(PDF) Multi-Resolution and Multi-Temporal Satellite Remote Sensing Analysis to Understand Human-Induced Changes in the Landscape

Article Multi-Resolution and Multi-Temporal Satellite Remote Sensing Analysis to Understand Human-Induced Changes in the Landscape for the Protection of Cultural Heritage: The Case Study of the MapDam Project, Syria Nicodemo Abate 1 , Diego Ronchi 1, * , Sara Elettra Zaia 1 , Gabriele Ciccone 1 , Alessia Frisetti 1 , Maria Sileo 1 , Nicola Masini 1 , Rosa Lasaponara 2 , Tatiana Pedrazzi 1 and Marina Pucci 3 1 2 3 * Institute of Heritage Science, National Research Council, Via Cardinale Guglielmo Sanfelice 8, 80134 Napoli, Italy;

(N.A.);

(S.E.Z.);

(G.C.);

(A.F.);

(M.S.);

(N.M.);

(T.P.) Institute of Methodology of Environmental Analysis, National Research Council, C.da S. Loja sn, 85050 Tito Scalo, Italy;

Department of History, Archaeology, Geography, Art and Entertainment, University of Florence, Via S. Gallo, 10, 50129 Florence, Italy;

Correspondence:

Abstract Academic Editor: Faccini Francesco Received: 15 October 2025 Revised: 5 November 2025 Accepted: 7 November 2025 Published: 11 November 2025 Citation: Abate, N.; Ronchi, D.; Zaia, S.E.; Ciccone, G.; Frisetti, A.; Sileo, M.; Masini, N.; Lasaponara, R.; Pedrazzi, T.; Pucci, M. Multi-Resolution and Multi-Temporal Satellite Remote Sensing Analysis to Understand Human-Induced Changes in the Landscape for the Protection of Cultural Heritage: The Case Study of This study presents a multi-resolution and multi-temporal remote sensing approach to assess human-induced changes in cultural landscapes, with a focus on the archaeological site of Amrit (Syria) within the MapDam project. By integrating satellite archives (KH, Landsat series, NASADEM) with ancillary geospatial data (OpenStreetMap) and advanced analytical methods, four decades (1984–2024) of land-use/land-cover (LULC) change and shoreline dynamics were reconstructed. Machine learning classification (Random Forest) achieved high accuracy (Test Accuracy = 0.94; Kappa = 0.89), enabling robust LULC mapping, while predictive modelling of urban expansion, calibrated through a Gradient Boosting Machine, attained a Figure of Merit of 0.157, confirming strong predictive reliability. The results reveal path-dependent urban growth concentrated on low-slope terrains (≤5◦ ) and consistent with proximity to infrastructure, alongside significant shoreline regression after 1974. A Business-as-Usual projection for 2024–2034 estimates 8.676 ha of new anthropisation, predominantly along accessible plains and peri-urban fringes. Beyond quantitative outcomes, this study demonstrates the replicability and scalability of open-source, data-driven workflows using Google Earth Engine and Python 3.14, making them applicable to other high-risk heritage contexts. This transparent methodology is particularly critical in conflict zones or in regions where cultural assets are neglected due to economic constraints, political agendas, or governance limitations, offering a powerful tool to document and safeguard endangered archaeological landscapes. the MapDam Project, Syria. Land 2025, 14, 2233. https://doi.org/10.3390/ land14112233 Keywords: remote sensing; SAR; optical; shoreline changes; land use land cover changes; archaeology; machine learning Copyright: © 2025 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and 1. Introduction conditions of the Creative Commons Anthropogenic action, characterized by the uncontrolled expansion of urban areas, acts of war, illegal excavations, or public works into surrounding rural lands, poses significant challenges to the preservation of cultural heritage. The intersection of remote sensing Attribution (CC BY) license (https://creativecommons.org/ licenses/by/4.0/). Land 2025, 14, 2233 https://doi.org/10.3390/land14112233 Land 2025, 14, 2233 2 of 25 technology with Land Use Land Cover Changes (LULC) studies has emerged as a critical area of research, providing innovative methodologies for monitoring and managing the impacts of human-induced changes on areas of high cultural (archaeological) potential. Remote sensing technologies, including satellite imagery and aerial surveys, have been key to monitoring urban expansion and its effects on cultural heritage [1–3]. Researchers emphasized the importance of using Earth observation data to assess damage to cultural heritage in Europe, highlighting the establishment of initiatives such as the Copernicus program for cultural heritage (https://www.copernicus.eu/en/news/news/ observer-copernicus-preservation-global-cultural-heritage-sites, accessed on 4 November 2025), which aims to harness remote sensing for the protection of cultural sites. The results of this program underscore the growing recognition of remote sensing as a vital tool for cultural heritage management, particularly in the context of urban sprawl, land use (land reclamation), and water management [4]. Remote sensing in this area has proven to be extremely effective in analyzing archaeological sites and the landscape in which they are located. This approach not only helps to document cultural heritage, but also facilitates risk assessment, enabling proactive or mitigative measures to be taken. Integrating remote sensing data with traditional archaeological methods also improves understanding of how urbanisation affects archaeological landscapes, providing a comprehensive view of cultural heritage at risk [5]. The combination of remote sensing with advanced analytical techniques has proven beneficial for cultural heritage preservation. Some works illustrated the use of infrared thermography and UAV digital photogrammetry for the protection of rock heritage sites in Georgia, arguing the need for a site-specific and interdisciplinary approach to heritage conservation. This methodological framework highlights the potential of remote sensing to inform conservation strategies that respond to the unique characteristics of cultural heritage sites, particularly in urban settings where sprawl threatens their integrity. Monitoring landscape changes due to urbanisation is another critical application of remote sensing in the cultural heritage sector [6]. SAR (e.g., Sentinel-1 Synthetic Aperture Radar) imagery was used to assess landscape changes such as in the Paphos area of Cyprus following seismic events, demonstrating the usefulness of remote sensing in monitoring cultural heritage sites threatened by both natural and anthropogenic factors [7–10]. Open-source satellite data have been successfully used to understand the risk induced by anthropogenic changes to the nearshore water management system in archaeological sites on the northern Egyptian coast [11], as well as significant changes produced to the coastline in the southern Italian peninsula [12–17]. This capability is essential for developing effective risk management strategies that consider the dynamic nature of urban environments and their impact on cultural heritage. In addition to monitoring changes, remote sensing facilitates the identification of unstable areas that may pose a risk to cultural heritage. Multiplatform remote sensing techniques are used to identify potentially unstable areas, highlighting the importance of rapid and cost-effective solutions for urban and landscape management [18,19]. Such assessments are critical to safeguarding cultural heritage sites located in vulnerable urban settings, where sprawl can exacerbate risks associated with natural hazards. The role of remote sensing in detecting urban change was further explored using multispectral remote sensing and SAR data to monitor urban change [2,20,21]. Their results underscore the importance of integrating different remote sensing technologies to gain a comprehensive understanding of urban dynamics and their implications for cultural heritage. This multifaceted approach enables the identification of urban growth patterns that may encroach on culturally significant areas, providing valuable data for urban planners and heritage managers. Urban expansion in culturally rich regions, such as Southeast Asia or India, poses unique challenges for heritage conservation. Many scholars focus their attention on analysing contexts that are difficult to monitor, highlighting the negative Land 2025, 14, 2233 3 of 25 effects of urban sprawl on UNESCO World Heritage sites. Their study highlights the need for effective monitoring and management strategies that incorporate remote sensing data to protect cultural heritage from urban development pressures [22–24]. The integration of remote sensing with geographic information systems (GIS) has improved the analysis of urban sprawl and its effects on cultural heritage. For example, researchers discussed the assessment of opportunities for systematic condition assessment of cultural heritage sites using high-resolution multispectral imagery from Copernicus Sentinel-1 and 2 [8,24–26]. This integration enables detailed spatial analysis, allowing researchers to visualize and quantify the impacts of urbanisation on cultural landscapes. In order to address this need, remote sensing often uses techniques to map land use coverage and changes. These operations can be carried out using maps already provided by various providers (free of charge or otherwise), such as (i) ESA (European Space Agency) with Corine Land Cover, or ESRI [27]; and (ii) maps produced periodically from satellite data (e.g., Sentinel-2) using land use change classifications based on multispectral and SAR satellite data and machine and deep learning algorithms [7,20,28–33]. In addition, the use of remote sensing in disaster risk management for cultural heritage has gained attention. Satellite remote sensing has a crucial role in addressing the various hazards that threaten cultural heritage sites, emphasizing the need for multidisciplinary cooperation in disaster risk management. This approach is particularly relevant in urban areas, where rapid development can increase vulnerability to disasters, necessitating effective monitoring and response strategies [19,34–41]. This research focused on the acquisition and processing of medium- and highresolution satellite data with the specific purpose of addressing