Investigating the Evolution of Active Deformation Areas (ADAs) in the Veneto-Friulian Plain Using Multi-Platform SAR Data

Highlights

What are the main findings?

• The 30-year multi-platform SAR data has proven effective in subsidence evolution from strong early deformation to slower, more stable trends.
• ADAFinder’s workflow helped in isolating coherent deformation clusters, and the ADA Quality Index (QI) helped in separating reliable ADAs from noisy ones.

What are the implications of the main findings?

• Subsidence is not purely natural; construction loads significantly amplify subsidence over compressible Holocene deposits.
• The Multi-Platform SAR dataset allowed for consistent tracking of ground motion across decades, providing evidence for prioritizing inspections and mitigation measures.

Abstract

Coastal alluvial plains underlain by unconsolidated deposits are prone to land subsidence, a geohazard that can damage infrastructure and alter drainage patterns. One such example is the Venetian–Friulian coastal plain (NE Italy), where natural sediment compaction and anthropogenic activities have led to ground deformation across multiple zones. From this perspective, this study presents a 30-year analysis of land subsidence across the Venetian–Friulian plain, particularly highlighting municipalities such as Portogruaro, Concordia Sagittaria, San Stino di Livenza, Eraclea, and Caorle. The dataset comprises multi-source SAR data from ERS, Envisat, COSMO-SkyMed (CSK), Sentinel-1, and the European Ground Motion Service (EGMS), covering the period from 1992 to 2021. The study integrates multi-platform SAR observations with ADAFinder-based extraction of Active Deformation Areas (ADAs), data quality evaluation using the Quality Index (QI), building-scale analysis based on LOS-derived vertical displacement time series, and orthophotos to confirm the building’s presence and evolution. By using the adopted extraction thresholds, a total of 57, 16, 83, 33, and 72 ADAs were identified from the ERS, ENVISAT, COSMO-SkyMed, Sentinel-1, and EGMS datasets, respectively. The result suggests that the strongest deformation occurred during the earlier observation periods in Zones 1 to 3, then progressively stabilized, whereas some parts of Zone 4 remained active and showed renewed deformation during the later periods. The research highlights the importance of conducting long-term analysis using multi-platform interferometric datasets to refine and personalize outcomes in geohazard monitoring. The findings from this research offer invaluable insights into the ongoing surveillance of geohazards, which are progressively related to urban development and planning.

1. Introduction

Land subsidence in coastal alluvial plains constitutes a notable geohazard with the potential to damage infrastructure and modify established drainage patterns [1,2]. These areas are often underlain by recently deposited, unconsolidated sediments that tend to compact, especially due to urban expansion or fluid extraction activities [3]. The Venetian–Friulian coastal plain, located in northeastern Italy, exemplifies this phenomenon, where both natural consolidation processes and anthropogenic influences (such as groundwater extraction, land reclamation, and construction loads) contribute to triggering land settlement [4,5].

The phenomenon of subsidence has been measured through a diverse array of methodologies, including leveling, GNSS, and interferometric techniques [6]. Since the later part of the 20th century, satellite-based A-DInSAR (Advanced Differential Interferometry Synthetic Aperture Radar) systems have emerged as highly effective tools for detecting, monitoring, and analyzing ground motion associated with geohazards [7,8]. The applications of satellite datasets are extensive and differ, encompassing seismic and volcanic hazards [9,10], subsidence phenomena [11,12], landslide incidents [13,14,15], debris-flow susceptibility mapping [16,17], aquifer depletion [18], and the surveillance of civil infrastructure [19]. A-DInSAR processing services, along with the European Ground Motion Service (EGMS), are currently being utilized effectively for the detection and monitoring of ground deformation [20].

This efficacy can be attributed to several factors: (i) A-DInSAR processing is less time-consuming, (ii) it does not require expert user intervention, and (iii) it enables the acquisition of results over vast areas covering tens of thousands of square kilometers [21]. The Persistent Scatterer Interferometry (PSI) methodology, characterized by extensive post-processing, assesses both the spatial and temporal dynamics of subsidence, thereby enabling correlation with geomorphological attributes and alterations in land utilization [22]. Interferometric data provide a powerful means of effectively monitoring the subsidence that is induced by the consolidation processes occurring within clayey soils when subjected to the loads imposed by buildings, thereby offering invaluable insights that contribute to our understanding of the long-term behavior of urbanized areas that extend well beyond the confines and limitations of both traditional laboratory analyses and in situ testing methodologies [23].

At the same time, the increasing availability of dense PSI datasets creates an interpretation problem. Raw deformation point clouds are powerful, but at the regional scale, they are not always easy to analyze, especially when the goal is to identify the most significant deforming sectors and distinguish reliable patterns from noisy or isolated measurements. To address this issue, Barra et al. [24] proposed the concept of Active Deformation Areas (ADAs) within the ADAtools framework. In this approach, neighboring active points are clustered into spatially coherent deformation areas, and each cluster is assigned a quality class based on the temporal and spatial consistency of the associated time series. The main advantage of this procedure is that it reduces the complexity of wide-area PSI outputs and supports a more operational interpretation of deformation maps. Subsequent studies have confirmed the usefulness of ADA-based workflows for geohazard screening, urban subsidence analysis, and building-oriented monitoring applications [25,26,27].

