This study assesses Hurricane Iota’s impact on Providencia island’s reef environments, using Google Earth Engine, Satellite Derived Bathymetry, and machine learning to calculate a supervised classification process to delineate six geomorphic reef units. Results reveal dynamic changes, including erosion in the Lagoon unit (4.47% pre-Iota, 2.27% post-Iota), loss on the Back Reef (38.14%), and Rock Terrace (6.15%). Reef Ridge showed minimal change, acting as an effective wave barrier. Back Reef and the deep Rock Terrace experienced significant erosion (−3 to −14 m) to the northeast, with sedimentary dynamics observed in deeper units (up to 22 m). The high thematic accuracies found (Kappa 99%) illustrate the effectiveness of the assessment to (i) map the reef rapidly, (ii) provide tools for long-term monitoring of changes over time and (iii) improve management strategies and decision-making.
1 Introduction
Shallow water coral reefs provide valuable protection to coastal infrastructure from storm surge and waves. However, the same reefs that provide protection can also be damaged by storm-related wave energy. As the environment continues to change through both natural and human-influenced means, catastrophic events such as hurricanes are projected to have much larger devastating effects on small island territories. The intensity, frequency, and duration of hydrometeorological events such as hurricanes increase the risk of a grave danger by storm surge and waves substrate erosion, or sediment deposition (Verfaillie et al. 2009; Goes et al. 2019; Kumar et al. 2021). In addition to the socioeconomic risk, the spatiotemporal disturbance affects the habitat suitability for benthic flora and fauna (Post 2008; Lecours et al. 2015).
A novel approach introduces Nature-based Solutions (NBS), inspired by nature, and more efficient cost-effective management strategies to mitigate the increasing risk from hydrometeorological hazards (HMHs). For marine spaces, the NBS strategy focuses on developing large-scale bathymetry and geomorphological preliminary approaches to monitor the complexity of seafloor changes caused by HMHs that may generate socioeconomic and environmental losses (UNISDR 2009). Although the technologies developed in the field of remote sensing for marine environments have included mechanisms for geomorphic mapping and classification using strong computational approaches (Kennedy et al. 2021), machine learning classification and cloud computing processing offer a unique opportunity to access petabytes of free-access data generated from moderate-resolution satellites such as Sentinel 2 (Gorelick et al. 2017).
Anzeige
Hurricane Iota made landfall on the island of Providencia as a strong category 5 storm on November 17, 2020, between 4:00 and 7:00 a.m., passing within 10 km north of Providencia with sustained wind speeds >250 km/h and gusts of 270 km/h (INVEMAR 2021). Hurricane impacts on coral reefs can come in many forms, from broken pieces missing from branching coral species to entire colonies dislodged, cracked, or shattered into multiple pieces or fragments. They can also strip off the superficial reef framework, deposit loosened material onto beaches or cays above sea level, or propel them into deeper sub-reef environments. However, the assessment of these impacts by traditional field surveys is time-consuming and expensive to investigate the full extent and magnitude of the reef geomorphic changes, in order to support effective post-hurricane management approaches such as the emerging areas of NBS and climate change adaptation. Thus, this powerful and rare weather event provides an opportunity to examine the effects of extreme physical forces on coral reefs and their impact on reef geomorphology.
Accordingly, this study is focused on a rapid assessment of Hurricane Iota’s impact on the coral reef geomorphology in Providencia, located in the Seaflower Biosphere Reserve that offers the ideal conditions to generate accurate maps using remote sensing techniques, due to its clear water conditions and shallow depth platforms. Calibrated and corrected Sentinel image composites for the entire island were generated using Google Earth Engine (GEE) for a comparable pre-Iota and post-Iota timeframe that accounted for reef geomorphic zonation. Our results demonstrate how open-source satellite imagery (Sentinel 2) enables efficient analytic processes of changes during the investigated event. Results show dynamic variation in the geomorphic units following a major hurricane event and could lead to improved management strategies such as (i) restoration efforts, (ii) monitoring activities for the analysis of the timing and nature of recovery initiation after impact, (iii) the spatial prioritization of conservation activities such as the reefs with the best chance of survival with the increasing frequency of extreme events, and (iv) to improve the implementation of NBS approaches and climate change adaptation strategies.
