Makowski, C.; Finkl, C.W., and Vollmer, H.M., 2017. Geoform and landform classification of continental shelves using geospatially integrated IKONOS satellite imagery.
Geomorphological characterization of coastal environments along continental shelves depends on accurate interpretation of mesoscale lithic and clastic benthic geoforms and landforms. Using the Geospatially Integrated Seafloor Classification Scheme (G-ISCS), cognitive visual interpretations of seafloor geoforms and their associated landforms were conducted along a diverse segment of the southeast (SE) Florida continental shelf. GeoEye IKONOS-2 satellite imagery provided the remotely sensed visual medium on which interpretations were based. With ESRI ArcGIS® ArcMap software, classification maps were created from the cognitive interpretations to show spatial distribution results of geoform and landform features throughout the study area. Additionally, smaller-scale “call-out figures” documented specific geomorphological associations among the classified units. Analysis attribute tables compared and contrasted the abundance (i.e. number of classifying vector polygons) and calculated areas for each geoform and landform classified. It was determined that classification of geoform and landform benthic features along continental shelves can be achieved where water clarity conditions allow for the cognitive visual interpretation of such seafloor formations. Future studies may build upon the classification of continental shelf geoforms and landforms to integrate more transient biogeomorphological features of the marine environment (e.g., sediment distribution, biological species identification, density of flora and fauna present), thus creating a more detailed and inclusive classification of a selected coastal region.
One of the standard components of classifying continental shelf coastal environments is the demarcation of spatially distributed geomorphological signatures along the seafloor, commonly known as geoforms and landforms. The interpretation of these general-to-specific benthic features provides a “structural framework” of the seascape topology by which a hierarchical census of biophysical units can be compiled for a given shelf region (see, for example, Fairbridge, 2004; Finkl, 2004); the ultimate result then allowing for hierarchical classifications to be extrapolated over large coastal areas and applied to corresponding mapping units. This, in return, would potentially elucidate spatial relationships between geoforms and landforms of a given area by showing how exposed lithic structures, unconsolidated clastic materials, and other geomorphological substrates available to sessile biological communities are universally distributed and interconnected. By accurately interpreting geoform and landform-based formations over a select region of continental shelf, the delineation of coastal environments and creation of a comprehensive geospatially referenced classification map can be achieved (e.g., Achatz, Finkl, and Paulus, 2009; Finkl and Banks, 2010; Finkl, Benedet, and Andrews, 2004, 2005a,b; Finkl and Vollmer, 2011; Lidz, Reich, and Shinn, 2003; Lidz et al., 2006; Makowski, 2014; Makowski, Finkl, and Vollmer, 2015, 2016; Steimle and Finkl, 2011).
In the past few decades, multispectral satellite senor images have been among the most preferred mediums by which visual interpretation of coastal environments are carried out (e.g., Andréfouët et al., 2001, 2003; Bouvet, Ferraris, and Andréfouët, 2003; Dial et al., 2003; Dobson and Dustan, 2000; Finkl, Makowski, and Vollmer, 2014; Finkl and Vollmer, 2011; Hochberg, Andréfouët, and Tyler, 2003; Klemas, 2011; Klemas and Yan, 2014; Makowski and Finkl, 2016; Makowski, Finkl, and Vollmer, 2015, 2016; Manson et al., 2001; Mumby and Edwards, 2002; Palandro et al., 2003; Steimle and Finkl, 2011). Specifically, GeoEye IKONOS-2 satellite images can offer an enhanced view of benthic geomorphological features along those continental shelf regions where optimal water clarity permits it. In conjunction with specialized spatial analysis software applications, such as ESRI's ArcGIS® 10.3 ArcMap program, IKONOS-2 images provide the visual means to cognitively identify, delineate, and classify seafloor geoforms and landforms along coastal shelf environments.
