The Use of Aerial RGB Imagery and LIDAR in Comparing Ecological Habitats and Geomorphic Features on a Natural versus Man-Made Barrier Island

The Mississippi (MS) barrier island chain along the northern Gulf of Mexico coastline is subject to rapid changes in habitat, geomorphology and elevation by natural and anthropogenic disturbances. The purpose of this study was to compare habitat type coverage with respective elevation, geomorphic features and short-term change between the naturally-formed East Ship Island and the man-made Sand Island. Ground surveys, multi-year remotely-sensed data, habitat classifications and digital elevation models were used to quantify short-term habitat and geomorphic change, as well as to examine the relationships between habitat types and micro-elevation. Habitat types and species composition were the same on both islands with the exception of the algal flat existing on the lower elevated spits of East Ship. Both islands displayed common patterns of vegetation succession and ranges of existence in elevation. Additionally, both islands showed similar geomorphic features, such as fore and back dunes and ponds. Storm impacts had the most profound effects on vegetation and geomorphic features throughout the study period. Although vastly different in age, these two islands show remarkable commonalities among the traits investigated. In comparison to East Ship, Sand Island exhibits key characteristics of a natural barrier island in terms of its vegetated habitats, geomorphic features and response to storm impacts, although it was established anthropogenically only decades ago.


Introduction
Barrier islands front approximately 10% of the world's open-ocean shorelines, with 35% of those barrier islands in North America [1][2][3][4]. They facilitate resiliency in coastal ecosystems by buffering mainland shores from storms and daily wind and wave energies. However, barrier island ecosystems are susceptible to saltwater intrusion with continuing sea-level rise due to their low elevation, presumably making them reliable markers of global climate change [3,5]. It has been generally accepted that the presence of a freshwater lens indicates the overall sustainability of a barrier island and drives habitat development [6,7]. With access to fresh water, bare sand is colonized by pioneering species, leading to foredune growth and gradual transitions to shrubland and maritime forest [3,8,9]. On barrier islands in the northern Gulf of Mexico, the extent of total land area and drier habitat types, which form at higher elevations, such as shrubland and woodland, is declining relative to lower-elevation, wetter habitat types, such as marsh, at a time scale of a few decades [3,10,11]. This appears to be a combined effect of relative sea level rise and navigational shipping channels impeding natural littoral drift, ultimately affecting sediment replenishment along these islands [11]. In order to better understand the natural and anthropogenic effects on succession pathways, short-term succession in these environments needs to be better understood. A better understanding of sub-decadal vegetation succession on these islands would yield an improved perspective on these recently-documented, decadal-scale trends.
The Mississippi (MS) barrier island chain ( Figure 1) is comprised of several naturally-formed islands (Cat, East and West Ship, Horn and Petit Bois) plus one island (Sand) that was created over recent decades through artificial deposition of dredge material from navigation channels. The natural barriers are thought to have formed from the slowing of sea-level rise during the Holocene Transgression [12]. Their habitats and especially geomorphic features have been studied since the early and mid-20th century [3,5,[8][9][10][11][12][13]. However, to our knowledge, no prior research on the MS islands has focused on short-term habitat and geomorphic change on man-made versus naturally-formed islands. Of particular interest was a comparison between Sand and East Ship because their land areas are similar and least among the Mississippi barrier islands. to be a combined effect of relative sea level rise and navigational shipping channels impeding natural littoral drift, ultimately affecting sediment replenishment along these islands [11]. In order to better understand the natural and anthropogenic effects on succession pathways, short-term succession in these environments needs to be better understood. A better understanding of sub-decadal vegetation succession on these islands would yield an improved perspective on these recently-documented, decadal-scale trends. The Mississippi (MS) barrier island chain ( Figure 1) is comprised of several naturally-formed islands (Cat, East and West Ship, Horn and Petit Bois) plus one island (Sand) that was created over recent decades through artificial deposition of dredge material from navigation channels. The natural barriers are thought to have formed from the slowing of sea-level rise during the Holocene Transgression [12]. Their habitats and especially geomorphic features have been studied since the early and mid-20th century [3,5,[8][9][10][11][12][13]. However, to our knowledge, no prior research on the MS islands has focused on short-term habitat and geomorphic change on man-made versus naturally-formed islands. Of particular interest was a comparison between Sand and East Ship because their land areas are similar and least among the Mississippi barrier islands. The purpose of the present study was to: (1) compare habitat type composition, geomorphic features and elevation between East Ship and Sand Island; and (2) evaluate the extent to which habitat types changed on the two islands over a time period of less than one decade. Remotely-sensed and ground data were used to determine habitat type coverage, elevation and land area of each island and their extent of change from summer 2007 to summer 2012. The purpose of the present study was to: (1) compare habitat type composition, geomorphic features and elevation between East Ship and Sand Island; and (2) evaluate the extent to which habitat types changed on the two islands over a time period of less than one decade. Remotely-sensed and ground data were used to determine habitat type coverage, elevation and land area of each island and their extent of change from summer 2007 to summer 2012.