The ongoing natural consolidation of loosely packed shallow sedimentary layers is recognized as a main factor in the overall displacements in Italy’s coastal areas. This phenomenon is particularly evident in regions such as the Venetian–Friulian Plan [28,29] and the Po River Delta, which have experienced rapid development over the last few centuries [4]. Italy’s coastline extends approximately 7500 km, encompassing the mainland and surrounding islands. This extensive coastal area is home to approximately 18% of the nation’s population, emphasizing the significant role that coastal regions play in Italy’s demographic distribution [30,31].

Previous work by Floris. et al. [23] in the Portogruaro–Concordia sector using multi-temporal A-DInSAR has already shown that it can delineate differential subsidence associated with buried post-LGM incised valleys filled with unconsolidated Holocene sediments. In that study, higher subsidence rates were observed within the valley fill than in the surrounding older plain deposits, and displacement time series were used to interpret the possible role of recent loading and shallow hydrogeological conditions. That contribution is important because it demonstrated the geological significance of differential subsidence patterns in this sector of the Venetian–Friulian Plain. However, it was focused on a more limited area and on the relationship between deformation and buried geomorphological boundaries.

Therefore, the main objective of this study is to conduct a comprehensive analysis of subsidence dynamics in the Veneto-Friulian Plain, a coastal area located within NE Italy. In order to achieve our research objectives, we implemented a methodological framework based on the ADA Finder tool by using multi-platform interferometric A-DInSAR datasets: ERS (June 1992–2000), Envisat (April 2003–July 2010), COSMO-SkyMed (CSK) (February 2012–January 2016), Sentinel-1 (S-1) (December 2014–July 2017), and the European Ground Motion Service (EGMS) (February 2015–December 2021), a service by European Environment Agency (EEA) under the Copernicus Land Monitoring Service.

The aim is to improve the long-term interpretation of subsidence patterns. The specific objectives are: (i) to reconstruct the spatial and temporal evolution of subsidence across the study area using successive SAR observation periods with limited temporal gaps; (ii) to detect and evaluate ADAs and their reliability through the Quality Index framework; and (iii) to analyze the representative building scale deformation rate, supported by historical orthophotos, in order to distinguish sectors characterized by stronger early settlement from those showing persistent or locally renewed deformation. In doing so, the study provides a practical basis for identifying urban sectors that may require closer inspection, further ground investigation, or future in situ monitoring.

2. Study Area

The study area covers about 500 km² in the Venetian–Friulian plain, NE Italy (Figure 1). The area of interest was divided into five sub-areas, hereafter referred to as “Zone” to facilitate and enhance the effectiveness of SAR image processing.

Zone 1 (54.45 km²) comprises the municipalities of Portogruaro and Concordia Sagittaria, while Zone 2 (120.35 km²) includes the municipality of San Stino di Livenza. In addition, Zone 3 (127.95 km²) consists of Eraclea, Zone 4 (116.56 km²) includes Caorle, and Zone 5 (93.90 km²) encompasses Torre di Mosto (Figure 1).

Geographically, the study area lies within the lower Veneto-Friulian Plain. It is characterized by a great variety of environments, including lagoons, marine environments of the resurgence area, and plain environments, which confirm the geological history of the late Pleistocene and Holocene, i.e., the last 150,000 years [32]. The Quaternary evolution of the plain is closely linked to the response of alluvial systems to climate and sea-level changes. However, since the Bronze Age and with increasing intensity, human impact has also played a significant role from Roman times onward. This human influence is particularly evident from the Middle Ages onward, when the Republic of Venice implemented extensive river diversions and reclamation works across the coastal plain. During the last century, artificial embankments and the reclamation of large coastal sectors further strengthened the anthropogenic imprint on the landscape [32]. The exploitation and subsequent artificial modification of coasts have resulted in the flattening of existing coastal dunes, thereby altering the natural appearance of the landscape [33]. Finally, in terms of the river environment, this region is characterized by its bumps, abandoned riverbeds, and incisions, the majority of which have been filled with sediment and are no longer visible at the surface. In particular, the municipality of Portogruaro lies within the alluvial system of the Tagliamento River and exhibits a fan-like geometry when viewed from an aerial perspective. This structure, once referred to as an alluvial fan, is now regarded as a megafan due to its considerable size [34].

The geology of the Venetian–Friulian plain is closely linked to the Quaternary evolution of the Isonzo, Tagliamento, Piave, Brenta, and Adige Rivers and has undergone multiple course changes, progressively affecting larger regions and, consequently, creating sedimentary systems with a distinctive fan shape, referred to as alluvial megafans [34,35]. The sediments forming the surface deposits originated from the Last Glacial Maximum and were deposited between the final phases of the Pleistocene and the late Quaternary, approximately over the last 20,000 years [32].

The Pre-Quaternary sedimentary sequence in the study area exhibits considerable thickness (up to 1500 m) and is derived from erosional processes acting on the Dinaric orogenic belt and Alpine geological formations, predominantly comprising clay–silt strata with thin sandy layers [36,37]. The Quaternary sedimentary deposits in the area are characterized by consolidated sandy and silty-clay layers, which are predominantly associated with alluvial and marine–lagoonal environments and exhibiting a thickness that varies significantly and ranges from several hundred meters to 3000 m as one transitions from NE (i.e., the Friuli region) to the southwest (i.e., the Venezia lagoon) [38,39]. Within the Venice lagoon region, sandy sediments are primarily concentrated in the central sector, whereas the northern and southern regions are predominantly composed of more compressible clay sediments [40].

The unconsolidated Holocene deposits reach thicknesses of several tens of meters and show a heterogeneous mix of fluvial sand, fluvial silt, clay, and peat, indicating the combined influence of marine transgression and fluvial processes [39] (Figure 2).

