk management, an accurate identification of the hazard is required in coastal populations exposed to hurricanes, as in the case of the archipelago (Committee on Homeland Security and Governmental Affairs 2006).
A TC hazard assessment implies the estimation of the hazards associated with extreme winds, coastal and urban flooding (Abtew 2019; Rezapour and Baldock 2014). Previous studies have evaluated the impact of hurricanes on different regions in the world, specifically through intensity, hurricane-induced waves, and storm surge modeling using historical and synthetic TC events (Kowaleski et al. 2020; Tian and Zhang 2021; Vickery et al. 2009; Yin et al. 2021). Marsooli and Lin (2020) showed the impact of the surface waves and the extent of coastal flooding induced by several selected hurricanes in Jamaica Bay, New York, while Lin et al. (2010) estimated the multiple hazards from a specific event, Hurricane Isabel, with an atmospheric and oceanic circulation model. Extreme waves and water levels related to the passage of TC on both the Pacific and Atlantic Mexican coasts were simulated by Meza-Padilla et al. (2015). They found that the Caribbean Sea and the northern coast of the Gulf of Mexico were the areas most exposed to the highest waves, and the northern part of the Yucatan Peninsula to the highest flood levels.
A recent study of the flooding caused by Hurricane Iota on the archipelago has shown that the storm surge and wave setup generated a flooded area corresponding to 3.7% of the total area of Providencia, with maximum storm surge values of 1.25 m at the east side of the island (Rey et al. 2021). Along with the east of San Andrés, this region in Providencia coincided with the areas most likely to be flooded in the archipelago. Furthermore, Hurricane Iota flood levels modeled with a 1D model evidenced the importance of including the wave contribution (wave setup and swash) to correctly estimate the seawater level during extreme conditions (Rey et al. 2021). Despite these results, an individual estimation of the hazard related to winds, waves, and flooding is still needed in the archipelago to obtain a comprehensive analysis of hazards associated with hurricanes in the archipelago.
This chapter aims to reconstruct the passage of Hurricanes Eta and Iota in terms of the intensity of winds and waves, and the associated coastal and urban flooding effects by using numerical modeling. These events and the study zone are briefly described in Sects. 2 and 3, respectively. The materials and methods are described in Sect. 4. The results are shown in Sect. 5. Finally, the discussion and main conclusions are presented in Sect. 6.
Anzeige
2 Hurricanes Eta and Iota (2020): An Overview
Eta and Iota were the two most powerful TCs of the 2020 Atlantic hurricane season (Blunden and Boyer 2021). Eta was a category 4 hurricane that passed close to the archipelago on November 2, 2020. Moreover, only two weeks after, the category 4 Hurricane Iota affected the archipelago on November 16. Although both systems were classified as category 4, Iota presented minimal central pressure and sustained wind speed higher than Eta. A brief description of the evolution of each TC is presented below.
2.1 Eta Patterns Description
Eta started as a tropical wave on October 22, 2020, which moved across the tropical Atlantic until it reached the Eastern Caribbean and became a tropical depression about 350 km southwest of the Dominican Republic at 18:00 UTC on October 31. Six hours later, the depression became a tropical storm and turned toward the west due to a high-pressure center located to its north. By 06:00 UTC on November 2, Eta was already a 36 m/s hurricane with its center located 500 km to the south of Grand Cayman. A continuous rapid intensification generated a category 4 hurricane with roughly 59 m/s winds at around 18:00 UTC that day, that is, over a period of just 12 h it underwent an increase of 23 m/s in its intensity. Additionally, this strengthening occurred while passing to the north of the archipelago. The intensity peak reached 69.5 m/s winds and passed about 101 km northeast of Nicaragua. Subsequently, Eta became a tropical depression during its landfall and passage over the Gulf of Honduras. A re-intensification kept Eta as a tropical storm due to the interaction with the Loop Current at around 12:00 UTC on November 11, until it transformed into an extratropical cyclone by 12:00 UTC on November 13 (Pasch et al. 2021). Although Eta’s center passed away from the archipelago, sustained surface winds and gusts of wind of 10 m/s and 20 m/s, respectively, were measured at San Andrés airport at 09:00 UTC on November 3, while winds up to 16 m/s were recorded on Johnny Cay by a weather station operated by the Marine Research Institute of Colombia (INVEMAR).
2.2 Iota Patterns Description
On October 30, 2020, a tropical wave in front of the western coast of Africa started moving westward until it became a low-pressure system close to the southwest of the Dominican Republic. Subsequently, the convective structure improved, resulting in a tropical depression formed over the south-central Caribbean Sea at around 12:00 UTC on November 13. Only six hours later, when the system was located northwest of Aruba, a strengthening in the depression led it to become a tropical storm. During a 42-h period, between 18:00 UTC on November 14 and 12:00 UTC on November 16, Iota underwent a strong deepening with an 80 mb decrease in its central pressure and a rapid intensification with a 46 m/s (from 23 to 69 m/s) increase of the 10 min averaged maximum wind speed. The deepening rate of Iota has been the third largest in the Atlantic basin since 1965. When Iota reached its peak of intensity, it was located about 37 km northwest of Providencia and Santa Catalina, after which it gradually weakened due to passing over a cool wake created by Hurricane Eta a few weeks earlier, and it made landfall in less than 24 h along the eastern coast of Nicaragua. At 18:00 UTC on November 17, Iota became a tropical storm. Finally, it dissipated over western El Salvador 6 h later than its passage across the mountains of southern Honduras (Stewart 2021).
