**1. Introduction**

Coastal marine ecosystems are among the most diverse and productive in the world, and they provide critical habitats for a wide variety of plants, fish, shellfish, and other wildlife [1–3]. Coastal and near-shore marine ecosystems are facing unprecedented pressures from land use modification. Many studies have analysed change dynamics in wetland ecosystems due to the utilisation of remote sensing techniques [4–6] resulting from a combination of two factors: (1) greater open access to longer time series of image archives and their derived products and (2) more easily accessed tools for using remote sensing data and their products to monitor change from local to global scales. The Landsat satellite archive has provided new opportunities for assessing historical changes in landscapes [7], including coastal ecosystems.

Estuarine wetlands are located at the interface of land and sea and are essential support mechanisms in the marine and terrestrial systems. The inter-realm connectivity of coastal wetlands features strongly in integrated conservation planning approaches [8]. An improved understanding of land–sea connectivity dynamics is crucial to the health of coastal fisheries species' populations [9]. However, connected ecosystems are traditionally studied as separate entities, despite the potential for interactions between them to have consequences for their health and functioning [10]. Estuary-dependent fisheries species are important because they contribute 75% of the total value of Australia's commercial fisheries catches and 90% by numbers of Australia's recreational fish catch [11]. Coastal wetlands that support fisheries are a diverse assemblage of marshes, mangroves, forested wetlands, and estuaries. Mangroves are coastal forests with unique adaptations to saline conditions, and they form a characteristic vegetation zone along sheltered bays, tidal inlets and estuaries in the tropics and subtropics, globally [12,13]. These wetland types fulfil critical roles in ecosystem functions, and they provide many highly valued ecosystem services: raw materials and food, coastal protection, erosion control, water purification, the maintenance of fisheries, carbon sequestration, tourism, recreation, education, and research [14]. Now evident is a drastic decline in ecosystem services on which human society depends from changes in land use and land cover within coastal wetlands [15]. Coastal land use and land cover (LULC) change is illustrated by clearing and modifying coastal habitats and artificial barriers to flow. For example, one of the highest risks to the Great Barrier Reef that has been identified by the Australian Government is the degradation to coastal habitats and connectivity impairment as a result of land use changes affecting the region's ecosystems [16].

The major threats to coastal wetlands are climate change, clearing (through urban areas, ports, and industry development), dieback, changes in hydrology (e.g., the restriction or alteration of flows), and pollution [17–19]. Additionally, overfishing, cattle grazing, pest animals, the use of recreational vehicles, and fire have had impacts on some components of wetland systems [20]. Pressures can be subtle but may result in considerable changes in ecosystem functioning. These threats are often related to, for example, hydrological change (including the development of ponded pasture) may significantly alter water quality, and heavy and sustained grazing pressure of marine grasslands can dramatically alter ground cover. Thus, the modification of ecosystems a ffects both habitat value and the filtration and retention capacity of those areas [21]. On Queensland's east coast, agriculture and the urban development of infrastructure with berms, ponded pasture, dams, seawalls, and roads on coastal plains impose threats to the resilience of mangroves and associated wetlands [22]. For example, in the Mackay region of Queensland, Pioneer River mangroves have been reclaimed on average by 5 ha each year over the last 50 years [23], with a total loss of 26% since European settlement [24]. Mangrove–fishery links are well-recognised [25], but, to expedite conservation e fforts, it is necessary to quantify the spatio–temporal scales of change in mangrove habitats (e.g., disturbance, loss, and regrowth) [26,27]. Indeed, the array of benefits that are o ffered by wetlands makes it critical that they are monitored, maintained, and restored where and whenever possible [28]. Paradoxically, within Great Barrier Reef coastal provinces, ecosystem e ffects and cumulative impacts on fishery resources are poorly understood [29]. Moreover, disparate jurisdictional responsibilities hinder assessment e fforts. With the ongoing loss of these systems, Australia's commercial and recreational fisheries are becoming depleted nation-wide [30].