Study Area
East Ship Island, initially formed ca. 4.5 ka 14 C [14], is located approximately 15 km south of the Mississippi mainland shore (30˝14 1 13"N,´88˝53 1 13"W). The island is composed mostly of quartz sand with finer silts and clays on its lagoonal side. As of 2012, it was approximately 3.5 km in length (E-W) and 400 m wide (N-S). The island is characterized by two low elevation (<0.5 m) east and west spits extending from a higher elevation (>0.5 m) central core. The core, which has remained relatively intact through multiple historical tropical storms, has been the most stable portion of the island [15]. The east and west spits support mainly beach grasses and algal flats [15]. The core includes a tidally-influenced pond, with muddy peat soils and relict dune ridges vegetated with beach morning glory (Ipomoea imperati (Vahl) Griseb.), salt meadow cordgrass (Spartina patens (Aiton) Muhl.), marsh elder (Baccharis halimifolia L.), saw palmetto (Serenoa repens (W. Bartram) Small) and sand live oak (Quercus geminata Small). East Ship is currently managed by The National Park Service (NPS), Gulf Islands National Seashore (GUIS).
Disposal Area 10, known locally as Sand Island, is located approximately 11 km off the Mississippi mainland shore (30˝13 1 26"N,´88˝31 1 19"W) on the western edge of the Pascagoula ship channel and is approximately 1.7 km in length (E-W) and 580 m in width (N-S). The island began formation in the 1960s from an underwater shoal created through the deposition of dredge material from the neighboring Pascagoula ship channel [12]. The island emerged around 1971 through channel maintenance and the natural reworking of the ebb-tidal sediment, combined with deposition from littoral drift [11,16,17]. Notably, among the islands in the chain, Sand has the highest elevation of approximately 9 m North American Vertical Datum 1988 (NAVD88). It is composed mostly of quartz sand mixed with calcareous shell fragments. The island is geomorphically characterized by highly-elevated east-west recurved and truncated beach ridges [11,16]. As with East Ship, Sand has a stable core supporting a freshwater pond and vegetated habitats. Vegetation includes wax myrtle (Morella cerifera Small), sea oat (Uniola paniculata L.), gulf bluestem (Schizachyrium maritimum (Chapm.) Nash) and salt meadow cordgrass. The region has a humid subtropical climate with mean air temperatures of 12˝C and 27˝C in the winter and summer, respectively. Peak rainfall occurs in July through September with mean annual amounts of approximately 140-160 cm [18].
From 2007-2012, two tropical events affected the study area with sustained winds >20 m/s [19,20]. Hurricanes Gustav (1 September 2008) and Isaac (29 August 2012) both passed to the west of the MS islands on similar northwesterly tracks. According to Beven and Kimberlain [20], storm tides for Hurricane Gustav were 2.03 m at Point Cadet in Biloxi, Mississippi (30˝23 1 23"N,´88˝51 1 28"W) and 2 m at the Port of Pascagoula (30˝20 1 43"N,´88˝34 1 02"W). Wind gusts of approximately 30 m/s were also reported in these areas during Gustav. In 2012, Hurricane Isaac produced storm tides of 2 m at Point Cadet and 1.81 m at the Port of Pascagoula [20]. Wind speeds were slightly higher than Gustav's at these sites, with gusts >32 m/s. Hurricane Isaac was very similar to Gustav with respect to wind, storm tides, trajectory and total projected area; however, Isaac's forward speed was hindered due to a mid-level blocking ridge northwest of the storm [19,20].