3. Materials and Methods

3.1. Materials

As mentioned earlier, the study analyzes long-term land subsidence evolution using five descending-orbit SAR-derived deformation products covering successive observation periods with minimal temporal gaps: ERS (1992–2000), Envisat (2003–2010), COSMO-SkyMed (2012–2016), Sentinel-1 (2014–2017), and the European Ground Motion Service, EGMS (2015–2021). These datasets differ in wavelength, spatial resolution, revisit interval, and processing workflow; therefore, they were not merged into a single homogeneous deformation product. In addition, a descending acquisition geometry was preferred to reduce geometric inconsistencies across the multi-platform comparison (

Table 1. Main features of the available SAR datasets used in this study.

Satellite Mission	Orbit	Period	N. of Images	Revisit Time (Days)	Band/ Wavelength (cm)	Resol. Az./Range (m)	LOS Incidence Angle, θ	LOS Azimut, α
ERS	Desc.	14 June 1992–13 December 2000	63	36	C/5.6	6/24	~23°	~274
ENVISAT	Desc.	2 April 2003–14 July 2010	71	36	C/5.6	6/24	~23°	~274
COSMO–SkyMED	Desc.	18 February 2012–12 January 2016	66	12	X/3.1	2.5/2.5	~33°	~277
Sentinel-1	Desc.	23 December 2014–22 July 2017	91	6/12	C/5.6	5/20	~37°	277
EGMS	Desc.	17 February 2015–30 December 2021	353	6/12	C/5.6	5/20	~37°	277

).

ERS and Envisat products were accessed through Geoportale Nazionale (PCN) within the framework of the “Extraordinary Plan of Remote Sensing” (http://www.pcn.minambiente.it/mattm/en/, accessed on 20 February 2026). In similar regional applications based on the same national archive, these products are described as PSI-derived deformation datasets provided by the Italian Ministry of the Environment for regional subsidence studies. The COSMO-SkyMed and Sentinel-1 deformation products used in this study were processed using the PSI module implemented in SARScape software developed by Sarmap SA. Finally, EGMS data was obtained from the EGMS web platform (https://egms.land.copernicus.eu/, accessed on 20 February 2026), a service of the Copernicus Land Monitoring Service (CLMS) implemented on behalf of the European Union by the European Environment Agency (EEA).

3.2. Methods

All SAR-derived layers were incorporated into the GIS environment using a consistent spatial reference framework. WGS 1984/UTM Zone 33N was used as the working projection for the project, whereas EGMS products originally use the ETRS89 reference system and therefore require reprojection before cross-comparison. The deformation point densities differ strongly among datasets, with much lower densities in the ERS and Envisat products and far denser point coverage in COSMO-SkyMed, Sentinel-1, and the EGMS. These differences were considered during the interpretation stage and are one of the reasons why the datasets were compared as successive periods rather than treated as directly interchangeable measurements. This reduced the risk of over-interpreting differences caused by wavelength, density of persistent targets, incidence angle, and revisit frequency [41].

3.2.1. LOS Interpretation and Estimation of Vertical Deformation

In this study, vertical ground displacement rates were estimated using a simplified projection of Line of Sight (LOS) velocities, since the movement in the area is mainly vertical [23]. The vertical component of ground motion was derived by projecting the LOS velocity onto the vertical axis using the radar incident angle ($\theta$). The relationship is expressed as

$$V_{\mathrm{vertical} }={\displaystyle \frac{V_{\mathrm{L}\mathrm{O}\mathrm{S}, \mathrm{descending} }}{\cos \theta }}$$

(1)

where $V_{\mathrm{L}\mathrm{O}\mathrm{S}}$ is the radar LOS velocity obtained from descending-orbit measurements and $\theta$ is the incidence angle.

The LOS to vertical conversion was applied solely to obtain a first-order estimate of subsidence magnitude. This step assumes that ground motion in the study area is mainly vertical, which is a reasonable approximation for settlement processes in a flat coastal plain, where horizontal displacements are expected to be much smaller than downward motion. However, this is still an approximation, not a full motion decomposition. Because only one viewing geometry (descending) was used, any horizontal component aligned with the radar Line of Sight (LOS) cannot be separated from the vertical one and may therefore affect the converted values. For this reason, the derived vertical velocities should be interpreted as approximate proxies of dominant subsidence trends, useful for comparing the relative intensity and spatial distribution of deformation rather than as exact three-dimensional ground-motion measurements [42,43,44].

3.2.2. ADA’s Extraction and Processing Parameters

This step can be summarized in two primary phases: (i) the automated extraction of the most dependable and relevant Active Deformation Areas (ADA) (Figure 3), (ii) the assignment of a Quality Index (Q-I) attributed to each ADA (Figure 4).

3.2.3. (i) The Automated Extraction of the Most Reliable Active Deformation Areas (ADAs)

The aim of creating the ADA map is to enable the rapid identification of the most active deformation zones. The resultant map serves as a definitive input for validation and is combined with supplementary datasets (e.g., geohazard inventories and ground-truth information) to evaluate the characteristics of active deformation and produce the Geohazard Activity Map. The ADA map was developed by adapting the methodologies proposed by [24].