the issue of the impact of anthropogenic action on the archaeological area (e.g., land use change and Urban Sprawl, and induced changes to the coastline). Satellite data assessment and acquisition procedures were implemented, as well as artificial intelligence-based processing aimed at understanding the landscape evolution of the southern coast of Syria, in an area of 1700 km2 (170,000 hectares) approx., to understand the impact of human activity over time near the archaeological area of Amrit, in the framework of MapDam (Mapping Archaeological Damage) project. The methodological approach used in this work aims to show a completely opensource flowchart, through the use of open and big data and cloud processing platforms such as Google Earth Engine, which can be useful for monitoring an area considered to have strong archaeological potential. Furthermore, this approach can also be useful for managing the archaeological area, in this case, the Amrit site. In fact, for this site, there is a lack of large-scale information that would allow us to understand the evolution of the anthropised landscape and the consequent changes made by humans in the past, present and, predictably, in the future. This study aims to reconstruct how human activities have altered the Amrit cultural landscape over the past four decades and to demonstrate how predictive modelling can support proactive cultural heritage risk assessment in regions affected by instability or rapid urban growth. These objectives provide the conceptual and methodological foundation for linking remote sensing evidence to broader socio-environmental processes and for advancing a data-driven framework for preventive heritage management. 2. Materials and Methods 2.1. Study Area The Syrian coastal region, and particularly the area around Tartous (Figure 1), presents a rich and complex interplay of geomorphological, archaeological, and historical elements. Land 2025, 14, 2233 4 of 25 Figure 1. (a) Study area with a zoom on the area of interest (dashed lines) also shown in its drawn representation (from [42]); (b) Phoenician ma’adeb © CNR ISPC; (c) the so-called meghazil © CNR ISPC. Geographically, the region is structured into three major geomorphological units: the coastal plain, the plateau, and the mountainous hinterland. From an archaeological perspective, the area has been the focus of both historical and recent investigations. Notably, during the 1990s, a Syro-French mission undertook systematic surveys in the Tartous region, with particular attention to its prehistoric occupation and geomorphological evolution [43]. These efforts were resumed and expanded in the following decades by the Syrian Directorate-General of Antiquities and Museums (DGAM), particularly around the site of Amrit [44]. In recent years, two international projects have significantly contributed to our knowledge of the region’s cultural and environmental heritage. The EAMENA project (Endangered Archaeology in the Middle East and North Africa, https://eamena.org/, accessed on 4 November 2025) has catalogued terrestrial heritage sites through remote sensing and database integration, while the Honor Frost Foundation’s initiative has focused on the maritime archaeology of the Syrian coast. Archaeological interest in the region dates to the earliest phases of research in the Levant. In 1860–1861, Ernest Renan’s Mission de Phénicie included exploratory work at Amrit and Tartous. However, systematic excavations at Amrit were only initiated in 1926 by Maurice Dunand and continued in 1954 by Nassib Saliby [45,46]. Further investigations were resumed in 2010 under the direction of Michel al Maqdissi. Earlier, in 1938, R. Land 2025, 14, 2233 5 of 25 Braidwood had conducted soundings at the sites of Tabbat al-Hammâm and Tell Simiriyân, south of Amrit [47], while salvage excavations at Tell Ghamqa, near Tartous, have also contributed valuable data [48]. The most comprehensive evidence for the Late Bronze Age in the region derives from the site of Tell Kazel—identified with ancient Simyra/Sumur—located along the Nahr el-Abrash, and extensively excavated over several decades [49]. By the 8th century BCE, a Phoenician settlement had developed on the island of Arwad, assuming a prominent role in regional maritime trade [50]. Inland settlements of the same period, while likely dependent on Arwad, appear to have maintained varying degrees of autonomy—some even possessing their own coinage [51]. Among the most significant material attestations of Phoenician presence in the region is the sanctuary of Amrit, located approximately 7 km south of Arwad. Probably constructed in the 7th or 6th century BCE, the sanctuary features a large court carved into the bedrock and bordered on three sides by a portico; at its center stands a well-preserved naos; nearby are two funerary monuments known as meghazil (“spindles”), interpreted as elite burial structures [52]. In the 19th century, Amrit was visited by travelers and scholars such as Ernest Renan, who dedicated several pages of his monumental work Mission de Phénicie to the site and its architectural remains. Systematic archaeological excavations at Amrit began only in the 1920s and were later continued in the 1950s by Maurice Dunand and Nassib Saliby [42]. More recently, further investigations have been carried out by the Directorate-General of Antiquities and Museums (DGAM) of Syria, under the direction of Michel al Maqdissi, contributing significantly to our understanding of the site’s stratigraphy, architectural evolution, and its broader cultural context [53]. The sanctuary of Amrit, known as the maabed, has been described as “the most singular building in all of Phoenicia”. Indeed, it represents an innovative architectural creation—not for its individual features, but rather for its overall conception. The architectural design of the Amrit sanctuary is therefore without precedent, yet it is the product of deep and long-standing cultural interactions—with Egypt, the broader Near Eastern world, as well as Cyprus and Greece [54]. 2.2. Large Spatial and Temporal Scale Analysis To analyze the described area near the archaeological site of Amrit (Figure 1), different satellite datasets were processed to understand the different phenomena that affected the area. The workflow used for this study is described in Figure 2. Several satellite datasets and ancillary data were considered between the years 1970 and 2024. Specifically, (i) Landsat data (TM, ETM+, OLI & TIRS, OLI-2, and TIRS-2), between 1984 and 2024; (ii) KH; (iii) NASADEM (Table 1); and (iv) OSM data. Table 1. Used satellites. Satellite Instruments Resolution Landsat 5 Landsat 7 Thematic Mapper (TM) Enhanced Thematic Mapper Plus (ETM+) Operational Land Imager (OLI) and Thermal Infrared Sensor (TIRS) OLI-2 and TIRS-2 KH-9 Digital Elevation Model 30 m (MS), 120 m (IRT) 30 m (MS), 15 m (Pan) Landsat 8 Landsat 9 KH NASADEM 30 m (OLI), 100 m (TIRS) 30 m (OLI), 100 m (TIRS) 6–9 m 30 m The Landsat programme is a collaboration between the U.S. Geological Survey (USGS) and the National Aeronautics and Space Administration (NASA), is the world’s longestrunning civilian Earth observation programme. Since the launch of its first satellite in 1972, Landsat has provided a continuous stream of global data on the Earth’s surface, Land 2025, 14, 2233 6 of 25 becoming a key resource for studying environmental change, managing natural resources, and monitoring landscape alterations, both natural and man-made. Landsat missions have provided continuous data for Earth observation since 1972. Landsat 5 used the Thematic Mapper (TM) sensor with 7 bands and 30 m resolution. The next Landsat 6, which unfortunately failed, was supposed to introduce a panchromatic band. Landsat 7 brought the Enhanced Thematic Mapper Plus (ETM+) with a 15 m panchromatic band and a 60 m thermal band. The Landsat 8 and 9 missions revolutionised the programme with OLI and TIRS sensors, introducing new bands for aerosols and cirrus clouds and increasing radiometric resolution from 8-bit to 12-bit (Landsat 8) and 14-bit (Landsat 9), greatly improving accuracy and spatial resolution [20,30,55–62]. Figure 2. Flowchart. The Keyhole (KH) satellite system KH-9 (Project Name Hexagon) operated between 1971 and 1984. The KH-9 program was designed to support mapping requirements and the exact positioning of geographical points for the military. The KH-9 panoramic cameras captured high-resolution (0.61–1.22 m) (Declass 3 2013) and moderate resolution (6–9 m) (Declass 2 2002) terrain images. High-resolution images were acquired on 16.5-cm-wide variable length film. The moderate resolution terrain camera acquired images that were printed to 23-cm-wide variable-length film. Almost all of the imagery from these cameras was declassified in 2011 as a continuation of Executive Order 12951, the same order that declassified CORONA (Declass 1 1996). The declassified images were transferred to the U.S. Geological Survey’s Earth Resources Observation and Science (EROS) Center, and it has been processed to make them available on the USGS website. The process is still ongoing (from the USGS website https://www.usgs.gov/centers/eros/science/usgs-eros-archivedeclassified-data-declassified-satellite-imagery-3 accessed on 9 October 2025) [63–69]. Several ancillary data (e.g., structures, roads, railway networks, waterways) were retrieved from OSM (Open Street Map). Open Street Map is a project involving the collaborative implementation of georeferenced cartographic data, which can be used free Land 2025, 14, 2233 7 of 25 of charge by anyone. This data can be consulted within a GIS environment (via plug-ins) or externally. The data obtained through OSM were combined with that obtained from satellite processing [70–72]. NASADEM represents a major advance in the field of global Digital Elevation Models (DEMs). It is not a simple update, but a complete modernisation of Shuttle Radar Topography Mission (SRTM) [73] data through a sophisticated reprocessing process and the integration of auxiliary data. The main improvements include significantly higher vertical accuracy and a significant reduction in gaps (areas with no data). A crucial aspect of the reprocessing procedure is the role of ICESat GLAS data [74,75]. The new improved SRTM elevations in NASADEM are derived from better vertical control of each SRTM data strip through reference to ICESat elevations [76,77]. NASADEM was accessed directly from the Google Earth Engine (GEE) database [78]. The analysis of this data was used to: (i) calculate the change in land use between 1984 and 2024; (ii) calculate the change in the shoreline for the period 1970–2024; (iii) combine data from satellites, their derivatives, and ancillary data in order to predict future scenarios of anthropogenic expansion in the area and changes in the shoreline. The analysis of this data was carried out solely and exclusively for the purpose of recognising human induced changes in the area of interest and on the site of interest, as shown in Figure 1. 