The objective of this study was to cognitively interpret, classify, and map geoforms and associated landforms along a segment of the southeast (SE) Florida continental shelf (Figure 1) using IKONOS-2 satellite imagery and the Geospatially Integrated Seafloor Classification Scheme (G-ISCS) method (Makowski, Finkl, and Vollmer, 2015). By doing so, analysis of geomorphological attributes can aid in determining how the “structural framework” over a given shelf area is spatially distributed and interrelated. Overall, the goal is to potentially expand upon the results presented in this paper, as well as other studies (e.g., Makowski, Finkl, and Vollmer, 2016), in an effort to create an inclusive classification of this continental shelf segment by integrating more transient biogeomorphological features of the marine environment (e.g., sediment distribution, biological species identification, density of flora and fauna present) with larger-scale physiographic realms and morphodynamic process zones.
GeoEye IKONOS-2 (i.e. IKONOS-2) multispectral satellite images were acquired, processed, and visually interpreted using the study methods derived from the Geospatially Integrated Seafloor Classification Scheme (G-ISCS), as developed by Makowski, Finkl, and Vollmer (2015). Specialized processing software (IDRISI© Taiga, Clark Labs, Worcester, Massachusetts, U.S.A.; Environmental Systems Research Institute's [ESRI] ArcGIS 10.3 ArcMap geographical information system program, Redlands, California, U.S.A.) allowed for appropriate digital enhancement techniques to the IKONOS-2 images, which then provided the required raster-based GIS and image processing modules for cognitive interpretation.
Using the G-ISCS method, each geoform and associated landforms were interpreted in tandem to describe the benthic substrate framework by which surface sediments accumulate upon and biological communities grow (Table 1). For example, in the case of the Coral Reef Geoform, which contains ridge-like structures throughout the Florida Reef Tract (FRT) Physiographic Realm built from living coral, coral skeletons, calcareous algae, mollusks, and protozoans, the corresponding landforms of Barrier Reef, Patch Reef, Aggregated Reef, Coral Apron, and Reef Gap were found (Banks et al., 2007; Cronin et al., 1981; Finkl, 2004; Finkl and Andrews, 2008; Finkl, Benedet, and Andrews, 2005a; Finkl et al., 2008; Hoffmeister, 1974; Jaap, 1984; Lidz, 2004, 2006; Lidz, Hine, and Shinn, 1991; Lidz et al., 1997). Hardbottom Geoforms, conversely, are exposed Pleistocene limestone (e.g., bedrock, boulders, rubble), including reef debris and dredged spoil piles, that are found intermittently in the Atlantic Ocean Morphodynamic Zones, Hawk Channel Physiographic Realm, Biscayne Bay Physiographic Realm, and along the leeward and windward sides of the Florida Keys Physiographic Realm (Chiappone and Sullivan, 1994; Finkl, 2004; Finkl et al., 2008; Lidz et al., 2006). Associated Hardbottom landforms include flat, generally featureless, continuous platforms in the form of Subtidal Pavement and conically shaped Rubble Fields and Dredged Spoil Piles (Buddemeier, Smith, and Kinzie, 1975; Finkl et al., 2008; Lidz et al., 2006; Lirman and Fong, 1997; Selkoe et al., 2009). When Islands associated with the coral and mud-based Florida Keys Physiographic Realm archipelago are classified as the predominant geoform, the resultant landforms include Bay Key, Karst Island, and Salina (Finkl, 2004; Hoffmeister, 1974; Lidz, Hine, and Shinn, 1991; Lidz et al., 1997; Nunn, 1994). Sediment Flat Geoforms, which constitute extensive, unconsolidated seafloor sediments composed of sand and mud combinations, are divided among the landforms of Intertidal Sand Flat and Planar Bed Forms and Ripples (Davis, Hine, and Shinn, 1992; Duane and Meisburger, 1969; Finkl, 2004; Finkl and Andrews, 2008; Finkl and Warner, 2005; Finkl, Benedet, and Andrews, 2005b; Finkl et al., 2008; Hoffmeister, 1974; Lidz, Hine, and Shinn, 1991; Lidz, Robbin, and Shinn, 1985). The geoform of Dune and Beach is a geomorphologically coupled coastal system consisting of Holocene dune fields, foredunes, and beach berms. The associated landforms include Bay Beach and Ocean Beach (Finkl, 