The processed datasets were then analyzed using the AdaFinder (v24.10.30) tool to delineate and identify Active Deformation Areas (ADAs). In the AdaFinder environment, stability thresholds were applied (4σ when Vm < σ and 3σ when Vm > σ) to ensure that only reliable deformation signals were considered during the investigation. For example, when the mean velocity was very close to the noise level, the deformation signal could not be easily distinguished from the measurement. In such cases, a stricter (4σ) rule was applied, meaning only points where the deformation exceeded 4σ were classified as active to reduce the risk of misinterpreting noise as a real deformation signal.

On the other hand, when the mean velocity exceeded the uncertainty, the deformation trend was clearly distinguishable from noise. In such cases, a less strict (3σ) threshold was sufficient to confirm the deformation as active (Figure 3).

For each identified ADA, the following corresponding parameters were evaluated:

• The total count of aggregated active points (APs).
• The mean, maximum, and lowest values of the velocities of the APs.
• The average value of the total deformations of the APs. To mitigate the substantial impact of atmospheric or digital elevation model inaccuracies, we calculate the final accumulated deformation as the mean of the accumulated values from the last four acquisition instances of all APs within the ADA. The categorization of velocity, which classifies the ADA based on its maximum velocity $\left(v_m\right)$, is established as 1 when |$v_m$| is greater than 1 cm/year or 0 if it falls within the range of $3{\sigma }_{\mathrm{m}\mathrm{a}\mathrm{p}}$ < |$v_m$| < 1 cm/year.

3.2.4. (ii) ADA Quality Index (Q-I)

Although the ADA map utilizes filtered data, a final quality assessment is essential for each individual ADA to ensure the automation process is effective. The level of noise in the time series and the reliability of deformation estimation can differ significantly from point to point [24]. Therefore, each ADA was assigned a Quality Index (Q-I) to evaluate the reliability of the associated time-series information. The Q-I combines two complementary parameters: the (i) Temporal Noise Index (TNI) and (ii) Spatial Noise Index (SNI). In this study, the original threshold values were slightly revised to better reflect the characteristics of the analyzed multi-platform datasets and to assess the overall quality of our datasets.

The TNI evaluates the temporal stability of the active-point time series through the median first-order autocorrelation, Med(ρ). Higher Med(ρ) values indicate stronger temporal consistency and lower relative noise. In this study, slightly stricter thresholds than those proposed in the original methodology were adopted to better reflect the characteristics of the analyzed multi-platform datasets. Accordingly, TNI class 1 corresponds to Med(ρ) > 0.85, class 2 to 0.75–0.85, class 3 to 0.65–0.75, and class 4 to Med(ρ) < 0.65 and therefore represents poor spatial consistency (

Table 2. Final classification of Temporal Noise Index (TNI).

Med(ρ)	Noise-Velocity Ratio (%)	Class
>0.85	<15	1
0.75–0.85	15–25	2
0.65–0.75	25–35	3
<0.65	>35	4

).

The SNI evaluates the spatial consistency of the active points within each ADA. Accordingly, class 1 corresponds to cumulative frequency values greater than 75%, indicating high spatial consistency and the upper quartile. Class 2 corresponds to values between 25 and 75%, representing the intermediate quartiles (2nd and 3rd) and therefore indicating moderate spatial consistency. Class 3 corresponds to values between 2 and 25%, representing the lower quartile (4th) and representing low spatial consistency. Class 4 corresponds to values below 2%, representing poor spatial consistency; therefore, ADAs falling in quartile 4 and beyond were excluded from further time series analysis due to low coherence and temporal and spatial stability (

Table 3. Final classification of the Spatial Noise Index (SNI).

Med(ρ)	Cumulative Frequency (%)	Class
>0.85	>75 (1st quartile)	1
0.75–0.85	25–75 (2nd and 3rd quartiles)	2
0.65–0.75	2–25 (4th quartile)	3
<0.65	<2 (Poor spatial consistency)	4

).

Figure 4 illustrates the relationship between temporal autocorrelation and the random noise to deformation velocity ratio used to support the classification of the Temporal Noise Index (TNI). The plotted points show that temporal autocorrelation decreases progressively as the relative noise increases for both deformation velocities of 5 mm/yr and 10 mm/yr. This inverse trend is summarized by the negative linear regression, which provides the basis for linking Med(ρ) to approximate noise-to-velocity conditions. The figure shows that higher Med(ρ) values correspond to lower temporal noise and therefore to more reliable ADA time-series behavior, whereas lower Med(ρ) values reflect progressively noisy signals resulting in higher temporal noise (Figure 4).

Finally, the Quality Index (Q-I) is generated by integrating the Temporal Network Index (TNI) with the Spatial Network Index (SNI) and serves as a standard for assessing the consistency of all identified ADAs. The classifications (Q-I) assigned to each TNI-SNI pairing are illustrated in the matrix shown in (Figure 5). The Q-I spectrum ranges from class 1 to 4. Class 1 and class 2 ADAs are generally reliable because they are supported by high-quality time series data, while class 2 remains dependable with only minor limitations in time-series quality. Class 3 still represents a dependable ADA, but its associated coherent points often yield noisy or incomplete time series. As class 4 shows the highest level of noise relative to classes 1, 2, and 3, it is therefore regarded as an unreliable ADA, with a very noisy time series and very low coherence.

4. Results

4.1. PS Velocity Maps

The multi-platform SAR observations clearly demonstrated that subsidence did not remain constant through time but rather evolved across successive periods. By examining the mean velocity maps, clear patterns of subsidence towards gradual stability are obvious across various locations.