2.3. Land Use Land Cover Changes over Time The analysis of the expansion of anthropogenic activity was conducted using the Google Earth Engine (GEE) platform. GEE is a powerful open-source platform developed by Google that provides a web interface and interactive development environment (IDE) to access and work with a wide range of datasets, covering more than four decades of global data. These include satellite data from missions such as MODIS, ALOS, Landsat, and Sentinel, as well as other useful resources such as digital elevation models, shapefiles, meteorological data, and land cover information [78]. With its high-performance computing capabilities and handling of large volumes of data, GEE has established itself as a benchmark tool in remote sensing and big data analysis [79,80]. The platform’s popularity has grown in many disciplines, as evidenced by the increasing number of scientific publications using it. Indeed, researchers have adopted GEE in various fields, including the study of vegetation, land use and land cover, hydrology, climate, and cultural heritage analysis. In addition, the spread of GEE has led to the creation and sharing of numerous free tools that can be accessed directly from the GEE website [81–83]. Satellite data were classified to create a land use map based on three classes: (i) water, (ii) permanent or temporary anthropogenic structures (e.g., greenhouses), and (iii) other use (e.g., bare soil, cropland, forest). The analysis involved the creation of a JavaScript code useful for: (i) select satellite data related to the Landsat missions (5, 7, 8, 9) for the entire area, with filtering based on cloud percentage set to a maximum threshold of 5%; (ii) application within the images of the collection of masks for clouds and snow, according to the functions provided by GEE for these datasets; (iii) creation of vegetation indices; manual selection of training and validation points for water classes, anthropogenic structure, other; (iv) selection of training and test datasets; (v) training of Random Forest (RF) algorithm [20,84] for classification; (vi) estimation of “Hyperparameter Tuning for the number Of Trees Parameters”; (vii) generation of classifier error matrix and estimation of Train and Test Accuracy and Kappa, and the weight of the input data used in the classifier; (viii) generation of the classified image. More than 1000 satellite images (Table 1) were used taken for this activ- Land 2025, 14, 2233 8 of 25 ity to create Land Use Land Cover maps for the years 1986 (±2 years), 1994 (±2 years), 2004 (±2 years), 2014 (±2 years); 2024 (±2 years). The vegetation indices created for the analysis are listed in Table 2. Table 2. Used indices. Index References ( Nir − Red) NDV I = ( Nir+ Red) ( Nir − Red) SAV I = ( Nir+ Red+ L) [85] ∗ (1 + L ) ( Nir − Red) EV I = G ∗ ( Nir+C ∗ Red−C ∗ Blue+ L) 1 2 Nir −1 GCV I = Green ( Nir − RedEdge) NDRE = ( Nir+ RedEdge) √ 2 ∗ Nir +1− (2 ∗ Nir +1)2 −8 ∗ ( Nir − Red) MSAV I = 2 ( Green− Nir ) NDW I = (Green+ Nir) (Swir + Red)−( Nir + Blue) BSI = (Swir+ Red)+( Nir+ Blue) ( Nir −Swir ) BU I = ( Nir+Swir) (Swir − Nir ) NDBI = (Swir+ Nir) [86] [87] [87] [2] [86] [88,89] [88] [21] [90] Training datasets containing the classes (i) water (125 items); (ii) Anthropic (220 items); and (iii) other (220 items) were used for the data. The training dataset was divided into 70% training and 30% testing for the classifications. RF classification was performed using Hyperparameter Tuning [91] for the number Of Trees Parameters with any number of trees from 0 to 150, to determine the best number in the computation-time-accuracy ratio. The data were classified using 110 trees in the RF algorithm, it turns out to be the best number of trees. This activity made it possible to: (i) generate multi-band satellite imagery from 1984 to 2024, of the area of interest; (ii) generate land use maps from 1984 to 2024, of the area of interest, on a 10-year cadence. The data extracted in this way were correlated with the DEM and slope calculated from NASADEM data to understand the evolution of anthropogenic expansion in relation to land slopes. 2.4. Shoreline Changes To ensure a consistent long-term shoreline/coastline analysis, satellite images were used, comprising the following platforms: KH-9 Hexagon images, Landsat 5 Thematic Mapper (TM), Landsat 7 Enhanced Thematic Mapper Plus (ETM+), Landsat 8 Operational Land Imager (OLI), and Landsat 9 OLI-2. The preliminary analyses for understanding the dynamics of interest on the coastline involved (i) extracting the coastline from KH data; (ii) extracting the coastline from Landsat data; (iii) calculating the change rate of the shoreline using Ordinary Least Squares and End Point Rate methods. KH-9 Hexagon images were used. In particular, the KeyHole satellite system KH-9 (Hexagon), missions 1210-5 and 1209-1, collected in 1974, were downloaded from the U.S. Geological Survey website (https://earthexplorer.usgs.gov/, accessed on 10 July 2025). The process of extracting the coastline from satellite images was carried out: (i) (ii) First, by improving contrast and reducing noise and the NDWI index for Landsat data; Then, performing unsupervised classification to separate the two classes of water and land in the KH-9 images, while creating a binary mask for water bodies in Landsat (using NDWI values as thresholds); Land 2025, 14, 2233 9 of 25 (iii) Finally, converting the raster of water bodies into vector polygons and finally extracting the coastline [11,14]. Temporal subsets of Landsat images were selected to represent five key time intervals: 1984–1989 (Landsat 5), 1994–1996 (Landsat 5), 2004–2006 (Landsat 7), 2014–2016 (Landsat 8), and 2023–2024 (Landsat 9). All images were filtered by cloud cover (<5%) and clipped to the boundaries of the defined area of interest. Radiometric corrections were applied to all scenes using GEE-native scale factors for surface reflectance. Thermal bands were also recalibrated to Kelvin using standardized coefficients, although they were not used in the subsequent shoreline analysis. For each image collection, the Normalized Difference Water Index (NDWI) was computed to enhance open water features. Median NDWI composites were generated for each temporal window to reduce noise and minimize ephemeral water artifacts. To delineate coastlines, each NDWI image was thresholded using a fixed value (NDWI > 0), thereby isolating persistent water bodies. The resulting binary water masks were converted into vector polygons, with 30-m spatial resolution and 8-connected neighborhood logic. The resulting geometries represent the generalized extent of the coastline during each time period (Figure 3). Figure 3. Binary classification from NDWI. The coastlines obtained from both methods were then treated as vector files in a GIS environment: (i) First, transects were calculated along all coastlines. An offshore baseline parallel to the coast was added, while an onshore baseline was added a few kilometers inland. The former served as the starting point for the transects, to include all coastlines. The latter is the internal boundary for the calculations. The transects were set at a distance of 50 m from each other. (ii) The Extract Intersection Points tool was used to calculate the points of intersection between the transects and the shorelines, while the perpendicular distance from each point of intersection to the baseline was extracted. (iii) The Ordinary Least Squares (OLS) tool was used to calculate the Linear Regression to determine the shoreline change rate for each transect, according to (1): Land 2025, 14, 2233 10 of 25 Distance = (Slope × Year) + Intercept + ε (1) where: slope of this equation represents the shoreline change rate (meters per year); intercept is the shoreline position at Year equal to 0; ε represents the residual error. In OLS, the explanatory variable (independent variable) is the year, and the dependent variable is the distance from baseline (in meters). (iv) In addition to OLS, End Point Rate (EPR) was used to study the shoreline changes (erosion and accretion). EPR is a method to calculate the rate of shoreline change by dividing the net shoreline movement (distance between the oldest and newest shorelines) by the time elapsed between those shorelines. It is a straightforward calculation, and it assumes a linear rate of change between the two measured shorelines. 