Remote Sensing and Ground Data
Remotely-sensed data products for this investigation included: NAIP imagery was used because it is collected every two years [21]. Starting in 2012 for the state of Mississippi, NAIP imagery consisted of three visible (blue: 400-580 nm, green: 500-650 nm and red: 590-675 nm) and one near infrared (675-850 nm) band. However, prior to 2012, NAIP imagery acquired in Mississippi did not collect near-infrared. Thus, this study employed the blue, green and red bands for all analyses to ensure image comparability between years. Image data were comprised of a Brightness Value (BV) (0-255) referenced to NAD83 and projected in Universal Transverse Mercator (UTM) Zone 16 North [21]. All NAIP imagery was post-processed and quality checked by the data collector. Post-processing included removal of bidirectional reflectance effects, vignetting and other artifacts [21].
The 2010 acquisition was chosen as the base image for both islands because ground data were collected in that year. BV histograms for each band (blue, green and red) and year were stretched to match the 2010 BV histograms to correct for BV intensity differences and to provide a baseline for calibration to surface reflectance. Image BVs then were scaled to surface reflectance using the empirical line method (ENVI v4.3, Exelis Visual Information Solutions, Boulder, CO, USA). Assuming that island bare sand surface reflectance remains consistent year after year, Regions-of-Interest (ROI) consisting of 3000 bare-sand pixels were selected in each image to provide sample data for the empirical line method. In situ data used for this calibration were the spectral reflectance of beach sand, recorded using a portable spectroradiometer (ASD-FS, Boulder, CO, USA) on neighboring Horn Island. An image-to-image geometric registration was performed for all East Ship and Sand Island imagery. Images for 2007, 2009 and 2012 were georectified to 2010 with no less than 10 Ground Control Points (GCP) for each image with a total RMS error < 0.5, as per the registration methods outlined by Jensen [22]. Due to the lack of anthropogenic features and hard targets on both islands, GCPs consisted of individual patches of vegetation that remained stable and identifiable in each image.
In 2010-2011, the Gulf Coast Geospatial Center conducted a field vegetation survey for the MS-AL barrier island chain [23]. Ground surveyed data were taken using randomly-selected points at a mean density of one point per 6 ha of island area. The method was based on Lucas and Carter's [24] previous study, which utilized a similar sampling density for line transects. Sample points were positioned randomly (ENVI v4.3) from SPOT-5 (10 m) Ground Sampling Distance (GSD) multispectral coverage (April-July 2010 image acquisitions, North American Data Purchase, USGS EROS Data Center, Sioux Falls, SD, USA). Geographic centers were then visited using a Trimble Geoexplorer 6000. If a sample point fell in a body of water, it was moved to the nearest above-water area. A range pole was then planted in the ground to mark the sample point location, and notes were made concerning primary and secondary vegetation species in the immediate area. Additionally, cardinal-directional photographs were taken at each point with the range pole being in the center of the field-of-view using a Nikon model D-60 SLR with a Nikon 18-55 mm f/3.5-5.6G VR AF-S DX lens, set at 18 mm to achieve the widest field-of-view [23]. Photographs were taken clockwise at magnetic compass headings of 0˝, 90˝, 180˝and 270˝. Habitat classifications for each sampled location were then made through the visual inspection of photographs and documented primary and secondary species. The classification scheme was based on those used in the Mississippi Natural Heritage Program 2006 [25].
A geodatabase was constructed (ArcGIS v10.1, ESRI, Redlands, CA, USA) to house all sample point information. All point data share the same datum and projection as the NAIP imagery. Attribute fields for point data included site identification number, site visitation date, time, horizontal precision, geographic coordinates of the site, primary species, secondary species, general notes pertaining to the site and habitat classification type. Sample site photographs were attached as binary large objects in the geodatabase for data management and ease of use.