However, detailed temporal interpretation was limited to Zones 1–4, where the more urbanized sectors provided sufficiently stable reflectors to support consistent deformation mapping. In contrast, Zone 5 (Torre di Mosto) did not yield a PS dataset robust enough for reliable multi-temporal comparison, as its predominantly agricultural land cover resulted in sparse and discontinuous PS coverage across the different SAR acquisitions. This limits the extraction of stable ADAs that could be compared confidently through time and reduces the continuity of the deformation pattern. Therefore, to avoid introducing bias related to uneven data quality into the comparative analysis, Zone 5 was not included in the detailed interpretation and is shown only in the velocity maps (Figure 6).

The first signs of subsidence were mainly around ancient riverbeds and low-lying floodplains at the beginning of our ERS observation period (1992–2000). During the mid-phase (2003–2016, ENVISAT & COSMO-SkyMed), deformation continued, with some areas exhibiting accelerated deformation due to new construction and the reconstruction of existing buildings. However, some regions in recent years (2014–2021, Sentinel-1 and the EGMS) have stabilized; others continue to show ongoing deformation due to urban development and the reconstruction of old buildings on soft Holocene deposits.

The total number of point targets achieved from ERS reached 10,322, resulting in a calculated point density of 20.11 PS/km². Envisat reached 17,703 points, yielding a calculated point density of 34.49 PS/km². However, COSMO Sky-Med achieved a much higher number of target points due to its X-band capabilities, with high-resolution imagery and short revisit times compared with ERS and Envisat. In total, CSK gathered 693,435 points, producing a density of 1350.80 PS/km². The targets recorded by Sentinel-l reached 197,514 points with a calculated density of 384.76 PS/km². Lastly, for the EGMS, 215,781 point targets were tracked, yielding a calculated point density of 431.56 PS/km² (Figure 6).

4.2. ADA Quality Index (Q-I) Derivation Based on the Data Used

The Quality Index (Q-I) is structured as a four-class matrix system in which each class corresponds to a defined range of combined temporal and spatial coherence values, as has already been explained and depicted in (Figure 5). Since this classification is already briefly described in the Section 3.2 above, we will therefore move directly to our findings achieved through our datasets by applying defined thresholds for the Q-I matrix. Based on sources and datasets with varying resolutions and revisit times, ERS showed excellent results in the Q-I matrix system by Adafinder, followed by the EGMS, with some clusters falling into quartile 2 and very few into quartile 3. Furthermore, Envisat also performed well but registered a low number of ADAs based on the assigned thresholds. Nonetheless, in the case of CSK, some clusters fell into quartile 3. However, due to its high point density and higher resolution, we were able to select the most coherent and reliable points within each cluster for subsequent time-series analysis. In contrast, several Sentinel-1 ADAs fell into quartile 4, which represents the lowest quality class for our area in the Q-I matrix due to noisy time series; therefore, they were excluded from further analysis (Figure 7).

4.3. Temporal Evolution of ADAs

The evolution of land subsidence within the study area indicates a systematic stabilization of subsidence trends across most regions. However, Zone 4 continues to undergo active deformation in certain areas, which we will explain in subsequent parts.

A total of 57 ADA clusters were identified during the ERS observational period. This number decreased to 16 during the ENVISAT observation phase, then increased to 83 ADAs during the CSK period. In Sentinel-1 33 ADAs were detected, while in the most recent EGMS dataset 72 ADAs were identified (Figures 8, 10, 12 and 15). For each ADA, the information listed in

Table 4. Attributes associated with each ADA.

Field	Description	Units
Join Count	Number of unstable points grouped in the hotspot	-
F1	Geographic Latitude	°
Lambda	Geographic Longitude	°
E	X-coordinate (Easting)	m
N	Y-coordinate (Northing)	m
H	SRTM Height	m
Accumulated Deformation	Accumulated deformation of PS	mm
Velocity (Mean)	Mean velocity of the hotspot	mm/year
Velocity (Max)	Maximum velocity of the PSs	mm/year
Velocity (Min)	Minimum velocity of the PSs	mm/year
QI	Quality index of the ADA	-
Class	Classification of the hotspots based on the max velocity	-

is compiled to form the attribute table for the related polygonal shapefile.

Starting from Zone 1, this zone is characterized by channel deposits, predominantly consisting of gravels, clay, gravelly sands, and, less frequently, silts and sandy silts [34,45,46]. It is known as the Torresella Unit in the geological map of Venice [47]. Furthermore, this pattern is consistent with the geological setting of the area, where the urban sector lies over the buried Valley of Concordia and related post-LGM incisions, which are filled by younger, more compressible Holocene deposits, already reported in [23].

The color-coded velocity (Figure 8) presents a visual representation of subsidence intensity. Red and orange colors (<−8 mm/yr) highlight high deformation zones, primarily observed during the early ERS and ENVISAT phases. On the other hand, the gradual shift toward yellow and green in the following datasets (Sentinel-1 and EGMS) reflects a trend of stabilization or very minimal movement in regions that were once highly active (Figure 8).

In the ADA map (Figure 8), the transition across different acquisition periods is clearly noticeable. Deformation was higher in the early observations; however, it progressively slowed and attained a certain level of stabilization in subsequent acquisitions. This pattern indicates progressive consolidation, a phenomenon that gradually expels pore water, compresses the soil, and causes settlement that typically slows over time. In doing so, we analyzed vertical displacement time series from each acquisition to verify whether the same trend could be observed at the building scale.