Although the center of Iota’s eye did not cross Providencia and Santa Catalina, the southern eyewall directly impacted them, where it was estimated that sustained category 4 wind speeds of at least 59 m/s winds occurred, while hurricane-force winds persisted for approximately 7 h. In San Andrés, 10-min average wind speeds of 17 m/s and gusts of 22.5 m/s were measured, and tropical-storm-force winds occurred for at least 14 h (Stewart 2021). The Colombian Institute of Hydrology, Meteorology and Environmental Studies (IDEAM) also reported that hurricane winds of 50 m/s and 17.5–32.5 m/s were expected in Providencia and San Andrés, respectively and waves higher than 3 m (IDEAM 2020). Iota destroyed Providence’s electric plant, and its 5,000 inhabitants were entirely left without communications and electricity for about 24 h (Stewart 2021). Three people died and six people were injured, about 80% of buildings were destroyed, while another 20% were severely damaged (UNGRD 2020). Meanwhile, in San Andrés, communications were lost during the storm, the high-intensity winds (greater than 25 m/s) caused damage to several homes and one person died. A flood level of around 0.15 m generated a temporary shutdown of the island’s international airport, according to Stewart (2021).
3 Study Zone
The archipelago is located in the Caribbean Sea, and it is the largest island region in Colombia. It is composed of a group of islands, lesser islands, atolls, and cays (Fig. 1a). It is located 110 km east of Nicaragua, and roughly 720 km northwest of the Caribbean coast of Colombia. San Andrés and Providencia are the two most populated and largest islands in the archipelago, with surface areas of 27 km2 and 17km2, respectively. The other lesser formations in the archipelago occupy 8.5 km2, approximately. A mountain range crosses San Andrés from north to south, with maximum elevations up to 85 m above sea level, while Providencia has a steeper topography with a mountainous inner region with maximum elevations up to 360 m. Coastal ecosystems surround both islands (e.g., coral reefs, mangroves, and seagrasses) that foster a vast variety of marine flora and fauna (CORALINA-INVEMAR 2012) being one of the top tourist travel destinations in Colombia. The estimated population of the archipelago in 2022 was around 58,817 inhabitants (DANE 2020), and it is concentrated in a few flat areas associated with beaches. Previous studies (Ortiz-Royero 2012) have suggested that the archipelago is the zone most likely to be affected by storms in the Colombian Caribbean, highlighting the importance of a better understanding of the physical processes that take place during the passage of a TC in terms of coastal/urban flooding as well as preparing the community to face these events (Ortiz-Royero et al. 2015).
According to Ricaurte-Villota et al. (2017), the climate in the archipelago is strongly influenced by the displacement of the intertropical convergence zone (ITCZ), generating two seasons of maximum winds during December–January–February (DEF) and June–July–August (JJA), with values between 3.8 and 6 m/s, while periods of weak winds occur during March–April–May (MAM) and September–October–November (SON), ranging between 3 and 5 m/s. This climatic pattern in the Caribbean Sea has been reported by several authors who research wind speed and its connections with the El Niño Southern Oscillation (ENSO) (Alexander and Scott 2002; Alfaro 2002; Enfield and Mayer 1997; Giannini et al. 2000, 2001a, b, c; Ruiz-Ochoa and Bernal 2009).
Additionally, due to the archipelago being highly influenced by trade winds, the wind direction does not change significantly during the year, showing semi-permanent winds blowing from the northeast. Regarding the mean wave distribution in the archipelago, there is a dominance of the waves from the 0–90° direction throughout the year with some southward variation during the lesser winds season and the highest values of the significant wave height occurring during the first season (DEF) with values between 1.6 and 2.0 m (Osorio et al. 2016). The tidal regime in the archipelago is mixed diurnal and its range is 0.31 m (IDEAM 2017).
The average annual precipitation in San Andrés and Providencia shows a unimodal monthly distribution with a dry season from January to April with minimum values in March (23.1 mm in Providencia and 25.3 mm in San Andrés) and a wet season from May to December with maximum values in October (344.5 mm in Providencia and 315.5 mm in San Andrés). In general, the precipitation and wind speed regimes are higher in San Andrés than in Providencia (Ricaurte-Villota et al. 2017).
4 Materials and Methods
4.1 Field Measurements
A field campaign was conducted (March 6–19, 2021) to measure hourly wave conditions at six locations around Providencia (Fig. 1b). Outside the reef barrier, an acoustic Doppler current profiler (AWAC 600 kHz from Nortek instruments) was installed at the sea floor at 22.6 m, followed by an AWAC 1000 kHz at 9.1 m, a pressure sensor (RBRduo from RBR Lda.) at 5.1 m and three pressure sensors (AQUAlogger P520) from AQUATEC Ltd. (see AQ3, AQ2, AQ1 in Fig. 1b). The transect of instruments was heading the line north-east considering the main flow direction and wave characteristics during the field campaign in Providencia island. The information recorded was used as a boundary condition for the calibration of the XBeach model, which in turn was used for the local modeling of coastal flooding in the archipelago during the hurricane events.
The topo-bathymetric model used as an input for the coastal and urban flood-level modeling was composed of the Digital Elevation Model (DEM) and the Digital Surface Model (DSM) supplied by the government of San Andrés and Providencia in 2020 and obtained from aerial photographs, LIDAR, and base mapping of the archipelago. Additionally, beach profiles were measured using differential GPS in Real Time Kinematic mode (RTK) along 7 km of coastal beaches in San Andrés and 1.6 km in Providencia and Santa Catalina, these recordings were used as contour conditions for the coastal flooding simulations along several beach profiles. Moreover, a mapping of the drainage system was built based on earlier design reports and in-situ inspections of the hydraulic structures (e.g., box-culverts, channels, ditches, pipes) to be considered in the urban hydrological model.
