Drought ranks first with regard to the number of people affected due to a natural hazard. Similarly, droughts have significant negative impacts on livelihoods and economy of Kazakhstan. Despite existing demand in spatial information on droughts extent and severity, in Kazakhstan in particular, there has been so far a very limited research conducted on drought hazard assessment. Moreover, there is a lack of studies on diverse responses of vegetation and LULC to droughts as well as the effects of droughts on both environment and socio-economic situation in the region. In this study, we aimed to close up the existing informational gap by analyzing and characterizing drought hazard in Kazakhstan using satellite Remote Sensing time series for the monitoring period between 2000 and 2016. First, we calculated Vegetation Condition Index (VCI) and Standardized Enhanced Vegetation Index anomalies (ZEVI) based on 250 m Moderate Resolution Imaging Spectroradiometer (MODIS) Enhanced Vegetation Index (EVI) time series. Second, we assessed vegetation cover changes for the observation period. Third, we analyzed different characteristics of the drought hazard as well as spatial distribution of the drought-affected areas within the country. The results confirmed that drought is one of the biggest challenges for the country’s agricultural sector and people’s livelihoods as during analyzed period of time there was no single year without occurrence of drought events. The obtained maps showed that drought hazard conditions were observed every year, though the areal coverage of the drought conditions largely varied between the analyzed years, while the longest sequence of the droughts were observed within grasslands and dryland forests. The calculated drought indices indicated that in years 2000, 2008, 2010, 2011, 2012, and 2014, more than 50% of the area of the country were affected by drought conditions of different severity with the largest droughts in terms of the areal spread occurring in 2012 and 2014. The effective drought impact mitigation and management for the country should include various strategies targeting improvement of crop and livestock productivity, agricultural diversification, and more effective irrigated water management. At the same time, measures that aim to reduce people and environmental vulnerability to droughts, such as poverty reduction and land rehabilitation practices, should be also considered. We concluded that the pre-requisite of successful implementation of drought hazard and risk mitigation strategies is availability of spatially-explicit, timely and reliable information on drought hazard. This suggests the necessity of incorporation of Remote Sensing based drought information, as was demonstrated in this paper, in the national drought monitoring system of Kazakhstan.

Area A - Board 60

Coarse spatial resolution NDVI data are widely used to monitor vegetation conditions in near real-time (NRT) for food crises early warning purposes. Different types of NDVI anomalies are typically employed to assess the current state of vegetation as compared to previous years. Timeliness and accuracy of such anomalies are critical factors to an effective monitoring. Temporal smoothing can efficiently reduce noise and cloud contamination in the time series of historical observations, where data points are available before and after each observation to be smoothed. With NRT data, smoothing methods are adapted to cope with the unbalanced availability of data before and after the most recent data points. These NRT approaches provide successive updates of the estimation of the same data point as more observations become available. Anomalies compare the current NDVI value with some statistics (e.g. indicators of central tendency and dispersion) extracted from the historical archive of observations. With multiple updates of the same datasets being available, two options can be selected to compute anomalies, i.e. using the same update level for the NRT data and the statistics or using the most reliable update for the latter.
In this study we assess the accuracy of three commonly employed 1 km MODIS NDVI anomalies (standard scores, non-exceedance probability and vegetation condition index) with respect to (1) delay with which they become available and (2) option selected for their computation. We show that a large estimation error affects the earlier estimates. With regards to the preferable option to compute anomalies, we empirically observe that it depends on the type of application (e.g. averaging anomalies value over an area of interest vs. detecting “drought” conditions by setting a threshold on the anomaly value) and the employed anomaly type. Finally, we map the spatial pattern in the magnitude of NRT anomaly estimation errors over the globe and relate it to average cloudiness.

Area A - Board 61

Germany – like most parts of Europe – experienced several dry and hot spells in the recent past. Also, the year 2018 was characterized by extraordinarily extreme low rainfall rates and high temperatures, which resulted in heavy yield losses. Time series analyses of earth observation data enable the characterization of past drought events over large temporal and spatial scales. Within this study, MODIS Enhanced Vegetation Index (EVI) (MOD13Q1) 250m time series of the vegetation period (March to August) were investigated for the period 2000 to 2018 in order to compare the spatial and temporal drought development and characteristics in 2018 to other dry and hot years, like the drought year 2003. For each 16-day composite the difference from the long-term median was calculated. These were then averaged to investigate monthly and seasonal anomalies. For the analysis, only vegetated surfaces according to the CORINE land cover map 2012 were included. Areas, for which the land cover type changed within the last 18 years, were excluded. In addition, the data was masked and only good quality pixels according to the MODIS pixel reliability layer and positive EVI values were retained. The temporal and spatial inter- and intra-annual patterns of the anomalies were analyzed for all land cover types combined and for cropland, forest and grassland individually. The year 2018 showed negative anomalies in a similar magnitude to 2003, 2006 and 2013 especially in the summer months, which was particularly obvious within grassland and cropland areas in Germany. The spatial pattern of 2018 showed high regional variation with north-eastern Germany mainly affected in June, north-western parts in July and western Germany in August. The temporal development of the vegetation condition in 2018 revealed positive anomalies in April and May and strongly negative anomalies in June, July and August. In contrast to that, the year 2003 showed negative anomalies at the beginning and end of the vegetation period. The study reveals that the year 2018 was an extremely dry and hot year in Germany and amongst the worst within the study period. However, the spatial pattern and temporal development between the drought years differ. Including more datasets, like precipitation and soil moisture, might enable a more complex assessment of droughts and drought effects.

Area A - Board 62

By the launch of Sentinel 2A and 2B, and their suitable revisit time, new opportunities has just opened for examining damages of vegetation on an impressive spatial resolution. These characteritics of Sentinel 2 makes it possible to analyse the consequences of a specific damage with a relatively short response time and to monitor the post-damage status of the area in discrete time points. In this way we can perform a time series analysis of the damage consisting of: monitoring the status of the location before the damage, analysing the effects right after the damage and monitoring the effects after the damage until the end of the vegetation period.

In our analysis we present a case study for the effects of hail in an agricultural area. We used various spectral indices in order to characterize the changes between the status of vegetation before and after the damage. In order to gain information about the number of actually useful images throughout the vegetation period, we examined how many images were available in the area of interest that is free from any influencing factors (for instance, cirrus or clouds). Then we analysed a time series of useful images taken before the exact time of damage and another time series of useful images taken after the exact time of damage. In this way we could compare the development pattern of the vegetation before and after the damage in true color images and in images of various spectral indices. We also compared the biomass status of the images taken before and after the damage as close to the date of the damage as possible.

Our aim is to present an effective method of estimating extent of the area affected by the damage and the grade of the effect. These estimations can be valuable inputs in the workflow of any research projects or organizations dealing with the effects of natural hazards, like professionals in insurance industry or environmental analysis.

Area A - Board 63

In this summer, global high temperature has causes many wildfires over the world, even Nordic Sweden also experienced several forest wildfires. Due to its high temporal resolution and global coverage, Sentinel-1 SAR data has been wildly used in land, sea and forest monitoring. SAR also has been successfully exploited to estimate the fire affected areas and their effects on the local environment because it is sensitive to the changes in forest vertical structure, and it is independent from cloud cover or solar elevation. In this study, an innovative method is developed to estimate radar burn ratio (RBR) based on atrous Convolutional Neural Network (AtrousCNN), which is a powerful tool in dense prediction tasks. Atrous convolution makes it possible to explicitly control the resolution at which feature response and computed with deep CNN, and it has the advantage to effectively enlarge the field of view of filters to consider larger context without increasing the number of parameters or the amount of computation.

Thomas wildfire is taken as the study area, and a pair of SAR GRD data are used to estimate the RBR, including one pre-fire date (20171204) and one post-fire data (20171228). For training an atrousCNN model, we take optical difference normalized burn ratio (dNBR) as referenced Pseudo labels, and both two dates of SAR-VV and VH data are stacked as the input. The training samples are randomly selected to learn an astrousCNN model, which then is tested on the whole image. The experimental results are listed in attached figure, in which (a) is VV-logRt map, (b) is VH-logRt map, (c) is referenced dNBR map and (d) is the estimated burn probability by atrousCNN based on VV and VH. From which, we can find that atrousCNN
highlights the burn areas and suppresses the optical dNBR map, and outperforms than VV or VH-based log-ratio difference operator. And atrousCNN keeps the boundaries of burnt area well and can greatly restrain the false-alarm pixels, which is critical in mapping burnt area. In this research, we demonstrate the effectiveness of atrousCNN in estimating RBR and further confirm the potential of radar backscatter on forest wildfire detection.

Area A - Board 64

As fire-prone conditions are predicted to increase resulting in record forest fire seasons, it is critical to understand wildland fire patterns and create monitoring systems for these events. This is becoming the major focus of governments around the globe and requires substantial effort towards obtaining reliable targeting and time sensitive intelligence related to wildfire behaviour. Addressing the amount of fuel being consumed is a critical element in developing the information necessary to carry out action on the wildfire. We can begin to estimate fuel consumption through calculated fire indices such as Fire Radiative Power (FRP), which is related to the measure of radiant energy output from the source of the fire along with the smoke generated, and aids in calculation of total Fire Intensity (FI).

Remote sensing allows for multi-scale (air to space) data to be captured over naturally-occurring and controlled-fires to better understand wildland fire behaviour and build fuel consumption models. There have been international endeavours to deploy sensors on airborne and space platforms (e.g. Sentinel-3 SLSTR) to measure these fire characteristics. The National Research Council of Canada (NRC) Flight Research Laboratory has carried out airborne campaigns to collect thermal imaging of wildland fires in Northern Ontario from NRC’s Twin Otter aircraft. The campaigns were able to successfully image multiple Being Observed Burns (BOB) fires using onboard MidWave InfraRed (MWIR) broadband data sensors. These data sets were all collected under well-constrained conditions. From this data we will show calculated FRP taking into account a variety of conditions (changing winds, humidity, day vs. night, etc.) and how weather-based fire indices can be used to provide a corrected FRP. Further, the variances of the FRP are investigated as a function of spatial size where single pixel calculations and standard deviations are compared against image and mosaicked image data sets. The scalability of the FRP as a function of size is fundamental to show how this method may be applicable to orbital platforms with MWIR bands such as the Sentinel 3. This validation helps support top-down wildfire monitoring as satellites provide large-scale and time-controlled observations, to ultimately calculate better fuel-consumption rates.

Area A - Board 65

Prolonged heat and absence of rainfall drive vegetation into water stress conditions that lead to an increase of its temperature. The latter can be detected by remote sensing measurements in the thermal infrared as a deviation of observed land surface temperature (LST) from climatological values, that is as an LST anomaly. Since stressed vegetation is more prone to fire, a relationship is expected between LST anomalies and fire characteristics such as burned area and duration. A wide number of static and dynamic factors control fire characteristics, including topography, land cover, climate, weather and anthropic activity. A few of them are drivers of vegetation canopy temperature. The construction of specific predictive models is based on the selection of those factors relevant to a specific geographic area/ecosystem, usually identified at national level for fire danger management purposes. However, it would be relevant to investigate the role that remote sensing measurements might have in such models as an individual parameter, i.e. whether LST anomalies are a covariate of fires burned area and duration.
This research approached the outlined knowledge gap through the analysis of a dataset of forest fires in Campania, Italy (13595 km2) against estimates of LST anomaly. Campania is one of the most fire-affected regions in Europe, and economic losses due to forest fires put an additional pressure to its depressed economy. A collection of daily diurnal Aqua-MODIS LST data from 2003 to 2017 was used for this research (product MYD11A1, collection 6). The Harmonic Analysis of Time Series (HANTS) algorithm was initially used on the entire dataset to construct a pixel-wise daily climatology of LST based on the longest available series of complete annual data. HANTS was also used to compute individual annual models of LST data, with this minimizing the effect of varying observation geometry, cloud contamination and sensor inaccuracies while retaining yearly variation. LST anomalies where thus quantified as the difference between LST annual models and LST climatology.
Fire data were intersected with LST anomaly maps to associate each fire with the LST anomaly value observed at its position on the day previous to the event. Further to this step, the closest probability distribution function describing burned area and fire duration were identified against a selection of parametric models through the maximization of the Anderson-Darling goodness-of-fit. Results show that in the study area log-transformed burned area is described by a normal distribution, whereas log-transformed fire duration is closer to a generalized extreme value (GEV) distribution. Parameters of the identified distributions conditional to LST anomaly where then determined along their confidence intervals.
Observed average LST anomaly is 1.3 K. 77% of fires occur when a positive LST anomaly is observed. The parameters of the normal distribution of burned area were evaluated in each of the ten decile bins of LST anomaly values, i.e. conditional to LST anomaly values. These show a clear trend with increasing LST anomaly; significance of this observation was verified through a likelihood ratio test. A similar result was observed with the parameters of the GEV distribution of fire duration. This confirmed that LST anomaly is a covariate of both burned area and fire duration. As a consequence, conditional probabilities of extreme events were observed to increase with increasing deviations of LST from its climatology values.
This investigation confirms, in the study area, the stated hypothesis that LST anomalies affect forest fires burned area and duration, clearly identifying an opportunity to model the statistic distribution of these two characteristics, and predict the probability of extreme events.