1.1 Study Area
In the Archipelago of San Andrés, Providencia and Santa Catalina, among the oceanic islands, atolls and banks, the island of Providencia—also known as Old Providence—is found, which extends 7.2 km across from north to south. The barrier reef—the second largest in the Southeastern Caribbean—is an extensive calcareous platform stretching over 32 km (Diaz et al. 2000; Sanchez et al. 1998). Situated at 25 m deep, there is a submerged elongated ridge in a shelf-margin position, which may be a drowned shelf-edge barrier reef. Geister (1992) and Geister and Diaz (2007) clearly describe the reef complex’s geomorphology: the lagoon platform is occupied by extensive semi-closed and gently sloped terraces up to 14 m deep with areas 2–6 m wide that are occupied by an extensive shallow lagoonal terrace. Front Reef is a fore-reef terrace in front of the shallow peripheral reef, it is up to several meters wide and slopes gently to the Rock Terrace. A significant part of the barrier reef is formed by a wide belt made up of numerous patch reefs, mostly of the pinnacle type, which rise from the seafloor at −6 to −8 m and reach the low-tide level. Occasional storms with westerly or northwesterly winds reaching speeds over 20 m/s do occur, mostly in the second half of the year (Geister and Diaz 2007). The mean annual air temperature is 27 °C, with a 1 °C range between monthly values, while rainfall is irregular and varies greatly from year to year. According to Geister and Diaz (2007), sedimentation processes in the area are controlled by the surface persistent northward flow of the Caribbean Current through large gaps and narrow open seaways across the top of the Nicaraguan Rise.
2 Methodology
2.1 Image Pre-processing
Moderate-resolution satellite imagery from GEE JavaScript-API was selected from Optical Sentinel 2-MSI Multispectral imaging bands and 10 m of spatial resolution, as shown in Table 1. Using image metadata (cloud < 10%) and reducing filtering in GEE libraries, we selected atmospherically corrected scenes from “COPERNICUS/S2” level-2A of surface reflectance before and after the Iota HMHs event (November 17, 2020). Scenes of the shallow and clear sea bottom in Providencia were reduced to annual averaged mosaics with few clouds between 2019 and 2021 (Table 1). We selected all bands used in the visible regions: red, green, and blue (0.45–0.68 μm) in the image, because they provide the spectral attenuation differences in seafloor structures at different depths as a function of wavelength. In addition, the near-infrared band (0.78–0.90 μm) was included to reduce the solar brightness on the sea surface, to mask the land and the wave crest (Fig. 1).
Table 1
Sentinel 2 MSI scenes in Providencia
Date
Image ID
Cloud cover (%)
2019-09-10 to 2019-09-20
20190910T155752_T17PMQ-PMR
20190920T155836_T17PMQ-PMR
<10
2020-09-08 to 2020-09-19
20200909T155527_T17PMQ-PMR
20200919T155527_T17PMQ-PMR
200914T155529_T17PMQ-PMR
<10
2021-01-27 to 2021-02-01
20210201T155525_T17PMR-PRM
20210201T155525_T17PMQ-PMR
<10
A methodological diagram mainly includes 4 phases. 1. Optical preprocessing. 2. Optical processing and water column correction. 3. Benthic terrain modeler. 4. Machine learning. It generates geomorphology maps and D T M maps.
Fig. 1
Methodological phases of optical preprocessing, water column correction and machine learning in Sentinel 2
×
Anzeige
2.2 Image Processing
2.2.1 Deglint Correction
Water outflow reflectance can be difficult to observe due to the reflection of direct sunlight at the air–water interface in the satellite direction. In addition, specular reflection of the incident radiation occludes the benthic component in optical remote sensing and confounds the visual identification of the bottom feature that could influence the image classification. As a result, brightness in the surface water pixels was removed using the algorithm proposed by Hedley et al. (2005) (Table 2 and Fig. 1). To do so, a regression was performed in GEE between the near infrared (NIR) brightness and the visible bands using open-water pixels free of the NIR bottom reflectance and following the indications of ESA (2019). The outcome of this correction is illustrated in Fig. 2.