4.2 Numerical Modeling
Figure 2 shows the workflow of the methodology, where wave and hydrodynamic models (WAVEWATCH IIITM and SWAN) were forced with a parametric model for hurricane winds to estimate storm surge together with wave setup and run-up, and complemented with urban flood using a storm water management model (SWMM). The following subsections provide more detailed information on each model.
A workflow diagram outlines the methodology, illustrating how data is channeled into WaveWatch 3 and S W M M. The parametric hurricane wind model feeds into WaveWatch 3, which subsequently supplies data to S W A N and XBeach for further analysis.
Fig. 2
Workflow methodology of the numerical modeling process
×
4.2.1 Atmospheric Parametric Modeling
Several free wind field databases, such as the European Center for Medium-Range Weather Forecast Reanalysis (ECMWF, ERAinterim, and ERA5), the North American Regional Reanalysis Center (NARR), and the National Center for Environmental Prediction and Atmospheric Research reanalysis (NCEPR1), among others, underestimate wind speed and do not adequately represent the spatial distribution of winds near the eye of the hurricane (Cavaleri and Sclavo 2006; Montoya et al. 2013; Ruti et al. 2008; Sharma and D’Sa 2008). These databases have low spatial and temporal resolution, which do not allow an accurate capture of the evolution of the phenomena next to the eye of the hurricane (maximum speed, asymmetry, maximum wind radius, and trajectory), even though they use the values of magnitude and wind direction of oceanographic buoys in their assimilation process. To improve this issue, different authors have proposed methodologies to calculate the wind field under extreme conditions (Lizano 1990; Visbal and Ortiz 2006; Willoughby and Rahn 2004). In this study, numerical wind field simulations were conducted considering the methodology proposed by Montoya et al. (2013), combining wind data from ERA5 reanalysis and a parametric model of hurricane winds developed by Roldán-Upegui et al. (2022), in order to better estimate the magnitude, asymmetry, and spatial distribution of the wind field under extreme conditions compared to the mentioned reanalysis databases.
4.2.2 Wave Modeling
The third-generation wave model WAVEWATCH III (WWIII) (WW3DG 2019) was used to propagate waves from deep water in the Caribbean Sea to intermediate and shallow water near the archipelago during the conditions of Hurricanes Eta and Iota. The WWIII model solves the random phase spectral action density balance equation and was set up to consider the source term package (ST2) (Tolman and Chalikov 1996), the Discrete Interaction Approximation (DIA) (Hasselmann et al. 1985) for nonlinear wave-wave interactions, the source term to model the bottom friction (Hasselmann et al. 1973) and the parameterized linear input (Cavaleri and Rizzoli 1981). Additionally, sea ice dissipation and reflection were disabled, and a third-order propagation scheme was used (Tolman 2002). The model was executed considering two nested domains with horizontal resolutions of 1/3° (37.1 km) and 1/12° (9.3 km) respectively. The frequency-direction space was discretized in 72 directions (5°) and 30 frequencies, varying from 0.042 to 0.65 Hz with an increment factor of 1.1. The bathymetric conditions were obtained from 1 arc-minute Gridded Global Elevation Data (ETOPO-1) available at https://www.ngdc.noaa.gov/mgg/global/. Available wave data from buoys 42,056, 42,058, and 42,060 from the NDBC (National Data Buoy Center: https://www.ndbc.noaa.gov/) was used to calibrate and validate the significant wave height (Hs) simulated by WWIII during Hurricanes Dean 2007 and Matthew 2016 (not shown here). Furthermore, considering the best configuration parameters and mesh details, the regional model was used to simulate the extreme wave conditions during Hurricanes Eta and Iota.
For shallow water modeling, the Simulating Waves Nearshore model (SWAN) (Booij et al. 1999) was used to represent the nearshore wave conditions around the archipelago. This model also solves the action density balance equation considering shallow processes that affect wave propagation. The wind wave growth parametrizations proposed by Komen et al. (1984) and Cavaleri and Rizzoli (1981) were used as well as the white capping source term proposed by Komen et al. (1984). Nonlinear quadruplet and nonlinear triad wave-wave interactions were modeled with the schemes of Hasselmann et al. (1985) and Eldeberky and Battjes (1984), respectively. Bottom friction (Hasselmann et al. 1973) and depth-induced breaking (Battjes and Janssen 1978) source terms were also included. Similar to WWIII, the wave direction was discretized in 72 directions (5°) and considered 30 frequencies, varying from an initial value of 0.042 Hz with an increment factor of 1.1. The SWAN model was executed using a nesting scheme, where the initial modeling domain was delimited by the red dots in Fig. 1a, which correspond to the location of the directional wave spectra obtained from WWIII and used as boundary conditions for local wave propagation. Subsequently, 2 additional meshes with a spatial resolution of 200 m and 50 m (green and red boxes in Fig. 1a, respectively) were nested to allow for improving the spatial wave resolution results around each island.
The bathymetry in the nearshore areas of the archipelago (Fig. 1b, c) was obtained through interpolations using the Inverse Distance Weighting (IDW) method by combining data from the topo-bathymetric survey of the beaches (as described in Sect. 4.1), using differential GPS (DGPS) with RTK (real-time kinematic) + PPK (Post Processed Kinematic) techniques, photogrammetric height points, and contour lines every 2 m obtained from the Agustín Codazzi Geographic Institute (IGAC), nautical charts from the Center for Oceanographic and Hydrographic Research of the Caribbean (CIOH) and information available from the General Bathymetric Chart of the Oceans (GEBCO) (https://download.gebco.net/).
4.2.3 Coastal Flood Modeling
The coastal flood level due to waves was estimated considering the sea level anomalies, the astronomical tide (AT), the meteorological tide or storm surge (SS), and the wave run-up (R2), as described below.