Area A - Board 66

Four earthquakes with Mw > 6 and numerous smaller earthquakes struck Southern Mexico in September 2017. Another Mw > 7 struck in February 2018. The sequence started with the 8 September Mw 8.2 Tehuantepec Earthquake beneath the Gulf of Tehuantepec offshore Chiapas and Oaxaca. We study the ruptures of these earthquakes with geodetic and seismic methods to examine the rupture extents and locations. The Caltech-Jet Propulsion Laboratory (JPL) Advanced Rapid Imaging and Analysis (ARIA) data system performed rapid automatic processing of the Sentinel-1A and -1B synthetic aperture radar (SAR) data to interferogram products. We also looked at whether this M8.2 earthquake triggered the four subsequent large M>6 quakes in southern Mexico to improve understanding of earthquake interactions and time-dependent risk. All four large earthquakes in September 2017 were extensional despite the subduction of the Cocos plate at the convergent plate boundary. In contrast, the 16 February 2018 Mw 7.2 Pinotepa earthquake near Pinotepa Nacional in Oaxaca was a thrust event on the subduction interface. The traditional definition of aftershocks: likely an aftershock if it occurs within two rupture lengths of the main shock soon afterwards. Two Mw 6.1 earthquakes, half an hour after the M8.2 beneath the Tehuantepec gulf and on 23 September near Ixtepec in Oaxaca, both fit as traditional aftershocks, within 200 km of the main rupture. The 19 September Mw 7.1 Puebla earthquake was ~600 km away from the M8.2 shock, outside the standard aftershock zone. The Pinotepa earthquake was about 350 km away from the Tehuantepec rupture but was in an area of Coulomb stress decrease from the M8.2 quake, so it seems unlikely to be a regular aftershock.

Geodetic measurements from interferometric analysis of synthetic aperture radar (InSAR) and time-series analysis of GPS station data constrain finite fault static slip models for the M8.2, M7.2, and M6.1 Ixtepec earthquakes. We include open-ocean tsunami waveforms for the M8.2 inversions. We analyzed InSAR data from Copernicus Sentinel-1A and -1B satellites and JAXA ALOS-2 satellite. Our Bayesian (AlTar) static slip model for the M8.2 quake shows significant slip extended > 150 km and possible 220 km NW from the hypocenter. There is a high probability that the slip extended to depths of at least 70 km indicating slab pull stress state. Our AlTar slip model for the M7.2 Pinotepa thrust earthquake is similar to the USGS NEIC FFM with all of the slip confined to a very small (10-20 km diameter) rupture. The Pinotepa earthquake ruptured a portion of the Cocos megathrust that has been previously mapped as partially coupled and shows that at least small asperities in that zone of the subduction interface are fully coupled and fail in high-stress drop earthquakes. The previous 2012 Mw 7.4 Ometepec earthquake is another example of asperity in the partially coupled zone but was not imaged by InSAR so the rupture extent is not so well constrained. Inversions for the M6.1 Ixtepec normal quake confirm shallow depth in the upper-plate crust and show centroid is about 30 km SW of the preliminary NEIC epicenter but consistent with cluster relocations. Similarly, the preliminary NEIC epicenter for the Pinotepa earthquake was about 40 km away (NE) from the rupture imaged by InSAR. The NEIC updated epicenters and Mexican SSN locations are closer to the InSAR-constrained location for both earthquakes, because they use regional seismic data.

Area A - Board 69

North-East of Baghdad, at the Iran-Iraq border a magnitude 7.3 earthquake took place in the evening of 12 November 2018. With more than 600 persons killed, it was the deadliest earthquake of 2017. The earthquake was located in the zone of continental collision between the Arabian and Eurasian Plates.
In our work we used Sentinel-1 data to map the co-seismic displacement. Our focus is on the assessment of different methods to retrieve the east-west, up-down, and north-south components using differential interferometry as well as offset-tracking.
To avoid phase jumps between subsequent bursts in the co-seismic differential interferograms it was necessary to adjust the co-registration for the co-seismic displacement. To unwrap the differential interferograms, a minimum-cost flow algorithm was used. In order to use the resulting line-of-sight (LOS) displacement component to retrieve the East-North-Up (ENU) displacement vector, the look vector was calculated for each pixel of the geocoded result in map coordinates. In addition, offset-tracking was applied to estimated the range and azimuth offset fields for both the ascending and descending orbit co-seismic pairs. To optimize the quality of the estimated offsets and avoid aliasing effects, relatively large matching windows were used and an oversampling of the SLC data in both range and azimuth direction was applied. The six displacement components that were estimated were then used in different combinations to estimate the ENU displacement vector:
1) Using the ascending and descending orbit DInSAR LOS displacements.
2) Using the ascending and descending orbit DInSAR LOS displacements and the ascending and descending orbit azimuth offsets.
3) Using the ascending and descending orbit range and azimuth offsets without DInSAR LOS displacements.
The relation between the components determined using DInSAR and offset-tracking, the vectors indicating the direction of these components, and the ENU displacement vector can be expressed as
[A][x] = [b]
The rows of the matrix A are the look-vectors for each of the observation points, [x] is the solution in ENU coordinates, and [b] contains the LOS deformations. The displacement vector in ENU coordinates was estimated for each point in the scene based on the selected displacement components using Singular Value Decomposition (SVD). The Least-Squares solution permits estimation of the sensitivity to measurement errors and can reveal inconsistency between partially redundant measurements. Estimated errors (one standard deviation) for the 3 cases are indicated in Table 1.

Area A - Board 70

In 2017, two Mw 7.3 and Mw 6.2 earthquakes hit southeast and northwest of Iran, respectively. The Mw 7.3 Sarpol-e Zahab earthquake occurred on 12 November 2017 close to Iran-Iraq border and was the strongest seismic event to ever hit the region in the past century. The Mw 6.2 Hojedk earthquake occurred on 12 December 2017 in the Kerman region between Kuhbanan and West Lut faults. The goal of our study is to better understand the regional tectonics in NW and SE Iran by studying source parameters of these two earthquakes.
Coseismic deformation fields of both events were well constrained by interferograms constructed from Sentinel-1 TOPS and ALOS-2 ScanSAR data. SAR interferometry is applied on the available SAR data to derive detailed pattern of coseismic deformation. The combination of offset tracking and burst overlap interferometry is also exploited to resolve the displacement map in areas affected by strong decorrelation due to large gradient. The Bayesian inversion is then applied on InSAR measurements to derive source parameter and slip distribution of the coseismic rupture planes.
Sentinel-1 and ALOS-2 observations of the Mw 7.3 Sarpol-e Zahab earthquake showed a wide deformation field of about 90x70 km in an area between High Zagros Fault (HZF) and Mountain Front Fault (MFF) with a peak uplift and subsidence of 70 cm and 35 cm, respectively, in the line-of-sight (LOS) direction from the satellite to the ground. Geophysical inversion suggests that the earthquake was generated by a blind ENE oblique thrust faulting with maximum slip of approximately 5 m at a depth of 17 km. The Hojedk earthquake caused a zone of clear surface rupture with the length of 8 km, causing decorrelation in Sentinel-1 interferometric measurement in the near field area. ALOS-2 L-band interferogram, however, were less affected by decorrelation near the fault rupture. Combination of ALOS-2 and Sentinel-1 help us resolve the geometry and slip distribution of this oblique thrust earthquake in SE Iran. The main rupture was caused by a shallow rupture concentrated at the depth between 1 and 6 km with a maximum slip of 6 m. We further complement coseismic source parameter analysis by investigating approximately one year of post-seismic deformation, analyzed through time-series analysis of Sentine-1 interferograms to derive the relation between coseismic slip and afterslip.

Area A - Board 71

Sinkholes are widespread geologic hazards that are mainly formed in water soluble carbonate bedrocks such as limestone, dolomite or gypsum. Sinkhole formation causes the surface to subside or collapse without any warning. Delineating sinkholes is important for understanding hydrological processes and mitigating geological hazards in karst areas.
In this study, we use observations from multi-sensor SAR satellites to analyze morphological features and ongoing deformation related to cover collapse sinkholes in Hamedan, Iran. First using time-series of TerraSAR-X (TSX) / TanDEM-X (TDX), we develop a methodology to delineate and explore geomorphologic features of sinkholes. An adaptive sinkhole analysis methodology is developed by using bistatic interferometry, wavelet analysis, Canonical Correlation Analysis (CCA) and watershed algorithm to analyze geomorphometric characteristics of sinkholes. Bistatic interferometry is initially applied on multi temporal TSX/TDXs data to produce DEM. A wavelet based refinement approach is implemented during the interferometric processing to reduce the baseline bias effects and align the interferometrically derived DEMs together. The multi temporal DEMs are then successfully stacked using CCA to reconstruct a higher quality DEM. Finally, feature extraction using the watershed algorithm is applied to precisely delineate depression areas. We evaluate the performance of our method over 11 sinkholes in Hamedan plain in western Iran and show that it enables us to effectively distinguish sinkholes from other depression features of the basin. Comparison of information extracted for the shape and depth of sinkholes using TSX/TDXs data shows good consistency with field observations (Std < 1 m)
Second, we supplement the above analysis by applying the Small BAseline Subset (SBAS) InSAR time series technique to assess ongoing deformation in areas prone to sinkhole formation. All the available SAR data acquired from a variety of sensors including Envisat, ALOS, TerraSAR-X, and Sentinel-1 are entered for SBAS processing. The results demonstrate that the formation of lots of sinkholes is influenced by land subsidence, affecting the region over 100 km with the maximum rate of 4-5 cm/yr from 2003 to 2018.

Area A - Board 72

Gas extraction from oil fields may lead to induced earthquakes in the adjoining regions. This happens due to a gradual build-up of stress due to rock compaction, which causes an earthquake if the sheer stress goes beyond a critical limit and the stressed fault fails [1]⁠. Although the magnitude of such induced earthquakes is relatively smaller as compared to high magnitude earthquakes caused due to tectonic plate movements, they still cause movement of the land surface which leads to hazards such as settlement and facade fractures of individual buildings. Moreover, the gradual buildup of the stress is responsible for land surface deformations which precede the occurrence of the earthquake event. These land surface movements can be measured efficiently and precisely using the geodetic estimation techniques of Synthetic Aperture Radar Interferometry (InSAR) [2]⁠ and time series InSAR [3]⁠.
The Groningen gas field in North-Eastern Netherlands has been known to be one such example of this phenomenon. This study focuses on identifying deformation patterns in relation to induced earthquakes such that the identified patterns can be used to detect precursory changes, assess post-disaster damage and even predict impending disasters. For this purpose, we employ a stack of SAR images, generate deformation time series of all InSAR measurement points, and attempt to categorize those points in terms of different deformation models. The InSAR deformation time series is attempted to be categorized using a combination of a machine learning based method, i.e. multivariate time-series pattern recognition technique [4]⁠ and multiple hypothesis testing (MHT) method based on a library of physical deformation models [5]⁠. In the first part of the analysis, the deformation time series and time series of other contextual datasets, e.g. ground water table and surface temperature, is used for breaking the combined time series into segments according to some statistical situations. The segments are then clustered together spatially and temporally based on similar context. MHT is then used to first identify a unique model to fit into the data points of individual clusters and hence, to identify anomalous points lying in individual clusters which do not follow the identified model. In this way, this method is intended to classify InSAR measurement points, and detect spatial and temporal anomalies and the associated deformation patterns with respect to the earthquake events.