Table 2
Column water correction and image processing algorithms
\(R^{\prime}_{i}\) is the deglinted pixel in band i. \(R_{i}\) is the reflectance from a visible band i. \(b_{i}\) is the regression slope. \(R_{NIR}\) is the NIR band value. MinNIR is the minimum NIR value of the sample
Z is the derivate depth. m1 is a tunable constant to scale the ratio to depth, n is a fixed constant for all areas, and m0 is the offset for a depth of 0 m
Zannulus is the mean elevation value of all cells within an annulus-shaped neighborhood
Two satellite images of a beach area. A. The land and water areas are visualized. B. The land area is dark, and the water is more shiny.
Fig. 2
Glint (a) and Deglint (b) correction in east Providencia shallow water setting
×
2.2.2 Land, Wave Crest, and Cloud Shadow Masking
Masking the land, white wave crest, and cloud shadow is an essential processing step to separate bright features that can be identified by high reflectance in the NIR. These areas interfere with the spectral response in the classification; therefore, it is important to mask them. NIR wavelengths do not penetrate the water, so after deglint, clear areas of water appear very dark. To solve this, an operation threshold was applied in NIR < 0.05 over the deglinted image (Table 1). The resulting mask was vectorized and the image clipped from the shallow bottom to the intermediate depth up 20 m. In addition, the dark shadows of clouds were also clipped using a threshold (B2 > 0.01) in the blue band.
2.2.3 Depth Invariant Index
To remove the confusing influence of variable water depth we applied the method provided by Lyzenga (1978) that compensates for the variable effect of depth when mapping bottom features. The first processing linearizes the effect of depth on reflectance with a natural logarithm (Table 2). To establish the depth-invariant indices, we compute the ratio between Sentinel-2 bands 1–2 in GEE over the same bottom type at different depths following the indications provided by ESA (2019).
2.2.4 Satellite-Derived Bathymetry (SDB)
The satellite-derived bathymetry (SDB) implemented was based on the principle that the water column attenuation coefficients differ between spectral bands and that the ratio between the two visible bands will change with depth (Stumpf et al. 2003). To determine the ratio between the two bands, we used GEE to apply the empirical regression method for mapping shallow waters using the algorithm proposed by Stumpf et al. (2003) (Table 2). The Z data comes from in-situ bathymetric data collected by Consorcio Dragado Providencia and provided by Aqua & Terra S.A.S (Fig. 3). The data were collected from a multi-beam High Frequency (200 kHz) MBI ODOM echosounder. Finally, to better observe the geomorphic features of the reef, we built a Digital Terrain Model (DTM) in Surfer 17.1.288.
Two satellite images of a beach area. A cluster of dark dots presents the bathymetry points. The two regions are illustrated as very shiny.
Fig. 3
Bathymetry points in Providencia collected by Consorcio Dragado Providencia and provided by Aqua & Terra. The data were collected with a multi-beam High Frequency (200 kHz) MBI ODOM echosounder system
×
2.2.5 Benthic Terrain Modeler (BTM)
We employed the derived factors such as the Bathymetric Position Index (BPI) broad and fine, and the slope calculated by the Benthic Terrain Modeler (ArcGIS 10.8 toolbox) and the satellite-derived Bathymetry (SDB) as input layers for the machine learning processes.