Sea Level Anomalies and Astronomical Tide
The sea level anomalies were obtained from altimeter data from the Copernicus Marine Environment Monitoring Service (CMEMS) and the Copernicus Climate Change Service (C3S) (https://cds.climate.copernicus.eu/), considering a spatial resolution of 1/4° (27.8 km) and daily temporal resolution. The astronomical tide was obtained from the TPXO8 database (https://www.tpxo.net/regional). TPXO models allow the estimation of the amplitudes and phases for M2, S2, N2, K2, K1, O1, P1, Q1, M4, MS4, and MN4 harmonics.
Storm Surge
The storm surge was estimated from wind variations and atmospheric pressure at the sea surface. To determine the contribution of each component to the total flood level, empirical formulations were used (Benavente et al. 2006; Genes et al. 2021; Isobe 2013; Li et al. 2020). Sea level rise by winds \(d\xi_{w}\) was calculated with the expression proposed by Bowden (1983):
where \(\rho_{a}\) is the air density (1.25 kg/m3), \(\rho_{w}\) is the water density (1025 kg/m3), g is the gravitational acceleration (9.81 m/s), h represents the wave relative depth, U10 is the wind speed at 10 m height obtained from ERA5 Reanalysis data and the parametric model of hurricane winds, the drag coefficient Cd was estimated for extreme wind conditions according to Peng and Li (2015), and the fetch (F) was obtained from Isobe (2013) considering the significant wave height Hs simulated by SWAN model.
The storm surge due to atmospheric pressure variations was estimated according to Benavente et al. (2006), as \(d\xi_{p} = {{\left( {\Delta P_{a} } \right)} / {\left( {\rho_{w} g} \right)}}\), which considers that a decrease in atmospheric pressure by 1 mbar implies a 1 cm increase in sea level. The gradient \(\Delta P_{a}\) represents the atmospheric pressure difference between the minimum pressure in the eye of the hurricane at time t, and the pressure at different points of interest around the archipelago. The minimum pressure data in the eye of the hurricane were obtained from the NHC database HURDAT2, and the pressure at the points of interest were obtained from the ERA5 database with an hourly temporal resolution and a spatial resolution of 1/4° (27.8 km).
Wave Run-Up
Wave run-up, defined as the maximum onshore elevation reached by waves, was calculated by subtracting the tidal components (obtained by the tidal component explained in the section above) from the free sea surface elevation simulated with the two-dimensional model Xbeach (Roelvink et al. 2010), originally designed to simulate the hydrodynamic and morphological processes and their impact on the coast. Xbeach was executed in non-hydrostatic and non-stationary mode due to all hydrodynamic processes involved in the non-linear shallow water equations. Directional wave spectrums simulated from SWAN were used as boundary conditions at six (6) bathymetric profiles placed around San Andrés (Fig. 1c) and five (5) at Providencia and Santa Catalina (Fig. 1b).
The calibration process was done by comparing the model results (significant wave height) with the wave data from Doppler (AWAC 600, AWAC1000) and pressure sensors (RBR, AQ3, AQ2, AQ1) recorded during 13 days at 6 points along a bathymetric profile in Providencia (Fig. 1a) and adjusting the non-dimensional wave friction coefficient (Cf), which has been used for model calibration under similar seafloor configurations (Roelvink et al. 2021). Cf values between 0.01 and 0.9 were obtained (not shown here) and were associated with the different bottom characteristics (sand, coral, coral rubble) to find the zones of the model domain where each Cf value applies. The calibrated Cf values were used to simulate the coastal flooding during the hurricane periods.
4.2.4 Urban Flood Modeling
The Storm Water Management Model (SWMM) (Rossman 2015) is a one-dimensional model that solves the Saint–Venant equations and it was used to simulate the urban flood level during Hurricanes Eta and Iota by analyzing storm sewer and other drainage systems. The main parameters required in the SWMM model for the sub-catchment elements are area, width, percentage slope, percentage of impervious areas, the manning coefficient for pervious and impervious areas, and depression storage for impervious and pervious areas. For San Andrés, the SWMM model was composed using two different configurations: (i) Areas with a storm sewer network obtained from the Plan Maestro de Alcantarillado, or Sewer Master Plan (Consorcio Plan Vial Caribe 2007) located along the road system and conceptualized as a dual drainage model, and (ii) Areas with roads without a storm sewer network and conceptualized as artificial channels generating flooding along them. For both cases, afferent areas to the storm sewer channels (rectangular in San Andrés) and streets for areas without the storm sewer system were estimated based on the Euclidean distance technique (Fig. 3a). The sub-catchments and their afferent areas in Providencia and Santa Catalina (Fig. 3b) were estimated based on the box culverts located along the perimeter of the islands and the rural basins between each of them. The main sub-catchments are conceptualized in the model as rural catchments draining to the box culverts.
A geographic map highlights San Andrés, exhibits road connections, canals, and surrounding areas. Key locations include the airport, schoolhouse neighborhood, swamp avenue, maritime terminal, and Forbest Landing. An aerial view reveals box culverts, roads, and nearby areas such as Santa Catalina Bay, Old Town, Fresh Water Bay, and Providencia Airport.
Fig. 3
a Storm sewer system and road system with afferent areas for San Andrés and b Perimetral road configuration and afferent areas for Providencia. Information available from Plan Maestro de Alcantarillado (Consorcio Plan Vial Caribe 2007)
×
The main parameters of the sub-catchments were estimated as follows: The percentage of impervious areas was obtained from the maps of land use and vegetation cover for the different sub-basins, the depression store was obtained as an initial approximation using the values proposed in the SWMM model user manual (Rossman 2015) according to the land uses and vegetation cover for San Andrés. For Providencia and Santa Catalina, given their high proportion of rural areas, constant values were selected for the entire area (5 mm for permeable areas corresponding to grass), the Slope was obtained automatically for each sub-catchment from the approximate surface model (MDS) that accounts for all the natural terrain modifications due to the construction of urban infrastructure, and from the DEM for those rural basins interacting with urban drainage (a filter was employed for slopes greater than 100% to avoid unrealistic slopes). The manning coefficients for the different sub-basins in San Andrés were obtained for impermeable and permeable urban areas based on land use and vegetation cover. The values were selected from the SWMM manual (Rossman 2015). For Providencia, an average value for dense grass of 0.27 was selected.