References:

[1] T. Candela, B. Wassing, J. ter Heege, and L. Buijze, “How earthquakes are induced,” Science (80-. )., vol. 360, no. 6389, pp. 598–600, 2018.
[2] M. Moro et al., “New insights into earthquake precursors from InSAR,” Nature, vol. 7, no. 12035, pp. 1–11, 2017.
[3] L. Chang, O. Ku, and R. F. Hanssen, “Identification of deformation pattern changes caused by enhanced oil recovery (EOR) using InSAR,” Int. J. Remote Sens., vol. 1161, pp. 1–11, 2018.
[4] S. Spiegel, J. Gaebler, A. Lommatzsch, E. De Luca, and S. Albayrak, “Pattern Recognition and Classification for Multivariate Time Series,” in Proceedings of the fifth international workshop on knowledge discovery from sensor data (SensorKDD’11), San Diego, USA, 2011, pp. 34–42.
[5] L. Chang and R. F. Hanssen, “A Probabilistic Approach for InSAR Time-Series Postprocessing,” IEEE Trans. Geosci. Remote Sens., vol. 54, no. 1, pp. 421–430, 2016.

Area A - Board 73

The improvement of processing algorithms, the enhancement of computing capabilities and the development of several satellite missions belonging to different national and international space agencies are currently offering a wide spectrum of applications, providing effective solutions for hazard prevention, mapping and monitoring.
Currently, Satellite SAR interferometry has proven to be an operational monitoring tool at different stages of the life cycle of a mine (i.e. prospecting and exploration, development, extraction and closure/reclamation) and for different applications, such as: detection and control of hazard-prone areas and risk management.
In fact, satellite-based techniques has clear peculiarities over classical monitoring methods; the first one is the capability to perform historical analyses investigating past phenomena, the second is the capability to investigate wide areas with high accuracy, inferring the complete displacement pattern for the investigated scenario.
This work is focused on the application of different remote sensing techniques for the identification and monitoring of slope instability processes in mining areas exploiting SAR satellite images. Specifically, the results obtained on two case studies are presented: the Chuquicamata Mine (Calamata, Chile) and the Bingham Canyon Mine (Salt Lake City, Utah), two of the largest man-made excavations and open pit copper mine in the world.
Chuquicamata mine has been analysed with A-DInSAR technique using a stack of Sentinel-1 images acquired between 2017-2018. The high performances in terms of temporal resolution and the C-band sensor of Sentinel-1 constellation allow reaching a high spatial distribution of measurement points, allowing the accurate identification of unstable areas inside the excavation site.
The past deformations affecting the Bingham Canyon Mine have been analysed using a stack of 44 high-resolution RADARSAT-2 images acquired between October 2009 and April 2013. The activities have been performed in the framework of the SOAR-EI Project “Landslides forecasting by means of satellite InSAR” (between the Canadian Space Agency and Sapienza University of Rome). The change detection analysis, based on the processing of amplitude SAR images, was performed to detect instability phenomena and excavation phases, while A-DInSAR analysis was carried out to detect the spatial-temporal deformational pattern before the occurrence of the 10th of April 2013 massive landslide (65 million cubic metres ).
The results obtained represent interesting examples of exploitation of both open and commercial EO data as a valuable resource of information for mining companies and people engaged in the risk management.

Area A - Board 74

The Hengchun ridge is a major concern for the Taiwan Natural Hazards as the Nuclear Power Plant N°3 is settled above the southern tip of the Hengchun Fault and faced a major Taiwan earthquake in december 26th, 2006 (depth: 44km, ML=7.0). It is consequently needed to much better constrain the active Hengchun Peninsula tectonic structures and their evolution through time in order to prevent major destructions in the near future.
Consequently, we propose herein firstly to up-date the so numerous and different geological and geomorphological mappings of the Hengchun Peninsula using the High Resolution remote sensing images and compare them with pre-existing works. We used as basic references both the CPC and CGS pre-existing geological maps, and on the other side GIS geological mapping issued from classical photo-interpretation of our very high resolution newly acquired UAS ortho-photos (<10cm GR) and photogrammetrically processed HR Digital Terrain Model characterized by a planimetric (X,Y <10cm), and vertical precision (<50cm) validated by our field studies. In order to complete the gaps we fill-in our geological interpretation with Sentinel 2 and SPOT images.
Secondly, our new synthetic up-dated geological/structural mapping of the Hengchun Peninsula is then compared to JAXA ALOS-1 multi-temporal SAR Interferometry results (PSInSAR and SmBAS e.g. Champenois, 2011, Deffontaines et al., 2017, 2018). The latter reveal interseismic velocities of the tectonically active structures. It leads us to much better analyse, characterise and quantify the active tectonic displacements toward the LoS during the ALOS-JAXA monitoring time-series (2007-11). The latter had been validated by levelings comparisons and three GPS fixed stations.
We then acquired new geological, geomorphological and tectonics observations on the Hengchun Peninsula at a global and local scales e.g. Hengchun and Kenting active Faults, or the general tilting of the whole Peninsula (West uplifted, East. subsidence) contrasting to the published previous studies (Giletycs, 2016 and Giletycz et al., 2017).

Area A - Board 75

Acquisitions of the Sentinel-1 satellite are processed and comprehensively analyzed to investigate the ground displacement during a three-year period above two gas storage sites (Lussagnet and Izaute) in Southwestern France (Figure 1). Despite quite low vertical displacements (between 4 and 8 mm) compared to the noise level, the local displacements reflect the variations due to charge and discharge during summer and winter periods, respectively. A simplified mechanical model can explain these displacements at both storage sites. Nonetheless, despite a general consistency, some features in the Lussagnet case that cannot be explained by the reservoir pressure variations are further investigated. The 10-day SMOS surface soil moisture (SSM) products for descending overpasses between 18 October 2014 and 26 October 2017 in the grid cell of the studied zone (Figure 1) are used to calculate the average of the median, minimum and maximum SSM values for each Sentinel-1 period (12 days). Using a wavelet-based analysis, we show there is an uplift in the Lussagnet zone that contrasts both in phase and period with the seasonal deformation due to the gas exploitation and that is related both to the rainfall and to the surface soil moisture. This transient surface deformation has a similar amplitude than the cyclic deformation due to gas exploitation (about 5 mm). This last result is consistent with the swelling of the clay layer described in the geological logs extracted from the BRGM subsurface database (BSS). This work provides therefore new insights on the ground deformation using a three-year integrated monitoring of a gas storage site. As a perspective of this work in the framework of future European Ground Motion Services (EU-GMS), this method coupling both satellite acquisitions (Sentinel-1 and SMOS) and the BSS database may be generalized to improve the global French shrink/swell hazard map based on the 1:50 000 geological map.

Area A - Board 76

Jharia coalfield, located in Dhanbad district of Jharkhand state, India, is the largest coal reserve in India and one of the most extended in the world. This area lies in Damodar valley along the north of Damodar river. It is well known not only for prime coke coal production but also for the numerous coal fires which frequently affect this area. Both the mining activity and coal fires activating in open-pit and underground mines are causing dangerous collapses here.
The present work is focused on the analysis of Jharia coalfield through space-born Synthetic Aperture Radar (SAR) data, exploiting the frequent acquisitions and the wide extension of the data. In detail, freely available Sentinel-1 data have been processed over a study area of approximately 600 km2. Both ascending and descending datasets have been considered, spanning the time interval October 2016 – August 2018, with a revisiting time of 12 days. Integration of different data analysis techniques was done, covering both the amplitude and the phase of the acquired images. Analysing multi-temporal changes of the signal intensity and coherence, it was possible to distinguish between active and non-active open mines. Moreover, the used multi-temporal interferometric technique (SBAS) allowed to identify surface deformations affecting the area. Thank to the availability of data acquired with ascending and descending orbits, deformation data have been projected along the vertical and east-west directions. It was used to classify displacements according to their different genetics e.g. sliding materials, area of active surface mining subsidence and sinkholes induced by underground cavities. In addition, we identified excavated areas and waste material storages, using the elevation correction retrieved from SBAS processing in respect to the low resolution reference DEM,. During post processing analysis, results obtained from the different approaches have been correlated. This could lead on detecting deformation affecting open mines and underground mines. Exploiting the spatially extended and temporally dense results could help on reducing the hazard affecting this sensitive areas, reducing possible catastrophic events.

Area A - Board 78

In the framework of the DRAGON-4 Project, funded by the European Space Agency, the National Institute of Geophysics and Volcanology of Rome (INGV, Italy) and the Northeastern University of Shenyang (China) collaborate to study the surface movement over industrial regions in Northeast China. The traditional heavy industrial base, especially in the Benxi-Anshan-Shenyang-Fushun (BASF) region, is playing an important role in the economic development of the region, although severe consequences on the local environment are taking place due to the continuous mining activities. Various geo-hazards, such as subsidence, landslides, ground breakage and building inclinations, have been occurring for decades. The continuous monitoring of the effects of the mentioned phenomena is thus of great importance for the safety of the local population. Taking advantage of the availability of dense remote sensing data sets is possible to analyze the geo-hazards and their environmental impacts in the region; and then making forecast about their occurrence in the future and providing support for disaster prevention and damage reduction.
We adopted multi-temporal InSAR methodologies able to analyze the spatial and temporal deformation over large areas. In this study, we use time-series InSAR results from multiple stacks (from ascending and descending orbits) and different sensors to monitor gravitational deformations and subsidence in urban areas.
We adopted both the Persistent Scatterers Interferometry (PSI, [1]) and Small Baseline Subset (SBAS, [2]) techniques, for processing SAR data stacks acquired from Sentinel-1A/B (C-Band) and COSMO-SkyMed (X-Band).

Fig.1a – Fushun open pit mine area: CSK ground velocity map retrieved by
the PSI technique along the descending orbit

Fig.1b – Fushun area, CSK descending displacement time series

In Fig.1, the descending CSK ground velocity map for the Fushun coal pit mine and an example of the displacement time series retrieved using the PSI technique are shown. We considered 40 images acquired between 2015 February and 2017 May. The SRTM DEM (90m posting) was used to remove topographic contribution.
We also studied the same areas (Shenyang city and Fushun pit mine) analyzing the Sentinel-1 datasets provided by ESA, along both the tracks. The considered time interval spans from June 2015 to September 2018 for the descending orbit and from April 2017 to September 2018 for the ascending case, respectively.

Fig.2 – Fushun area, descending mean ground velocity from S1 data

The retrieved mean ground velocity maps confirm the results from CSK data. Subsidence phenomena are still ongoing reaching values higher than -120 mm/yr inside the mine and -60 mm/yr at its edges. Some small areas (localized pixels) show positive values (uplift) probably due to stockpile of excavation debris or processing waste material.
We used the CSK data even to monitor the ground movements occurred close to the newly-built Shenyang South Railway Station and the nearby high-speed railway tracks (Fig.3). The data used in this study area are 18 X-band SAR images acquired from two different orbits from March 2016 to October 2016. The deformation rate maps in Fig.3(a) and Fig.3(b) are obtained from the two different SAR frames, overlapping each other on the area of interest, which show very good consistency. As we can see, the largest deformation happens on the rail tracks near the southern part of the platform, with velocities varying from -51mm/yr to -16mm/yr. Instead, the deformation of other parts of the stations is smaller, with deformation rate varying from -16mm/yr to 0mm/yr. This is because the construction of Shenyang South Railway Station was completed in August 2015, therefore the accumulation of deformation should have slowed down during the time span of our monitoring.