1993, 2004; Finkl and Restrepo-Coupe, 2007; Finkl, Benedet, and Andrews, 2006; Hoffmeister, 1974; Wright and Short, 1984). Ridge Fields are geoforms where continental shelf sand waves display parallel long-axis alignment on sandy seafloors and contain landforms that are designated as either Discrete Ridges or Complex Ridges (Ashley, 1990; Duane and Meisburger, 1969; Finkl, 2004; Finkl and Andrews, 2008; Finkl, Andrews, and Benedet, 2006; Finkl, Benedet, and Andrews, 2005b; Finkl et al., 2007, 2008; Stapor, 1982). Channel Geoforms are low relief, elongated tidal conduits cut into seafloor bank sediments or karst bedrock. The associated landforms are Neochannel, Paleochannel, and Seafloor Channel (Davis and Fitzgerald, 2004; Finkl, 2004; Finkl and Andrews, 2008; Finkl, Benedet, and Andrews, 2005b; Finkl et al., 2008; Hoffmeister, 1974; Lidz, 2006; Lidz, Hine, and Shinn, 1991; Lidz, Robbin, and Shinn, 1985; Lidz et al., 1997, 2006). Deltas are depositional geoforms produced from the shoaling sedimentation of paleochannel tidal passes. Representative landforms include Ebb-Tidal Deltas and Flood-Tidal Deltas (Davis and Fitzgerald, 2004; Davis, Hine, and Shinn, 1992; Finkl and Andrews, 2008; Finkl, Benedet, and Andrews, 2005b; Finkl et al., 2008; Lidz, Hine, and Shinn, 1991; Schwartz, 2005). The final geoform identified was categorized as the Peninsula and Coastal Plain, which includes the low-lying marshlands of the Everglades within the Southeast Distal Florida Physiographic Realm and such morphodynamic zones as Tree Island Biohydrologic Systems, Marsh Prairie Regimes, Mangrove Forest Biomes, and Everglades Swampland Systems (Makowski, Finkl, and Vollmer, 2016). Landforms include Intertidal Mud Flat, Supratidal Mud Flat, and Anthropogenic Modified Coastal Plain (Finkl, 1994; Finkl and Makowski, 2013; Finkl and Restrepo-Coupe, 2007; Finkl et al., 2008; Gorsline, 1963; Hoffmeister, 1974; White, 1970). Locations, definitions, and descriptions of all interpreted geoforms and associated landforms along the continental shelf study area are provided in Table 1.
Definitions table for geoforms and associated landforms along the SE Florida continental shelf, including locations, definitions, descriptions, and references.
For the purposes of onscreen cognitive interpretation of geoforms and landforms, enhanced IKONOS-2 satellite images were imported into ESRI's ArcGIS 10.3 ArcMap program and displayed on a 1.2 m interactive SmartBoard® overlay system. This allowed for the real-time, onscreen digitization of geoform and landform features based on varying color tones, saturations, textures, and relative spectral reflectance signatures. Closed vector polygons were drawn around areas of similar visual composition, thereby cognitively delineating the boundaries of specific geoforms and landforms as defined in Table 1. A nominal scale of 1:6000 was selected when cognitively digitizing boundaries, as no minimum mapping unit (MMU) was designated for this study. Individual geoform and landform thematic-layered maps, along with associated legends, were created, with vector polygons being assigned a specific classifying color that corresponded to the appropriate cognitive interpretation mapping unit.
Attribute tables of all interpreted geoforms and landforms were complied for the purpose of analysis and directly corresponded to the cognitively digitized polygons. By doing so, a multitude of spatially queried information, including the areal extents of each classified geoform and landform, could be stored for further investigation and analysis.
Application of the previously described methods resulted in a classification of possible geoforms and landforms occurring along the SE Florida continental shelf. In addition to providing full-extent area maps showing the distribution of cognitively interpreted geoforms and landforms, those classifying units that recorded the highest and lowest number of vector polygons, as well as the greatest and smallest calculated areas, are reported. Furthermore, smaller-scale call-out figures are shown to provide side-by-side results of IKONOS-2 image interpretation by comparing annotated images with those containing superimposed color coded classifying units.