Supervised Classification of Habitats
The Maximum Likelihood (ML) supervised classifier was used because of its wide acceptance, and it has provided the greatest accuracy in comparison to other classifications used on the MS islands [3,22,24,26]. Classifications were created for East Ship and Sand Island initially for 2010 because the ground survey took place that year. The habitat classes used for East Ship Island were: algal flat (Salicornia sp. and matted organics), bare sand, beach dune herbland (sea oat, gulf bluestem, and beach morning glory), marsh shrubland (smooth cordgrass, wax myrtle, marsh elder and sand live oak) and water. The same classes were used for Sand Island, except it did not include algal flat. Sampled pixel data were generated from the 2010 ground survey GPS locations with habitat type information and photographs attached. Additionally, training data pixels (Table 1) were selected in 1-m ROI increments based on estimations from field photographs. To ensure classification of only island land areas, masks were created (ArcGIS v10.1) by digitizing shorelines (1:600 scale) for each island image. The idealized shoreline is very dynamic, requiring the use of shoreline indicators to determine the boundaries [27]. Thus, shorelines were visually interpreted by the discernable wet-to-dry sand interface. Any pixels outside of the masked shoreline for classifications were given the designation sub-aqueous. This differed from the water class in that the water habitat type (non-vegetated pond) fell within the shoreline boundaries. Sample pixel data were selected in 1-m increments for 2007, 2009 and 2012 images, based on mean surface percent reflectance threshold values from each habitat in the 2010 ML. Post-classification reports were generated, consisting of a confusion matrix to determine the overall accuracies and errors of commission and omission [22]. Commission errors represented the percentage of pixels that are assigned to a class, while belonging to another class. These errors apply to the training data used in the ML classification. Omission errors were represented by the percentage of pixels that should have been included in a class, but were not [28].

Change Detection
A post-classification change detection compared initial and final state classification maps on a pixel-by-pixel basis using Geographic Information System (GIS)-based raster algorithms (ENVI v4.8) to produce a change detection matrix [22,29]. Classifications assign pixels a digital number for their respective habitats. As classifications are stacked, the algorithm assigns pixels a new value based on change or no change. Special attention was given to ensure proper image registration and the overall accuracies of the classifications used in a post-classification change detection, as errors in both maps would affect the final output [22].

Elevation Analysis
LIDAR data were post-processed, and Digital Terrain Models (DTM) of East Ship and Sand Island for 2012 were constructed (ArcGIS v10.1). Data were delivered by the vendor in Laser file exchange format (LAS) v1.2, following criteria set by the American Society for Photogrammetry and Remote Sensing [30]. LAS files were selected for East Ship and Sand by ensuring full island coverage based on a polygon tile scheme. Each island was represented by six LAS files. LAStools (University of North Carolina, Chapel Hill, NC, USA) software, was used to initially filter and remove noise by extracting values above 50 m in elevation. Bare earth point clouds were made by filtering all points not classified as 2 (bare earth). Water surface elevations were determined from partial reflection. Elevation rasters were then produced using the mean elevation of all bare earth points within a 1-m GSD; this generally consisted of 0-3 returns. A neighborhood raster calculator function then was used to interpolate elevation rasters to produce a smooth surface without data voids based on mean values within a 5ˆ5 moving window. This method was chosen to mimic the collection of the NAIP imagery. A total of three iterations was performed to fill data voids in DTM rasters. Comparisons of field real-time network GPS observations with the derived elevation rasters confirmed the reported accuracy of the original LIDAR dataset [18]. Elevation rasters were then clipped using digitized shorelines from the 2012 ML mask. DTM rasters were referenced horizontally to NAD83 and vertically to NAVD88 and registered to the 2012 ML classifications. Habitat type classifications for 2012 were sampled on a pixel-by-pixel basis against the 2012 DTMs only.