Fig.3 CSK mean ground velocity maps over Shenyang South Railway Station: (a) eastern frame, (b) western frame.

Our results confirm that the heavy industrial exploitation of mines and water pumping in the BASF region of Northeast China cause clear and strong ground deformation effects of high potential impact on the local infrastructures and population. The use of multiple stacks, from different sensors, of InSAR data allows to monitor these phenomena with an accuracy and a temporal sampling not possible otherwise.
By now, the use of EO products plays a fundamental role to monitor natural and man-induced hazards and to support Disaster Risk Management providing an important tool for local and national organizations.

Area A - Board 79

High relief active volcanic areas are frequently affected by surface changes due to eruptive product as lava flow or fall out deposit as well as morphological changes. Customary methods of deriving elevations and a rapid mapping of surface changes may be prohibitive. To overcome this, two pairs of stereo Pleiades imagery collected in July 2015 over Mt. Etna volcanic area (Italy) were considered. The Pleiades stereo pairs (1&2 and 2&3) were processed using the NASA Ames Stereo Pipeline (ASP), a suite of open source automated geodesy and stereo-photogrammetry tools intended to process stereo imagery with or without accurate camera pose information. The workflow involves image pre-processing, feature matching and 3D reconstruction. The image processing step prepares the data for feature matching by updating the sensor ephemeris / attitude information for two or more input images. The feature matching procedure prepares the data for 3D point cloud generation. Each pair of stereo satellite imagery was run with two algorithms, one using a local search window (LSW), and the other using a more global matching (MGM) methodology, a modification of the semi-global matching (SGM) algorithm. The first algorithm estimates the apparent motion of the scene points, or disparity, by comparing local windows around each pixel in the stereo pair. This method cannot cope with ambiguities in the image texture such as smoothness or repetitive patterns. The second algorithm compensates for ambiguities by using the smoothness of the disparity map and reducing the two-dimensional smoothness constraint to the average of the one-dimensional line optimization problem. The last step generates a 3D point cloud through triangulation by combining spacecraft ephemeris / attitude information, sensor model and the refined and filtered disparity map of left and right pixel matches. The planimetric root mean square error (RMSE) of all derived 2-meter spatial resolution DEMs ranges from 1 to 2 pixels. The elevations are positively biased of 1.2 m. The best results for both planimetry and altimetry are obtained for pair 2&3.

Area A - Board 80

Since May 3, 2018, Kilauea Volcano in Big Island, Hawaii, had been erupting, causing many fissures and lava flows in Leilani Estates. There was also a Mw 6.9 earthquake occurred in south flank. The eruption and earthquake caused serious damage to life and property in eastern part of Big Island. Requested from the International Disaster Charter, emergency observations by ALOS-2/PALSAR-2 were conducted.
To examine surface changes due to the eruption and the earthquake, Interferometric SAR (InSAR), Multiple Aperture Interferometry (MAI), offset tracking, and intensity analysis using a series of ALOS-2/PALSAR-2 data were performed. PALSAR-2 has been utilized for disaster monitoring, environmental monitoring, and many other fields of applications with the advantages of SAR such as all-weather and day-and-night observation capability. Especially, PALSAR-2 uses L-band microwave, which has great advantage of high interferometric coherence over vegetation area.
The first InSAR and the MAI results using the Stripmap images acquired on February 27 and May 8, 2018 show one meter or more dyke-opening deformation over the Leilani Estates. We also created the polarimetric color-composite image to reveal distributions of lava flow and fissures in Leilani Estates. The image clearly captures some fissures in Leilani Estates in the early stage after the eruption. The second InSAR result was derived from ScanSAR images obtained on January 20 and May 12, 2018, showing the major crustal deformation areas in Leilani Estates, East rift zone and Kilauea Caldera. Especially, the surrounding of Kilauea Caldera significantly subsided due to the decrease of the lava level, which caused the collapse of Halemaumau Crater in Kilauea Caldera. Three-dimensional displacement of the collapse was calculated from the two-pair results from applying offset tracking technique to Spotlight images. The total subsidence was greater than 100 m, which corresponds to the result from LiDAR by USGS. Time-series of polarimetric color-composite images show the spatial and temporal changes in the distribution of lava flows, fissures over Leilani Estates and land expansion due to ocean entry of lava. In last August, the lava drains seemed to terminate.

Area A - Board 81

In July 2011 a volcanic unrest started in El Hierro (Canary Islands). An anomalous seismic activity started and thousands of earthquakes migrated from north to south. The island experienced uplift of several centimeters until the onset of the eruption on 10 October 2011. Seismicity, surface deformation and petrological data indicate the existence of a magma reservoir at about 10–12 km depth where magma evolved until the initiation of the eruption, contributing to the observed island-wide uplift processes. Geodetic studies also point to the existence of some shallower and ephemeral magma accumulation at the flank of the southern rift (4-5 km deep). The characteristics or even the existence of this shallower magma accumulation remains unclear. Here, we conduct a joint interpretation of pre-eruptive GPS and InSAR data to better constrain the magmatic plumbing system at El Hierro. This is the first work about El Hierro eruption which carries out a 3D finite element model inversion, incorporating an irregular topography/bathymetry and mechanical heterogeneities constrained by seismic tomography. The integration of our results with previous geodetic, geophysical and petrological ones opinó to the existence of such as ephemeral and shallower magma accumulation. Furthermore, the new data allow us to constrain the onset of such accumulation phase just before the initiation of the eruption. This study highlights the importance of remote sensing techniques to manage volcanic hazard response during a crisis.

Area A - Board 82

The explosive volcanic eruptions eject a big quantity of pyroclastic products in the atmosphere that need a continuous monitoring because may cause large damages in rural and metropolitan areas and disrupt the surface and air traffic. Remote sensing observations proved to have remarkable potentialities to track the evolution of pyroclastic eruptions as well as to detect slope instability processes induced by volcanic activity.
A remote sensing methodology was applied for observing and quantifying eruptive phenomena of Mount Etna (Italy), developed through processing, integrating and comparing radar data and optical images acquired from different satellite platforms. The approach also integrates images acquired from in-situ visible and thermal video cameras, processed and orthorectify through a dedicated tool. The proposed methodology was tested by analysing data acquired during the 2011-2015 Etna eruptive activity for mapping the evolution of the lava fields emitted from the New South East Crater. Lava flows evolution maps, derived from the analysis of satellite data, are compared and combined with the maps obtained from the ground-based images collected by the permanent ground NEtwork of Thermal and VIsible Sensors located on Mt. Etna (Etna_NETVIS).
This methodology can be applied also to 3D rendering and quantifying of the plume-forming explosive eruptions, in particular to the Galeras vulcano (Colombia). In this case the approach is mapping the evolution of the ash column to calculate different parameters as height, density and direction. The combination of different satellite sensors optimizes and enhances the observational capability of standard surveillance activities. The data acquired from ground- based sensors allows to downscale the information derived from satellite data and to integrate the satellite datasets in case of incomplete coverage or missing acquisitions. The performed analysis enables to assess the integrated capability of ground-based and satellite sensors to monitor the evolution of fast evolving pyroclastic flows propagating along the volcano slopes, also tracking the ash clouds carried by the wind. This activity can contribute significantly to the management of the volcanic crises and to improve the risk assessment.

Area A - Board 83

Pioneristic ground deformation monitoring at Mt. Etna started in the 70’s by mean of Electro-optical Distance Measurements on three networks located on the three main rift zones of the volcano. Later, GPS technique provided a big enhancement of the monitoring, allowing the 3D displacements to be measured over a unique network covering the entire volcano; during the last decades, the initial GPS network has been continuously improved and implemented for achieving higher coverage and precision. From 2001, continuously GPS running stations have been installed and now they covers the entire volcano, providing daily to 1Hz ground deformation data. Since the 90’s InSAR technique has been applied for studying high-detail ground displacements on the volcano; initially, C-Band ESA ERS1/2 allowed the technique to be tested and continuously improved, passing from classic DInSAR interferograms to A-DInSAR processing (pixel time series). They also provided the very first areal information of ground displacements, allowing better constraints in modelling of the magmatic sources and the main structural features to be precisely tracked. Later on, ASAR sensor mounted on board the ENVISAT space vehicle allowed this kind of monitoring to be continued and further improved with a similar 35-days revisiting time with the same orbit (ascending or descending); finally, Sentinel-1A and Sentinel-1B are currently providing valuable data, improving the detail in time, with a revisiting time of 6 days with the same orbit.
The availability of both multiple SAR datasets and GPS stations over Mt. Etna during the last three decade, allowed us to apply the SISTEM integration in order to capture a more complete figure of the ground deformation affecting the volcano.
Here, we summarize the record and the results of the InSAR and especially the GPS and InSAR data integration applied to study and monitor the ground deformation at Etna volcano. The improvements in the feeding and structural framework and some new features revealed by the most recent Sentinel-1A/B constellation.

Area A - Board 84

European forests provide an essential set of services that contribute to climate mitigation and human well-being. At the same time, forests are vulnerable systems because the long life-span of trees limits their ability to rapidly adapt to drastic environmental changes. Climate-driven hazards for forests, such as drought, wildfires, and insect outbreaks, are expected to rise drastically with global warming. As a result, key forest services, such as carbon sequestration and the supply of wood products, could be profoundly affected in the near future.
Quantifying the future risks for forest ecosystem services due to natural disturbances requires the integration of hazard, exposure and vulnerability components. Despite the relevance and urgency of the issue, little is known about the risks multiple climate-related hazards pose to European forests. Substantial knowledge and methodological gaps have hampered so far a quantitative assessment of the effects of future natural disturbances of forest ecosystems.
• Hazard: unsatisfactory representation of natural disturbances in land models. The current generation of land models incorporates a mechanistic representation of biogeochemical and biophysical processes in coupled climate-vegetation systems. However, models do not include a mechanistic representation of natural disturbances and the effects of extreme climate events, thus they can only partially predict the effects of climate hazards on forests. Furthermore, their level of aggregation in plant functional types does not allow capturing tree species-scale processes relevant for impact assessments.
• Exposure: poor integration of human and natural drivers in forest species projections. Future forest ecosystems services depend on where individual tree species will grow. Projections of future forest composition have been provided by employing environmental predictors into statistical models to explain the presence/absence of individual tree species. However, these methods do not account for the pace at which ecosystems will migrate following their natural ecological niches, neither incorporate socio-economic processes, such as forest management practices.
• Vulnerability: limited availability of data and incomplete understanding of forest vulnerability. Despite the recent efforts in developing shared and standardized databases of observed forest disturbances, the coarse resolution at which data are usually recorded (e.g., country level) strongly masks the spatial variability of the phenomena. This has inevitably limited the quantitative evaluation of tree species vulnerability to climate-related hazards. To date, no analytical functions reflecting the degree of susceptibility of trees have been documented in literature. The few pioneering studies on the topic have not been supported by empirical evidences.
To fill this knowledge gaps we aim to investigate the evolution of forest disturbances under various climate scenarios. To do so we will model the long-term changes in species composition, whereas forest vulnerability will be assessed with satellite observations and records of disturbances. By integrating these elements we will produce a range of plausible risk estimates to support the definition of effective adaptation strategies for forests, with the goal of increasing the resilience and long-term stability of these fundamental elements of the European landscape.

