The analysis of the vertical movements of the soil in the
Po River plane of the Emilia-Romagna Region (Italy) was updated through an
interferometric analysis referred to the 2011–2016 time-span. This activity
is a continuation of previous studies on the state of knowledge of vertical
soil movements in the same area, analyzed firstly by levelling and GNSS and
more recently by SAR interferometry for the periods 1992–2000, 2002–2006,
2006–2011, on behalf of the Emilia-Romagna Region. The survey area analysed
was approximately 13 300 km
This work focuses on the alluvial plain of the Po River valley, Emilia
Romagna region (Italy). This sedimentary basin has been affected by a
widespread land subsidence of both natural and anthropogenic origin, studied
since several decades (Caputo et al., 1970; Arca and Beretta, 1985).
Ongoing natural factors (i.e. tectonic, consolidation, oxidation, organic
soils shrinkage) still contribute a few mm yr
Since the 1950s, different agencies handled the subsidence monitoring in the Emilia Romagna region by geodetic levelling surveys. The campaigns have been initially performed in the localized areas where the phenomenon had become particularly evident, without the benefit of a consistent geodetic network design at regional scale. Over the years, different approaches were adopted, comprising three main techniques: geodetic levelling, GNSS and SAR interferometry.
In 1999, ARPA-Emilia-Romagna (Regional Agency for Environmental Prevention in Emilia-Romagna Region), in collaboration with University of Bologna, designed and instituted a network of 2300 levelling benchmarks, connected to 60 GNSS stations; it was the first integrated regional-scale monitoring geodetic network (Bitelli et al., 2000). Both the levelling and the GNSS networks were surveyed in 1999; the GNSS survey was repeated in 2002.
During the successive campaigns, different techniques have been integrated.
In 2005, the high precision geodetic levelling of a subnet of the 1999
network (more than 1000 benchmarks), and the radar interferometric analysis
(PSInSAR™ – Permanent Scatterers SAR Interferometry, by TRE –
Tele-Rilevamento Europa) have been integrated to update the subsidence
measurements (Bissoli et al., 2010). PSInSAR™ analysis was
conducted using European Space Agency's (ESA) ERS1 and ERS2 data for the
interval 1992–2000; data from ESA's Envisat and Canadian Space Agency's
Radarsat (RSAT) satellites have been processed for the 2002–2006 period.
The SqueeSAR™ technology (developed by TRE ALTAMIRA) was
further applied on RSAT ascending data for the 2006–2011 campaign. No
geodetic levelling surveys were performed during this period. Seventeen
GNSS-derived positioning time series were processed both for inserting the
whole survey in an international geodetic datum and for the calibration and
validation of the SqueeSAR™ results. After sub-sampling on a
The workflow process is described in Fig. 1; the same approach has been adopted, with some fine-tuning, in the processing of the 2011–2016 campaign.
The workflow from data acquisition to final map production.
The interferometric analysis for the 2011–2016 period, covering the
territory of the Emilia-Romagna regional Po River plain (red polygon
in Fig. 2), was carried out by TRE ALTAMIRA using the
SqueeSAR™ technology. The radar dataset includes imagery from
RADARSAT-2 (RSAT) and COSMO-SkyMed (CSK) in ascending geometry with a
resolution of 20 m
The 6 tracks of processed radar with their area (km
The area covered by the 6 tracks of processed radar data.
Radar data calibration was performed using measurements from GNSS stations, following the basic approach of the previous campaign. A network of 16 permanent GNSS stations within the study area was used to perform an accurate SAR calibration. An additional 6 permanent GNSS stations were used as control. The movements of the 22 permanent reference stations (inside the blue area of Fig. 3b) have been processed within a reference European network, depicted in Fig. 3a.
Prior to the generation of the subsidence map, a thorough statistical analysis was performed in order to identify outliers in the acquired dataset. Any potential outlier could degrade the subsidence estimation. In this scenario, an outlier is an anomaly of a single radar target, or a small group of radar targets, with a low spatial correlation to surrounding radar targets. Therefore, their vertical movement is deemed to be unrelated to subsidence phenomena, and ascribed to problems related to interferometric processing or local deformation such as thermal deformation or the construction of new structures, buildings or roads. Outlier detection is both a crucial and delicate process, carried out through the application of statistical procedures and an objective approach. The process needs the evaluation and control of an experienced operator.