Quantifying LULC changes is not only crucial for the evaluation of services but also the protection of coastal wetland ecosystems, and remote sensing technology provides one of the most useful ways to monitor wetland dynamics [6]. Protected areas such as national parks often depend on the landscapes surrounding them and their hydrologic connections to maintain flows of organisms, water, nutrients, and energy. Park managers have little authority over the surrounding landscape, although land use changes and hydrology alterations can have major impacts on the integrity of a protected area [31,32]. In Queensland, public and private lands are surveyed by the Statewide Landcover and Trees Study (SLATS), which uses satellite imagery to monitor woody vegetation clearing in native vegetation including mangroves and estuarine regions [33], but no studies have focused on using remote sensing to map biome variability and change dynamics in Great Barrier Reef catchments on a landscape scale.

The fundamental broad objectives of this research are to assess the regional drivers of wetland degradation in order to assist in maintaining the values that underpin estuarine ecosystem integrity within and outside the boundary of a protected area. The research presented here can aid the prediction of responses under future change scenarios (e.g., climate shifts/disturbance). The change detection process identifies di fferences in the state of an object or phenomenon by observation at di fferent times [34]. Possible classification inaccuracies and a lack of consensus in regional-scale LULC approaches necessitates the employment of more than one method for comparative purposes and to aid validation, particularly for change detection in complex landscapes such as coastal wetlands [35].

Our study focuses on change dynamics by using four methods of change analysis: post-classification change analysis with a supervised classification technique, visual interpretation, thematic change dynamics, and trend analysis. There has been an increased use of supervised classification techniques in comparison with unsupervised techniques in the last decade [36]. The supervised pixel-based maximum likelihood ML classification is the most common method in remote sensing image data analysis, and it is often applied as a benchmarking algorithm [37–39]. Supervised change analyses for wetland mapping in remote sensing studies have previously focused on coast line dynamics or reclamation activities [40,41]; few studies have examined wetland change dynamics within and surrounding the border of a protected area and adjacent to a World Heritage Site. We provide a description of observed changes through maps generated for a fourteen-year time period (2004–2017) on a two/three-yearly basis. The research contributes to the fields of land cover characterisation, landscape dynamics, and conservation planning.

The objectives of this study are: (1) to quantify how the coastal landscape (mangroves and associated communities) has spatially and temporally changed in a period of 14 years (2004–2017) within a region that is subjected to intense commercial and recreational fishing; (2) to assess the implications of landscape change to biodiversity within and outside the boundary of a national park; (3) to inform regional land planning, conservation e fforts, and policy-makers. In summary, the study addresses the important question: has significant, human-induced change occurred in the coastal landscape, resulting in altered ecosystem function that could have possible repercussions for the fishery resource?

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

#### *2.1. Study Area and Data Sources*

Our study area is located within the northeast coast drainage division of the Central Queensland coast, specifically the Plane Creek Basin catchment of the Mackay Whitsunday Natural Resource Region, Central Queensland, Australia. Rocky Dam Creek and Cape Palmerston National park are positioned in the Ince Bay Receiving Waters adjacent to the World Heritage listed Great Barrier Reef. The primary intensive land use is the cultivation of sugar cane, making up 18% of the catchment area, with Mackay being the largest sugar-producing region in Australia [42]. Grazing is also an important land use, accounting for 42% [21]. The region's estuaries directly support several commercial fisheries, e.g., East Coast Inshore Fin Fish Fishery, East Coast Otter Trawl Fishery, and Coral Reef Fin Fish Fishery [43]. Additionally, recreational fishing is a considerable activity in the region, with 24.8% of the population participating in fishing for recreation, far greater than the state average of 15.1% [44]. The Mackay Whitsunday Natural Resource Region supports extensive areas of estuarine and mangrove wetlands, these being dominant features of the coastal landscape [45]. Mangroves and associated communities cover 62,094 ha of tidal land in the region, with nine wetland areas recognised as nationally important [46]. The total area of the Rocky Dam Creek sub-catchment is 53,697.5 hectares. Cape Palmerston National Park is listed as a category II protected area on the International Union for Conservation of Nature (IUCN) World Database on Protected Areas [47] and covers 7200 hectares (Figure 1). Ten ecosystem types listed as endangered in the IUCN Red List are present (Table 1 and Figure 2). The areal extent of the sub-catchment and coastal zone used for land cover classification has 53,302.05 hectares of a variety of land cover types. The study area is located between latitude 21◦27–21◦37S and longitude 149◦17–149◦26E.