Results
Due to spectral similarities between algal flat and beach dune herbland on East Ship, separate classifications were made for the spits and core sections of the island. These were mosaicked to produce a single seamless classification map for each year (Figure 2). Overall accuracies for individual classifications for East Ship and Sand ( Figure 3) ranged from 82%-95% and 90%-96% respectively. Errors of commission and omission were generally limited to the misclassification of bare sand classed as beach dune herbland or algal flat classed as bare sand. Post-classification reports of total land area and habitat type coverage for East Ship Island indicated overall land area growth with an exception from 2009-2010, with fluctuations in habitat type coverage for all years ( Table 2). Total land area on Sand Island also increased, except from 2010-2012, with fluctuations in individual habitat type coverage in all years (Table 2).
Change detections for East Ship and Sand Island were expressed as the total land area (m 2 ) that changed or remained stable within a given time period (Tables 3 and 4        The 2012 DTMs for East Ship and Sand indicated a maximum elevation of 4.50 m and 9.34 m and a minimum elevation of´0.50 m and´0.28 m NAVD88, respectively ( Table 5). The presence of several overwash lobes, overwash channels, elongated fore and back dunes and a pond could be noted on East Ship Island. Whereas on Sand, only the presence of elongated fore and back dunes and a pond could be delineated (Figures 4 and 5). Minimum elevations for common habitat types were similar for both islands with between island differences of 22 cm for bare sand, 6 cm for beach dune herbland, 6 cm for marsh shrubland and 31 cm for water. Maximum elevations of common habitats were considerably higher on Sand Island, with the exception of water. Although mean elevations for common habitats were slightly higher on Sand Island than East Ship, the differences between them were approximately 1 m or less.

Discussion
On islands, bare sand provides the parent material needed for pioneering grass species that eventually are replaced by shrub and woodland habitats via succession [31]. However, bare sand is very unstable and may be found among a variety of habitat types. Substrate elevation and elevational differences, which might be considered trivial in other ecosystems, largely control vegetation composition on barrier islands [3]. This study compared two islands within the same barrier chain,

Discussion
On islands, bare sand provides the parent material needed for pioneering grass species that eventually are replaced by shrub and woodland habitats via succession [31]. However, bare sand is very unstable and may be found among a variety of habitat types. Substrate elevation and elevational differences, which might be considered trivial in other ecosystems, largely control vegetation composition on barrier islands [3]. This study compared two islands within the same barrier chain, one man-made and one occurring naturally, and examined the changes in geomorphic and biological features over a seven-year time period. Additionally, ranges in the elevation of the corresponding habitat types were compared using the 2012 habitat classification and a 2012 Digital Terrain Model (DTM) constructed from LIDAR. After investigating the aforementioned subject matter, both islands were found to exhibit remarkably similar geomorphic characteristics and vegetation composition. With the exception of algal flat found on East Ship Island, both islands contained the same habitat types and were found to exist at comparable ranges in elevation ( Figure 6). For example, on both islands, plant species, such as sea oat, beach morning glory and gulf bluestem, were indicative of the beach dune herbland habitat type, and salt meadow, smooth cordgrass, marsh elder and wax myrtle were symbolic of the marsh shrubland habitat type. Additionally, changes in vegetation coverage and geomorphology were similar over the period of investigation. For example, bare sand transitioned mainly to and from algal flat and beach dune herbland through overwash and aeolian burial of vegetation. Identifying an ecosystems' climax species allows the definition of successional stage [32]. Sand Island, because it is much younger than East Ship Island, lacks woody vegetation, such as slash pine (Pinus elliottii var. elliottii Engelm) and sand live oak that sparsely inhabit East Ship Island. Although these two species constitute the climax community vegetation on East Ship Island, their coverage is not sufficient in differentiating them from the shrub species that make up the marsh shrubland habitat type. The climax community species found on Sand Island are wax myrtle (Morella cerifera) and smaller shrubs, such as marsh elder (Iva frutescens). Absent on Sand Island, the algal flat habitat on East Ship was expanded by Hurricane Gustav. Overwash from Gustav's shore normal tidal surge in 2008 aided in the creation of a large intertidal zone, extending from the island's core down the entire length of the northeast spit (Figure 7). Due to the low mean elevation of beach dune herbland, major transitions occurred in the habitat, with as much as 90% loss in its total land area from 2007-2009. Approximately half of