Figure 9 illustrates the LOS-derived vertical time series of the four selected buildings in Zone 1, located in the urban areas of Portogruaro and Concordia Sagittaria, and highlights the relative temporal evolution of deformation across the different SAR observation periods. Figure 9A represents the position of these buildings, while Figure 9B shows the LOS-derived vertical displacement time series of these selected buildings calculated by using Equation (1). One of the main reasons for selecting these buildings was that they were constructed just before or around the 1990s, so in our case scenario, they were the right ones to be further evaluated for vertical time series analysis. The selected buildings showed considerable deformation in the beginning. However, in the later stages of the observation period, the subsequent datasets showed signs of stabilization (Figure 9B).

Subsidence evolution for Building (A) showed an overall displacement of approximately −150 mm, with peak subsidence rates of −8.5 mm during the ERS period. In the long run, the velocity decreased to −3.8 mm during the EGMS period and −3.5 mm in COSMO-SkyMed, clearly indicating a gradual decrease in deformation intensity, a phenomenon often observed with progressive consolidation. Building (B) underwent a more significant displacement of about −200 mm, with the highest velocity observed during the ERS period at −9.9 mm. Once again, a clear decline in displacement patterns can be seen. However, the difference between S-1 at −3.1 mm and the EGMS at −3.6 mm is very minimal, suggesting that secondary consolidation is still ongoing. Similarly, in Building (C), a deformation trend similar to that of Buildings (A) and (B) is observed, with an overall displacement of approximately −150 mm. Its displacement velocity gradually decreased to −4.6 mm in the EGMS. This transition clearly indicates that primary consolidation is nearly towards completion. Finally, Building (D) in the southern part of Concordia Sagittaria recorded a displacement rate of −7.0 mm during the ERS period. However, later it decreased to −2.2 mm, as shown by the EGMS data, further reinforcing the evidence of ongoing stabilization (Figure 9B).

Similar to the trends observed in Zone 1, the signs of deformation in Zone 2 (San Stino di Livenza) are more evident throughout the entire ERS and Envisat observational period. Nevertheless, subsequent satellite missions showed a significant reduction in subsidence velocity (Figure 10).

The deposits within “San Stino Di Livenza” include a range from coarse to fine gravels and sandy gravels, which can create very deep incisions, as well as medium-to-fine and very fine sands that define channel deposits, along with silts, clayey silts, and peaty silts that are associated with floodplain deposits [48,49]. It is part of the Malamocco Unit and consists of alluvial deposits from the Upper Pleistocene to the Holocene [47] (Sheet 107, Portogruaro, CARG Project).

The ADA map of Zone 2 showed a high level of deformation at the beginning of our observational period, consistent with the area of “San Stino Di Livenza” falling within a concealed paleochannel [47], which further solidifies the evidence that the construction of buildings significantly contributes to displacement within this geomorphological feature, as the fill comprises more recent sediments still undergoing consolidation. The presence of an ancient Paleochannel in San Stino di Livenza has already been explained and supported by the following articles [34,35,50].

In Figure 11B, Buildings (A) and (B) underwent considerable deformation, especially during the initial observation period (ERS 1992–2000), with rates exceeding −8 mm and −14 mm, respectively. Over the next period, the rate of displacement for both buildings decreased gradually, with results from Sentinel-1 and the EGMS indicating considerably lower velocities: −2.3 mm for Building (A) and −3.4 mm for Building (B), respectively.

Zone 3, Eraclea, is situated above the floodplain and is primarily composed of silt, clayey silt, and clay [50]. This area is part of the Torcello Unit and consists of deposits from the post-Roman Holocene alluvial environment [47] (Sheet 108, Portogruaro, CARG Project).

Although PS density is usually lower for ERS and Envisat due to their longer revisit times and lower point density, the industrial nature of the area provided numerous stable man-made scatterers, allowing us to still obtain a reasonable number of coherent samples. Therefore, we still obtained a reasonable number of coherent points for ADA mapping. We identified multiple deformation zones during the ERS phase; however, over time, they showed signs of stabilization during the Envisat acquisition phase. However, during CSK, Sentinel, and EGMS acquisitions, the region continues to exhibit low to moderate deformation, as illustrated by the ADA map of EGMS (Figure 12).

Similarly, Figure 13 illustrates the LOS-derived vertical ground displacement time series trends in Zone 3 (Eraclea). In this zone, two buildings (A and B) located in different parts of the area were also being monitored to observe their vertical displacement. Figure 13A shows the location of these buildings, while Figure 13B demonstrates the time series of vertical displacement for these buildings. Remarkably, a very similar pattern was observed in this zone, consistent with the deformation behavior identified in the previous zones. For Building (A), the total deformation reached about −220 mm during the ERS period, with an abrupt increase, and a velocity of −12.9 mm. However, in subsequent years, the rate slowed significantly. The EGMS shows only a −2.0 mm decrease, suggesting the area is gradually stabilizing due to ongoing consolidation. Building (B) shows a similar pattern, with a total displacement of about −190 (mm/yr). Initial subsidence was also at its peak during the ERS period (−11.4 mm) and later reduced to −3.7 mm during the EGMS period.

All buildings considered for vertical displacement time-series analyses (Figure 9, Figure 11 and Figure 13) were verified using old orthophotos and accessed through Geoportale Nazionale (PCN). A Web Map Service (WMS) from the Italian Ministry of Environment and Energy Security (MASE) (https://gn.mase.gov.it/portale/it/web/geoportale-mase/servizio-di-consultazione-wms, accessed on 3 January 2026) was used to independently verify building presence and to cross-check the reference targets selected for the vertical deformation time-series analysis in Zones 1–4.

