radar remote sensing for the angkor world heritage...
TRANSCRIPT
Outline
• Research objectives • SAR data • Extraction of potential flood-influenced regions • Monument motion monitoring & collapses • The sustainability of Angkor site
Kulun Mountain
Angkor World Heritage site It is well-known for the
Temple of Angkor Wat, Angkor Thom, the Bayon Temple and many other monuments of the different capitals of the Khmer Empire from the 9th to the 15th century.
Research objectives
The conservation of this site has stepped into another stage; that is, from the monument restoration to the environmental analysis on the whole site
Flood Collapse
Impact assessment from flood & human activities
Decision making
Facing challenges
Forest ecosystem (deforestation/ fire）
Water system （flood/soil erosion）
Ground subsidence （loosen/
displacement ）
Climate system（wind/
temperature/ Precipitation）
Human Activity （tourism/
urbanization）
Spatial database of Angkor Site
Policy recommendations for
heritage protection and sustainable
development
3D simulation and GIS
Optical image Radar data LiDAR data
Satellite platform Ground collection
Multi-platform remote sensing data
Basic geodata Auxiliary data
Training of heritage management personnel
Soils
Rock Types
Backscattering sensitivity to Waters
Interferometry
All-weather, all day
Advantages of SAR
SAR data
46 scenes of Stripmap TerraSAR data (3 m ground resolution) acquired in the span of February, 2011 to December, 2013 were used for the flood hazard evaluation and deformation time series analysis
Extraction of flood-influenced regions Mixed-threshold for flood extraction
Two-dimensional maximum entropy approach—t0 water feature A and background B
Otsu method for sub-classification of A and B—t1 water feature A1 and background B2
Otsu method—t2 water feature A2 and background B1
Flood monitoring and analysis
Using the acquisitions of dry season & flood season, the performance of three two-fold classification approaches, were compared
Comparison of Otsu, two-dimensional
maximum entropy and mixed threshold in
water body extraction using the dry and flood season SAR images in
2011
Acquisition
Otsu Two-dimensional maximum entropy mixed threshold
Grey Threshold
Extracted water body pixels
Grey Threshold
Extracted water body pixels
Grey Threshold
Extracted water body
pixels
20110310 36 35885918 133 76884520 54 36443974
20111016 60 47552026 110 61524376 83 53679328
Performance comparison of Otsu, two-dimensional maximum entropy and mixed threshold methods
An optimal result was derived according to the optimized tradeoff between the omission detection and false-alarm detection
The relationship between the dynamics of water body (expansion/shrinkage) and the precipitation was further investigated using
time series analysis to assess the significance of driving forces. • the seasonal or annual
variation of water bodies in Angkor was impacted by the precipitation, demonstrating a positive low correlation
• There are other factors linked to the occurrence of floods in Angkor site, such as the downstream runoff of Mekong River and the water level rise of Tonle Sap Lake
Monument motion monitoring & collapses • Prevailing notion: (right or wrong, is still unknown) on-going
decline of the ground-water table, due to increasing demand for water to meet the needs of the management of Angkor as well as that of resident communities and a growing number of visitors, was a significant cause that increases the vulnerability of monuments to collapse.
• Totally 46 scenes of TerraSAR data acquired in the span of February, 2011 to Dec, 2013 were used for the surface deformation estimation and time series analysis. The acquisition of June 3, 2012 was selected as the reference image for the PS-InSAR and Tomo-PSInSAR analysis to minimize the potential decorrelation induced by spatial-temporal baselines as well as the Doppler centroid difference.
A two-scale InSAR approach combining PSInSAR and Tomo-PSInSAR was developed PSInSAR: In order to mitigate outliers from infrastructure constructions and farming activities, the point-like targets of buildings and ancient monuments were the focus for the PSInSAR procedure. PSInSAR measures the movements, including motion velocity and accumulated time series motions with reference to the first acquisition, in the slant sight of radar observed. The PSInSAR prototype (Ferretti et al., 2000) with S-estimator and ridge regression were applied. Tomo-PSInSAR: Taking unfavorable shadowing/layover on SAR images and the occurrence of sparse spatial density of PS points into account, the Tomo-PSInSAR was applied for the structural instability monitoring and defect diagnosis. Tomo-PSInSAR overcomes this limitation due to its capability for separation of multiple PS points in overlaid layover areas. To improve the computational efficiency of sparse matrices, we developed an enhanced, compressed sensing algorithm based on weighted L1 norm and Tucker decomposition model for super-resolution tomographic imaging.