the beach dune herbland habitat type on Sand Island transitioned to bare sand following Gustav, due to run-up overwash on fore and back dunes. The storm also caused a foredune blowout responsible for the partial filling of East Ship's pond. While losses did occur in the marsh shrubland habitat type on East Ship Island, they were not as profound as they were in other habitat types. This was likely due to Absent on Sand Island, the algal flat habitat on East Ship was expanded by Hurricane Gustav. Overwash from Gustav's shore normal tidal surge in 2008 aided in the creation of a large intertidal zone, extending from the island's core down the entire length of the northeast spit (Figure 7). Due to the low mean elevation of beach dune herbland, major transitions occurred in the habitat, with as much as 90% loss in its total land area from 2007-2009. Approximately half of the beach dune herbland habitat type on Sand Island transitioned to bare sand following Gustav, due to run-up overwash on fore and back dunes. The storm also caused a foredune blowout responsible for the partial filling of East Ship's pond. While losses did occur in the marsh shrubland habitat type on East Ship Island, they were not as profound as they were in other habitat types. This was likely due to protection from its fore and back dunes. The impact of Gustav on East Ship added total land area, while reducing vegetated habitat coverage. The effects on Sand Island were limited mainly to the lower-elevation beach dune herbland (Figure 3a,b). Some spit reworking and land growth on the island's southeast flank were visually identifiable in the NAIP imagery. Major changes on both islands from 2010-2012 are attributed to the impact of Hurricane Isaac. The slower forward speed of Isaac led to prolonged periods of storm tide run-up overwash and high southeast wind and wave fetch, shaping both islands geomorphically and ecologically [34]. Approximately 23 ha of bare sand were added to East Ship's southeast spit, completely reshaping the island and skipping the algal flat successional phase (Figure 2c,d). Additionally, the majority of algal flat on the northeast spit transitioned to bare sand through sand overwash and burial (Figure 2c,d).  In evaluating the similarities and differences between habitats and their respective elevations on both islands, it was noted that the range of water habitat on East Ship is larger and exists at lower elevations than water on Sand Island. This is presumably due to East Ship's pond being periodically tidally influenced, whereas Sand Island's pond is not. Although algal flat exists on East Ship and not on Sand Island, the habitat spans the entire range of elevation that the water habitat occupies on both islands. This is a function of a low elevation gradient and frequent tidal inundation found on the low elevated spits of East Ship Island. Sand Island lacks these particular geographic features. Interestingly, although Sand Island has a higher overall elevation by approximately 4.8 m, the majority of its climax community species, which fall under the marsh shrubland habitat, exists at a lower range of elevation than the comparable community and habitat on East Ship. Currently, we attribute the higher range of elevation in marsh shrubland on East Ship to the differences between elevation gradients between the two islands. However, this subject warrants its own study. Finally, the ranges and overall elevation of bare sand and beach dune herbland were greater on Sand Island ( Figure 6). The high between-island variability in elevations for these habitats is a function of Sand Island being created through artificial deposition of dredge material. Major changes on both islands from 2010-2012 are attributed to the impact of Hurricane Isaac. The slower forward speed of Isaac led to prolonged periods of storm tide run-up overwash and high southeast wind and wave fetch, shaping both islands geomorphically and ecologically [34]. Approximately 23 ha of bare sand were added to East Ship's southeast spit, completely reshaping the island and skipping the algal flat successional phase (Figure 2c,d). Additionally, the majority of algal flat on the northeast spit transitioned to bare sand through sand overwash and burial (Figure 2c,d). Presumably, the higher mean elevation of beach dune herbland on Sand allowed for it to remain relatively stable during Isaac. This was not the case on East Ship, as nearly all beach dune herbland reestablishment post-Gustav was eliminated by Isaac (Figure 2c,d). It is postulated that wind, wave and surge forces were not shore-normal on Sand as they were for East Ship during Isaac, resulting in approximately 11 ha of land loss on the island's southwest flank (Figure 8). Average water depths of 3 m flank Sand Island to the south and southwest, dropping steeply to about 5 m-6 m to the west and northwest of the island [35]. Sediment eroded from the island was most likely lost due to the sharp increase in depth. Depths between East and West Ship average 1.5 m and are gently sloping [36]. Further, sediment transport continues between West and East Ship Islands eroded tidal inlet via littoral drift. East Ship's land growth in 2012 could have been affected by the relatively shallow depths, allowing for sediment accretion [37].

