Analyzing subsidence in the Netherlands with attribute-enriched InSAR data

Abstract

Subsidence is affecting different parts of the Netherlands. The strongest subsidence is observed in the province of Groningen due to ongoing natural gas extraction. Subsidence is also observed at several locations in the province of South Holland, where the processes of peat oxidation, soil compaction and the withdrawal of groundwater are at the root of the problem. In South Limburg, the after-effects of coal mining are seen in the surface deformation, which is characterised by ground heave due to rising mine water, and the potential risks for sinkhole and local subsidence due to near-surface mining. Not only does subsidence cause damage to the natural and built environment, but it also increases the vulnerability towards flooding. Considering the rise in sea level, subsidence forms a pressing issue for low-lying countries, such as the Netherlands. The combined efforts of interferometric synthetic-aperture radar (InSAR) data, global positioning system (GPS) points and gravity measurements have led to the recent Bodemdalingskaart (2018), which represents the nationwide subsidence in three statistically inferred products. These products are the the total surface deformation, the deformation caused by shallow subsurface processes, and the deformation caused by deep subsurface processes. This has triggered the discussion on how the deformation from 'shallow' and 'deep' processes can or should be better understood in a physical sense. The aim of this research is to 'enrich' Sentinel-1 persistent scatterer interferometry (PSI) points with contextual information, or 'attribute-enrichment', in order to better understand the origin of the observed deformation. Classifications from the Dutch 'basisregistraties' (base registries) are assigned to the PSI points, which are stored in a spatial database. The classifications include information on the Dutch soil types and geomorphology. The classifications from Bodemkaart are used in an analysis of nationwide extent, focused on the deformation behaviour of different soils and their groundwater levels. The nationwide average deformation of all track results suggest subsiding trends of about -1.24 mm/yr for marine clay soils and -0.48 mm/yr for peat soils, whereas opposite trends of uplift are observed of, on average, +0.77 mm/yr for river clay soils and +0.54 mm/yr for sand. The soil and geomorphology datasets are also used in combination with the geographic classifications of the built environment to study the distinction between deep and shallow-induced deformation in Groningen and South Holland. Here, the 2 x 2 km grid cell representation of deformation makes it possible to characterise areas with overall low PSI point density. In the localised deformation cases, the behaviour of classes from different attributes can be directly compared, e.g. road polygons intersecting with soil type polygons. For South Limburg, risk-assessment-based classifications are used to enrich the PSI points in the study the coal mining after-effects in the identified risk areas. The risk-deformation results suggest an overall slowing trend in the ground heave of the potential impact areas. The results from the case studies highlight the efficiency of attribute-enriched PSI datasets for the interpretation of deformation, based on both direct and indirect physical classifications.