2.2.6 Machine Learning (ML) and Land Change Modeler (LCM)
We selected the Random Forest (RF) algorithm in GEE to perform a supervised classification image of Providencia reef geomorphic zones. As indicated above, we employed the 2.3.5 BTM data, plus the visible bands (Blue, Green, and Red) of the Sentinel 2 image. In addition, we directly photo-interpreted the corrected image in GEE to produce the polygons as training data. The resulting supervised classification was then imported to IDRISI Land Change Modeler (LCM) toolbox, to perform a multitemporal analysis with which to evaluate the spatial changes in the reef geomorphic zonation during the 2019–2020 and 2020–2021 study periods. The Land Change Modeler (LCM) was developed (Eastman 2006) as a change projection tool to support a wide range of planning activities. This modeler has been designed for REED projects. (Reducing Emissions from Deforestation and forest Degradation), but its applications can be observed in different earth change modeling investigations using CA-Markov (Areendran et al. 2013; Halmy et al. 2015; Eastman et al. 2018), modeler the Earth Trends (Fuller et al. 2012), Forestry changes (Holmer et al. 2001; Hill et al. 2003) and biodiversity and Habitats (Poirazidis et al. 2006; Bino et al. 2008). LCM examines each of the historical classification’s pixel transitions, change pixels and number of persistence pixels. The resulting models show the units of change without subjective intervention and are an alternative to geo-statistical techniques.
3 Results
3.1 Geomorphic Classification
Using the optical properties of Sentinel 2, the SBD bathymetry, and the factors derived from the BTM, the Bathymetric Position Index (BPI) and the slope, we delineated six geomorphic units (GU) occupying different percentages of surface in shallow waters (<25 m): Lagoon (44.2%), Rock Terrace (40.15%), Back Reef (6.3%), Reef Crest (4.4%), Sand Terrace (3.3%), and Front Reef (1.5%).
The 3D Digital Terrain Model (Fig. 4) of the Satellite-Derived Bathymetry (SDB) shows a heterogeneous and discontinuous, 26 km long stretch of reef complex in the eastern and western part of the island with an extensive Reef Crest and pinnacles segments to the north (4–10 km) and to the windward east (10 km). An extensive lagoon area towards the north of the island showed the presence of irregularly distributed patch reefs and a broad shallow marine terrace covered by fine sediments. In the deepest areas, satellite information allowed us to identify the outer reef Rock Terrace (18 m) and the Sand Terrace up to 20 m deep, demonstrating the capabilities of the sensor reflectance response in the deepest zone of oligotrophic and ocean waters.
A 3-D digital terrain model of the satellite-derived bathymetry. It presents a heterogeneous and discontinuous, long stretch of reef complex. The emerged island is highlighted. The north direction is marked.
Fig. 4
Digital terrain model (DTM) showing the reef complexity and the reef extension at Providencia
×
The slope layer showed very steep slopes (between 6° and 7°) at the Rocky Terrace and the Sand Terrace corresponding to the Fore Reef and the outer slope at depths between 11 and 15 m. In shallow areas, the Lagoon, the Back Reef, and the Reef Crest were the most gently sloped areas with 0.1°–1.8° between 1 and 10 m deep (Table 3). The BPI and the slope allowed to distinguish the reef landscape structures (e.g., plains and barriers). In particular, the slope position accounted over two scales for: (1) negative and near to zero values in the Lagoon, the Rock Terrace, and Sand Terrace, and (2) positive or positive to negative values in the Reef Crest, the Back Reef, and the Front Reef (Table 3). These differences of the BPIs layers (broad and fine), the slope, and the SBD served as data inputs to run the Machine Learning process and to delineate the GU in the reef as follows:
Table 3
Upper and lower factors resulting from the BTM processing 2019–2021 in Providencia
Class
Zone
BroadBPI_Lower
BroadBPI_Upper
FineBPI_Lower
FineBPI_Upper
Slope_Upper
Depth_Lower
Depth_Upper
1
Reef Crest
235
413
372
426
1.2
−4
−1
2
Lagoon
−120
−31
−35
118
0.5
−10
−8
3
Back Reef
57
324
−35
118
1.4
−6
−3
4
Front Reef
225
146
−35
118
2.6
−8
−6
5
Rock Terrace
−35
−31
−120
−35
6
−14
−11
6
Sand Terrace
−120
−31
−35
118
7
−15
−13
Reef Crest: Covers 11.2 km2 of the survey area; this intertidal zone represents the shallowest or emerged part of the reef due to the presence of live coral mounds or pinnacles that reach the surface with a steep slope. The SBD-derived model showed extensive breaker zones located to the north and east with interruptions up to the main reef ridge that is slightly more continuous and curved to the east up to 4 m in depth (Fig. 5).