2.5. Spatially-Constrained Probabilistic Urban Growth Modelling with Multi-Scale Validation 2.5.1. Data Harmonisation and Pre-Processing Multi-temporal land-use/land-cover (LULC) classifications at 30 m spatial resolution were used for the epochs 1984, 1994, 2004, 2014, and 2024 (class 1 = urban) [92]. A binary validity mask identified pixels with consistent labels across epochs. All ancillary data (slope, DEM) and OSM-derived vectors (roads, building footprints, hydrography) were co-registered to the LULC-2024 grid [93,94]. To increase robustness and mitigate label noise, all data were aggregated to 90 m support (3 × 3 of 30 m) with a majority rule for binary layers and block statistics for continuous layers. For a binary class map Ut (x) ∈ {0, 1} at epoch t and location x, the 90 m urban mask Ut 90 (2) [95]:  Ut90 ( x ) = 1 ∑ Ut x x ′ ∈ N3 ( x ) ′  ≥ θ (2) For a continuous layer Z(x), the 90 m values are given by (3) [95]: Z90 ( x ) = 1 Z x′ ∑ | N3 ( x )| x′ ∈ N ( x) (3) 3 where x is the pixel center; Ut is the urban mask; N3 is the 3 × 3 pixels window; θ is the majority threshold, i.e., at least 5 out of 9 pixels must be urban at 30 m for the 90 m block to be considered urban; and Z is any continuous (non-binary) raster, e.g., slope (in degrees) or DEM (in metres). OSM roads, buildings, and hydrography were rasterised onto the analysis grid using Euclidean distance transformations, which provided pixel-by-pixel distances to the nearest feature. To create a binary support raster B(x) marking OSM feature (4) was used: DB ( x ) = min ∥ x − y ∥2 y : B(y) = 1 (4) Access and attraction scores were normalized to [0, 1] using (5) [96,97]: AB (x) = 1 − DB ( x ) max2 DB ( x ) (5) where B is the binary mask of a vector (e.g., roads, buildings, water); DB is the distance to B; ∥·∥2 is the Euclidean normalisation; AB is the normalized accessibility. In order to establish predictors and use the 2014 and 2024 LULC maps as validators, predictors were computed on the 2014 and 2024 bases at 90 m, using [98]: Land 2025, 14, 2233 11 of 25 (i) Proximity to existing urban (distance to the urban edge inverted and normalized) and local means with 3 × 3 and 5 × 5 filters (PU,3 , PU,5 ) to encode short-range spillovers (6), where P is the local means; and d(x, ∂Ut ) is the distance to urban boundary [99]: PU ( x ) = 1 − (ii) d( x, ∂U t ) max x d( x, ∂U t ) (6) Terrain suitability (slope preference) (7), where S(x) is the slope value in degrees [29,100,101]: S pre f ( x ) = 1 − min{S( x ), 30◦ } 30◦ (7) (iii) Accessibility/density to networks (Aroads , Abldg , Awater ) from (5) [72]; (iv) Binary indicators for near roads and buildings (≤180 m) were averaged with a 5 × 5 window to obtain local densities ρroads , ρbldg [102]. 2.5.2. Learning Target and Sampling Domain To forecast future non-anthropised to anthropised (e.g., 2024–2034) with calibrated probabilities, the model was trained on a fully observed historical interval (2004–2014) and validated out-of-sample on a disjoint interval (2014–2024). The data used were those described in 2.5.1. The analyses were conducted in Python [103] using various libraries (e.g., Scikit-learn [104], Rasterio [105], Geopandas [94,106]). The supervision and sampling design is crafted to (i) prevent temporal leakage, (ii) suppress edge noise due to georegistration errors, and (iii) control class imbalance. Two classes were established to train the model used: (i) Positive class: Non-anthropised pixel in 2004 but anthropised in 2014; and (ii) Negative class: Non-anthropised pixel in both 2004 and 2014. The label at location x is given by (8) [107]: y04 → y14 ( x ) = 1(¬U2004 ( x ) ∧ U2014 ( x ) ∧ V ( x )) (8) where V is a validity mask, ∧ and ¬ are logical AND and NOT. The training domain (D) based on pixels that are non-urban during 2004 was given by (9) [108]: D = { x : ¬U2004 ( x ) ∧ V ( x ) ∧ f in( X2004 ( x ))} (9) where X is the predictor vector at epoch t, and fin means that all predictors are finite (e.g., no NaN). In order to avoid misprediction and address class imbalances, negative classes were randomly down-sampled to a ratio of ≤3:1 (negative: positive) [109]. 2.5.3. Prediction of Future Scenarios and Evaluation Metrics After the described operation, the study area was then divided into 64 × 64 pixel tiles (at 90 m/pixel). Then a 60/20/20 split by tiles was set, defined as: (i) training; (ii) calibration; (iii) and spatial hold-out test. A histogram-based Gradient Boosting Machine (GBM) [110] φθ was trained on D. Probabilities were calibrated by an isotonic regression ϱiso applied to calibration tiles (10): ρ04 → ρ14 ( x ) = ϱiso ( φθ ( X2004 ( x ))) (10) where ρ is the calibrated transition probability. The GBM outputs were passed through the isotonic calibrator to obtain calibrated propensity maps for 2004 to 2014, 2014 to 2024, and 2024 to 2034. Pixel-wise propensities were converted to binary change maps through a regionalized Top-K allocator [22]. The domain was tiled; each tile received a quota proportional to its Land 2025, 14, 2233 12 of 25 historical share of change. Within each tile, the top-ranked non-urban pixels were selected subject to eligibility constraints (validity mask, slope ≤ 15◦ , distance to water > 50 m, and non-urban status at the start year). This reduces over-concentration in single hotspots and preserves mesoscale heterogeneity consistent with observed patterns. Several evaluation criteria were applied to understand the results obtained by the model. Ranking quality, probabilistic reliability, spatial allocation approach and quantity and scale sensitivity approach were used: (i) Figure of Merit (FoM) on change (strict and buffered with a one-cell tolerance); (ii) Pontius decomposition of error into quantity and allocation disagreement; (iii) ROC–AUC (Area Under the ROC Curve) and PR (PrecisionRecall)–AP (Average Precision) computed over the 2014 non-urban evaluation domain; (iv) reliability via Brier score and calibration plots; (v) multi-scale scoring by majorityaggregating maps to 3 × 3 and 5 × 5 windows; and (vi) spatial block bootstrap (64 × 64 tiles, 300 replicates) to derive 95% confidence intervals for FoM and AUC [95,107,108,111–113]. 2.5.4. Business-as-Usual (BAU) To estimate the change in land use from non-anthropised to anthropised, as a baseline for estimating the “natural/no policy” change, the BAU (Business-as-Usual) method was used for the period 2024–2034. For the calculation, no constraints were imposed on the predictive model for the development of anthropogenic actions, even in the archaeological area of Amrit, to understand how, in the absence of protection measures by the competent authorities, the new anthropogenic fabric could be arranged in the archaeological area and in the surrounding landscape. BAU projection is adopted as a transparent, policy-neutral baseline. The approach anchors the overall amount of future urbanisation to empirically observed non-anthropised rates, rather than to speculative shocks or policy changes, which keeps assumptions verifiable. It also enhances comparability and interpretability, since BAU is a standard reference in land-change research and is readily understood by planners across regions and periods [114]. Methodologically, BAU is parsimonious yet uncertainty-aware: it carries forward what is best supported by data (historical rates and spatial patterns) while explicitly propagating variability (in rates and regional heterogeneity) to maps and area totals. Finally, BAU provides the quantity target and spatial apportionment rules. This modular design limits overfitting and clarifies whether differences in outcomes stem from the probabilistic model or from the scenario’s quantity/allocation assumptions [2,19,102,115–119]. 3. Results 3.1. Analysis of Anthropogenic Impact on the Study Areas The results of the classifier used on Landsat data from the 1980s to the present show high statistical reliability in identifying pixels belonging to the selected classes (Figure 4): (i) (ii) (iii) (iv) Train Accuracy: 0.9935; Train Kappa: 0.9896; Test Accuracy: 0.9367; Test Kappa: 0.8893. The accuracy of the classification and the creation of thematic LULC maps have made it possible to extract useful information about the development of human activity in the area (Figure 5). Anthropisation in the area has increased over the years, counting 200,598 pixels (18,053.82 ha) in 1984, the starting point of the analysis, reaching 885,023 pixels (79,652.07 ha) in 2024, the last year considered (Table 3). Land 2025, 14, 2233 13 of 25 Figure 4. Land Use maps from 1984 to 2024. Respectively: satellite RGB (above), 3-class Land Use Land Cover (below). Figure 5. Land Use Land Cover changes over time: (a) whole study area, (b) zoom on Amrit Site area. Land 2025, 14, 2233 14 of 25 Table 3. Spread of anthropogenic structures over time. Period Pixel Count Hectares (ha) 1984 1984–1994 1984–2004 1984–2014 1984–2024 200,598 366,628 630,827 776,257 885,023 18,053.82 32,996.52 56,774.43 69,863.13 79,652.07 In the area immediately surrounding the site (Figure 5, Amrit Archaeological Site), diachronic analysis of land cover classifications clearly indicates a progressive and accelerated pattern of human expansion over the last three decades. Between 1994 and 2004, the built-up area increased moderately, from approximately 7.0 to 9.2 hectares (+30.8%), reflecting limited and localised development in the vicinity of existing infrastructure. A similar trend continued in the following decade (2004–2014), reaching 11.6 hectares (+26.5%), in line with gradual densification rather than large-scale construction. In contrast, the period 2014–2024 reveals a sharp increase in anthropisation, with the anthropisation pixel class expanding from 11.6 to 51.2 hectares, corresponding to an increase of approximately +39.6 hectares (+341%) within the area surrounding the site. The acceleration of anthropic transformation observed over the last decade marks the transition from slow and linear urbanisation to a more intense and spatially invasive dynamic, increasing direct and indirect pressures on archaeological remains. The use of LULC maps combined with NASADEM and the slope obtained from it has made it possible to understand urban settlement preferences in the area over time. Anthropisation based on terrain confirms a persistent topographical signal. New urbanisation is present at ≤5◦ (particularly 0–2◦ ), well distributed also at 5–10◦ , and underrepresented above 10–15◦ . This monotonic decline with slope is stable over decades and aligns with the model’s learned propensities, which increase near existing urban areas/roads/buildings and decrease with slope (Figure 6). Figure 6. Settlement preferences based on slope, over time: (a) cartographic representation; (b) numerical representation (1984 to 2024). Land 2025, 14, 2233 15 of 25 3.2. Shoreline Changes over Time The performed analyses reveal a substantial change in the shoreline between 1974 and subsequent decades, indicating a major coastal regression after 1974, which remained fairly consistent thereafter. In fact, the shoreline of 1984–1994–2004–2014–2024 shows little to no change. Multiple factors, such as the resolution of the satellite data, could have emphasized this discrepancy. Landsat resolution is 30 m, while high-resolution Hexagon images are within 1 m. Moreover, while Landsat imagery was georeferenced automatically, Hexagon imagery underwent a manual georeferencing process, which only partially allowed for rectifying the distortion of the image due to the lens and relative position of the camera when acquired. Nevertheless, the results show a high standard deviation. The standard deviation is a statistical measure of data variability, indicating how spread out individual data points are from the mean of a dataset. A low standard deviation indicates that values are clustered around the mean, while a high standard deviation suggests that values are more spread out over a wider range. The results show changes indicating an unsteady process, with a strong coastal regression between 1974 and the following years (calculated between 0 and 60 m). The coastal changes after 1974 also involve the construction of additional minor harbor structures along the coast, south of the main jetty, which is the segment interested by the major regression (Figure 7). Figure 7. Cont. Land 2025, 14, 2233 16 of 25 Figure 7. Shoreline changes over time: (a,b) shoreline changes derived from KH and Landsat; (c) transect changes; (d) OLS and EPR values. 