For each zone, the mapped building locations and enlarged insets clearly show the building footprints and the nearby road network, thus confirming that these structures were already present during or just before the start of our study period (Figure 14 and Figure 16).

Zone 4 (Caorle) is composed of sandy, sandy-silty, and silty-sandy sediments, which are typical of river hump areas. In fact, it is located above the Livenza River’s uplift [51,52]. The ADA evolution map for Zone 4 (Figure 15) clearly shows that most construction or reconstruction took place during the later periods. For this reason, the deformation behavior in some parts of Zone 4 differs from that observed in the previous three zones. It is evident that Zone 4 has been affected by deformation from the CSK period onward throughout the EGMS acquisition interval. Specifically, for Building (A) shown in Figure 16A, the 1996 orthophoto confirms that the building was already present at the beginning of the observation period and showed only minor movement. In the 2006 orthophoto, corresponding to the ENVISAT interval, the movement is still relatively low because the building had not yet been reconstructed. However, the 2012 orthophoto clearly shows that the building was reconstructed on the same footprint. This reconstruction imposed an additional load on the underlying soil, and the building began to exhibit more rapid deformation from the CSK period onward, continuing through the EGMS acquisition period. This provides clear evidence that the building had already passed its initial consolidation stage and had started to deform again after reconstruction, as shown in Figure 16B(A).

Yet, Building (B) showed a different trajectory. It showed stronger early deformation, followed by an obvious slowdown during the ENVISAT period. However, in later stages the recurrence of deformation was observed although at considerably lower rates, even though the orthophotos do not indicate major nearby reconstruction (Figure 16B). These two representative cases clearly demonstrate that the later-stage activity in Caorle cannot be reduced to a single behavior: in some locations it is associated with renewed loading after reconstruction, whereas in others it reflects continued or renewed settlement on susceptible coastal deposits.

5. Discussion

The results presented above revealed a clear shift in subsidence evolution in the Venetian–Friulian Plain: initially characterized by early deformation intensity around historical urban hubs, followed by a progressive transition toward coastal and newly developed areas, as shown in the ADA’s evolution maps (Figure 8, Figure 10, Figure 12 and Figure 15). In this study, the use of EGMS data, combined with independent PSI processing of Sentinel-1 data, along with CSK, Envisat, and ERS, enabled high-quality validation of deformation signals and aided in identifying zones requiring long-term monitoring. The ADA Quality Index (Q-I) (Figure 7) has been proven to be valuable in distinguishing reliable deformation clusters from isolated or noisy signals.

It is also worth mentioning that the ADA map serves as a valuable tool for regularly updating geohazard inventories and facilitating risk management initiatives. A fundamental aspect of using such a methodology is its reproducible workflow, which can be customized to align with the requirements of each particular case study or end user [53,54]. The implementation of the methodology is dependent upon the site’s characteristics and the core objectives of monitoring, especially regarding the spatial and temporal dimensions of expected deformations, which necessitate the modification of elements such as the D-InSAR processing technique, the minimum requisite number of points for the extraction of the ADA, or the defined stability threshold.

5.1. Geological Control on the Spatial Distribution of Subsidence

The results indicate that the long-term distribution of subsidence in the Veneto-Friulian Plain is not random but strongly conditioned by the shallow geological and geomorphological architecture of the coastal plain. In the Portogruaro–Concordia sector, the main deformation clusters coincide with the buried post-LGM fluvial incisions and their Holocene infill, where soft alluvial, swamp, lagoonal, organic, and peaty deposits overlie older and mechanically more competent LGM deposits [47,48,51].

This relation is consistent with previous work showing that the Valley of Concordia and related buried incisions contain thicker, younger, and more compressible sediments than the adjacent plain and therefore respond differently to loading and drainage changes [23,37,48,51]. The same first-order control explains why deformation in Zone 1 and Zone 2 is concentrated in sectors associated with paleochannels and fine-grained floodplain fills, whereas more stable sectors tend to correspond to older or less compressible deposits [34,47]. A similar interpretation applies to the other zones, although their local geological settings vary. In Zone 3, Eraclea, the active sectors are associated with low-lying Holocene alluvial deposits dominated by silt, clayey silt, and clay, which are prone to settlement under load due to sediment compaction [32,49]. Zone 4 (Caorle) shows a different evolution from Zones 1 to 3 because it is developed in a coastal-lowland sector above the Livenza river-hump system, where sandy, sandy-silty, and silty-sandy deposits are laterally associated with young coastal and lagoon sediments [34,49]. This more heterogeneous shallow setting produces a less uniform deformation pattern than that observed in the buried-incision and paleochannel sectors. In addition, the orthophotos and ADA evolution maps revealed that much of the construction or reconstruction in Zone 4 occurred during the later phases. Therefore, some areas did not follow a simple trend towards progressive stabilization but remained affected by deformation from the middle phase of the observation period onward, with continued movement observed from CSK to the EGMS.

5.2. Anthropogenic Controls

The construction of new buildings imposes additional loads on the underlying fine-grained soil, leading to significant changes in its physical properties due to consolidation [23,46]. During consolidation, water trapped in pores gradually expels, resulting in a reduction in soil volume and an increase in soil density [50,51,52]. This phenomenon is characterized by an early rate of vertical displacement that can reach as much as −10 mm during the primary consolidation phase. Subsequently, this rate reduces to approximately −4 mm and then further reduces to −3 mm during the secondary consolidation phase [55], as illustrated in Figure 9, Figure 11, Figure 13 and Figure 16. On this background, local construction loading, urban expansion, drainage regulation, and groundwater withdrawal can accelerate subsidence dramatically where compressible sediments are present [23,51]. This combined effect is especially convincing in the Portogruaro–Concordia sector, where previous geological and hydrogeological studies show that this area is characterized by shallow buried incisions filled with soft fine-grained sediments and peat, while deep confined aquifers between about 480 and 600 m have also been intensively exploited [56]. The widespread presence of deep wells, piezometric decline in the deep aquifer system, and the large number of water drawdowns in the south of Concordia indicate that aquifer depletion may significantly contribute to this phenomenon [52].