A S B D derived model eastern reef of Providencia. The different zones are marked as Lagoon, Back Reef, Front Reef, and Rock Terrace. The back reef break, crest line, reef front break, and mid-self-break line are marked.
Fig. 5
Geomorphic units (GU) in the eastern reef of Providencia. Sentinel 2 corrected image date February 1, 2021
×
Lagoon: This zone was the most extended class (112.6 km2) with highly depositional environments in the flat plains between −8 and −10 m. These areas included patch reefs observed in the images with darker spots and are easy to differentiate from sandy bottoms with higher reflectance. Additionally, the lagoon shows large coral heads and pinnacles in shallow areas (Fig. 5). The lagoon is variable in extension from the east coast between 0.2 to 2 km to 10 km in the flat plain to the North (Fig. 6).
3 geomorphic units present geomorphic cover changes. These units highlight the reef crest, lagoon, back reef, front reef, rock terrace, and sand terrace.
Fig. 6
Geomorphic units pre-Iota 2019 (a), 2020 (b), and post-Iota 2021 (c) at Providencia
×
Back Reef: This unit is located behind the Reef Crest, covers around 16 km2 and shows a gently sloping surface adjacent to a Reef Flat. The area is sheltered by sediments dominated by coral rubble and broken reef material over a variable bottom depth (6–3 m). The extent of the Back Reef to leeward is greater on the exposed eastern side of the island, averaging one kilometer in width (Fig. 5). The presence of debris detected in the image is different from the sand which shows more brightness reflectance. The debris area is located next to the Reef Crest, while the sand area is located towards the Lagoon. In both cases, the depositional material comes from the Reef Crest and the Front Reef.
Front Reef: This is a narrow and sloping area extending from the Reef Crest at the windward margin towards the Rock Terrace. This area covers 3.8 km2. It is characterized by high wave exposure (Fig. 5).
Rock Terrace: This feature is the second most extensive GU in the area, covering 102.3 km2, and represents a division between the Front Reef and the deeper Fore Reef in a sloping shallow area between 11 and 14 m. It shows a dark contrast in the satellite images probably due to the attenuation factor of the spectral response of the visible bands with increasing depth. Between the Rock Terrace and the Sand Terrace, the Mid-Shelf Break is a widespread slope break delimiting the edge of the inner Front Reef shelf and occurs in depths between 10 and 15 m.
Sand Terrace: The sandy surface contrasts with the Rock Terrace in some sectors of the survey area between 13 and 15 m and occupies an area of 8.42 km2. The sandy terrace forms channels parallel to the reef crest about 2 km offshore (Fig. 5). The sandy formations are at the detection limit of the satellites, where the bathymetric models lose accuracy, and it is difficult to interpret the geomorphology of these structures.
3.2 Cartography Accuracy
The evaluation of the thematic accuracy in the models generated from the ML process resulted in high global accuracy and kappa index metrics in all years of the multitemporal analysis as follows: The producer’s accuracy ranged from 99.64 to 100% accuracy, which evidences the suitability of the Random Forest algorithm to separate the pixels between the GU’s using the BPI bands and the Sentinel 2 visible bands (Table 4). The user’s accuracy showed high percentages between 98.04 and 100% representing an adequate selection of training samples associated with the established polygons in each of the GU assessed.