The width parameter was obtained using the approximated formulation based on geomorphology proposed by Babaei et al. (2018) for higher values C = 1.28. This formulation was employed for both islands. Finally, the curve number CN, from the Soil Conservation Service (SCS) method for runoff abstractions, proposed by the Soil Conservation Service (1972) was employed. This CN value was obtained from Chow et al. (1994) based on the conditions for the influential variables and information on land use and soil type for the archipelago.
For the storm sewer system and the associated joints in San Andrés, the main parameters such as length, bottom elevation, hydraulic section, and roughness, among others, were obtained from information provided by CORALINA (Consorcio Plan Vial Caribe 2007). The details of the streets located in areas with and without storm drainage systems acting in the SWMM model as channels during extreme rain events were estimated from the DSM.
Dual Drainage System
For the dual model in San Andrés, the runoff is transferred from the streets (main system) toward the secondary drainage system corresponding to the storm sewer system. For areas without the existence of a storm sewer system, the drainage is composed of areas related to the road sections and its direct discharge towards the representative nodes of road intersections. The flow is discharged directly towards the roads represented in the SWMM model by channels with a defined cross section. To determine the cross sections on the roads, a review of the field campaign information and the orthophoto was performed. A typical street cross section of 8 m roadway width, 2 m sidewalk, 0.2 m elevation above the roadway, and outer wall edifications of 3 m were defined for the street systems of San Andrés. For Providencia and Santa Catalina, only the roadway width was modified to 7 m.
For discharges from the main drainage system (streets or roads) toward the secondary system (pluvial sewer system), the equations for INOS gratings on slope type I and type II presented by Rincón and Muñoz (2013) were used. For the roads of San Andrés and Providencia, a transverse slope of the road equal to zero is assumed from what was observed in the field campaigns. The Manning roughness coefficient used was 0.018 for concrete.
Extreme Storm Rainfall During Eta and Iota
To determine the precipitation in each sub-catchment of the archipelago, rain information from the Integrated Multi-satellitE Retrievals for GPM (IMERG) (Huffman et al. 2020) was downloaded (with an original spatial resolution of 0.1° × 0.1°) and modified to a higher resolution (300–500 m) through a downscaling method. Those events with the highest rainfall intensity for each of the islands were associated with high water elevations over urban roads and with the highest potential damage across the islands. Figure 4 shows the rainfall intensity during the storm duration (black line) and the range of variability associated with all the main sub-catchments (gray shaded area) in San Andrés (22 drainage sub-catchments) for Hurricanes Eta, while for Providencia and Santa Catalina, just one integrated catchment was assumed as representative since not significant variation between sub-catchments was obtained. Hence, the flood modeling was carried out considering hurricane Eta as the most intense event for San Andrés and hurricane Iota for Providencia and Santa Catalina (see Fig. 4).
A double line graph displays rainfall against rainstorm duration, featuring two lines: San Andrés E T A and Providencia I O T A. Both lines exhibit fluctuations followed by a decrease in rainfall. The peak rainfall recorded in San Andrés is 20 millimeters, while Providencia reaches a maximum of 50 millimeters.
Fig. 4
Rainfall intensity during the storm duration occurrence of Hurricane Eta for San Andrés (starting at 2020.11.04—00:30:00 UTC) and Iota for Providencia and Santa Catalina (starting at 2020.11.15—19:30:00 UTC), including the storm peak date. The gray-shaded area around the black solid line represents the range of rain variability for all the main sub-catchments in San Andrés
×
It is observed that the accumulated precipitation in Providencia during the passage of Iota was around 247 mm, and maximum intensities of almost 49.24 mm/h were reached. For San Andrés, significantly lower accumulation was reached for Iota (maximum intensities of around 19 mm/h and accumulated precipitation of 175.5 mm). During Hurricane Eta, around 159.2 mm of accumulated rain was obtained, with maximum intensity values of up to 17.2 mm/h on average for all the urban areas.
5 Results
For Hurricanes Eta and Iota, Fig. 5 shows the spatial patterns occurring during the most extreme wind speed near San Andrés and Providencia and the time series of wind speed for points located in the middle of both islands. The specific time for each hurricane was selected based on the approximate distance between the hurricane eye and the center of the respective island. For Hurricane Eta, the maximum and nearest wind speed spatial pattern (Fig. 5a) occurred during 12:00 UTC on November 2, with a maximum wind speed of around 40 m/s and an average distance of 162 km for Providencia and 257 km for San Andrés. Time series of wind speed in the middle of both islands show maximum wind speed reaching values of around 16.5 m/s during 18:00 UTC on November 2 for Providencia, and slightly lower values of around 15.4 m/s for San Andrés.
Two contour graphs plot latitude against longitude, highlighting wind speed over time for San Andrés and Providencia. Both graphs exhibit a fluctuating trend, followed by a decrease in wind speed.