Conclusions
In evaluating the similarities and differences between habitats and their respective elevations on both islands, it was noted that the range of water habitat on East Ship is larger and exists at lower elevations than water on Sand Island. This is presumably due to East Ship's pond being periodically tidally influenced, whereas Sand Island's pond is not. Although algal flat exists on East Ship and not on Sand Island, the habitat spans the entire range of elevation that the water habitat occupies on both islands. This is a function of a low elevation gradient and frequent tidal inundation found on the low elevated spits of East Ship Island. Sand Island lacks these particular geographic features. Interestingly, although Sand Island has a higher overall elevation by approximately 4.8 m, the majority of its climax community species, which fall under the marsh shrubland habitat, exists at a lower range of elevation than the comparable community and habitat on East Ship. Currently, we attribute the higher range of elevation in marsh shrubland on East Ship to the differences between elevation gradients between the two islands. However, this subject warrants its own study. Finally, the ranges and overall elevation of bare sand and beach dune herbland were greater on Sand Island ( Figure 6). The high between-island variability in elevations for these habitats is a function of Sand Island being created through artificial deposition of dredge material.

Conclusions
This study compared short-term (less than one decade) change in habitat type coverage with respective elevation, geomorphic features and total land area between naturally-formed East Ship Island and man-made Sand Island. Habitat types and species composition were the same on both islands with the exception of algal flat existing on the lower elevated spits of East Ship. DTMs showed very similar geomorphic characteristics on both islands, such as elongated fore and back dunes and ponds. However, relict dunes, which were indicative of sand live oak, were found only on East Ship. These biogeographic features were absent on the much younger Sand Island. While Sand Island has a higher overall mean elevation, common habitat types were found to inhabit like ranges of elevation, with between-island differences of about 1 m or less. Change detection analyses showed that these two islands are also succeeding similarly, with dominant shifts occurring to, and from, bare sand to beach dune herbland habitat types. This was evident after Hurricane Gustav, in which beach dune herbland was buried by sand and then quickly reestablished on both islands. Given the two storm impacts within the study period, the marsh shrubland habitat type was the most stable. On East Ship, Hurricanes Gustav and Isaac could be considered beneficial rather than detrimental because both storms added land area. Conversely, land area on Sand Island was lost likely the result of the interaction among island orientation, angle of wave approach and bathymetric slope during storm impacts. Previous long-term ecological and geomorphic investigations on the MS barrier islands have accredited change to subsidence and sea-level rise, sediment starvation, storm impacts and a warming climate. However, during this investigation, it was found that major vegetation and geomorphic change was attributed mostly to the impacts of Hurricanes Gustav and Isaac. While it is fairly common to attribute short-term change on barrier islands to tropical events, these two islands in particular responded similarly with respect to vegetation loss and reestablishment. Although they responded alike in terms of vegetation, this was not the case with geomorphic change. Interestingly, Sand Island behaved like a natural system when compared to East Ship Island. For example, both islands have a central core, which remained relatively stable during the time period studied. Additionally, the presence of Sand Island's stable fresh water pond coupled with freshwater-dependent vegetation further exhibits island stability, response and behavior in a natural system. In the brief time Sand Island has evolved, it appears to have responded to the natural biological and geomorphic drivers of change in much of the same ways as East Ship Island. Although vastly different in age, these two islands show remarkable commonalities among the traits investigated. In comparison to East Ship, Sand Island exhibits key characteristics of a natural barrier island in terms of its vegetated habitats, geomorphic features and response to storm impacts within the time period investigated. Although future studies are needed, this presents an interesting quandary that must be considered when determining dredge material borrow and placement sites. Sand Island has become a viable location for the growth and stability of flora and fauna mirroring the natural East Ship Island in many ways. In consideration of the future of the MS barrier island chain, Sand Island offers ecological and economic value to which barrier island protection and restoration may be compared.