Area A - Board 85

At P band, the electromagnetic waves can go through the canopy and interact with the trunks and the ground. In particular, for flat surfaces, it can give double bounce between the trunk and the ground. These double bounces give strong returns on monostatic radar images. They are therefore of a significant interest.
Deforestation means a suppression of the trunks, whatever the cause (fire, cut for culture expansion etc.). This suppression will in turn suppress the double bounce trunk-ground.
The future BIOMASS mission will offer full polarization images of the Earth at P-band. As a preparation to this mission, ESA has commissioned ONERA for the TROPISAR campaign. It was an airborne campaign, at P band, with full polarization and a good resolution (better than BIOMASS). We have thus used these data to assess our technique.
We will first show that studying the double bounce component of the Pauli vector at P-band is a first good indicator of forested versus non-forested area.
In a second step, we will add cross-polarization. It has been shown that HV presents the best sensitivity to variation of forest biomass [LeToan]. We have previously shown that the use of HV is relevant to detect and classify rotated buildings [Guinvarc’h]. In that case, the crosspolarization values were used to discriminate forests (so random scatterers) and rotated cities (characterized by rotated dihedrals, so coherent scatterers). For both cases, the HV signature is significant (with respect to the signature of the sea). However, they relate to two different physical mechanisms. For the rotated buildings, the contribution is then stronger than for the forests.
Here, we highlight the same kind of difference with the signature of the double bounce trunk-ground versus the signature of non-forested areas. The HV response is then split into two sets, one being associated to double bounce. We then use this HV part to enhance the double bounce response.
Finally, we also make use of the double bounce of the border of the forest (so surrounding the deforested part) and the normal forest. A result is shown in the figure, to be compared to the optical image (which is not at the same date, as most of the optical images are covered by clouds). It presents deforestation due to flooded area and to manmade cuts.

We are thus able to detect deforested area with a single SAR image at P band.

R. Guinvarc’h and L. Thirion-Lefevre, "Cross-Polarization Amplitudes of Obliquely Orientated Buildings With Application to Urban Areas," in IEEE Geoscience and Remote Sensing Letters, vol. 14, no. 11, pp. 1913-1917, Nov. 2017.
Le Toan, T.; Beaudoin, A.; Riom, J.; Guyon, D., "Relating forest biomass to SAR data," in Geoscience and Remote Sensing, IEEE Transactions on , vol.30, no.2, pp.403-411, Mar 1992

Area A - Board 86

Advanced Differential Interferometric Synthetic-Aperture Radar (A-DInSAR) technique has advanced rapidly in the last years. It is a useful technique for detecting and monitoring ground surface deformations due to landslides. Identification of the areas affected by ground motion through A-DInSAR data is generally based on the visual inspection and hotspot or cluster analysis of the average displacement rates. However, interpreting A-DInSAR time series of a particular area provides a better indication of the real trend of displacement. Moreover, it can be identified the possible moment of acceleration of the deformation process as well. For these reasons, a novel methodology is proposed for identifying different typologies of ground motion areas mainly related to landslide phenomena at a regional scale, by means of A-DinSAR data at high spatial and temporal resolutions. This methodological approach was tested and validated in Piedmont region (northern Italy), by means of RADARSAT and COSMO-SkyMed satellite data. Linear (constant in time displacement) and non-linear (acceleration or deceleration in the displacement rate) trends were recognized, allowing characterization of the kinematic pattern of a landslide or a portion of it. Local and site-specific scale analyses, performed in an Alpine valley and in some hillslopes representative of the main geological and geomorphological contexts of the study area, validated the results obtained at regional scale. This supported the interpretation of the driving mechanism for such known landslides, or the other geological processes which can cause the ground motion along slopes. The developed procedure can allow one to specify the priority areas for prevention activities, in order to optimize the costs and benefits of designing a plan to monitor instability phenomena at regional and site-specific scales. Moreover, ground motion areas identified by different sensors in the same landslide help in the characterization of the state of activity of this phenomenon, identifying also the possible moments of re-activation.

Area A - Board 88

Landslides represent major threats in mountainous terrains. Given access difficulty, field mapping remains a challenge, which can be partially offset by exploiting continuous remote sensing monitoring data. The open and free policy of the Copernicus Sentinel-1 mission offers a unique opportunity for detailed insight into hillslope extent and motion rate of gravitational deformations. The purpose of this contribution is both to present a demonstration of ESA's online GEP processing platform and a case study applied to the French Pyrenees Mountain chain where inventoried landslides are scarcer than reality.
In the Ossau valley, along the Axial Zone of the French Western Pyrenees, the hillsloppe north-west of the willage of Laruns hosts both a little known deep-seated landslide affected by superficial gravitional phenomena. The deep-seated structure affects a triangular mountain wedge shape with a 3-km-long base, 2-km-long height, 1.4-km elevation drop from hilltop to valley bottom. Geomorphological and geological field mapping carried out the slope demonstrated its recent activity, scarps staircases, backscarps, meter-wide open cracks, tilted geological bedding planes, active rockfalls. Though, beyond this morphological situation, present-day deformation rates have remained elusive. The hillslope has not yet proven worth, given the underlying cost, to be instrumented and monitored in-situ. Yet, in order to gain some insight of rock slope dynamics, a multi-temporal InSAR analysis of Sentinel-1 archive was considered.
Sentinel-1 images in descending mode between 09 June 2016 and 09 October 2018 were processed using FASTVEL PSI service available via the ESA Geohazards Exploitation Platform (GEP) to measure mean ground displacement rates over the area of interest. Preliminary PSI processing results show that the context of Laruns does not allow to delimit the deep-seated landslide. Expectedly forests covering about half the hillslope contributes to the decorrelation of the InSAR signal. Further, despite less than 1900m in elevation, seasonal snow cover driven by regular humid oceanic air masses incursions affects also the InSAR measurements. Despite these limits, several clusters of pixels anchored on buildings such as farms and barns, as well as rock outcrops and open grassland pastures demonstrate motion within a range of 1 to 5 cm/yr. An increasing downslope motion gradient can also be noticed. Some localized motion can be attributed to superficial instabilities, while the rest can putatively be attributed to a broader mountainside scale deformation. Further analyses, on longer time-scales, excluding snow-covered images are expected to refine our results.

Area A - Board 89

The mountainous northwestern U.Ss is prone to heavy rainfall in winter, which results in hundreds of landslides, human casualties, and billions of dollars of expense to federal and state budgets for fixing railways, highways, and dams. Precipitation is a major hydrologic variable and a landslide trigger. The linkage between precipitation and landslides has been very well studied over many decades, but mainly in location-specific studies. Studies covering broad areas are typically after-the-fact assessments correlating storm intensity, topography, and landslide occurrence.
This talk highlights our new initiative funded by NASA IDS Program that aims to characterize the major hydrology-driven landslides in the northwestern US (Washington, Oregon) at a large scale with high resolution and accuracy, using radar remote sensing techniques – time-series InSAR (Interferometric Synthetic Aperture Radar) methods, TRMM (Tropical Rainfall Measuring Mission) precipitation product, and SMAP (Soil Moisture Active Passive) soil moisture. In this project, we will bring these fragmented observations and measurements to a single framework for the modeling of rapidly accelerating and slow-moving landslides. InSAR time-series displacement with dense scatterers on the slope will give an indication of landslide occurrence and help characterize the slide-body volume and basal slip surface based on geomechanical modeling. Precipitation and soil moisture sensed from space will contribute to creating hydrogeologic models with topographic features. These crucial parameters will be tracked through slope stability analysis and iterative numerical simulations, and then plugged into state-of-art landslide dynamics and run-out models.
We’ll present our initial results, and the road map to accomplish this project that will ultimately enhance our understanding about the inter-relationships between topographic and geologic settings, hydrologic variables, landslide mechanisms and kinematics inferred from time-series measurements and landslide modeling on a regional scale.

Area A - Board 90

Despite landslide maps being compiled every year in the world at different scales, limited efforts are made to critically compare landslide maps prepared using various techniques or by different investigators. Event-based landslide inventories indicate the location, distribution, and detected boundaries of landslides caused by a single event such as earthquake and rainstorm. Event-based landslide inventories are essential for landslide susceptibility mapping, hazard modelling and further management of risk mitigation. Moreover, such these inventories have deployed for recording the magnitude of the triggered reason of the landslide. In this research, we compare the statistical characteristics of existing landslides inventories. Landslide inventories were analysed based on the methodology, quality and completeness levels. They are analysed to determine the quality based on the direct comparison, from cartographic matching and mismatch between the maps. Also, statistical properties of landslide areas were defined, and the correlation between the frequency area statistics distribution of landslides was carried out. In Nepal, there have been several attempts to map landslides after the Gorkha earthquake 2015. Especially, after the main event of 25 April, 2015 researchers around the world mapped the landslides induced by Gorkha earthquake. In this research, we use four inventories obtained from different sources which are prepared through visual interpretation of remote sensing imageries.
Figure 1 illustrates the landslides in the study area which mapped by four different studies. It shows that the number of mapped landslides for the same area differs per interpreter which depends upon the personal perspective in mapping and method of mapping adopted. Interpreter A mapped 65 landslides for the common area using Planet scope satellite imagery. Interpreter C and Interpreter D also mapped landslides using Google Earth TM images and mapped 25 and 43 landslides respectively, and. The other result shows the 102 landslides were mapped by interpreter B which is the result of using high-resolution Digital Globe’s WorldView-3 imagery. The main reason for differences among the mapped landslide inventories is the personal mapping preference of the researchers and the detail of delineating landslides.
The minimum landslide area mapped by different researchers ranges from 35.50m² to 121.19m² for the study area. Also, the most massive landslide mapped ranges from 118805 m² to 764038 m². There is a zigzag pattern of the plotted landslide probability density to the inverse gamma fit. Differences in the probability density distribution and inverse gamma fit can be the result of gaps related to mapped landslides for given inventories which means that some landslides are missing or not mapped by respective interpreters due to personal mapping preferences. The rollover points for landslides mapped by interpreter A and interpreter D are almost similar around 290, but for landslides mapped by interpreter C, the rollover point is 223 which is nearly 70 points less than other two inventories. For inventory generated by interpreter B, the rollover point is 85.13 which is small in comparison to other inventories as the distribution of small landslides is given in inventory. Interpreter B Mapped smaller landslides in detail but for larger and medium size landslides distribution is scattered. For an inventory of interpreter C, there is a large gap between the rollover point and the smallest mapped landslide in the inventory that is 95.49 m². Between the smallest mapped landslide in the inventory and the rollover point, there are many small landslides in the region, but they are not mapped because of amalgamation, which is smaller landslides forms parts of bigger landslides polygons.
We compared four inventories qualitatively and quantitatively using different techniques. We also discussed the reasoning for the differences between the inventory maps using statistical methods degree of matching and mismatch between the inventories. Results of our work give an overview of the impact of methodology selection and outline the limitations and advantages of different remote sensing and mapping techniques.

Area A - Board 91

The project «J-CROSS, Joint Cross Border Cooperation for Securing Societies Against Natural and Man Made Disasters», which is funded by the European Committee in the frame of “INTERREG IPA CBC Programme”, aims at minimizing the (increasing) severity of risks in the two neighboring regions, i.e. Western Macedonia from Greece and Pelagonia from F.Y.R.O.M., with the joint design, development and implementation of drastic initiatives. Its subject is to jointly develop cross-border action plans, which will lead to joint assessments and common risk classification, countermeasure financing, joint preparedness exercises and training activities. Within the project it is planned to monitor selected landslides in the area of Florina (Greece) and Vitola (F.Y.R.O.M.) exploiting satellite geodesy, such as real-time GNSS and temporal interferometry with active remote sensing data. The project team exploits experience cumulated from past relative projects in the Peloponnese (Greece) region. In particular, a permanent meteorological station and a GNSS geodetic station are being installed within the landslide body providing real-time data, with the GNSS station being resolved in real time for relative positioning based on a second base GNSS station less than 10 km away. At the same time, 20 ascending and descending high-resolution active remote sensing acquisitions (DLR TERRASAR-X) are being acquired in order to calculate the temporal ground deformation. The acquisitions are of very high resolution (Spotlight mode) and the incidence angles are selected to be optimal based on the coverage, the local inclination and orientation of the main landslide. Ten artificial angular reflectors are being installed inside and outside the landslide body so that they can be properly utilized by remote sensing data. Relevant know-how is expected to be transferred to the neighboring country. The landslides included in the study areas are expected to be mapped accurately, quantitatively and spatially. The project is funded by the European Commission under Interreg-IPA CBC programme, priority axes Protection of Environment – Transportation.