Fig. 5
Spatial patterns a, b and time series c, d of the wind speed during the passage of Hurricanes Eta and Iota, respectively obtained using the methodology proposed by Roldan et al. (In preparation)
×
For Hurricane Iota the maximum and nearest wind speed spatial pattern (Fig. 5b) occurred at 12:00 UTC on November 16 with a maximum wind speed of around 52 m/s. The average distance between the eye of the hurricane and the centers of Providencia and San Andrés are around 22.32 km and 109.2 km, respectively. The time series of wind speed in both islands shows clearly how the maximum wind speed during the most intensive occurrence was obtained for Providencia and Santa Catalina with maximum winds of around 40 m/s when compared to San Andrés with maximum values around 28 m/s. These results agree with the damage reported by several governmental agencies and newspapers, among others, showing the most catastrophic situation for Providencia and Santa Catalina with infrastructure widely affected. According to the evaluations, there were approximately 6,300 people affected in Providencia and at least 700 families in San Andrés.
5.1 Wave Modeling and Coastal Flooding
Figure 6 shows the spatial patterns of Hs estimated by WWIII during the passage of Hurricane Iota (the Hs values for Hurricane Eta are not presented here since their magnitudes were smaller than Iota). Before the rapid intensification of Iota, leading it to become a category 4 hurricane, the maximum values of Hs were between 7 and 8 m, while from 07:00 UTC on November 16, when the cyclone was close to Providencia, they reached values of up to 10 m in the right forward quadrant (with respect to the direction of the hurricane translation). Even after passing Providencia, these Hs values were maintained near the eastern coast of Nicaragua.
Four contour graphs plot latitude versus longitude for November 15 at 11 U T C and November 16 at 5, 11, and 3 U T C. Arrows on the graphs indicate the direction of the waves.
Fig. 6
Evolution of significant wave height simulated by WWIII during Iota. Arrows indicate the wave direction. The dotted line represents the track of Iota
×
The highest Hs values seem to be located under the hurricane eyewall, suggesting that they are directly generated by the wind hurricane action and are not advected from other regions outside the simulation domain. From the temporal evolution, it can be noticed that the higher waves are in the right forward quadrant as expected. In this quadrant, the forward motion of the hurricane increases the wind speed and wave-growth processes, together with a partial resonance effect (Wright et al. 2001). In general, the temporal changes, the asymmetry, and the spatial distribution of Hs during these extreme conditions were well represented by the regional wave model. The directional wave spectrum across the whole domain boundary was provided by WWIII as boundary conditions (red dots in Fig. 1a) to simulate the spatial distribution of local waves (Hs) with the SWAN model around the archipelago (Fig. 7).
6 contour graphs plot latitude versus longitude for November 16 at 5, 11, and 17 U T C. Arrows on the graphs indicate the direction of the waves. The scale ranges from 0 to 8 meters.
Fig. 7
Significant wave height and wave direction during the passage of Iota over the archipelago between 05:00 and 17:00 UTC on November 16. Black arrows indicate the wave direction
×
During the approach and passage of Iota around the archipelago, the islands were affected by the left forward quadrant of the hurricane. In San Andrés, even though waves lower than 2 m were estimated numerically in a large part of the island at 05:00 UTC on November 16, waves roughly 4 m may have hit the west and south. At 11:00 UTC on November 16, Iota generated waves up to 6 m in the south of San Andrés. It can also be noted that the coral reef attenuated the waves from the north to northeast of San Andrés, where the Hs values decreased significantly from 5 m to 1–2 m. Regions like the southeast and west were impacted with waves of 4–5 m, approximately. After the passage of Iota, there are no significant changes on the east coast of San Andrés, although the waves far from the coast no longer exceed heights of 2–3 m. The west and south of San Andrés show the highest values (less than 4 m).
In Providencia, waves about 6 m far from the coast were estimated before the major impact of Iota, especially over the south of the island. A Hs change is evidenced close to shore, where wave heights are lower than 2–3 m. When Iota directly hit Providencia, waves of 7–8 m were simulated outside the coral reef that surrounds Providencia to the west. Despite the enormous decrease in height (~6 m), waves of 2.5 m impacted Providencia, especially in the west. The Hs estimated in the east, south, and north were around 1 m. At 17:00 UTC on November 16, wave heights outside the barrier reef continued to be higher in the west of the island (6 m) than in the rest of the island (2–4 m), while Hs values of 2–3 m were still present on the coast.
The values estimated are coherent and show good agreement with warning reports issued by IDEAM, so waves higher than 3–4 m were expected on November 15 and 16 (IDEAM 2020). However, the results are presented as indicative of the possible physical values sensed in the archipelago due to the lack of available information to corroborate the estimations.
5.2 Coastal Flooding
Figure 8 shows the correlation coefficient (Corr.), the Root Mean Square Error (RMSE), and de Mean Bias Error (MBE) between the XBeach model and Hs values from the AWAC1000 and RBR locations, where Hs values from AWAC600 were used as boundary conditions. The calibration was carried out considering the first 9 days of the 13 days of measurements and the best input parameters (e.g., bottom roughness and frictional coefficients) were later used to simulate the coastal flooding during the hurricane periods. These results show the high performance of XBeach in reproducing the hydrodynamic processes in the surf zone.
2 scattered graph plots the H s X beach versus H s measurement. The line of regression has a linearly increasing trend with data points is closely aligned to the line. In first graph, the R M S E is 0.0121, corr is 0.9319, and M B E is negative 0.0836. In 2nd graph, the R M S E is 0.0026, corr is 0.8182, and M B E is 0.041.
Fig. 8
Hs comparison between the XBeach model and field measurements in Providencia. Data from the AWAC100 and RBR devices are shown on the left and right plots, respectively. The locations of these measurements are indicated in Fig. 1b
×
The spatial–temporal variation of the maximum flood level or run-up (Rhigh = AT + SS + R2) and its components (AT = Astronomical Tide, SS = Storm Surge, and R2 = wave run-up) during Hurricane Iota are presented in Figs. 9 and 10 for San Andrés and Providencia (including Santa Catalina), respectively. From the run-up time series, the 98% percentile (Rhigh2) was estimated every hour during a simulation period of 72 h (36 h before and 36 h after the maximum flood peak recorded by each profile). Additionally, the time series of the meteorological tide or Storm Surge (SS) is presented, as well as the contribution to the total flood level given by its pressure (SP) and wind (SV) components.