3.3. Results of Predictive Model Out-of-sample validation over the period 2014–2024, using a model trained over the period 2004–2014, highlights the high reliability of the change mapping pipeline based on machine learning. The main indicator sensitive to quantity and suitable for evaluating rare events, the Figure of Merit (FoM), calculated on Top-K quantity-matched predictions, reaches a value of FoM = 0.157 for the natural to urban transition, when OSM street map data, built-up area, and hydrography are included as predictors in the model. Consistent with this result, the diagnostic metrics of ranking and calibration (ROC curves, Precision-Recall, and reliability diagrams) indicate a clear separability of classes and good calibration of predicted probabilities. The residual error is mainly of the allocative type (i.e., in the location of the change) rather than quantitative (i.e., in the extent of the change), as expected from protocols that set the quantity to be evaluated a priori. Multiscale checks, conducted using 3 × 3 and 5 × 5 majority windows, determine an increase in FoM, suggesting that much of the inconsistency between predicted and observed occurs at the local or sub-pixel scale, and not due to systemic patterns. Based on these results, a Business-as-Usual (BAU) projection was made for the period 2024–2034, anchoring the amount of urban transformation to the empirically observed natural to urban transitions on the 90 m grid, and allocating the changes according to the calibrated propensities derived from the model, constrained by an explicit eligibility mask (non-urban pixels in 2024, slope ≤ 15◦ , exclusion of coastal or hydrographic areas where relevant). The BAU analysis predicted a transformation rate of 0.05849, implying 10,711 pixels, or approximately 8676 ha. The target is fully compatible with the spatial constraints of Land 2025, 14, 2233 17 of 25 the eligibility mask. The spatial distribution of BAU transitions is concentrated almost exclusively in areas with a slope ≤ 5◦ , located near existing urban areas and along OSM infrastructure (roads and buildings), delineating consistent marginal expansion fronts, with negligible extension to steep areas (Figure 8). Figure 8. BAU results: (a) predicted 2004–2014; (b) predicted 2014–2024; (c) predicted 2024–2034; (d) 2034 new buildings; (e) Land Use in 2024 and predicted new buildings. 4. Discussion The results obtained from satellite data and ancillary data show a coherent and physically plausible picture of urbanisation processes in the study area, with monotonous and path-dependent growth along the edges of built-up areas and infrastructure corridors. Quantitatively, natural to urban conversions at 30 m are relatively stable over time, indicating a process of incremental expansion rather than episodic expansion. This trajectory is compatible with the progressive erosion of the highly accessible non-urban inventory and with a topographical selection that systematically favours areas with low slopes. Similarly, the anthropisation of the area seems to have a direct and indirect impact on the coastline, with noticeable changes observable close to urban centres (e.g., industrial and harbor areas), as well as downstream of the main watercourses for reasons of land reclamation linked to agriculture or the creation of centres close to the coast. These results underscore the diagnostic value of long-term pixel-based monitoring for heritage protection. The quantified increase in built-up areas provides a measurable indicator of the risk of encroachment, supporting the delimitation of priority buffer zones and the development of preventive conservation protocols aimed at mitigating further urban intrusions within the Amrit cultural landscape. Land 2025, 14, 2233 18 of 25 From the point of view of drivers, the analysis of preference conditioned on availability confirms a monotonic relationship with morphology: over-representation of new urbanisations on ≤5◦ (maximum in classes 0–2◦ ), proximity to proportionality on 5–10◦ , and under-representation beyond 10–15◦ . In parallel, proximity to OSM roads and buildings and to pre-existing urban areas emerges as the dominant suitability gradient. The combination of these factors, accessibility and slope, explains both the spatial consistency of the transitions observed and the distribution of the model’s residual errors along thin fringes of urban areas and linear tracks, where the 30–90 m geometry accentuates the effect of small planimetric deviations. On a predictive level, the ML pipeline validated outside the sample on 2014–2024 shows high reliability in mapping rare events of change: the Figure of Merit (FoM), calculated with quantity-matched Top-K to avoid imbalance bias, reaches FoM = 0.157, while ranking and reliability diagnostics (ROC/PR and Brier) show good separation and calibration of probabilities. The breakdown of the error attributes the prevailing share to allocation (the where), not to quantity (the how much), in line with an evaluation protocol that sets the volume of changes and with a deliberately parsimonious but physically informed set of covariates. Furthermore, multi-scale tests (3 × 3 and 5 × 5 aggregations) improve FoM, indicating that a significant part of the discordance is local/sub-pixel rather than systematic. The obtained Figure of Merit (FoM = 0.157) represents the proportion of correctly predicted transitions among all observed and simulated changes. Although the value may appear numerically modest, it is fully consistent with those typically reported for regional-scale land-change and urban growth models based on medium-resolution satellite data (≈0.10–0.25) [95,107,120]. The FoM quantifies the agreement restricted to change areas only, rather than total map accuracy, and therefore yields lower absolute values than traditional metrics such as overall accuracy or Kappa. In this context, a FoM of 0.157 indicates moderate yet robust predictive reliability, reflecting a realistic performance given the rarity of urban transitions and the 90 m spatial aggregation adopted. Overall, uncertainties are well contained and distributed in space, owing in part to the use of spatially aware (tile-based) validations and topographical/hydrographic eligibility constraints. The Business-as-Usual (BAU) projection for 2024–2034, anchored to recent empirical rates on a 90 m grid and allocated using calibrated propensities within constraints of slope ≤ 15◦ (and coastal/hydraulic exclusions where relevant), produces a realistic and traceable projected volume: 86.38–86.76 km2 . The feasibility of the target is guaranteed by the 2024 non-urban inventory within the constraints, while the spatial distribution of BAU focuses almost exclusively on highly accessible plains, forming perimeter expansion fronts consistent with historical dynamics. The spatial patterns of urban growth observed along the Syrian coastal plain surrounding Amrit reflect distinct historical and socio-economic phases. During the 1980s, the region experienced limited and localised development, largely confined to infrastructure improvements and small-scale agricultural settlements under centralised planning policies. Urbanisation remained moderate and the archaeological landscape retained much of its traditional spatial coherence. In the 2000s, gradual economic liberalisation, improved water and infrastructure management, and population growth in the province of Tartous stimulated increased land conversion along major transport corridors and the coastal strip. This period saw the consolidation of suburban centres and the first signs of pressure on archaeologically sensitive areas. The intensification of agriculture, particularly the expansion of irrigated fields, contributed to the progressive transformation of the coastal plain, indirectly reshaping the drainage systems and micro-topographies associated with ancient sites. The 2010s, however, marked a turning point. The combined effects of the Syrian Land 2025, 14, 2233 19 of 25 conflict, internal displacement and uncontrolled construction activity in peri-urban areas have generated a wave of rapid and largely unregulated construction. Evidence from remote sensing from 2014 to 2024 clearly reveals this acceleration, with an increase in built-up area. This expansion is not only a demographic phenomenon, but a structural transformation linked to changes in governance, land ownership and the collapse of urban planning mechanisms. Combined analysis of these phases highlights that the recent expansion is both the result of long-term demographic pressure and the consequence of short-term socio-political instability. 