Area A - Board 92

Landslides can have catastrophic consequences in many mountainous areas of the world and thus represent a major problem for society. In Central Asia, and more specifically in Kyrgyzstan, landslides are recurrent mainly due to the high seismicity of the region and as a consequence of intense rainfalls. In spring 2017, Kyrgyzstan suffered a massive activation event which caused 160 emergency situations. Koytash and Tektonik, located in the Mailuu-Suu Valley, in the North of the Fergana Basin, were two of the numerous landslides that collapsed during this event. They represent a major threat to the local population of the small town of Maily-Say and to the villages downstream. In this region, risks related to landslides are accentuated by the presence of uranium tailings, remnants of the former nuclear mining activity. The rupture of the Koytash landslide, one of the largest in the area, resulted in the damming of the Mailuu-Suu River which lasted several weeks. This deep-seated, 1 km long, landslide is characterised by a rotational movement. The neighbouring Tektonik landslide was only partially reactivated, mainly in its upper part. The main objective of this study is to identify the geological-geomorphic and meteorological factors which contributed to the massive failure of Koytash and the partial reactivation of Tektonik in spring 2017. Different remote sensing techniques, combined with an extensive meteorological analysis, were applied to identify the conditions that triggered these landslide movements. The meteorological analysis considered many variables (rainfall, temperature, and snow depth) which influence soil water saturation and thus sediment mobility, playing a key role in the landslide process. In parallel, radar and optical sensors, such as the Copernicus Sentinel 1 and 2, were combined in order to monitor changes and displacement rates within the deformation zones. Differential synthetic aperture radar interferometry (D-InSAR) and the computation of deformation time series generated interferograms allowing us to identify slow displacements, along the line-of-sight of the satellite, and displacement rates evolution. These results are complemented by the comparison of digital elevation models (DEMs; satellite and UAV imagery) that shows the changes in topography, before and after total failure. Furthermore, we used optical imagery, through the difference of NDVI (Normalized Difference Vegetation Index), to observe changes in landcover associated with sliding processes. The meteorological analysis revealed the important contribution of precipitations (rainfall combined with large snow accumulations followed by a rapid melting) as triggers of the landslide activation in 2017. Indeed, despite a general decrease in annual rainfall recorded in 2017 in comparison to the preceding years, April 2017 was characterised by heavy rains, including a major peak of rainfall exactly on the day of Koytash’s failure. The D-InSAR analysis identified slow displacements during the months preceding the reactivation, indicating the long-term sliding activity of Koytash, well before the reactivation in April 2017. This was confirmed by the computation of deformation time series, showing positive velocity anomalies (downstream and westward oriented) in the upper part of both landslides. The difference of DEMs analysis highlighted areas of depletion and accumulation, in the scarp and foothill zones, respectively. Moreover, the surface effects of these deformations were confirmed by the optical analysis which showed the efficiency of land cover changes and denudation as indicators of ground displacements. The multidirectional approach used in this study, justified by the complementarity nature of the techniques, enabled the gathering of complete and coherent results. The promising outcomes obtained for the cases of Koytash and Tektonik show the applicability of these techniques to monitor landslides activity. Therefore, similar studies have now been initiated to monitor other landslides around the Fergana Basin.

Area A - Board 93

The Northern part of Nepal is located on Himalaya Mountains and is prone to a lot of rain-induced landslides. Assessing the kinematics of landslides in mountainous areas is important to have a better understanding of the process and mitigate hazards associated with them. The study area in this research consists of 3 landslides in Annapurna range of the Himalayas close to the Nepal-China border. The aim of the study is to evaluate the potential of SAR data for deriving the kinematics of landslides in this region. In order to achieve this goal we used InSAR (Interferometric Synthetic-Aperture Radar) technique utilizing data from ALOS-I satellite between the years 2007-2011and Sentinel-I satellite between 2014-2016 years. Ground surface deformation is analyzed first by performing the PSI (Persistent Scatterers Interferometry) and SBAS (Small Baseline Subset) time-series analysis approaches. A sensitivity analysis is then performed in the velocity field considering viewing geometry of the SAR acquisitions and local topography. The measurement points with higher sensitivity to the satellite’s acquisition geometry are considered for in-depth ground surface deformation analysis. Finally, the seasonality variations in landslide velocity field is analyzed against precipitation data as a major triggering of landslides in this area.

Index Terms: slow-moving landslides, ALOS-I satellite, Sentinel-I satellite, InSAR, Time-series analysis, slope deformation, PSI, SBAS

Area A - Board 95

The Big Sochi is a lovely resort on the Black Sea coast of the Greater Caucasus. It has always been a region of high landslide risk due to widely spread clays and marls saturated by water from abundant precipitations. Besides, it is a region of rather high seismicity. In recent years landslide monitoring and risk assessment have become vital due to strongly increasing anthropogenic impact.
The Big Sochi area amounting to 3500 sq.km stretches along the Black Sea coast for about 150 km. Near the coast it is a highly populated area while in a few km from the beach there are only isolated villages in the mountains. A plenty of existing landslides are shallow, very slow (rates of movements up to several sm/Y) with maximum dimensions up to tens or a few hundred meters at the most. InSAR methods seem to be very promising for landslide remote sensing in the Caucasus region due to application of data of frequent regular inspection of large hard to reach areas, acquisitions during day and night almost independent from the meteorological conditions. Besides, existence of archived data makes it possible to follow historical changes in landslide activity. We discussed some aspects of InSAR application for landslides studying on the Black Sea coast of the Northern Caucasus using satellite radar imagery with different wavelengths in [1-3]. Nowadays usage of S-1 images is of particular interest because of short time-interval between acquisitions (6-12 days) and free data access in almost real time of sensing via Internet.
The Big Sochi area has been fully covered by Sentinel-1 acquisitions from ascending and descending tracks since 2015. SBAS ENVI SARscape processing of more than 200 acquisitions from 43A, 145A and 123D tracks for the time period 2015-2018 enabled us to develop detailed surface displacement maps. Based on the calculated surface displacement fields the alert maps were plotted showing places where average displacement rate in the line of site (LOS) projection exceeds the 10 mm/Y threshold. Joint analysis of the S-1 InSAR alert maps with GIS maps of the study area (topographic, maps of landslide phenomena and risk assessment [4], interactive map of current landslide events) made it possible to update the preexisting landslide inventories based on ground survey for several regions. In the Big Sochi area ground survey have problems with mapping of a plenty of small in dimensions and very slow moving landslides in highly populated areas where the most part of buildings are private houses with restricted entry. So, field works provide comprehensive data for undeveloped areas and incomplete picture for densely built-up territories. On the contrary, compact districts are very favorable for S-1 InSAR while rural slopes have lack of reflectors of radar signal. So, in the Big Sochi these methods well complement each other. About 100 new active landslide phenomena were revealed with S-1 InSAR in the Big Sochi. The time series analyses permitted us to conclude that for the majority of landslides the period of activation is usually between November and June, although many landslides move permanently.
Thus, S-1 InSAR in the coastal part of the Big Sochi area appeared to be a powerful and low cost tool improving the existing systems of landslide monitoring based on ground survey.
This study was performed within the state assignment on subject 0144-2014-00105.

1. Mikhailov, V.; Kiseleva, E.; Smolyaninova, E.; Golubev, V.; Dmitriev, P.; Isaev, Yu.; Dorokhin, K.; Hooper, A.; SamieiEsfahany, S.; Hanssen, R. PSINSAR monitoring of landslides in the Great Caucasus, Russia, using ENVISAT, ALOS and TerraSAR -X SAR images. Proceedings of ESA Living Planet Symposium, Edinburgh, UK, 9 - 13 September 2013, 9 p.
2. Mikhailov, V.O.; Kiseleva, E.A.; Smol’yaninova, E.I.; Dmitriev, P.N.; Golubev, V.I.; Isaev, Yu.S.; Dorokhin, K.A.; Timoshkina, E.P.; Khairetdinov, S. A. Some problems of landslide monitoring using satellite radar imagery with different wavelengths: Case study of two landslides in the region of Greater Sochi. Izvestiya, Physics of the Solid Earth 2014, Volume 50(4), pp. 576-587, DOI: 10.1134/S1069351314040107.
3. Kiseleva, Е.; Mikhailov, V.; Smolyaninova, E.; Dmitriev, P.; Golubev, V.; Timoshkina, E.; Hooper, A.; Samiei-Esfahany, S.; Hanssen, R. PS-InSAR monitoring of landslide activity in the Black Sea coast of the Caucasus. Elsevier, Proceedia Technology 2014, Volume 16, pp. 404–413, DOI: 10.1016/j.protcy.2014.10.106.
4. Vozhik A.A. Assessment of exogenetic geological risk in the course of State monitoring of the state of the geological . Proceedings of the VIII Vserossiiskii s"ezd geologov, Moscow, Russia, 26-28 Oct 2016. Available online: http://www.specgeo.ru/pdf/doklad_viii_geolog_04.pdf

Area A - Board 96

The landslide occurrences in mountainous regions are mainly triggered by the earthquakes, hydro-climatic conditions. Accurate landslide inventory mapping on a regional scale after the main event is still a challenging task. However, landslide inventory maps considered as the main factor for landslide susceptibility assessment and risk analysis. As most of the landslide susceptibility models apply inventory maps to investigate the importance of each landslide conditioning factors. Consequently, analyzing the factors based on their significance in landslide assurances to finding the more likely areas for future hazards. However, happening a landslide in nature cannot be considered as an important event. The importance of these landslides appears when conducted with human made infrastructures and especially settlement areas. In this respect, accurate detection of both landslide and settlement areas is considered as the main factors for the risk analysis and mapping.
This study presents an integrated approach of machine-learning (ML) models and Dempster-Shafer theory (DST), using RapidEye optical images and ALOS digital elevation model (DEM) from a case study area in higher Himalayas. The visible bands of RapidEye optical images and the topographic landslide conditioning factors of slope, aspect and plan curvature which derived from the ALOS DEM were considered as the model's input dataset. Support vector machines (SVM) and random forest (RF) which considered as the common ML models for image classification (Mezaal, Pradhan, and Rizeei 2018), were used for detection of landslide and settlement areas. The models were trained based on the training segment of a landslide inventory map which prepared by an intensive field working and improved by manual extraction from RapidEye optical images. In this study, to improve the accuracy of the resulted probabilistic outputs of the applied models, the final decision was performed using DST.
The DST is an extension of the Bayesian probability theory can improve the accuracy of the model's probability outputs based on three functions namely: belief assignments (BBAs), belief, and plausibility (Feizizadeh 2018). Three areas of disbelief, uncertain and belief were illustrated in the DST scheme (See figure 1). Since the probability outputs of models are usually uncertain, the DST is a practical approach for reducing the existing uncertainty among such these outputs. In this study, the confusion/error matrices were derived from a comparison between the corresponding inventory map and the outputs of the SVM and RF models. The Kappa coefficient was obtained based on the confusion matrix with the following equation (1):
Kappa coefficient (α- β)⁄(1-β) (1)
where α refers to the ratio of correctly detected areas. Whereas, β denotes the proportion of agreement predicted by chance. The confusion/error matrices were used for the fusion process for increasing the accuracy of the resulting final maps. The whole process was summarized in the flowchart of figure 2. Our study illustrated a fusion of ML models of detecting of landslide and settlement areas using the DST method. The integrated approach fused the outputs of the models to prepare a more accurate detection. The more precise maps were derived based on the integrated approach, which was specified by an improvement in landslide and settlement areas detection over SVM and RF models.
References
Feizizadeh, Bakhtiar. 2018. "A Novel Approach of Fuzzy Dempster–Shafer Theory for Spatial Uncertainty Analysis and Accuracy Assessment of Object-Based Image Classification." IEEE Geoscience and Remote Sensing Letters 15 (1): 18-22.
Mezaal, Mustafa, Biswajeet Pradhan, and Hossein Rizeei. 2018. "Improving Landslide Detection from Airborne Laser Scanning Data Using Optimized Dempster–Shafer." Remote Sensing 10 (7): 1029.