Table 4
Confusion matrix, overall accuracy, and Kappa index for 2019, 2020, and 2021
Coverage classes
Reef Crest
Lagoon
Years 2019/2020/2021
Rock Terrace
Sand Terrace
User accuracy (%)
Back Reef
Front Reef
Reef Crest
688/493/835
0/0/7
3/0/0
9/0/0
0/0/0
0/0/0
98.3/100/99.2
Lagoon
3/0/3
10,136/6069/6569
0/0/0
0/0/0
0/0/0
0/0/0
99.9/100/99.9
Back Reef
0/1/0
0/0/0
2645/844/2595
0/0/0
0/0/0
0/0/0
100/99/100
Front Reef
1/0/0
15/0/0
0/0/0
799/155/502
0/0/0
0/0/0
98/100/100
Rock Terrace
0/0/0
0
0/0/0
0/0/0
1281/1454/1041
0/0/0
100/100/100
Sand Terrace
0/0/0
0
0/0/0
0/0/0
0/0/0
120/188/308
100/100/100
Producer accuracy (%)
99.4/99.8/99.6
99.8/100/99.9
99.9/100/100
98.9/100/100
100/100/100
100/100/100
Global accuracy
0.99/0.99/0.99
Kappa index
0.99/0.99/0.99
3.3 Geomorphic Cover Changes
3.3.1 Pre-Iota 2019–2020 Cover Changes
Based on the ML processing with high precision cartography, we detected conspicuous changes comparing the pre-Iota 2019–2020 and post-Iota 2020–2021 GU classifications (Fig. 6). The multi-temporal analysis assessed the biannual changes under typical climatic conditions and, without the presence of extreme weather events (pre-Iota). Results showed growths of the deposited material (18.55%) around the Back Reef covering a new area of 3.65 km2. The Sand Terrace showed an increase of 22.44% covering a new area of 1.89 km2. Finally, the area showing the largest increases was the Front Reef, which increased 88.5% covering an estimated area of 1.81 km2. Finally, the GU that showed slight tendencies to lose material (0.01–4.74%) were the Reef Crest, the Lagoon, and the Back Reef (Fig. 7).
A 3-D bar graph compares the geomorphic changes of the reef crest, lagoon, back reef, front reef, rock terrace, and sand terrace, in years 2019, 2020, and 2021.
Fig. 7
Geomorphic changes in km2, pre-Iota 2019–2020, and post-Iota 2020–2021, at Providencia
×
The multi-temporal analysis assessed the biannual changes under typical climatic conditions and, without the presence of extreme weather events (pre-Iota), showed deposited material (18.55%) around the Back Reef covering a new area of 3.65 km2. The Sand Terrace showed a 22.44% covering a new area of 1.89 km2. Finally, the area showing growth was the Front Reef, with 88.5% covering an estimated area of 1.81 km2. Finally, the GU that showed slight tendencies to lose material (0.01–4.74%) were the Reef Crest, the Lagoon, and the Back Reef (Fig. 7).
3.3.2 Post-Iota 2020–2021 Cover Changes
After the passage of Hurricane Iota, between November 2020 and the first months of 2021 (Fig. 7), an opposite behavior was observed with a loss of coverage in the Back Reef of 5.4 km2 (38.14%) and a decrease in the Front Reef with 3.77 km2 (64.75%). Furthermore, there was a noteworthy increase in the area covered by the Sandy Terrace, 11.41 km2 (57.56%).
A detailed analysis of the spatial distribution of changes calculated with the LCM for persistence, loss, and gain of the area in the pre- and post-Iota periods, showed different responses: The Back Reef reduced its coverage in an extensive area towards the north and the east part of the reef. The material moved to areas such as the Lagoon and Reef Crest, with a gain contribution between 5 and 7 m in depth (Table 5 and Fig. 7). In comparison, a decreasing trend was observed in the post-Iota period in the Rock Terrace at the northeast of the barrier reef. In this case, the material shows a contribution to the Sandy Terrace at depths between 7 and 10 m (Fig. 8). Finally, the Lagoon Unit lost coverture in all periods (pre- and post-Iota), but this was more localized in the deeper area of the Rock Terrace (Fig. 8).
Table 5
Losses and gains (km2) pre-Iota 2019–2020 and post-Iota 2020–2021 in the geomorphic units (GU) at Providencia
Cover class
2019–2020
2020–2021
Loss/gain (km2)
Net
Loss/gain (km2)
Net
Reef Crest
A symbol of an arrow directing downwards.
−3.55/3.02
−0.53
A symbol of an arrow directed upwards.
−3.75/3.11
−0.64
Lagoon
A symbol of an arrow directing downwards.
−10.40/5.85
−4.55
A symbol of an arrow directing downwards.
−11.17/8.90
−2.27
Back Reef
A symbol of an arrow directed upwards.
−1.04/4.69
3.65
A symbol of an arrow directing downwards.
−6.71/1.27
−5.44
Front Reef
A symbol of an arrow directing downwards.
