**1. Introduction**

Salt marshes are defined by daily tidal inundation and dominated by halophytic vegetation. These ecosystems are the boundary between terrestrial and nearshore aquatic environments their unique location on the landscape and vegetation composition provides ecosystem services such as denitrification, filtration of pollutants, nursey habitat, coastal resilience, and carbon storage and sequestration [1,2]. Historically, salt marshes have displayed high rates of loss due to land reclamation and disturbances such as mosquito ditching [3,4]. Currently, salt marshes along the mid-Atlantic coastal region of the United States are at risk of loss due to sea level rise (SLR), eutrophication, nutrient enrichment, sediment availability, tidal range, and herbivory and human disturbances [5–12]. Recent studies have demonstrated regional and site-specific salt marsh changes including degradation in the Mid-Atlantic [13], proliferation of salt marsh pools in Maryland [14], loss coupled with increased Phragmites on Long Island [15], and loss driven by SLR in New England [16]. However, salt marsh change is a complex combination of persistence, migration, and loss. In the Chesapeake Bay, conversion of uplands to wetlands has mitigated past salt marsh losses [17]. Future salt marsh change is uncertain with some models predicting that salt marsh migration in response to SLR will result in increased salt marsh area [18]. Salt marsh monitoring is necessary for improved understanding of how these ecosystems are changing which in turn can inform their management.

Salt marshes are changing in a variety of ways necessitating a shared nomenclature to discuss these changes. In New England, four types of salt marsh losses have been identified channel widening, interior die-o ff, shoreline erosion, and loss in the bay head region [16]. These distinctions are dependent on the location of the change. In remote sensing and this study, two major categories were evident change along the edge and change in the interior of the salt marsh. Two types of interior salt marsh loss have been identified: sudden vegetation dieback and drowning. Sudden vegetation dieback in salt marshes is a rapid onset event that persists for a brief period (≈2 years) [19]. These die-o ffs are predominately located in the mid-marsh and have been documented across the eastern Atlantic coast [20]. In contrast, interior drowning driven by sea level rise is outside the scope of these rapid die-o ff events and represents a fundamental shift in the ecosystem [20]. Monitoring salt marsh change is further complicated by drowning appearing similar to ponding in microtidal salt marshes [21] and pools changing shape dynamically, draining, and revegetating [22]. Monitoring and di fferentiating between sudden vegetation dieback, drowning, and ponding necessitates high spatial resolution monitoring to assess expansion and recovery dynamics of interior salt marsh areas.

In this study, we focused on estuarine persistent emergen<sup>t</sup> vegetation, i.e. intertidal areas with perennial salt marsh vegetation [23]. We are interested in identifying pools and pannes and how they changed. Pannes are recessed areas of the salt marsh which drain at mean lowest low water (MLLW) and can be vegetated or nonvegetated, in this study we are referring to nonvegetated pannes unless otherwise stated. In comparison pools are those areas of persistent water. Ponding, pools, and pannes are natural elements of the salt marsh landscape, however long-term and widespread loss of vegetation is not. An in situ study of Plum Island estuary, Massachusetts, found pools were in equilibrium with vegetation regrowth occurring after a few years or at most a decade [24]. However, vegetation regrowth is not guaranteed with site-specific characteristics, such as low sediment input, small tidal range, and high regional SLR, contributing to lack of vegetation growth in pannes and slow filling of pools [25]. Identifying these changes in situ is time consuming and non-tenable for large geographic areas. This study presents a method for using satellite and aerial imagery to di fferentiate between drowning and ponding by monitoring for decades, a temporal period beyond the expected recovery time.

This study is focused on Fire Island, NY, a barrier island, salt marshes on barrier islands have limited land for salt marsh migration. As a result, salt marsh persistence in place is of particular concern, and is a key component of understanding the long-term stability of barrier island systems. Storm events are critical for shaping the geomorphology of barrier islands, e.g., Hurricane Sandy caused overwash on 41 percent of Fire Island depositing an estimated 508,354 m<sup>3</sup> of sediment [26] and breached the barrier island. Both overwash and inlet creation are essential sources of sediment accretion in bayside salt marsh environments [27,28]. Even thick overwash deposits (>50 cm) result in quick recovery of the salt marsh vegetation [29]. Mapping salt marsh change following a storm event and breach is critical for our understanding of salt marsh persistence on barrier islands. A previous breach on Fire Island was documented with radiometric dating of salt marsh cores, finding a connection between back-barrier salt marsh formation and inlet creation and the resulting processes [30]. Recent research on anthropogenic alterations of an inlet, i.e., jetty creation, demonstrated changes in local mean sea level (LMSL) and tidal range, which resulted in stabilization of salt marsh directly surrounding the inlet [31]. Similar changes to LMSL and tidal range could be a ffecting the salt marshes of Fire Island. The decision to allow the Fire Island breach to evolve naturally facilitates monitoring to understand the e ffect of this process on the surrounding salt marshes.

Remote sensing monitoring can quantify essential attributes of the salt marsh landscape. Remote sensing has been used to understand di fferences in salt marsh pond density and the total surface area between ditched and unditched marshes [32]. Imagery and Light Detection and Ranging (LiDAR) have been used to quantify pools and pannes and their landscape location [33]. Salt marsh change analysis has been used to identify complex patterns of spatial and temporal variation [14], dramatic conversions from high to low marsh [34], impacts of salt marsh restoration and Hurricane Sandy [35], and identify the e ffect of SLR on salt marsh communities [36]. The combination of very high resolution (VHR) satellite imagery and aerial imagery provided the necessary spatial and temporal resolution to understand the dynamic coastal environment.

This study mapped salt marsh habitat on Fire Island National Seashore (FIIS) utilizing object-based image analysis (OBIA) with VHR satellite imagery. Multitemporal OBIA was utilized to analyze the development and geospatial dynamics of pool and pannes with high spatial resolution aerial imagery from 1994, 2011, 2013, 2015, and 2017. Change to pannes and pools and edge erosion were analyzed to determine if they were significant components of salt marsh loss at FIIS. The use of remote sensing to determine how protected areas are changing with high accuracy is vital to better manage these areas. The objectives of this study are to (1) classify FIIS salt marsh with OBIA and VHR remote sensing imagery data; (2) determine change rates for interior salt marsh pannes and pools and edge erosion from 1994-2017; (3) determine the relationship between hydrological connectivity and panne expansion; and (4) determine if edge erosion or panne/pool expansion increased surrounding the Hurricane Sandy breach between 2011–2017.

#### **2. Materials and Methods**