An amalgamation of 14 GeoEye IKONOS-2 satellite scenes resulted in a platform of high-resolution, multispectral images for the purposes of classifying 1407.4 km2 of the study area. A total of 3888 cognitive vector delineations were interpreted from the IKONOS-2 imagery (Figure 2). The boundaries of nine (9) geoforms and 24 associated landforms, with corresponding legends, are shown with color-assigned visual representations exported from ArcMap (Figures 3 and 4).
Among all the classifying units, the Coral Reef Geoform and associated Patch Reef Landform recorded the highest number of cognitively delineated vector polygons, with 2823 and 2705, respectively. The lowest number of vector polygons were tallied by Delta Geoforms (n = 5) and the Ebb-Tidal Delta Landform (n = 2). Additionally, the Sediment Flat Geoform and associated Planar Bed Forms and Ripples Landform recorded the greatest total areas, with 692.3 km2 and 681.5 km2, respectively, while the Dune and Beach Geoform (1.4 km2) and associated Bay Beach Landform (0.1 km2) recorded the smallest total areas. Figures 5 through 14 illustrate specific small-scale results when applying geoform and landform classifying units to the raw, uninterpreted IKONOS-2 satellite imagery.
Analysis of interpreted geoforms along the SE Florida continental shelf showed that the Coral Reef Geoform was among the most optimal for cognitive recognition in the IKONOS-2 imagery. When compared with other geoforms identified, Coral Reef recorded the greatest quantity of vector polygons, with over 72% of the overall total, but only constituted less than 10% of the total study area (Table 2). This was possible because of specific recognition of color, tone, texture, pattern, and spectral reflectance from the Coral Reef Geoform benthic signatures allowing for a more precise delineation without grossly overestimating the size of such features along the seafloor.
Quantity of cognitively classified vector polygons (n) and calculated areas (km2) for nine geoforms interpreted from the IKONOS-2 satellite imagery. Each individual percentage from the total is listed in brackets below.
Further examination of the landforms associated with the Coral Reef Geoform shows that geomorphological recognition of Patch Reef was the most optimal. With nearly 70% of all the vector polygons drawn in the study area and less than 2% of the total area classified (Table 3), Patch Reef Landform interpretation can be considered precise due to specific tone, hue, and spectral reflectance differences when the sunlight hits the sand or seagrass halos that usually surround these isolated reef patches. The IKONOS-2 imagery provides the visual means to discern these differences when interpreting the various landforms associated with geoforms such as Coral Reefs (Figure 15).
Quantity of cognitively classified vector polygons (n) and calculated areas (km2) for seven selected landforms interpreted from the IKONOS-2 satellite imagery. Each individual percentage from the total is listed in brackets below.
When analyzing the Hardbottom Geoform, a greater recognition of structural textures and color saturations was shown for Subtidal Pavement Landforms. This is because distinct tones and spectral reflectance of the rigid exposed Pleistocene limestone outcrops allowed for a higher level of textural contrast and color saturation characteristics to be identified in the IKONOS-2 imagery and demarcated along the seabottom. Similarly, Bay Key and Karst Island Landforms, both associated with the Island Geoform, showed individual recognition from one another during cognitive interpretation. Specific visual cues in the IKONOS-2 imagery allowed for identification of small, compact mud islands (i.e. Bay Key) versus those main islands formed from oolitic grainstone, fossil coral reef rock, and coquina shell bedrock (i.e. Karst Island). However, analysis of landforms associated with the Sediment Flat Geoform showed that Planar Bed Forms and Ripples were grossly recognizable within the IKONOS-2 imagery mainly because of the general tone and spectral reflectance of the unconsolidated materials. This then provided a basis for large-scale differentiation of seafloor sedimentary deposits versus rigid, hard structures, such as reef or hardbottom. Further analysis of unconsolidated materials in the IKONOS-2 imagery allowed for the detection of shelf sand waves, also known as Ridge Field Geoforms, and the discernment between those landforms oriented in the same direction (i.e. Discrete Ridges) and those forming a crisscrossing pattern of ridge and swale topography along the seafloor (i.e. Complex Ridges). This is because the spectral reflectance of light and the contrasting pattern of shadows along the benthic plane allow for interpretation of Discrete Ridges, which are aligned along one axis, versus Complex Ridges, which form along more than one axis to create a compound field of ridges (Figure 16).