−2.37/0.57
−1.8
A symbol of an arrow directed upwards.
−0.47/4.24
3.77
Rock terrace
A symbol of an arrow directed upwards.
−5.73/5.74
0.01
A symbol of an arrow directing downwards.
−14.43/8.51
−5.92
Sand Terrace
A symbol of an arrow directed upwards.
−1.37/3.25
1.88
A symbol of an arrow directed upwards.
−1.34/12.75
11.41
12 Geomorphical maps the losses and gains pre-lota and post-lota. The Lagoon unit lost coverture in all periods, but this is more localized in the deeper area of the Rock Terrace.
Fig. 8
Geomorphic units’ losses, persistence, and gains pre-Iota 2019–2020 and post-Iota 2020–2021 at Providencia
×
4 Discussion
Our geomorphological characterization of recent changes in Providencia offers a synoptic view of the magnitude and power of a category 5 hurricane event and its impact on the tropical shallow coral reef.
4.1 Changes in Shallow Units (Reef Crest, Lagoon, and Back Reef)
The GU with the greatest geomorphological stability was the Reef Crest, where insignificant structure variations were observed (0.64 km2). The structural stability of the Back Reef could indicate that this unit performed well as a NBS to the beating by the hurricane, and acted as a good barrier against wave power. The role of the Reef Crest as a buffer for storm energy has been extensively documented in various studies describing how much (up to 86%) of the incoming wave energy is dissipated by this feature (86%) (Beck et al. 2018 and references therein). Beck et al. (2018) performed flooding model scenarios of storm events with and without a Reef Crest. The reef scenarios attain only a decrease of 1 m in the height and roughness of Reef Crest. The authors concluded that without reefs, annual damages on the coast in the United States would be more than double (118%) and land flooding would increase by 69%, affecting 81% of the population. Furthermore, the use of natural barriers for sea-level rise and storm surges is becoming increasingly popular as a source of nature-based risk reduction options. Accurate seafloor maps are essential for determining the wave degradation of benthic habitats such as coral reefs, with the necessity of current and repeatable observations of sediment stability and geomorphic complexity (Fourqueran et al. 2020) as the ones presented here.
Regarding the Lagoon, this unit showed overall stability in the extension of the central body, both pre- and post-Iota. However, erosion zones were observed at depths beyond 7 m at the edge of the Lagoon area following the passage of the hurricane. This effect was observed all around the island with some focus on greater change to the northeast, southeast, and southwest. In-situ evaluations conducted by INVEMAR (2021) showed a generalized disturbance of 65% in the Lagoon and the Back Reef. As part of the disturbance, large colonies of Orbicella annularis and Orbicella faveolata were observed overturned in settings shallower than −6 m. Highly impacted sites such as Marcela’s place (a monitoring point within the MPA) with overturned colonies, located in transitional zones between the Back Reef and the Lagoon, could be affected by the dynamics of sediments transported from the Reef Crest and the Back Reef to the Lagoon. Similar results were also found by Bries et al. (2004) who reported that larger colonies of a species were more prone to damage than smaller colonies of the same species. They also claimed that larger, more aged colonies are not necessarily more strongly attached to the substratum, because their attachment bases become more likely weakened through bioerosion, rendering them vulnerable to toppling under sufficient wave energy. In some cases, the bioerosion of the colonies can be observed on the reef forming strips, channels, or lobes such as the ones observed on the satellite images in the same GU.
Finally, after the hurricane hit, we also observed the loss of coverture in the Back Reef, we detected eroded zones of 5.4 km2 that then moved with a net contribution to the Lagoon. These changes contribute to formed lobes or channels detected on the leeward side of the Reef Crest mostly on the northern reef. For instance, in the Lagoon zone, we observed accretion lobes between 50 and 150 m in length and 3 and 7 m in depth. However, under pre-Iota conditions, an increase (3.65 km2) in the amount of sedimentary material was also observed in the Back Reef, where unconsolidated granular material was transported from the Lagoon. There is evidence that accretion processes on the reef are heterogeneous in space and time and during different intervals (Medina-Valmaseda et al. 2020), and that some sections of a reef, such as the Back Reef, represent a highly depositional system (Kennedy et al. 2021).