**Figure 1.** Study site—Rocky Dam Creek and Cape Palmerston National Park Central Queensland—Sentinel-2B composite image visualised by using the red, green, and blue wavelength bands, captured 31 January 2018 at 00:22:57 provided by United States Geological Survey (USGS).


**Table 1.** (IUCN)-listed endangered ecosystems occurring at the study site [20]—Rocky Dam Creek/Cape Palmerston National Park.

*Remote Sens.* **2020**, *12*, 197


#### **Table 1.** *Cont.*

**Figure 2.** Sentinel-2B image captured 31 January 2018 with IUCN-endangered ecosystems [20]—Rocky Dam Creek/Cape Palmerston National Park.

Seven broadly recognised mangrove communities occur throughout the region. Within the high rainfall areas of the Central Queensland coast bioregion, estuarine wetlands are about equally dominated by saltpan and samphire flats along the high intertidal area; yellow and orange mangroves (*Ceriops tagal* and *Bruguiera* spp.) dominate along the mid-intertidal area; and the stilted mangrove (*Rhizophora stylosa*) dominates in the lower intertidal area [45]. Twenty-three tree and shrub species of mangroves are present [48]. Landscape elevation ranges from 238 m to sea level; therefore, the study site is not solely within the legislative constraints of the defined coastal area of the Queensland Government (i.e., 5 km from the coastline or where land reaches the height of 10 m; Australian Height Datum [29]).

We used the Sentinel-2B image captured on 31 January 2018 (spatial resolution of 10 m) to overlay the major surface water features of the Plane Creek Basin catchment from the Australian Hydrological Geospatial Fabric (Geofabric) [49] (Figure 3). The map provides a hydrological visualisation of topographically consistent spatial surface water features and stream connectivity. Geospatial stream data are useful for natural resource managers, as streams can be traced upstream and downstream to identify drainage networks and water movement within the catchment area. Here, the map is included to provide information on water connections and hydrology throughout the sub-catchment.

**Figure 3.** Sentinel-2B image captured 31 January 2018 overlaid with the Australian Hydrological Geospatial Fabric (Geofabric) [49] showing the drainage networks and hydrological connections of Rocky Dam Creek/Cape Palmerston National Park.

For change detection, we used the Landsat satellite archive images captured in April, August, and September for the years 2004, 2006, 2009, 2013, 2015 and 2017 (Table 2). We acquired images from the United States Geological Survey (USGS) Earth Explorer Landsat Archive at level 1T, (except for 2009) which has systematic radiometric and geometric correction applied to the data by incorporating ground control points and topographic accuracy by utilizing a digital elevation model. The 2009 image

was derived from USGS Collection 1 and processed at Tier 1. The 2004, 2006 and 2009 images were taken from the Landsat 5 Thematic Mapper (TM). The 2013, 2015 and 2017 images were taken from Landsat 8 OLI (Operational Land Imager) and TIRS (Thermal Infrared Sensor). Data from Landsat satellites are spatially and geometrically consistent, and they comply with UTM projection [50]. We derived maps from imagery acquired in winter and early spring, as cloud cover inhibits image availability in warmer months [51]. Tidal information corresponding to each image date and time are from the Bureau of Meteorology [52] (Table 2). Medium-resolution satellite imagery is suitable for mapping mangrove and wetland areas on a regional scale [53]. There are two reasons for selecting Landsat imagery: (1) It is acquired at regular intervals, and (2) it is freely available from USGS. We acquired data from two independent sources for use as ground truth data: (1) Queensland Herbarium from 2003, 2006, 2009, 2011, 2013, 2015, and 2017 [54] and (2) Google Earth images from 2005, 2009, 2013, and 2016. Local expert knowledge was included for validation, as this has become an important component of mapping methodology [55].


**Table 2.** Image dates and observed sea levels for Hay Point tidal gauge, Central Queensland.