Area A - Board 97

Despite landslides have strong effect on the society worldwide, landslide information is lacking. When they are rainfall induced, the availability of optical images might be rare due to cloud persistence and landslides might remain unknown, or unnoticed for long time, preventing studies and hampering the emergency phase management.
The Sentinel-1 high revisit time, the relative high spatial resolution of the images, the global coverage, the availability for free, including processing software offer an unprecedented opportunity to use the all weather measures of amplitude to investigate and monitor land surface changes phenomena due to natural hazard for a large-scale mapping
In this work we present the result of a systematic use of the Sentinel-1 SAR C-band images for landslide detection and mapping by interpreting measures of changes of amplitude.
The scheme to pre-process the SAR images was the same for all test cases and included radiometric, and geometric corrections of the amplitude to obtain Beta Nought radar brightness coefficient in slant coordinates, and measures of changes of between images acquired before and after the landslide occurrence using the natural logarithm of the ratio of Beta Bought post-event and Beta Nought pre-event.
We analyzed more than thirty cases worldwide of fast moving landslides differing in size, exposition and Geo-environmental factors including Geology, lithology, land cover, orography and climatic regime. In most of the cases the signs left by the landslides are clear enough to detect the event, but they do not allow for the production of inventory maps, while in some cases, a delineation of the landslide borders (i.e., mapping) seems to be possible.
The unprecedented availability of images provided by the Sentinel-1 constellation and their photo-interpretation promise then to increase the availability of information on landslides with advantages for weather-related emergencies management, model tuning, and landslide forecast model validation, in particular when accurate mapping is not required.

Area A - Board 98

As space-borne Remote Sensing enters its “golden age”, an ever-increasing wealth of EO data with high spatial- and temporal resolution is available, taken by a variety of different sensors. These data can be exploited to detect and monitor natural hazards, such as slope instabilities and landslides, in unprecedented detail. In recent years, a technique called Digital Image Correlation (DIC) has been more and more applied to detect and monitor surface displacements. DIC uses digital images of a site of interest at different points in time to reconstruct the displacement that happened between the acquisitions. DIC is capable of reconstructing even large displacements if the displaced mass maintains a coherent appearance, making it a complex, yet quick, flexible, and low-cost solution for slope instability detection & monitoring. In contrast to specialized methods, DIC can be applied on remote sensing data collected from active and passive sensors, including radar, optical, and thermal images. Due to its versatility and capability, DIC could help to exploit the ever-increasing stack of available EO data produced by various platforms in order to further improve recognition and monitoring of natural hazards and associated risks.
Here we show the results of our implementation of a DIC algorithm that operates in the frequency domain of the input images. We validated the DIC results in direct comparison to GNSS and in situ displacement data taken at two slope instabilities in the Swiss Alps, by considering input images obtained from satellites, Sentinel 1 (radar) & 2 (multispectral), and Landsat, but also by manned and unmanned airborne acquisitions, as well as ground-based cameras. We show that DIC results are accurate and consistent within the expected limitations of the technique. Additional sets of routines were developed to support the correlation procedure, such as input image preprocessing routines, including sub-pixel co-registration and dynamic contrast enhancement, as well as output data postprocessing, including spatial and RMSE filters that aim to reduce the correlation noise content. The compiled Matlab toolbox is freely available for research under the MIT license on GitHub at: https://github.com/bickelmps/DIC_FFT_ETHZ

Area A - Board 99

The Java Island (Indonesia) due to its location in tropical zone and within Pacific ring of fire is characterized by very dynamic environment. Its dynamics is manifested by numerous surface deformations of various genesis triggered e.g. by earthquakes, volcanoes, landslides or soil erosion. This study carried out within the ESA Project: Geoinformation Support for Integrated River Basins Management (Geo4IRBM) is concentrated on the deformations which take place within two watersheds: Cimanuk-Cisanggarung rivers watershed (East Java) and Jratunseluna river watershed (Central Java). As one of the main goals of the project is to assess the soil erosion within the watersheds, the deformation studies are performed using combined sets of both persistent and distributed scatters. The latter give the opportunity to measure deformation also outside of the built-up areas which is crucial for this project.
The studies are carried out on the basis of sets of Sentinel-1 A and B images derived from 5 different ascending and descending orbits. The data starting from the beginning of 2017 were taken into account. Before this period the Sentinel-1 data from Java Island were acquired with relatively low frequency (every 24 days or often less) which was not enough to obtain sufficient number of distributed scatters outside the built-up areas in this tropical, very humid and vegetated environment. Starting from January 2017 the data have been registered by Sentinel-1 A every 12 days. Unfortunately the images acquired with full potential temporal resolution using Sentinel-1 A and B have been available only for one descending orbit starting from the end of July 2017. The spatial distribution and density of persistent and distributed scatters obtained, based on the 6-days and 12-days sets of the images of the same region, were compared. The results show that frequency of data acquisition is crucial for obtaining the deformation signal from distributed scatters located outside the built-up areas in tropical zone.
The processing of the data was performed using the Gamma software. One of the biggest problems which have to be overcome is the correct phase unwrapping over the large areas, which is very difficult in this region due to the unfavorable combination of various factors: the spatial discontinuity of high quality persistent scatters, large deformations and very dynamic atmosphere, especially during the wet season.
First results obtained for the Jratunseluna river watershed show very large deformations, which take place in different parts of the study area. The biggest one is observed in Semarang City especially close to the fault located there. The relative deformations between various parts of the city observed using multitemoral interferometric techniques exceed 15 cm per year, which is in agreement with the information obtained directly from Indonesia, from the people working for the Asian Development Bank – the main end-user of this study. Many other deformations connected among others to landslides and soil erosion was also identified. The Sentinel-1 multi-temporal interferometry shows great potential for assessment of soil erosion in tropical zone, especially when 6-day’s datasets and distributed scatters are used for the study. The assessment of soil erosion was possible thanks to the favorable spatial mixing of built-up areas and agricultural lands on volcano slopes and flat areas in northern part of the Java Island. The soil erosion which was counted as a difference between the deformations obtained for built-up areas and surrounded natural and semi-natural surfaces exceeds in many cases 10 cm per year.

Area A - Board 100

The Adra River basin, located south of the Sierra Nevada Mountains (Inner Zones of the Betic Cordilleras, SE Spain) and lying between the provinces of Granada and Almería, was initially intended to be regulated by means of the Benínar Reservoir, a reservoir with water leaks from its construction. This basin extends over 746 km², the terrain is very rugged, with an altitudinal range of 2780 m, and frequent torrential water flow (García-López et al., 2009). The needs in agricultural and urban water consumption are increasing due to the daily increase of the large areas of hectares of greenhouses in the eastern and western parts of Almería. The aquifers have been and are being overexploited, concluding that the current system is not sustainable in the long term. This is a typical case in which the storage capacity of underground media and their possible contributions to the efficient management of resources have not been adequately taken into account.

The Benínar dam is located on the Grande de Adra River, in the Andalusian municipality of Benínar, in the Almería province. It was built with loose materials (rock fill with clay core), with distinct profile inclined impervious core and rock fill shoulders. It has a height of 87 m above the foundations and an enbankment length of 386 m with an enbankment elevation of 363 m a.s.l. The total reservoir capacity has a volumen of 68,2 hm³ although its practical capacity is 15 hm³, while the flooded river length is 5 km. Its catchment area (521 km²) affects the provinces of Granada and Almería on its eastern and western boundaries.

Construction works of the dam was initiated in 1974 and concluded in 1983, although it was not finally entered into service until 1988, following various projects to waterproof the base. Nowadays, after 30 years since its construction, it is clear that this dam has not satisfied expectations, due to severe leakage that has occurred through the reservoir bottom, resulting from complex, unfavorable geological characteristics, together with an initial concept that made little effort to integrate the possibilities of surface and groundwater resources, gave rise to this failure. In early 1991, the authorities revised the initial focus of the regulatory project and took action to tap the water thas was escaping into the underlying carbonate aquifer into which the reservoir water is draining (García-López et al., 2009).

The goals of the dam were to protect the coastal area of the Adra basin against catastrophic flooding and to supply 15 hm³/year for human consumption and 60 hm³/year for irrigation uses (mainly outside the basin, in Campo de Dalías). Today, while the first goal has been achieved satisfactorily, the second is far from a reality. In addition to the financial burden (construction, expropiations and exploitation), the social and environmental costs of the reservoir are very high, especially in view of the results obtained. The reservoir waters submerged the village of Benínar, a classical example of an Alpujarra-style village, attractive, and with valuable architectural features which had a population of about a thousand inhabitans. What is more, the project flooded the most fertile land in the center of the valley, on which tratitional market crops had been cultivated. Nowadays, even environmental groups as Ecologistas en Acción in Spain claim the demolition of this dam.

Despite little attention given to remote sensing technologies, the rapid development of space technology, occurred in the recent decades, has allowed the detection of the displacement of Earth surface from space with high precision and unexpected benefits for Earth observation and related global studies. In the new millennium, the use of satellite Synthetic Aperture Radar interferometry methods improved monitoring of ground and man-made structures which opened new possibilities for a more accurate interpretation of the observed displacement and driving mechanism. This progress has been possible thanks to microwave images obtained through SAR sensors mounted on satellites and the development of MT-InSAR techniques such as the most common and suitable methods known as Persistent Scatterers (PS) InSAR and Small Baseline (SBAS). These techniques have the potential to support the development of new and more effective means of monitoring and analysing the health of dams and to support and assist warning systems. With SAR Interferometry specific dams can be monitored to identify and investigate targets with suspicious displacement on a weekly time-scale. As a result, timely identification of potential problems can help mitigate their impact on structural health and lower infrastructure rehabilitation costs. Within currently available data it is possible to determine terrain deformations of theoretically any selected site worldwide (taking into account limitations such as vegetation cover, data availability, structure orientation, etc.) and also displacements of structures themselves occurring either in vertical or horizontal direction. While horizontal direction sensitivity is usually lower, satellite InSAR can detect vertical changes in the order of millimetric sensitivity. Though deformations in urban areas are well monitored and assessed already, InSAR should yield an additional value to optimize estimated subsidence/deformation models and potentially discover unknown dynamics thanks to its high revisit rate and unique opportunity of precise spatial information.

The Benínar dam has always had problems with water leaks. Even knowing that the land was not the most suitable, due to the frequent earth movements that took place in the town of Benínar, the dam was built. In this work, we process multi-temporal SAR data sets coming from the C-band satellites ERS-1/2, Envisat, and Sentinel-1A/B using MT-InSAR techniques, being able to monitor the deformation behavior of this dam for a long time period of more than 25 years, from 1992 to 2018.

Area A - Board 101

Tropical cyclones are major hazards representing 20% of the damage and casualties caused by natural hazards, all phenomena included. Hence, it is crucial to closely inspect them in order to monitor and predict them as accurately as possible. The first step is to detect and observe them, but as Tropical Cyclones develop over the oceans, observing tools are limited. While there are a variety of sources for Tropical Cyclones (TC) outer-core wind data, the only inner-core wind data that are routinely available at present are collected by airplanes and limited to Atlantic and East pacific hurricane seasons.

C-Band high resolution radar (SAR) is the only space-borne instrument able to probe at very high resolution O(1 km) and over all ocean basins the sea surface under extreme conditions. Thanks to the existing European Sentinel-1 constellation (Sentinel-1A and -1B), the Canadian Radarsat-2 mission and the Chinese Gaofeng-3 mission, four different C-band SAR systems can provide this unique view of the Tropical Cyclones. To note, the Canadian Radarsat Constellation (3 C-Band SAR system) mission is planned to be launched in 2019 and could strengthen this network. Then, the next Sentinel-1C and -1D are already planned by the Copernicus space component. This ensures continuity of this data until the end of 2030.