5. Conclusions This study demonstrates how multi-resolution and multi-temporal satellite remote sensing, integrated with advanced machine learning techniques and ancillary geospatial data, provides a robust and reproducible framework for analyzing anthropogenic pressures on cultural landscapes. Applied to the case study of Amrit within the MapDam project, the approach has revealed both the long-term trajectories of land-use change and the specific morphological and infrastructural drivers guiding urban expansion. The consistency of settlement preferences with slope and accessibility factors, together with the high predictive reliability of the calibrated models, underscores the capacity of the methodology to capture physically meaningful dynamics of landscape transformation. Equally significant is the clear evidence of coastal regression and localized shoreline modifications connected to human activities. These findings highlight the interdependence between terrestrial and coastal processes, stressing the need for integrated monitoring strategies that recognize the multiple scales at which cultural heritage is threatened. Beyond the specific results obtained for the Amrit area, the proposed workflow demonstrates strong potential for replication in other archaeological and cultural landscapes. Its modular design, based on open-access satellite archives, ancillary geospatial data, and reproducible Python and Google Earth Engine scripts, allows for adaptation to sites with different environmental, geomorphological, and historical settings. By adjusting input predictors (e.g., slope thresholds, proximity to roads, land-cover typologies) and spatial constraints, the same framework can be readily transferred to monitor urban pressure, land reclamation, or environmental degradation in other regions of the eastern Mediterranean or beyond. Nevertheless, several limitations should be acknowledged, such as (i) the spatial resolution of Landsat imagery (30 m) and the aggregation to 90 m inevitably reduce the sensitivity to small-scale or informal urban growth, which may be critical in heritage contexts; (ii) the scarcity of ground-truth data and the reliance on historical satellite archives constrain the capacity to validate transitions in areas affected by cloud cover, conflict, or limited accessibility; (iii) while the probabilistic model captures large-scale tendencies, it does not yet integrate socio-economic or policy variables that could refine scenario realism. Future developments should therefore combine higher-resolution datasets (e.g., Sentinel-2, PlanetScope, UAV imagery) and more context-specific drivers to enhance both spatial precision and interpretative depth. In addition, the BAU predictive model is conceptually stimulating, but although such models may prove useful for urban planning, engineering or regional development, their predictive validity for archaeological risk is limited, particularly in areas without comprehensive and up-to-date archaeological inventories. Furthermore, this approach can be integrated with predictive modelling techniques designed to simulate and anticipate the long-term evolution of archaeological landscapes [121–123]. By combining spatial analysis, remote sensing data, and computational modelling [124], it becomes possible to identify zones of potential archaeological sensitivity and areas most at risk from contemporary urban expansion. This integration supports Land 2025, 14, 2233 20 of 25 a forward-looking management perspective, enabling the definition of dynamic buffer zones that evolve in response to environmental and anthropogenic pressures. In doing so, predictive mapping becomes a practical tool for preventive archaeology and for the sustainable planning of heritage-rich territories. Despite these limitations, the approach offers a transparent and transferable template for data-driven monitoring of cultural landscapes, enabling heritage managers and planners to anticipate anthropogenic risks through reproducible and cost-effective methods. Particularly relevant is the applicability of this approach in politically unstable or high-risk contexts, such as zones of armed conflict or areas where governments are unable, or in some cases unwilling, to invest in the protection of archaeological and cultural assets. Author Contributions: Conceptualization, N.A., A.F., T.P.; methodology, N.A., G.C., S.E.Z., D.R.; software, N.A., A.F., G.C., S.E.Z.; validation, N.A., A.F., M.S., N.M., R.L.; formal analysis, N.A., A.F., and D.R.; investigation, N.A., T.P., M.P.; data curation, N.A., M.S.; writing—original draft preparation, N.A.; writing—review and editing, N.A., A.F., D.R., S.E.Z., G.C., M.S., N.M., R.L., T.P.; visualization, N.A., D.R.; supervision, T.P., M.P.; project administration, T.P., M.P.; funding acquisition, T.P., M.P. All authors have read and agreed to the published version of the manuscript. Funding: This research was funded by European Union—Next Generation EU, Missione 4 Componente 1 CUP B53D23001340006. Data Availability Statement: The raw data supporting the conclusions of this article will be made available by the authors on request. Conflicts of Interest: The authors declare no conflicts of interest. References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. Wang, J.; Wang, X.; Han, Z. Study on the Impact of Urbanization on Climate and Surrounding Habitat Quality of World Cultural Heritage Sites—A Case Study of Qingfu Tomb. Environ. Monit. Assess. 2025, 197, 1049. [CrossRef] Agapiou, A.; Alexakis, D.D.; Lysandrou, V.; Sarris, A.; Cuca, B.; Themistocleous, K.; Hadjimitsis, D.G. Impact of Urban Sprawl to Cultural Heritage Monuments: The Case Study of Paphos Area in Cyprus. J. Cult. Herit. 2015, 16, 671–680. [CrossRef] Luo, L.; Wang, X.; Guo, H. Transitioning from Remote Sensing Archaeology to Space Archaeology: Towards a Paradigm Shift. Remote Sens. Environ. 2024, 308, 114200. [CrossRef] Cuca, B.; Zaina, F.; Tapete, D. Monitoring of Damages to Cultural Heritage across Europe Using Remote Sensing and Earth Observation: Assessment of Scientific and Grey Literature. Remote Sens. 2023, 15, 3748. [CrossRef] Dana Negula, I.; Moise, C.; Lazăr, A.M.; Ris, cut, a, N.C.; Cristescu, C.; Dedulescu, A.L.; Mihalache, C.E.; Badea, A. Satellite Remote Sensing for the Analysis of the Micia and Germisara Archaeological Sites. Remote Sens. 2020, 12, 2003. [CrossRef] Frodella, W.; Elashvili, M.; Spizzichino, D.; Gigli, G.; Adikashvili, L.; Vacheishvili, N.; Kirkitadze, G.; Nadaraia, A.; Margottini, C.; Casagli, N. Combining InfraRed Thermography and UAV Digital Photogrammetry for the Protection and Conservation of Rupestrian Cultural Heritage Sites in Georgia: A Methodological Application. Remote Sens. 2020, 12, 892. [CrossRef] Tzouvaras, M.; Kouhartsiouk, D.; Agapiou, A.; Danezis, C.; Hadjimitsis, D.G. The Use of Sentinel-1 Synthetic Aperture Radar (SAR) Images and Open-Source Software for Cultural Heritage: An Example from Paphos Area in Cyprus for Mapping Landscape Changes after a 5.6 Magnitude Earthquake. Remote Sens. 2019, 11, 1766. [CrossRef] Cigna, F.; Lasaponara, R.; Masini, N.; Milillo, P.; Tapete, D. Persistent Scatterer Interferometry Processing of COSMO-SkyMed StripMap HIMAGE Time Series to Depict Deformation of the Historic Centre of Rome, Italy. Remote Sens. 2014, 6, 12593–12618. [CrossRef] Crosetto, M.; Monserrat, O.; Cuevas-González, M.; Devanthéry, N.; Crippa, B. Persistent Scatterer Interferometry: A Review. ISPRS J. Photogramm. Remote Sens. 2016, 115, 78–89. [CrossRef] Delgado Blasco, J.M.; Foumelis, M.; Stewart, C.; Hooper, A. Measuring Urban Subsidence in the Rome Metropolitan Area (Italy) with Sentinel-1 SNAP-StaMPS Persistent Scatterer Interferometry. Remote Sens. 2019, 11, 129. [CrossRef] Elfadaly, A.; Abutaleb, K.; Naguib, D.M.; Mostafa, W.; Abouarab, M.A.R.; Ashmawy, A.; Wilson, P.; Lasaponara, R. Tracking the Effects of the Long-term Changes on the Coastal Archaeological Sites of the Mediterranean Using Remote Sensing Data: The Case Study from the Northern Shoreline of Nile Delta of Egypt. Archaeol. Prospect. 2023, 30, 369–390. [CrossRef] Abdelhady, H.U.; Troy, C.D.; Habib, A.; Manish, R. A Simple, Fully Automated Shoreline Detection Algorithm for High-Resolution Multi-Spectral Imagery. Remote Sens. 2022, 14, 557. [CrossRef] Land 2025, 14, 2233 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 21 of 25 Himmelstoss, E.A.; Henderson, R.E.; Kratzmann, M.G.; Farris, A.S. Digital Shoreline Analysis System (DSAS) Version 5.0 User Guide; Open-File Report 2018–1179; U.S. Geological Survey: Reston, VA, USA, 2018. Mullick, M.R.A.; Islam, K.M.A.; Tanim, A.H. Shoreline Change Assessment Using Geospatial Tools: A Study on the Ganges Deltaic Coast of Bangladesh. Earth Sci. Inform. 2020, 13, 299–316. [CrossRef] Tamassoki, E.; Amiri, H.; Soleymani, Z. Monitoring of Shoreline Changes Using Remote Sensing (Case Study: Coastal City of Bandar Abbas). IOP Conf. Ser. Earth Environ. Sci. 2014, 20, 012023. [CrossRef] Tyagi, S.; Rai, S.C. Monitoring Shoreline Changes along Andhra Coast of India Using Remote Sensing and Geographic Information System. Indian J. Geo Mar. Sci. 2020, 49, 218–224. Guariglia, A.; Buonamassa, A.; Losurdo, A.; Saladino, R.; Trivigno, M.L.; Zaccagnino, A.; Colangelo, A. A Multisource Approach for Coastline Mapping and Identification of Shoreline Changes. Ann. Geophys. 2006, 49, 295–304. [CrossRef] Nolesini, T.; Frodella, W.; Bianchini, S.; Casagli, N. Detecting Slope and Urban Potential Unstable Areas by Means of MultiPlatform Remote Sensing Techniques: The Volterra (Italy) Case Study. Remote Sens. 2016, 8, 746. [CrossRef] Agapiou, A. Remote Sensing Heritage in a Petabyte-Scale: Satellite Data and Heritage Earth Engine© Applications. Int. J. Digit. Earth 2017, 10, 85–102. [CrossRef] Amini, S.; Saber, M.; Rabiei-Dastjerdi, H.; Homayouni, S. Urban Land Use and Land Cover Change Analysis Using Random Forest Classification of Landsat Time Series. Remote Sens. 2022, 14, 2654. [CrossRef] Elfadaly, A.; Attia, W.; Qelichi, M.M.; Murgante, B.; Lasaponara, R. Management of Cultural Heritage Sites Using Remote Sensing Indices and Spatial Analysis Techniques. Surv. Geophys. 