5.3. Considerable Limitations and Future Recommendations

The mechanism of subsidence imposes considerable limitations on the effective application of interferometry. InSAR technology quantifies surface displacement but cannot directly explain the geological dynamics of subsurface strata or differentiate which underlying layers and processes are responsible for the detected displacement in the absence of supplementary geological and geotechnical data. Therefore, it highlights the challenges in obtaining accurate geological insights. This limitation arises primarily from the inherent dependence of interferometric methodologies on the emergence and development of new structures and infrastructure, making these techniques particularly advantageous and efficient in densely populated urban environments.

Therefore, the effectiveness of these techniques is significantly reduced in sparsely populated areas, where the absence of man-made features limits the availability of coherent radar targets, as exemplified by Zone 5 in our case study. This poses a major challenge for assessing natural subsidence driven by long-term lithostatic consolidation, as it often lacks clear surface indicators. As a result, the applicability of interferometry is fairly limited in geological and geomorphological investigations, where detailed insights into subsurface behavior are essential for accurate interpretation and hazard assessment.

However, this challenge may be mitigated by using L-band (SAR) data, which offer superior penetration capabilities compared to C- and X-band data and exhibit reduced coherence loss in densely vegetated areas. Such advantages have been recently demonstrated in studies such as [57,58], which examined ground displacements in the southern sector of the Venetian–Friulian Plain using ALOS-PALSAR SAR data. A step forward is to regularly calibrate the time interval between successive updates to keep the ADA map up to date. This calibration should account for the specific deformation rates we aim to monitor, for instance, in scenarios where deformations occur more slowly.

Additionally, the hydrogeological context may also contribute to the observed deformation. Previous studies in the Portogruaro area suggested that, in addition to construction-related loading, groundwater conditions can influence subsidence, particularly where recent fills are water-saturated and susceptible to compaction, which has already been addressed in [23,51]. However, no piezometric or geotechnical monitoring dataset was available in the present study for direct comparison with the SAR time series. For this reason, groundwater-related processes should be regarded here as probable contributing factors that require further validation rather than as directly verified causes.

6. Conclusions

This study explored how land subsidence evolved over time across the Veneto-Friulian Plain by combining multi-platform SAR datasets from ERS, ENVISAT, COSMO-SkyMed, Sentinel-1, and the EGMS for the period 1992 to 2021, together with ADA-based post-processing with orthophoto verification. The integration of ADAfinder tool enabled effective clustering and quality assessment of deformation areas. The ADA Quality Index has also been shown to effectively differentiate reliable deformation clusters from isolated or noisy signals, further emphasizing the value of ADA maps for updating geohazard inventories and enhancing strategic risk management initiatives.

The findings suggest a significant transformation in subsidence activity over the observed period. At the beginning of our observational period, the rate of subsidence was markedly high, particularly in Zones 1, 2, and 3, due to the overwhelming urbanization over compressible Holocene sediments (clay, silt, and peat). These sediments are usually water-saturated and have a high void ratio, so they exhibit abrupt initial deformation and then settle over time under any added load. In contrast, most parts of Zone 4 (Caorle) are still experiencing ongoing subsidence due to coastal development and the reconstruction of old buildings in some areas.

Our findings also emphasize the importance of long-term multi-temporal monitoring in subsidence-prone regions. As urban development continues along coastal zones, understanding the temporal evolution of subsidence is essential for infrastructure planning, groundwater management, and climate resilience. At the same time, these findings should be interpreted within the limits of the available data. The SAR products represent successive observation periods acquired by different sensors and processing chains, rather than a single, fully homogeneous temporal record, whereas the LOS to vertical conversion provides only an approximate representation of the dominant subsidence component. In addition, no independent validation data, such as GNSS, leveling, piezometric observations, or geotechnical monitoring, were available for direct comparison in the present study. Even so, the combined analysis provides a consistent basis for describing the main spatial and temporal patterns of subsidence across the study area.

Future investigations should prioritize the cutting-edge application of interferometric methods for geological and geomorphological mapping in underdeveloped urban regions. Such research presents an opportunity to deepen our understanding of these sectors in contexts where conventional mapping methods may be inadequate.

Finally, it is also vital to incorporate independent validation and support data, particularly GNSS measurements, leveling surveys, piezometric records from the shallow phreatic aquifer, and, where possible, geotechnical monitoring. The integration of these datasets with the SAR observations would enable a more direct assessment of the reliability of the detected deformation patterns and help distinguish the relative roles of sediment compaction, groundwater-level fluctuations, and construction-related loading. In particular, piezometric data would be important for clarifying whether changes in groundwater conditions contribute to the temporal behavior of subsidence observed in some sectors, while GNSS and leveling data would provide an external basis for verifying the magnitude and spatial consistency of the interferometric results. A combined analysis of these complementary datasets would therefore provide a stronger framework for validating the present interpretation and for improving the understanding of subsidence processes and their implications for urban development in these vulnerable areas.