6 multi-line graphs display water levels over time for K18, S03, S04, K07, and K10, with lines labeled A T, S S, S P, S V, and A2, exhibit a gradual decrease. 6 double line graph plots Z against cross-shore distance for the maximum and flooded zones, demonstrating an increasing trend. Additionally, a map depicts ground elevation, ranging from 0 to 90 meters, with the highest elevation at the center.
Fig. 9
Spatial distribution of the total coastal flooding during the passage of Iota in San Andrés including percentage components and time series during the storm peak (72 h) of Astronomical Tide (AT), Storm Surge due to pressure (SP), Storm Surge due to wind (SV), wave run-up (Rhigh), 98th percentile of wave run-up (Rhigh2), and the run-up component subtracting sea level component (R2)
6 multi-line graphs display water levels over time for U16, U24, P04, P06, and U11 with lines labeled A I, S S, S P, S V, and R2, exhibit a gradual decrease. 6 double line graph plots Z against cross-shore distance for the maximum and flooded zones, demonstrating an increasing trend. Additionally, a map depicts ground elevation, ranging from 0 to 90 meters, with the highest elevation at the center.
Fig. 10
Spatial distribution of total coastal flooding during the passage of Iota in Providencia and Santa Catalina, including percentage components and time series during the storm peak (72 h) of Astronomical Tide (AT), Storm Surge due to pressure (SP), Storm Surge due to wind (SV), wave run-up (Rhigh), 98th percentile of wave run-up (Rhigh2), and the run-up component subtracting sea level component (R2)
×
×
A significant contribution (>62%) to the total flood level due to Pressure (SP) is observed in San Andrés (Fig. 9), especially at the northeast profiles (S01, S03) and between 45–52% for the rest of the island (S04, K07, K10, K18). On the other hand, the contribution of wave run-up (R2) to the total flood level is more noticeable towards the area of Sound Bay (K07) with a percentage around 54% with flood levels around 2 m, reaching more than 200 m of flooding towards the coast. In the area near the Old Point mangrove (S04), given the topo-bathymetric configuration and the low slope of the terrain, a 901 m advance of the seawater towards the coast and a maximum flood level, which exceeds 2 m during 5 consecutive hours, can be observed during the peak of the hurricane according to the numerical model.
In Providencia and Santa Catalina (Fig. 10), the contribution of the pressure component (SP) from the storm surge to the total flood level is dominant in P04 (Santa Catalina) and P06 (McBean Lagoon) with contributions of 65.9% and 71.3% respectively, reaching flood levels around 1.5 m and 472 m of coastal flooding at the P06 profile. The maximum run-up was obtained at U24 (profile located at the northwest of Providencia) where the wave run-up (R2) represented 70.8%, exceeding the effect generated by the storm surge (SS) and reaching a maximum flood level of 5 m above the mean sea level. In general, the contributions of the astronomical tide (AT) are not significant compared to the contributions of SP and R2 in the archipelago.
5.3 Urban Flooding
Figure 11 shows the SWMM maximum nodal levels for Hurricane Eta for San Andrés, and Hurricane Iota for Providencia and Santa Catalina.
An aerial view exhibits the water levels for San Andrés and Providencia and Santa Catalina. The flood level ranges from less than 0.10 to greater than 1.00.
Fig. 11
a Maximum extreme water levels for San Andrés and b Providencia and Santa Catalina during the most extreme hurricane events
×
The results obtained from the SWMM simulations for San Andrés show that the most critical values of maximum flood levels occur toward the north and center of the urban area near the airport. The most critical region is located in the vicinity of the El Isleño hotel, around the ecological park of San Andrés near the entrance to the airport runway. The critical zone extends around the airport toward the south through Swamp and Juan XXIII Avenues in the area of influence of the School House neighborhood. An area of high flooding is also observed near the Cartagena Alegre and Swamp Ground neighborhoods. These results are consistent with the flood sectors provided by CORALINA, historical photographic records, citizen reports, and even news records generated during the passage of Hurricane Eta (in early November 2020), showing strong floods in neighborhoods such as Natania, Serranilla, School House, Cocal, Santana, and Juan Avenue XXIII, areas of high population density in San Andrés (see Fig. 3a for neighborhoods).
Other regions with extreme values observed are found towards the north of the island in the San Francisco de Asís neighborhood. This area is not mentioned in the vulnerable areas reported in the information provided by CORALINA. Towards the southeast of the urban area of San Andrés, throughout the eastern coastal zone, areas with maximum flood values are observed near the maritime terminal and Forbes Landing, further to the northeast, Colombia Avenue near the Aquarium hotel and Las Américas Avenue are also reported as areas of high recurrence of flood events by CORALINA (see Fig. 3a for neighborhoods).
For Providencia, the results for the approximate drainage conceptualization show how the areas of greatest flooding are located toward the north of Santa Catalina Bay in the area known as Old Town, to the west in the region of the Agua Dulce (Fresh Water Bay) and to the east in the area surrounding the Providencia airport (see Fig. 3b for neighborhoods).
Since the SWMM model is a unidimensional model with several approximations and despite the fact that water flooding elevation was tested comparing with real values observed by the local community in some specific points, and that the SWMM model results are consistent considering the frequent flooding areas in San Andrés island reported by the Plan Maestro de Alcantarillado (Consorcio Plan Vial Caribe 2007), the flooding results during hurricanes Eta and Iota must be considered as indicatives.