2018, 39, 1347–1377. [CrossRef] Liu, J.; Ren, H.; Wang, X.; Shirazi, Z.; Quan, B. Measuring and Predicting Urban Expansion in the Angkor Region of Cambodia. Remote Sens. 2019, 11, 2064. [CrossRef] Lasaponara, R.; Abate, N.; Fattore, C.; Masini, N. Open Big Earth Observation Data and Artificial Intelligence for the Study and Preservation of UNESCO Natural and Cultural Heritage: The Case of Machu Picchu. In Machu Picchu in Context; Ziółkowski, M., Masini, N., Bastante, J.M., Eds.; Springer International Publishing: Cham, Switzerland, 2022; pp. 239–264, ISBN 978-3-030-92765-3. Tapete, D.; Cigna, F. InSAR Data for Geohazard Assessment in UNESCO World Heritage Sites: State-of-the-Art and Perspectives in the Copernicus Era. Int. J. Appl. Earth Obs. Geoinf. 2017, 63, 24–32. [CrossRef] Tapete, D.; Cigna, F.; Donoghue, D.N.M. ‘Looting Marks’ in Space-Borne SAR Imagery: Measuring Rates of Archaeological Looting in Apamea (Syria) with TerraSAR-X Staring Spotlight. Remote Sens. Environ. 2016, 178, 42–58. [CrossRef] Tapete, D. Remote Sensing and Geosciences for Archaeology. Geosciences 2018, 8, 41. [CrossRef] European Environment Agency. CORINE Land Cover 2018 (Raster 100 m), Europe, 6-Yearly—Version 2020_20u1, 2020. Available online: https://land.copernicus.eu/en/products/corine-land-cover/clc2018 (accessed on 4 November 2025). Li, Z.; He, W.; Li, J.; Lu, F.; Zhang, H. Learning without Exact Guidance: Updating Large-Scale High-Resolution Land Cover Maps from Low-Resolution Historical Labels 2024. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, WA, USA, 16–22 June 2024. Benabbou, A. Land Use Suitability Analysis Using an Approach Combining GIS, Regionalization, and PROMETHEE II. Ann. GIS 2024, 30, 121–136. [CrossRef] Deines, J.M.; Kendall, A.D.; Crowley, M.A.; Rapp, J.; Cardille, J.A.; Hyndman, D.W. Mapping Three Decades of Annual Irrigation across the US High Plains Aquifer Using Landsat and Google Earth Engine. Remote Sens. Environ. 2019, 233, 111400. [CrossRef] Wang, J.; Bretz, M.; Dewan, M.A.A.; Delavar, M.A. Machine Learning in Modelling Land-Use and Land Cover-Change (LULCC): Current Status, Challenges and Prospects. Sci. Total Environ. 2022, 822, 153559. [CrossRef] [PubMed] Jagannathan, J.; Divya, C. Deep Learning for the Prediction and Classification of Land Use and Land Cover Changes Using Deep Convolutional Neural Network. Ecol. Inform. 2021, 65, 101412. [CrossRef] Ebenezer, P.A.; Manohar, S. Land Use/Land Cover Change Classification and Prediction Using Deep Learning Approaches. Signal Image Video Process. 2024, 18, 223–232. [CrossRef] Agapiou, A.; Alexakis, D.D.; Hadjimitsis, D.G. Potential of Virtual Earth Observation Constellations in Archaeological Research. Sensors 2019, 19, 4066. [CrossRef] Rayne, L.; Bradbury, J.; Mattingly, D.; Philip, G.; Bewley, R.; Wilson, A. From Above and on the Ground: Geospatial Methods for Recording Endangered Archaeology in the Middle East and North Africa. Geosciences 2017, 7, 100. [CrossRef] Agapiou, A.; Lysandrou, V.; Hadjimitsis, D.G. Earth Observation Contribution to Cultural Heritage Disaster Risk Management: Case Study of Eastern Mediterranean Open Air Archaeological Monuments and Sites. Remote Sens. 2020, 12, 1330. [CrossRef] Al Kassem, A.; Hopper, K.; Ibraheem, I.M.; el Hajj, H.; Maier, A.; Ali, A.A.; Richter, J. Assessment of the Impact of the Syrian Conflict on Archaeological Sites in the Daraa Region. Levant 2024, 56, 425–446. [CrossRef] Cil, D.; Bassett, H.F.; Aronson, J.; Gunter-Bassett, M.; Welsh, W.; Hanson, K.; Wegener, C.; Daniels, B.I. Detecting Damage to Cultural Heritage during Armed Conflict Using the Case of Ukraine. Natl. Ethn. Polit. 2025, 31, 177–188. [CrossRef] Land 2025, 14, 2233 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 22 of 25 Sakai, M.; Freitag, M.; Sakurai, A.; Albrecht, C.M.; Hamann, H.F. Climatic & Anthropogenic Hazards to the Nasca World Heritage: Application of Remote Sensing, AI, and Flood Modelling. In Proceedings of the IEEE International Geoscience and Remote Sensing Symposium, Athens, Greece, 7–12 July 2024. Chavadi, B. From Sky to Screen: The Role of Remote Sensing and Digital Reconstructions in Preserving Endangered Archaeological Heritage. J. Res. Dev. 2025, 17, 94–98. Conolly, J.; Lake, M. Geographical Information Systems in Archaeology, 1st ed.; Cambridge University Press: Cambridge, UK, 2006; ISBN 978-0-521-79330-8. Dunand, M.; Saliby, N. Le Temple d’Amrith dans la Peree d’Aradus; Librairie Orientaliste Paul Geuthner: Paris, France, 1985; ISBN 978-2-7053-0259-7. Besançon, J.; Copeland, L.; Muhesen, S.; Sanlaville, P. Prospection Géomorphologique Et Préhistorique Dans La Région De Tartous (Syrie). Paléorient 1994, 20, 5–19. [CrossRef] Al Maqdissi, M.; Benech, C. The Spatial Organization of the Phoenician City of Amrith (Syria). ArchéoSciences 2009, 33, 209–211. [CrossRef] Saliby, N. Amrit. In Archéologie et Histoire de la Syria II: La Syrie de L’époqueachéménide à L’avènement de l’Islam; Dentzer, J.M., Orthmann, W., Eds.; Saarbrücker Druckerei und Verlag: Saarbrücker, Germany, 1989. Dunand, M. Recherches Archéologiques Dans La Région de Marathus, Note Préliminaire. Ann. Archéologiques Syr. 1953, 3, 165–170. Braidwood, R.J. Report on Two Sondages on the Coast of Syria, South of Tartous. Syr. Archéologie Art Hist. 1940, 21, 183–226. [CrossRef] Nunn, A. Michel Al-Maqdissi. Note d’Archéologie Levantine. XXI. Travaux Archéologiques à Tell Ghamqa Au Sud de Tartous En 1926. Abstr. Iran. 2013, 32–33, 100. [CrossRef] Chiti, L.; Pedrazzi, T. Tell Kazel (Syria), Area II. New Evidence from a Late Bronze/Iron Age Quarter. In Proceedings of the 8th International Congress of the Archaeology of Ancient Near East, 30 April–4 May 2012, University of Warsaw; Bielinski, P., Gawlikowski, M., Kolinski, R., Lawecka, D., Soltysiak, A., Wygnanska, Z., Eds.; Harrassowitz Verlag: Wiesbaden, Germany, 2014; pp. 67–84. Sader, H.S. The History and Archaeology of Phoenicia; Archaeology and Biblical Studies; SBL Press: Atlanta, GA, USA, 2019; ISBN 978-0-88414-405-2. Martinez-Sève, L. Corinne Bonnet, Les Enfants de Cadmos. Le Paysage Religieux de La Phénicie Hellénistique. Syria Archéologie Art Et Hist. 2017, 94, 408–411. [CrossRef] Garbati, G.; Pedrazzi, T. Eaux et cultes en Phénicie à l’époque perse. In Ancient Waterlands; Robinson, B.A., Bouffier, S., Fumadó Ortega, I., Eds.; Archéologies Méditerranéennes; Presses Universitaires de Provence: Aix-en-Provence, France, 2019; pp. 213–224, ISBN 979-10-320-0239-1. Al-Maqdissi, M.; Ishaq, E. Amrith in the Late Phoenician Period. Ash-Sharq Bull. Anc. East Archaeol. Hist. Soc. Stud. 2017, 1, 1–9. Rey-Coquais, J.-P. Arados et Sa Pérée Aux Époques Grecque, Romaine et Byzantine: Recueil Des Témoignages Littéraires Anciens, Suivi de Recherches Sur Les Sites, l’histoire, La Civilisation. Bibl. Archéologique Hist. 1974, 97, 24. Agapiou, A.; Alexakis, D.D.; Hadjimitsis, D.G. Spectral Sensitivity of ALOS, ASTER, IKONOS, LANDSAT and SPOT Satellite Imagery Intended for the Detection of Archaeological Crop Marks. Int. J. Digit. Earth 2014, 7, 351–372. [CrossRef] Agapiou, A.; Hadjimitsis, D.G.; Themistocleous, K.; Papadavid, G.; Toulios, L. Detection of Archaeological Crop Marks in Cyprus Using Vegetation Indices from Landsat TM/ETM+ Satellite Images and Field Spectroscopy Measurements; Michel, U., Civco, D.L., Eds.; SPIE Remote Sensing: Toulouse, France, 7 October 2010; p. 78310V. Acharya, T.D.; Wang, I. Exploring Landsat 8. Int. J. IT Eng. Appl. Sci. Res. IJIEASR 2015, 4, 4–10. Arekhi, M.; Goksel, C.; Balik Sanli, F.; Senel, G. Comparative Evaluation of the Spectral and Spatial Consistency of Sentinel-2 and Landsat-8 OLI Data for Igneada Longos Forest. ISPRS Int. J. Geo-Inf. 2019, 8, 56. [CrossRef] Bishop-Taylor, R.; Nanson, R.; Sagar, S.; Lymburner, L. Mapping Australia’s Dynamic Coastline at Mean Sea Level Using Three Decades of Landsat Imagery. Remote Sens. Environ. 2021, 267, 112734. [CrossRef] Du, Z.; Li, W.; Zhou, D.; Tian, L.; Ling, F.; Wang, H.; Gui, Y.; Sun, B. Analysis of Landsat-8 OLI Imagery for Land Surface Water Mapping. Remote Sens. Lett. 2014, 5, 672–681. [CrossRef] Nasiri, V.; Deljouei, A.; Moradi, F.; Sadeghi, S.M.M.; Borz, S. Land Use and Land Cover Mapping Using Sentinel-2, Landsat-8 Satellite Images, and Google Earth Engine: A Comparison of Two Composition Methods. Remote Sens. 2022, 14, 1977. [CrossRef] Hemati, M.; Hasanlou, M.; Mahdianpari, M.; Mohammadimanesh, F. A Systematic Review of Landsat Data for Change Detection Applications: 50 Years of Monitoring the Earth. Remote Sens. 2021, 13, 2869. [CrossRef] Fowler, M.J.F. Modelling the Acquisition Times of CORONA Satellite Photographs: Accuracy and Application. Int. J. Remote Sens. 2011, 32, 8865–8879. [CrossRef] Land 2025, 14, 2233 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. 86. 87. 88. 89. 23 of 25 Agapiou, A.; Alexakis, D.D.; Sarris, A.; Hadjimitsis, D.G. Colour to Greyscale Pixels: Re-Seeing Greyscale Archived Aerial Photographs and Declassified Satellite CORONA Images Based on Image Fusion Techniques: Colour to Greyscale Pixels Based on Image Fusion Techniques. Archaeol. Prospect. 2016, 23, 231–241. [CrossRef] Casana, J.; Cothren, J. Stereo Analysis, DEM Extraction and Orthorectification of CORONA Satellite Imagery: Archaeological Applications from the Near East. Antiquity 2008, 82, 732–749. [CrossRef] Beck, A.; Philip, G.; Abdulkarim, M.; Donoghue, D. Evaluation of Corona and Ikonos High Resolution Satellite Imagery for Archaeological Prospection in Western Syria. Antiquity 2007, 81, 161–175. [CrossRef] MacDonald, R.A. CORONA: Success for Space Reconnaissance, A Look into the Cold War, and a Revolution for Intelligence. Photogramm. Eng. Remote Sens. 1995, 61, 689–720. Lasaponara, R.; Yang, R.; Chen, F.; Li, X.; Masini, N. Corona Satellite Pictures for Archaeological Studies: A Review and Application to the Lost Forbidden City of the Han–Wei Dynasties. Surv. Geophys. 