An overview of the TerraSAR/TanDEM-X SAR data spatial coverage highlighted by the pink rectangle and the area within the Angkor World Heritage site highlighted by the green rectangle where detail studies reported were undertaken (Shuttle Radar Topography Mission data courtesy). Black circle indicates the location of NOT_015 well with groundwater level measurements in the period of 2009-2014.
Uncovers the decline of Angkor monuments
Field campaign
There is no land subsidence hole caused by groundwater exploitation
Mild uplift trends of ground surface occur in areas near water reservoirs (Barays) owing to the rehabilitation of the ancient hydraulic system since 2005
Local government and APSARA made an optimal site selection for the public pumping wells (a safe distance away from the heritage core area)
Deformation field derived by PSInSAR
Public and private wells for pumping groundwater: (a) a public well and (b) private well, imposed on the PSInSAR deformation map subsets. Moderate-severe subsidence values (shown by yellow to red color points marked by pink circles) were measured in areas immediately surrounding the wells.
Tomo-PSInSAR derived annual deformation rates in ancient temples of (a)Preah Khan, (b)Bayon, (c)Ph.Bakheng, (d)Ta Prohm, (e)Angkor Wat and (f)Bakong. Spatial motion heterogeneities are evident due to the enhanced PS targets extracted.
Structural motion anomaly on Angkor Wat by Tomo-PSInSAR -4~+4 mm/a Introducing motion heterogeneity
Validation of Tomo-PSInSAR derived motions utilizing the PS heights confirmed by the field verifications undertaken in 2014; (a) precise height estimation on Angkor Wat with temple-mountain shapes, marked by the white rectangle; (b) scatter-plot of estimated heights compared with the ground-truth on the terrace gallery, indicating a consistent trend supported by a significant correlation coefficient of 0.820 with a standard deviation of 0.44 m.
Correlation between groundwater levels (NOT_015), precipitation, and the Tomo-PSInSAR motion trend (ranging from -1.9 mm to +2.0 mm) can be observed; a mild rise and more stable groundwater table was detectable in Angkor Wat region after the optimization of water management by APSARA although facing to a decrease trend of annual precipitation in 2009-2013 (i.e. 1648.4 mm in 2009 regularly drops to 1037 mm in 2013).
The deterioration model shows that an entire rigid motion was prevalent in the early stages immediately after the temples had been constructed; however, the long-term material decay triggers structural instability which in turn aggravates the decay processes. As irregular and fragmentary motions become dominant, cracks and fissures in the monuments grow leading to severe structural instabilities and the potential final collapse.
Deterioration driving force is seasonal variations of the groundwater table instead of its decline.
The flood is no longer a trouble for the Angkor site since the ancient hydraulic system was rehabilitated and put into full operation since 2011. Using the two-scale monitoring scheme comprising PSInSAR and Tomo-PSInSAR, the factors that could result in the potential decline and eventual collapse of monuments in Angkor site were illustrated by the detection of quantitative pre-cursor movement anomalies.
It is now clear that water pumping either by public or private does not cause an immediate region-wide surface subsidence threatening the sustainability of monuments up to 2013; instead, imperceptible influence from seasonal variations of the groundwater table is a contributory factor to the long-term deterioration function as illustrated by our model.
The sustainability of Angkor site
The study highlights the need, prioritized by the ICC for Angkor since 2012, to shift the traditional temple-based conservation into an environmental approach to the conservation of the Angkor World Heritage site. As an entire geo-ecological system, impacts of geology, hydrology, climate change and anthropogenic activities interact and combine with one another in Angkor.
APSARA did a wise safeguarding measure (rehabilitation the ancient hydraulic system) for the sustainable development of the site.
Keeping a stable groundwater table by increasing replenishment sources and opportunities in areas surrounding the temples throughout the year should be seen as an approach that would minimize seasonal groundwater variations and their negative impacts on unbalanced structural motion of monuments over the medium-to-long term.