6 Discussion and Conclusions
The results show the potential of mathematical modeling and field measurements to explain the behavior of the atmosphere-ocean-land interaction, particularly extreme hurricane events, and their impacts on coastal areas. The passage of Hurricanes Eta and Iota through the archipelago showed the differentiated contribution of each hazard (rain, wind, and coastal flooding) on physical infrastructure, coastal ecosystems, and population. Particularly, for Iota the magnitudes of winds are influenced by the distance to the coast, thus the impact was not the same for San Andrés (around 103 km from the hurricane’s eye) as for Providencia (approximately 20 km) during the maximum development of the cyclone.
The wave results were validated (not shown here) for deep water with other hurricanes (Dean and Matthew) that have been reported in the literature using in-situ data available from the National Data Buoy Center (NDBC). In shallow waters for waves and storm surge levels, rainfall, drainage, among others, a direct validation was not possible since there is no monitoring system at the institutional level that permanently records these variables. Therefore, it is recommended to complement the information obtained with in-situ operational measurement systems (coastal buoys, water level sensors, and weather stations) and remote data provided by video camera systems, among others. All these measurements and modeling elements are part of the recommendations of the Sustainable Development Goals (SDGs) and the Decade of Ocean Sciences declared by the United Nations (UN).
For coastal flooding, the contributions of the storm surge (contributions of pressure and wind forcing) were larger for San Andrés, while in some areas of Providencia (e.g., the central area of Pueblo Viejo) the effect of the wave run-up was four times higher than the storm surge effect. The wave attenuation in distinct locations of the archipelago might suggest the importance of coastal ecosystems (e.g., coral reefs, mangroves, seagrasses, or beaches) for protecting coastal communities. In San Andrés, the profiles S01 and S03 showed the lowest contribution of wave run-up, probably due to the wave damping provided by the barrier reef at the northeast side of San Andrés, while in Providencia, the lowest wave run-up contribution was obtained along the main mangrove area of the McBean Lagoon National Natural Park.
Regarding urban flooding modeling for the archipelago, it is suggested to involve all possible existing hydraulic elements in the complex network of the archipelago in order to achieve a more comprehensive representation of the urban drainage process. Although the dynamic wave method to solve physical processes of urban drainage was used, the conceptualization carried out for the archipelago represents a simplification of reality and does not consider all the existing hydraulic elements in such complex networks. Considering the lack of information, further implementation of an automatic flow and level monitoring system is also recommended for the storm sewer network or other hydraulic elements (e.g., gutters and natural channels). Although the results were consistently adjusted to the visual perception of the inhabitants, it is recommended to keep an accurate record of the water sheet through a community monitoring system, including flood level and photographic records, among others. Such data could complement the information supplied by other entities in charge of risk management in the archipelago. The urban flood modeling showed the enormous potential to incorporate the results into urban planning schemes and the development of sustainable drainage systems.
Regarding atmospheric modeling, simple parametric methodologies were used to define flooding components such as storm surge (effect of wind and pressure). Although the results are robust based on historical reports, newspapers, and information provided by inhabitants, there is a need to employ methodologies that consider the coupling of physical-based numerical models such as the Regional Ocean Modeling System (ROMS) and Delft 3D, among others. This will allow for improvements in the coastal flooding representation, given that storm surge represents one of the most important components, even exceeding other essential components like sea level anomalies, astronomical tide, or wave run-up in the archipelago.
Topo-bathymetric changes are among the most relevant factors for wave propagation in shallow waters and coastal flooding. The results obtained from XBeach are strongly dependent on the available bathymetric information and bottom roughness. Hence, the influence of coral cover, grasses, mangroves, and dune vegetation, among others, should be incorporated into the models and strategies for planning Nature-based Solutions (Osorio-Cano et al. 2019). As such, it is recommended to apply satellite, drone, Remotely Operated Vehicle (ROV), and other techniques to make more robust digital terrain models.
These effects related to physical processes due to hurricanes and extreme events must be incorporated into territorial planning and decision-making regarding vulnerability and risk. Likewise, the reconstruction of infrastructure must be guided by the accurate diagnosis offered by numerical tools to predict the wind/wave climate and the interaction with the insular zones. The maximum winds determine the building material to be used and/or the ecosystems that can provide urban protection as an ecosystem service. The level of coastal flooding brings elements to define the retreat and/or shelter zones for evacuation, as well as the type of infrastructure that can be considered in these vulnerable zones. Hence, strategic coastal ecosystems (e.g., mangroves, corals, and seagrasses) are highlighted for providing coastal protection. The level of urban flooding allows the re-dimensioning of drainage systems and thinking about territorial planning solutions that include Sustainable Urban Drainage Systems (SUDS).
Nature shows us the way for future planning, enhancing the chance of Building with Nature (BwN) and not against it, in order to achieve protection against flooding and coastal erosion. Although the archipelago can be considered a region with scarce historical hurricane events, knowing the more susceptible areas in this region is useful information for policymakers as it encourages science-based decision-making and constitutes a key step towards the consolidation of a risk assessment in the archipelago.
Acknowledgements
The authors want to thank CORALINA for providing data and financial support. Thanks also to the collaborators Victor Saavedra, Jose Daniel Rios, Juan Pablo Ramirez, Alejandro Álvarez, Jhayron Pérez, Mauricio Zapata, Victor Rua, and Mariana Roldán for the technical support and numerical modeling. Thanks also to Alejandro Henao, David Quintero, Simon Acevedo, and Margarita López during field instrumentation and data processing. Thanks also to IDEAM, IGAC, and NOAA for allowing open access data. Finally, to CEMarin, which supports the publication of this chapter.
Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.
The images or other third party material in this chapter are included in the chapter's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