Using GeoEye IKONOS-2 satellite imagery and the Geospatially Integrated Seafloor Classification Scheme (G-ISCS) method, cognitive interpretation of continental shelf geoforms and associated landforms was conducted over a coastal segment off SE Florida. Classification of benthic components along continental shelves is predicated upon the interpretation of these standard geomorphological structural frameworks known as geoforms and landforms. The identification of such spatially distributed biophysical features potentially allows coastal researchers and resource managers to bridge the classification of larger-scale physiographic realms and morphodynamic process zones (Makowski, Finkl, and Vollmer, 2016) with more temporally influenced characteristics, such as unconsolidated sediment accumulations (e.g., sand, mud), sessile biological assemblages (e.g., scleractinian coral growth), and flora proliferations (e.g., seagrass and macroalgae cover). Hierarchical approaches to continental shelf classification prove to be most effective in locations where optimal water clarity is present, for example, along the coast of SE Florida. In addition to exhibiting favorable visual properties of the water column, the continental shelf off SE Florida contains a diverse mix of coastal environments (e.g., highly urbanized Miami, Florida; Florida Keys National Marine Sanctuary (FKNMS); Everglades National Park; Biscayne Bay Aquatic Preserve; Biscayne Bay National Park; John Pennekamp Coral Reef State Park) in which numerous geoforms and landforms could be identified and delineated along the seafloor.
Previous studies have attempted to interpret and classify benthic geoform and landform attributes in marine environments using various remotely sensed platforms (e.g., Ansari et al., 2014; Costello, 2009; Finkl and Andrews, 2008; Finkl, Benedet, and Andrews, 2005a,b; Finkl and DaPrato, 1993; Finkl and Vollmer, 2011; Finkl and Warner, 2005; Greene et al., 1999; Heap and Harris, 2008; Kouchi and Yamazaki, 2007; Lidz et al., 2006; Madden et al., 2008; Steimle and Finkl, 2011; Valentine, Cochrane, and Scanlon, 2003; Wedding and Friedlander, 2008). For example, Finkl and Andrews (2008) and Finkl, Benedet, and Andrews (2005a,b) interpreted discrete geomorphological features within a 600 km2 area of continental shelf along SE Florida using laser airborne depth sounder (LADS) imagery. They were able to successfully show geospatial relationships that included barrier coral reefs, nearshore bedrock, and morphosedimentary features by visually mapping the airborne laser bathymetry images as continuous sequences of geoform and landform signatures. By doing so, the geomorpholoical features of a continental shelf region were interpreted and classified in a way never before attempted with LADS imagery.
Other investigations involving the assessment of coastal marine environments have exclusively used IKONOS satellite imagery as the visual basis for their results (e.g., Andréfouët et al., 2003; Dial et al., 2003; Finkl, Makowski, and Vollmer, 2014; Finkl and Vollmer, 2011; Hochberg, Andréfouët, and Tyler, 2003; Klemas, 2011; Maeder et al., 2002; Makowski, 2014; Makowski, Finkl, and Vollmer, 2015, 2016; Mumby and Edwards, 2002; Palandro et al., 2003; Steimle and Finkl, 2011). Andréfouët et al. (2003) and Mumby and Edwards (2002) were among the first to examine the benefits of applying IKONOS satellite imagery for mapping shallow-water nearshore environments and concluded that such imagery could be used to accurately map at geomorphological spatial scales where coral, algal, and seagrass habitats exist. These findings helped to justify the use of IKONOS satellite images as a means to properly interpret and classify geomorphological features in marine environments. Finkl and Vollmer (2011) were able to apply that principle by incorporating multiple IKONOS satellite imagery scenes of the southern Key West National Wildlife Refuge in Florida, U.S.A., to identify over 90 mapping units that were defined in terms of geomorphologic base, geoform and landform zones (e.g., reef flats, forereef, patch reef, lagoon), biological communities (e.g., seagrass beds, macroalgae coverage, coral overgrowth), and the percentage of biological cover. Similarly, Maeder et al. (2002) used the blue, green, and red spectral bands from IKONOS images in one fixed location off of Roatan Island, Honduras, Central America, to map the biophysical features of a nearshore coral reef. Both of these studies proved that the texture, pattern, and structure of typical marine environments found in clear waters could be interpreted and mapped using high-resolution IKONOS imagery. Furthermore, IKONOS-2 satellite imagery has been considered among the most effective visual mediums when delineating the benthic seascape into major geoforms and smaller, individual landform units (Makowski, Finkl, and Vollmer, 2015, 2016; Mumby and Edwards, 2002).