2018, 39, 1303–1322. [CrossRef] Philip, G.; Donoghue, D.; Beck, A.; Galiatsatos, N. CORONA Satellite Photography: An Archaeological Application from the Middle East. Antiquity 2002, 76, 109–118. [CrossRef] OpenStreetMap Wiki. Available online: https://wiki.openstreetmap.org/wiki/Main_Page (accessed on 4 September 2025). Fitri, M.S.N.; Marena, O.; Hisam, O.A.; Hafiz, M.Y.M.; Izzati, A.K.N. Suitability of Open Street Map (OSM) for 1:50,000 Topographic Map. IOP Conf. Ser. Earth Environ. Sci. 2022, 1051, 012012. [CrossRef] Barrington-Leigh, C.; Millard-Ball, A. The World’s User-Generated Road Map Is More than 80% Complete. PLoS ONE 2017, 12, e0180698. [CrossRef] Bubenzer, O.; Bolten, A. The Use of New Elevation Data (SRTM/ASTER) for the Detection and Morphometric Quantification of Pleistocene Megadunes (Draa) in the Eastern Sahara and the Southern Namib. Geomorphology 2008, 102, 221–231. [CrossRef] Li, R.; Li, H.; Hao, T.; Qiao, G.; Cui, H.; He, Y.; Hai, G.; Xie, H.; Cheng, Y.; Li, B. Assessment of ICESat-2 Ice Surface Elevations over the Chinese Antarctic Research Expedition (CHINARE) Route, East Antarctica, Based on Coordinated Multi-Sensor Observations. Cryosphere 2021, 15, 3083–3099. [CrossRef] Brett, G.; Gardiner, N.; Rack, W. ICESat-2 Satellite Altimetry Measurement of Ice Shelf Water Influenced Fast Ice in Terra Nova Bay; ESS Open Archive, 2024. Available online: https://essopenarchive.org/users/563895/articles/970786-icesat-2-satellite-altimetrymeasurement-of-ice-shelf-water-influenced-fast-ice-in-terra-nova-bay?commit=0d5f889c1de806f94ffe059acc7bb56fc3df0087 (accessed on 4 November 2025). [CrossRef] NASA JPL NASADEM Merged DEM Global 1 Arc Second V001. 2020. Available online: https://cmr.earthdata.nasa.gov/search /concepts/C2763264762-LPCLOUD.html (accessed on 4 November 2025). NASADEM: NASA 30 m Digital Elevation Model|Earth Engine Data Catalog. Available online: https://developers.google.com /earth-engine/datasets/catalog/NASA_NASADEM_HGT_001?hl=it (accessed on 4 September 2025). Gorelick, N. Google Earth Engine; American Geophysical Union: Vienna, Austria, 2013; p. 11997. Kumar, S.; Arya, S. Change Detection Techniques for Land Cover Change Analysis Using Spatial Datasets: A Review. Remote Sens. Earth Syst. Sci. 2021, 4, 172–185. [CrossRef] Mutanga, O.; Kumar, L. Google Earth Engine Applications. Remote Sens. 2019, 11, 591. [CrossRef] Abate, N.; Lasaponara, R. Preventive Archaeology Based on Open Remote Sensing Data and Tools: The Cases of Sant’Arsenio (SA) and Foggia (FG), Italy. Sustainability 2019, 11, 4145. [CrossRef] Amani, M.; Ghorbanian, A.; Ahmadi, S.A.; Kakooei, M.; Moghimi, A.; Mirmazloumi, S.M.; Moghaddam, S.H.A.; Mahdavi, S.; Ghahremanloo, M.; Parsian, S.; et al. Google Earth Engine Cloud Computing Platform for Remote Sensing Big Data Applications: A Comprehensive Review. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2020, 13, 5326–5350. [CrossRef] Lasaponara, R.; Abate, N.; Masini, N. On the Use of Google Earth Engine and Sentinel Data to Detect “Lost” Sections of Ancient Roads. The Case of Via Appia. IEEE Geosci. Remote Sens. Lett. 2021, 19, 3001605. [CrossRef] Breiman, L. Random Forests. Mach. Learn. 2001, 45, 5–32. [CrossRef] Pervez, S.; McNally, A.; Arsenault, K.; Budde, M.; Rowland, J. Vegetation Monitoring Optimization with Normalized Difference Vegetation Index and Evapotranspiration Using Remote Sensing Measurements and Land Surface Models Over East Africa. Front. Clim. 2021, 3, 589981. [CrossRef] Huete, A.R. A Soil-Adjusted Vegetation Index (SAVI). Remote Sens. Environ. 1988, 25, 295–309. [CrossRef] Ceccato, P.; Flasse, S.; Grégoire, J.-M. Designing a Spectral Index to Estimate Vegetation Water Content from Remote Sensing Data: Part 2. Validation and Applications. Remote Sens. Environ. 2002, 82, 198–207. [CrossRef] Sithole, G.; Vosselman, G. Experimental Comparison of Filter Algorithms for Bare-Earth Extraction from Airborne Laser Scanning Point Clouds. ISPRS J. Photogramm. Remote Sens. 2004, 59, 85–101. [CrossRef] McFeeters, S.K. The Use of the Normalized Difference Water Index (NDWI) in the Delineation of Open Water Features. Int. J. Remote Sens. 1996, 17, 1425–1432. [CrossRef] Land 2025, 14, 2233 90. 24 of 25 Yuan, Y.; Lin, L.; Chen, J.; Sahli, H.; Chen, Y.; Wang, C.; Wu, B. A New Framework for Modelling and Monitoring the Conversion of Cultivated Land to Built-up Land Based on a Hierarchical Hidden Semi-Markov Model Using Satellite Image Time Series. Remote Sens. 2019, 11, 210. [CrossRef] 91. Probst, P.; Wright, M.N.; Boulesteix, A.-L. Hyperparameters and Tuning Strategies for Random Forest. WIREs Data Min. Knowl. Discov. 2019, 9, e1301. [CrossRef] 92. Wulder, M.A.; Loveland, T.R.; Roy, D.P.; Crawford, C.J.; Masek, J.G.; Woodcock, C.E.; Allen, R.G.; Anderson, M.C.; Belward, A.S.; Cohen, W.B.; et al. Current Status of Landsat Program, Science, and Applications. Remote Sens. Environ. 2019, 225, 127–147. [CrossRef] 93. Hunter, J.D. Matplotlib: A 2D Graphics Environment. Comput. Sci. Eng. 2007, 9, 90–95. [CrossRef] 94. McKinney, W. Data Structures for Statistical Computing in Python. In Proceedings of the 9th Python in Science Conference, Austin, TX, USA, 28 June–3 July 2010; pp. 51–56. 95. Pontius, R.G. Statistical Methods to Partition Effects of Quantity and Location during Comparison of Categorical Maps at Multiple Resolutions. Photogramm. Eng. Remote Sens. 2002, 68, 1041–1049. 96. Tellier, L.-N.; Quesnel, F.; Bur, J. Estimating Urban Sprawl Standards by Means of the Urban Metric System. Reg. Sci. Policy Pract. 2024, 16, 100131. [CrossRef] 97. Carot, J.M.; Villalba, A. A Quantitative Model of the City in 15 Minutes for Decision-Making. J. Sustain. Dev. Energy Water Environ. Syst. 2024, 12, 1–12. [CrossRef] 98. Yan, Y.; Zhang, Y.; Sharma, A.; Al-Amri, J.F. Evaluation of Suitability of Urban Land Using GIS Technology. Sustainability 2021, 13, 10521. [CrossRef] 99. Seto, K.C.; Fragkias, M.; Güneralp, B.; Reilly, M.K. A Meta-Analysis of Global Urban Land Expansion. PLoS ONE 2011, 6, e23777. [CrossRef] 100. Chen, C.; Judge, J.; Hulse, D. PyLUSAT: An Open-Source Python Toolkit for GIS-Based Land Use Suitability Analysis. Environ. Model. Softw. 2022, 151, 105362. [CrossRef] 101. Hassan, Z.; Shabbir, R.; Ahmad, S.S.; Malik, A.H.; Aziz, N.; Butt, A.; Erum, S. Dynamics of Land Use and Land Cover Change (LULCC) Using Geospatial Techniques: A Case Study of Islamabad Pakistan. SpringerPlus 2016, 5, 812. [CrossRef] 102. Shoyama, K. Assessment of Land-Use Scenarios at a National Scale Using Intensity Analysis and Figure of Merit Components. Land 2021, 10, 379. [CrossRef] 103. Canty, M.J. Image Analysis, Classification and Change Detection in Remote Sensing: With Algorithms for Python, 4th ed.; CRC Press/Taylor & Francis Group: Boca Raton, FL, USA, 2019; ISBN 978-1-138-61322-5. 104. Pedregosa, F.; Varoquaux, G.; Gramfort, A.; Michel, V.; Thirion, B.; Grisel, O.; Blondel, M.; Prettenhofer, P.; Weiss, R.; Dubourg, V. Scikit-Learn: Machine Learning in Python. J. Mach. Learn. Res. 2011, 12, 2825–2830. 105. Gillies, S. Rasterio: Geospatial Raster I/O for Python Programmers. 2013. Available online: https://sgillies.net/2013/11/24 /introducing-rasterio.html (accessed on 4 November 2025). 106. Jordahl, K.; Van den Bossche, J.; Wasserman, J.; McBride, J.; Fleischmann, M.; Gerard, J.; Bilogur, A. GeoPandas: Python Tools for Geographic Data. J. Open Source Softw. 2021, 6, 3021. [CrossRef] 107. Pontius, R.G.; Boersma, W.; Castella, J.-C.; Clarke, K.; de Nijs, T.; Dietzel, C.; Duan, Z.; Fotsing, E.; Goldstein, N.; Kok, K.; et al. Comparing the Input, Output, and Validation Maps for Several Models of Land Change. Environ. Plan. B Plan. Des. 2008, 35, 642–663. [CrossRef] 108. Pontius, R.G.; Millones, M. Death to Kappa: Birth of Quantity Disagreement and Allocation Disagreement for Accuracy Assessment. Int. J. Remote Sens. 2011, 32, 4407–4429. [CrossRef] 109. Chawla, N.V.; Bowyer, K.W.; Hall, L.O.; Kegelmeyer, W.P. SMOTE: Synthetic Minority Over-Sampling Technique. J. Artif. Intell. Res. 2002, 16, 321–357. [CrossRef] 110. Friedman, J.H. Greedy Function Approximation: A Gradient Boosting Machine. Ann. Stat. 2001, 29, 1189–1232. [CrossRef] 111. Fawcett, T. An Introduction to ROC Analysis. Pattern Recognit. Lett. 2006, 27, 861–874. [CrossRef] 112. Lahiri, S.N. Resampling Methods for Dependent Data; Springer: Berlin/Heidelberg, Germany, 2003. 113. Davis, J.; Goadrich, M. The Relationship Between Precision-Recall and ROC Curves. In Proceedings of the 23rd International Conference on Machine Learning, Pittsburgh, PA, USA, 25–29 June 2006; pp. 233–240. 114. Hall, C.M. Sustainable Tourism Beyond BAU (Brundtlund as Usual): Shifting from Paradoxical to Relational Thinking? Front. Sustain. Tour. 2022, 1, 927946. [CrossRef] 115. Egger, C.; Plutzar, C.; Mayer, A.; Dullinger, I.; Dullinger, S.; Essl, F.; Gattringer, A.; Bohner, A.; Haberl, H.; Gaube, V. Using the SECLAND Model to Project Future Land-Use until 2050 under Climate and Socioeconomic Change in the LTSER region Eisenwurzen (Austria). Ecol. Econ. 2022, 201, 107577. [CrossRef] 116. Lacher, I.; Fergus, C.; McShea, W.J. Modeling Alternative Future Scenarios for Direct Application in Land Use and Conservation Planning. Conserv. Sci. Pract. 2023, 5, e12940. [CrossRef] Land 2025, 14, 2233 25 of 25 117. Xiao, W.; Mills, J.; Guidi, G.; Rodríguez-Gonzálvez, P.; Gonizzi Barsanti, S.; González-Aguilera, D. Geoinformatics for the Conservation and Promotion of Cultural Heritage in Support of the UN Sustainable Development Goals. ISPRS J. Photogramm. Remote Sens. 2018, 142, 389–406. [CrossRef] 118. Xiao, D.; Lu, L.; Wang, X.; Nitivattananon, V.; Guo, H.; Hui, W. An Urbanization Monitoring Dataset for World Cultural Heritage in the Belt and Road Region. Big Earth Data 2022, 6, 127–140. [CrossRef] 119. Xie, Y.; Yang, R.; Liang, Y.; Li, W.; Chen, F. The Spatial Relationship and Evolution of World Cultural Heritage Sites and Neighbouring Towns. Remote Sens. 2022, 14, 4724. [CrossRef] 120. Evans, D. Airborne Laser Scanning as a Method for Exploring Long-Term Socio-Ecological Dynamics in Cambodia. J. Archaeol. Sci. 2016, 74, 164–175. [CrossRef] 121. Wheatley, D.; Gillings, M. Spatial Technology and Archaeology: The Archaeological Applications of GIS; 1. publ.; Taylor & Francis: London, UK, 2002; ISBN 978-0-415-24640-8. 122. Fernández, P.T. The Application of GIS Technologies in the Roman Period: Spatial Analysis, Historic Modelling and Reconstructions; Bloomsbury Publishing: London, UK, 2024; pp. 1–232. 123. Gillings, M.; Hacıgüzeller, P.; Lock, G. (Eds.) Archaeological Spatial Analysis: A Methodological Guide, 1st ed.; Routledge: London, UK, 2020; ISBN 978-1-351-24385-8. 124. Sarris, A. (Ed.) Best Practices of GeoInformatic Technologies for the Mapping of Archaeolandscapes; Archaeopress Publishing Ltd.: Bicester, UK, 2015; ISBN 978-1-78491-163-8. Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.