4.2 Changes in the Deepest Units (Sand Terrace and the Rock Terrace)
After the impact of Hurricane Iota, a high-sedimentary dynamic was observed in the deepest GU (the Rock Terrace and the Sand Terrace) up to −22 m. Turbulence generated by Iota probably caused that part of the Rock Terrace was covered with sediments from the Sand Terrace and upper slope deposits (11.41 km2) in the leeward northeast and southeast side of the study area. These sandbanks represented storm ridges and channels up to 7 km long along the Rock Terrace edge (Fig. 7c). The bathymetric model shows the formation of sand channels, 2–3 m deeper than the Rocky Terrace. Compared to windward settings, the northwest leeward side showed very little change in the Rock Terrace. Beyond −22 m, we could not identify changes since the area falls outside the boundary of the detection limit for bathymetry. Similar results were reported by Scoffin (1993) and by Bries et al. (2004), who found that after hurricane impacts, the Reef Crest and Front Reef coral debris accumulate as talus at the foot of the Front Reef slope and on submarine terraces and grooves. These studies also found that carbonate sand and mud move in deep off-reef locations in the Fore Reef. Blanchon and Jones (1995) mentioned the presence of a gently sloping Rock Terrace in the reef complex around Grand Cayman, which is either covered with coral spurs or is a bare rock ground, that has been sculptured by wave scour into low ridges and shallow furrows during the passage of hurricanes. Blanchon (2011) also stated that the Sand Terrace is frequently crossed by a system of widely spaced coral spurs between which thick deposits of skeletal sand and gravel accumulate. He also suggests that this material is only mobilized during storms, and this could explain why at Providencia just a few changes were observed in the Rock Terrace and the Sand Terrace in the pre-Iota period.
5 Conclusions
Hurricanes are short-lived but highly dynamic synoptic events that could intensively impact the sediment dynamics of coral reef ecosystems in the Caribbean Sea. Here, we adapted the factors provided by the BTM such as Bathymetric Position Index (BPI) and slope, incorporating the components of visible bands of the satellite image Sentinel 2, machine learning, and cloud computing in GEE to investigate the rapid response of the reef geomorphic changes on Providencia with the passage of Hurricane Iota in November 2020. To understand these changes, we also incorporated information about the pre-Iota period 2019–2020. Our results indicate that Hurricane Iota enhanced the delivery of the reef re-suspended sediment substantially to the northeast and seaward sediment transport. As a result, high sedimentary dynamics was observed in the deepest GU (the Rock Terrace and the Sand Terrace) up to 22 m. In contrast, the unit with the greatest geomorphological stability was the Reef Crest, where no significant structure variations were observed (0.64 km2), showing that this GU performed well against the beating of the hurricane, and acted as a good barrier against wave power, illustrating the importance of this GU as a protection service, and as an approach of NBS. The Lagoon zone also showed general stability in the extension of the central body, both pre- and post-Iota. However, erosion zones were observed beyond 7 m of depth at the edge of the Lagoon area after the passage of the hurricane. Thus, GUs changes due to the hurricane produced several distinct patterns based on reef site, reef depth, and colony size (such as those observed in Marcela’s place).
The temporal resolution of sensors such as Sentinel 2, the use of the Land Change Modeler and their easy accessibility, make the integration of these approaches a very interesting alternative for monitoring reef geomorphic units covered in extreme events such as hurricanes and where impact mitigation measures can be provided using quality mapping in a rapid manner. Its integration enables long-term monitoring by observing the evolution of changes through time. Thus, providing valuable information to coastal managers and stakeholders in the decision-making process.
Acknowledgements
This research was primarily funded by the Royal Academy of Engineering (IAPP 18-19\210 made through the Industry Academia Partnership Programme scheme) and by the Corporation Center of Excellence in Marine Sciences, CEMarin. We are thankful to Aqua & Terra S.A.S for providing the in-situ high-resolution bathymetry to calibrate our Satellite-Derived Bathymetry (SDB).
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.