For this study, classification maps were created in ArcMap to visually represent the spatial distribution of geoform and landform features along the continental shelf by superimposing color coded units over interpreted IKONOS-2 satellite images. As shown in this paper's analysis, IKONOS-2 image scenes provided optimal recognition qualities (i.e. color, tone, texture, pattern, saturation, and relative spectral reflectance) that permitted suitable cognitive interpretation of the marine benthos, especially when attempting to interpret those environments offshore in deeper water. Typically, other lower-resolution satellite imagery (e.g., LANDSAT Thematic Mapper) become less effective in depths exceeding 15–18 m because light penetration (i.e. attenuation of the reflected spectral signal) and underwater clarity (due to increased turbidity) becomes restricted throughout the water column (Makowski, 2014; Makowski, Finkl, and Vollmer, 2016; Mumby and Edwards, 2002). Conversely, the IKONOS-2 images visually penetrate in deeper waters offshore to interpret major geomorphologcial signatures of submerged environments without serious degradation of tone, texture, and pattern signatures. A prime example of such optimal visual properties was seen from the results of this study when the Coral Reef Geoform was identified and further delineated into the individual Barrier Reef, Patch Reef, Aggregated Reef, Coral Apron, and Reef Gap landforms (Figure 5). Overall, this study showed that IKONOS-2 images were suitable for the cognitive interpretation for geoforms and landforms along the continental shelf, as shown by the net result of 3888 vector polygons being drawn to demarcate the geomorphological framework of the region (Figure 2). Even though at deeper depths (i.e. greater than 50 m) the bottom reflectance signal is completely absorbed by the water column, IKONOS-2 images still provided enough detail for coastal (terrestrial) and benthic marine environment recognition along the continental shelf to complete the perceived geoform and landform classification over the study area.
As a census of geoform and landform attributes is interpreted, mapped, and ultimately compiled into a central database, it is postulated that a true hierarchical classification of a specific region begins to form when these attributes link larger-scale units (i.e. physiographic realms and morphodynamic zones) with smaller, more temporally influenced benthic features (e.g., dominant sediment, dominant biological cover). For this particular study area, Makowski, Finkl, and Vollmer (2016) have already reported an interpretation of physiographic realms and morphodynamic zones. Therefore, when the geoform and landform data from this study are added to those previous datasets, a true representation of the geomorphological framework along the continental shelf begins to take form. Future studies may contribute sediment and biological classification units, along with ranges in coverage, which would then provide an inclusive geo-referenced spatial database for the continental shelf region.
This study verified that cognitive interpretation, classification, and mapping of benthic geoforms and associated landforms can be accomplished along the SE Florida continental shelf using GeoEye IKONOS-2 satellite imagery. Through geomorphological attribute analysis, the spatial distribution, quantity of occurrence, and approximate area of each interpreted geoform and landform feature was determined. Using the Geospatially Integrated Seafloor Classification Scheme (G-ISCS) method in tandem with IKONOS-2 imagery and ESRI ArcGIS ArcMap software, classification maps helped conclude where the distribution of specific geoform and landform biogeomorphological signatures occurred along the seafloor throughout the continental shelf region.
Special appreciation goes to the Coastal Education and Research Foundation, Inc. (CERF) and to the GeoEye Foundation®, both of which provided logistical and financial support for this research. Individual acknowledgments are given to those external peer reviewers for their suggestions of improvement.