The dataset was preliminary filtered by the selection of a subset of points that present coherence stability in the time series. The density of those points is obviously higher in urban areas than in the vegetated ones. In our case, the selected coherence threshold was set to 0.7 after several analysis on study area and from previous experiences: any scatterer, permanent or distributed, with a coherence (COH.) lower than 0.7 was discarded from further processing (Table 2).
Comprehensive list of radar targets before (Origin) and after (
The total number of points preserved for further processing was 1 936 132 after having discarded 1.93 % of points. Considering each track, the largest percentage of points were discarded from the Parma track, which is the track with the highest density realized by CSK. No points were discarded from the Mirandola track (Tables 1 and 2).
Spatial distribution of 338 isolated outliers within the study area.
Spatial distribution of outliers in cluster within the area of the case study.
An automatic procedure was used to isolate the outliers from the filtered dataset. The procedure estimated a predicted displacement speed at each point and compared its value against the measured displacement speed. The analysis was performed with the use of Kriging algorithm: an exact interpolator for the optimal estimation of a measurement distributed over an area. All the values are estimated using the available observations and providing a reliability information for each known value. Then, the results obtained by the interpolation derived from the Kriging processing are used as input to a Cross Validation (CV) phase. CV is the procedure that computes the value of each point and relies on spatially adjacent data; the difference between the predicted value and the measured value is defined as a residual.
The procedure was performed iteratively in two subsequent steps, with the
results shown in Table 3. At each step, the points with the standard
deviation of the residual value greater than the defined 5
Outliers detection in 2 steps.
This analysis identified a total of 21 638 outliers, equal to 1.1 % of the input dataset. Two procedures were applied for further validation before discarding the outliers. A total of 338 outliers were identified as isolated, with less than 5 valid points within a radius of 500 m (Fig. 4). After a careful validation by a direct inspection, 24 of them were reintroduced in the output dataset.
A total of 1784 outliers were identified as grouped in clusters, potentially representing significant local phenomena such as new construction. After an automatic approach using density maps in a GIS analysis and further validation, about 33 % of them were reintroduced in the output dataset for a total of 1 915 122 points.
Updated subsidence map of Emilia-Romagna region with isokinetic isolines.
In a final step, 2341 points were removed from the dataset after a supervised analysis that identified local phenomena unrelated to subsidence. The final dataset is then composed by 1 912 781 points.
The 1 912 781 points of the final dataset were processed with geostatistical
interpolation methods (Kriging) to produce a dense regular grid of ground
vertical movements with a resolution of 100 m
Compared to the precedent 2006–2011 time interval investigated (Bitelli et al., 2014), subsidence has decreased in a large portion of the Emilia-Romagna plain, and in some places a small rate of uplift was indicated (green areas in Fig. 6). Quite impressive is the change in some parts close to Bologna area, where a strong subsidence has been recorded until the recent past (Bitelli et al., 2014). Uplift seems to be related to a large and fast rising in piezometric level, due to a reduction in pumping from water wells (ARPAE, 2018). Uplift recorded in the western part of the study area was also registered in all the previous surveys and is related to tectonic movements (Cenni et al., 2012).
The land subsidence results for 2011–2016 period, described in this paper, update the long history of measurements of vertical movements for the Po River plain that began in the 1950s. Different evolving approaches and technologies were adopted for these studies over this period. The use of modern SAR interferometry techniques, combined with accurate verification and validation procedures, demonstrate the strength of this approach over very large regions. In particular, the integration with GNSS was crucial for the calibration.
The results derived from this study indicate continued subsidence in some areas of the Emilia-Romagna region, whereas a reduction in the subsidence rate and even uplift was indicated in other areas.
Conceptualization of the study, GB, LV; interferometric analysis, SDC, FN; GNSS analysis and SAR data calibration, LV; outlier detection and geostatistical analysis, FF, AL; discussion and interpretation, GB, FB, PS, LV; writing-original draft, GB, FF, AL, LV.
The authors declare that they have no conflict of interest.
This article is part of the special issue “TISOLS: the Tenth International Symposium On Land Subsidence – living with subsidence”. It is a result of the Tenth International Symposium on Land Subsidence, Delft, the Netherlands, 17–21 May 2021.
The regional campaigns for subsidence monitoring have been financed by Emilia-Romagna Region. The authors gratefully acknowledge the review process for the helpful and constructive comments and suggestions.