The capability of these new generation C-band SAR to provide ocean surface wind field measurements for Tropical Cyclones including category-5 major hurricanes has already been demonstrated in recent literature. These works show the capability of such missions for providing key parameters related to the Tropical Cyclone wind structure. In particular, the eyewall and its vicinity is an important region with complex physical mechanisms related to the storm evolution and phenomena such as rapid intensification are still active subjects of research. More practically but equally important, during a storm event, these parameters are also used by forecasters in WMO (World Meteorological Organisation) TC centers to provide TC forecasts and to issue warning bulletins every 6 hours for people safety. In this presentation, the SAR capacity will be illustrated with validation against airplane measurements in the case of Irma category 5 hurricane.

However, these are sun-synchronus low-orbit missions with a strong constrain on the duty cycle. This prevents from continuous acquisitions everywhere. Moreover, to date, there is no operational and coordinated effort to optimize SAR acquisitions over Tropical Cyclones when they are still in open ocean, before landing to possibly devastate populated areas. In 2016, we proposed organising, in collaboration with ESA (SEOM Programme), a campaign called SHOC (Satellite Hurricane Observation Campaign), in order to gather extreme observations from Sentinel-1 over hurricanes. Initially intended to last two months during 2016 Tropical Cyclone season, SHOC activities are still on-going, leading to more than 100 hurricane hits (i.e. capturing the eye of hurricanes) by the Copernicus Sentinel-1 constellation, since 2016. As a matter of fact, more than 90% of the hurricane eyes captured by Copernicus Sentinel-1 have been obtained via SHOC. We will present the strategy set up developed with the ESA Sentinel-1 mission planning team to combine TC tracks forecasts and Sentinel-1 orbit, to resolve potential conflicts with other missions objectives and finally to optimize the acquisitions mission plan. SHOC now also includes Radarsat-2.

Results in terms of acquisitions obtained since 2016 will be reviewed for all basins and compared to the theoretical potential of all these missions, assuming no conflict with other users regarding the duty cycle, the acquisition mode etc...

Area A - Board 144

Figure 1: A) Sentinel 2 natural color image of (part of) the Tuxer Alps (Austria) on August 27th 2018, Innsbruck on the top left. [Sentinel Hub]
B) Example of a hydro-meteorological time series of the region. Precipitation, evaporation and runoff are indicative of the hydrologic processes. [Data: Earth System Data Lab and eHyd Austria]

Landslides, floods and droughts are major hazards to human life and business. In Europe alone, landslides cost 1370 lives in the period 1995–2004 and caused an average economic loss of 4.7 billion Euros per year [Haque et al., 2016]. With 1.3–3.6 million people living in areas susceptible to landslides and 8,000–20,000 km of transport infrastructure exposed to potential landslides, landslides form a serious hazard [Jaedicke et al., 2014].

Increased frequency of data acquisitions from satellite remote sensing platforms can complement existing terrestrial geohazard monitoring activities and provide opportunities for improving susceptibility estimates to an up-to-date hazard analysis. Geohazard susceptibility maps are traditionally made based on predisposing factors, such as strong relief, elevation, tectonic activity, vegetation characteristics and proximity to potential driving forces. The regional hazard assessment is then based on heuristics, either relying on expert analysis or trained on previous events, for example the global Landslide Hazard Assessment for Situational Awareness model (LHASA) by Kirschbaum and Stanley [2018].

Existing regional landslide susceptibility maps, for example form Günther et al. [2014], focus on static, predisposing factors and not on dynamic factors such as hydro-meteorological conditions. At the same time it is recognized that hillslope stability is strongly influenced by hydrology [Bogaard and Greco, 2016], and velocity changes in landslides have been shown to be directly coupled to hydrological changes [Benoit et al., 2015].

Our proposed solution is to directly exploit correlation between contributing factors. The increasing abundance of satellite and possibly in situ data allows for a data driven approach, enriched by local knowledge when available. So far our aim is to develop an integrated approach for data analysis available at different spatial scales combined with the multivariate pattern analysis of past events. Given enough of these patterns from the past prediction of future events becomes feasible.

A bucket model is a simple hydrologic water balance that can be compiled at various spatial scales, depending on the availability of the data. Most landslides are rainfall driven and water storage in the soil is key in this. The change in (ground)water storage over time (dS/dt) is the resultant of precipitation (P) input and evaporation (E) and discharge (Q) output fluxes: dS/dt = P − E − Q.

Data that can, for example, be used as input in such water balance is shown in Figure 1. Snow data is extracted from the ESA DUE GlobSnow program, land surface temperature data is extracted from ESA DUE GlobTemperature. Both are to be superseded by their ESA Climate Change Initiative (CCI) equivalents. Evaporation and transpiration are provided by the Global Land Evaporation Amsterdam Model (GLEAM). Local gauges provide an indication of the discharge. Due to the dynamic character of the model, resources can be replaced or complemented by other data sets and local ’ground truth’. Deformation rates (not shown) can be estimated from Sentinel 1 InSAR analysis, terrestrial total station or laser scanning measurements and provide information on the behaviour of the landslide under varying hydrologic conditions. The concept is tested in the Tuxer Alps in Austria. The method developed can be used for all geohazards that can be linked to hydro-meteorologic events, including floods and droughts.

This work is part of the H2020 OPERANDUM project on the impact mitigation of hydro-meteorological phenomena in risk-prone areas using nature-based solutions.


References
L. Benoit, P. Briole, O. Martin, C. Thom, J.-P. Malet, and P. Ulrich. Monitoring landslide displacements with the Geocube wireless network of low-cost GPS. Engineering geology, 195:111–121, 2015.
T. A. Bogaard and R. Greco. Landslide hydrology: from hydrology to pore pressure. Wiley Interdisciplinary Reviews: Water, 3(3):439–459, 2016.
A. Günther, M. Van Den Eeckhaut, J.-P. Malet, P. Reichenbach, and J. Hervás. Climate-physiographically differentiated pan-European landslide susceptibility assessment using spatial multi-criteria evaluation and transnational landslide information. Geomorphology, 224:69–85, 2014.
U. Haque, P. Blum, P. F. Da Silva, P. Andersen, J. Pilz, S. R. Chalov, J.-P. Malet, M. J. Auflič, N. Andres, E. Poyiadji, et al. Fatal landslides in Europe. Landslides, 13(6):1545–1554, 2016.
C. Jaedicke, M. Van Den Eeckhaut, F. Nadim, J. Hervás, B. Kalsnes, B. V. Vangelsten, J. T. Smith, V. Tofani, R. Ciurean, M. G. Winter, et al. Identification of landslide hazard and risk ‘hotspots’ in Europe. Bulletin of Engineering Geology and the Environment, 73(2):325–339, 2014.
D. Kirschbaum and T. Stanley. Satellite-based assessment of rainfall-triggered landslide hazard for situational awareness. Earth’s Future, 6(3):505–523, 2018.

Area A - Board 146

In recent years InSAR methods of estimation of surface displacements find ever-widening applications. Different techniques were developed to identify persistent scatterers (PSInSAR) in time series of SAR images (e.g., Feretti, 2001). However, density of PS in rural areas is usually low (<10-20 PS/sq.km). Along with PS techniques, methods for estimation of displacements of both coherent and decorrelated pixels have been developed. One of the possible approaches to reduce decorrelation is to analyze only interferometric pairs with small temporal or spatial baselines (e.g. SBAS, Berardino et al., 2002). The scatterers identified in such a way were called distributed scatterers (DS). The A. Hooper’s method (implemented in the StaMPS /MTI software) integrates advantages of both approaches. It is based on assumption of phase stability of scatterers as criterion for identification of persistent scatterers, thus, often increasing number of PS in rural areas.
Next step was development of techniques allowing to identify both PS and DS (e.g., SqueeSAR, ILS SM-phase estimation, see Ferretti et al.,2011; Lanari et al., 2013; Samiei-Esfahany et al. 2016; Wang et al., 2016, and others) employing all possible interferograms. This substantially increases spatial density of scatterers, thus, improving unwrapping procedure and accuracy of estimation of displacement fields.
The approach is based on the following. Scatterers on natural terrains are often identified in areas with low (for some periods of time) phase coherence. However, many neighboring pixels show similar reflectivity because they belong to the same natural object. Thus, similar to SqueeSAR (Ferretti et al., 2011) we first reveal spatially connected clusters of statistically homogeneous pixels (SHP). To estimate statistical homogeneity we employed the two sample Kolmogorov–Smirnov test. Phase filtering is performed using the Integer Least Squares method (Samiei-Esfahany et al. 2016) on the assumption of constant phase within the whole SHP range and expected phase consistency condition. The algorithm is implemented as a module compatible with StaMPS/MTI.
We applied the proposed approach to landslide investigation in the Caucasus (ALOS PALSAR и ENVISAT images) and seismic and volcanic events of the Kuril-Kamchatka subduction zone (ENVISAT images). Total number of PS+DS for 18 ALOS PALSAR was nearly the same as for interferograms without phase filtering while for the Envisat stack of images phase filtering provided the doubled number of distributed and persistent scatterers.
Authors are grateful to ESA for the ENVISAT images. This study was supported by grant of the Russian Ministry of Education and Science under contract №14.W03.31.0033.
1. Berardino P. et al, A new algorithm for surface deformation monitoring based on small baseline differential SAR interferograms. IEEE Trans. Geosci. Remote Sens., 2002, vol.40, no.11, pp. 2375–2383.
2. Ferretti A., C. Prati, and F. Rocca, Permanent scatterers in SAR interferometry, IEEE Trans. Geosci. Remote Sens, 2001, vol. 39, no. 1, pp. 8–20.
3. Ferretti A. et al., A new algorithm for processing interferometric datastacks: Squeesar, IEEE Trans. Geosci. Remote Sens, 2011, vol. 49, no. 9, pp. 3460 –3470.
4. Lanari R. et al., A full exploitation of the enhanced SBAS-DInSAR approach in volcanic and seismogenic areas, IGARSS, 2013, Proceedings, Melbourne 21-26 july 2013.
5. Samiei-Esfahany S. et al., Phase Estimation for Distributed Scatterers in InSAR Stacks Using Integer Least Squares Estimation, IEEE Trans. Geosci. Remote Sens, 2016, vol. 54, no. 10, pp. 5671-5686.
6. Wang M., Li T., Jiang L. Monitoring reclaimed lands subsidence in Hong Kong with InSAR technique by persistent and distributed scatterers, Natural hazards, 2016, vol. 81, no.1, pp. 541–543

Area A - Board 147

The Northern Black Sea coast suffers from significant landslide events as well as periodical floods as a consequence of the local geological and changing weather conditions, also due to the intensive construction activities in the tourism and leisure sector in the last twenty years. In the framework of the ESA-PECS funded project “Satellite-based Maritime Web-services for Bulgarian coastal area - SatWebMare”, we performed remote sensing observations over several test sites in the Northern Black Sea coast. The main goal of the project is the creation of Web-based System, integrated with Geo-portal, which aim is to provide information and services to end-users for maritime applications in the Black Sea coastal area. These actions follow the national needs and priorities set out in the National Strategy for Disaster Risk Reduction 2017-2030. To achieve the goal in georisk management domain we applied combination of several techniques with different time and spatial resolution. The advantages in dense time and wide space coverage of Sentinel 1A and B imagery are used to produce displacement maps and time series for the areas of testing sites. For the terrain deformation we used advanced PSI technology, the products of which are verified with GNSS products. Also we used optical Sentinel-2 imagery and UAV acquisitions to delineate and classify the areas under risk in the tested zones. Institutional provided datasets for the most active landslides in the study area were also used for validation of the achieved results. Based on the data from the National Cadastral administrative information system we performed evaluation of the vulnerability in the settlements in the areas of interest. Also an assessment of the agriculture units in danger is made based on the land cover maps, produced by Sentinel-2 classification procedure. The preliminary results from our investigation will be used as background in the construction of one of the engines in a multi-research platform which is the main objective of the SatWebMare project.

Area A - Board 148