We propose a novel processing chain based on modern spectral estimation algorithms to increase the Persistent Scatterer Candidates (PSC) density in Persistent Scatterer Interferometry (PSI). As many researchers showed, Capon's minimum-variance method (MVM) and Amplitude and Phase Estimation (APES) spectral estimation algorithms can be applied to SAR processing to improve the resolution and suppress the sidelobe level. However, these reprocessed SAR images cannot be imported into PSI directly. The main reason is that the resolution achievable by the reprocessing SAR algorithms depends on the statistical properties of the scene and is, therefore, spatially variant. Thus, the required grid spacing of the refocused SAR images is determined by the finest achieved resolution. Consequently, most observed targets are highly over-sampled, causing the traditional PSC selection algorithm to select pixel clusters rather than single pixels for each PSC. To tackle the problem, we introduce a PSC selection approach suitable to the reprocessed images and the processing chain. Including the adaptive resolution, we address two other features of the reprocessed SAR images: the peak jitter and the edge effect. Taking the features into consideration, we use three criteria to select PSC from the refocused SAR images: (1) the local peaks, (2) the normalized amplitude dispersion, and (3) the mean amplitude.

We evaluate the performance of the proposed algorithm by a quantity of simulation and a stack of TerraSAR-X data over Rotterdam, the Netherlands. Both the simulation and the real-data experiment show that the Capon algorithm outperforms others. With the Capon algorithm, we obtained a 58% increase of selected PSC w.r.t the traditional PSI processing for the real-life data. We also calculate the deformation velocity of the PSC by importing them into DePSI. The further comparison of the Capon algorithm and the traditional algorithm indicates the Capon algorithm can increase PS density significantly. The proposed algorithm can be used to improve the performance of PSI, especially when a higher PS density is needed.

Area A - Board 130

The South-to-North Water Diversion (SNWD) Project in China, is a large-scale inter-basin water transfer project aiming at alleviating the increasingly serious water shortage in Beijing, Tianjin and North China. In this manuscript, we will retrieve the dam deformation time series of Shuangwangcheng (SWC) Reservoir, one of the key storage reservoirs along the Eastern Route Project of SNWD. The designed maximum storage capacity of SWC Reservoir is 61.5 million cubic meters, with the highest water storage level of 12.5 m. The axis of the dam is 9.6 km. The project started in May 2010 and was completed in June 2013. During the filed investigation in September 2016, cracks are observed on the downstream face and the crest of the dam.
During the period of April 2017 and January 2018, an experimental network of GNSS was implemented on SWC reservoir dam. Three monitoring stations are selected on the crest of the dam near the dam-crossing culverts. The GNSS antenna of the base station was fixed to the top of a pillar with a reinforced concrete cushion cap and steel re-enforcement bars to a depth of more than 2 meters. 1-hour, 2-hours, 4-hours, 6-hours, 12-hours and 24-hours static baseline results from both GPS and BDS are evaluated for precision assessment. Deformation time series at the three monitoring sites were obtained and the correlation analysis of deformation and water level changes was conducted.
Different from the GNSS technology with high temporal resolution at spare monitoring points, InSAR provides much higher spatial density of deformation. 23 TerraSAR-X images with stripmap mode covering SWC reservoir were used for time series analysis. The pre-cast concrete slope protection is applied for the upstream face of the dam, and the upstream surface on the west side directly faces the side-looking SAR satellite. Thus, the reflection signal is strong, and more PSs were detected. Due to the turf cover, the downstream surface of the dam has poor interference coherence and we can hardly obtain effective signals. The top of the dam is flat, and the microwave travels approximate specular reflection. So, the echo signal is weak and appears dark on the SAR average intensity image. However, PS clusters appear near the lamp posts around the dam-top road because dihedral angle reflection was formed. With linear deformation assumption, mean velocities of PSs were retrieved, and the dam exhibits uneven subsidence characteristics.
Results are shown as graphics. Fig1: Photos taken at SWC reservoir while fieldwork: (A) View sight on the dam crest facing upstream; (B) Cracks found on the concrete stairs on the downstream face of the dam; (C) and (D) Cracks observed on the crest of the dam. Fig2: Layout of the GNSS stations established at SWC reservoir. Fig3: Deformation time series at the three monitoring sites (subfigure 1-3) and the water level changes (subfigure 4) during the GNSS observation session. Fig4: Correlation analysis of deformation and water level changes. Fig5: Coverage of the TSX image and the study area. Fig6: Mean velocity of the PSs on SWC reservoir dam.
Main conclusions can be drawn as follows: BDS can be used for deformation monitoring in Asian area where enough satellites available at current stage; the precision of BDS performs worse than that of GPS due to bad geometric distribution of the satellites, but the difference is not significant for longer session static baseline results; PS targets detected in time series InSAR analysis greatly improved the spatial information density, which helps to better reflect the deformation characteristics of the dam; uneven subsidence process occurred and may still be developing on SWC reservoir dam. We believe the manuscript will provide practical experience for similar projects.

Area A - Board 131

Decision makers in developing counties are tasked with managing a range of hazards and risks to their populations that include socio-economic, geopolitical, financial as well as geohazard risk. With the range of competing priorities and limited resources to invest in hazard and risk management, decision-makers are tasked to manage hazards affecting large populations without sufficient information. In challenging terrain, the budgets, expertise and personnel available to invest in data gathering in support of geohazard identification, assessment, monitoring and management is often very limited. Interferometric synthetic aperture radar (InSAR) derived from public domain data, such as Sentinel-1, has the potential to support decision makers in these environments through the provision of regular, high-quality ground deformation data that can support the identification of areas of potential instability and facilitate informed prioritization of the limited resources available to decision makers. This case study presents the benefits of Sentinel-1 data to support landslide hazard and risk assessment in a developing country and highlights the potential value for at-risk areas of including satellite data in the public domain.
The techniques used included an InSAR stack analysis of Sentinel-1 imagery to derive a point cloud with associated time series of deformation. The InSAR results were used to map the geohazards and identify areas at risks.
The presented work was completed by BGC Engineering Inc. as part of the philanthropic program, BGC Squared (BGC2). The program is a means for BGC staff to volunteer time, expertise and resources in support of communities and organizations worldwide. BGC2 projects are completed in partnership with academia, non-governmental organizations, communities and government agencies. These partnerships leverage the diverse skillset offered by each contributing organization to deliver on projects that consider the broader range of priorities affecting at-risk groups in developing countries. The work completed to date has been a proof-of-concept to assess the application of InSAR to landslide hazard and risk assessment in urbanized areas in developing countries. The next stage of the project will seek partnership with governmental agencies to evaluate the use and effectiveness of provision of InSAR data to trained professionals tasked with geohazard risk management.

Area A - Board 133

The wealth and quantity of Sentinel-1 data mean that we now have the potential to exploit time series and geostatistical techniques to monitor ground motions around critical infrastructure, slow landslides, and subsidence in near-real time. The challenges to automatic change detection revolve around the nature of the change itself. Subtle changes in the gradient require different statistical approaches to sudden changes in a pixel timeseries. Our on-online detection algorithm exploits the historical timeseries to assess if any new data points in the displacement time series are different to the historical trends. The algorithm raises an initial alert if the new data is deemed to be statistically different, while using additional SAR acquisitions to add additional confidence levels to the detection. Our detection threshold is given with a 95% confidence interval.

We test our algorithm on Bandung city in Indonesia where parts of the city has been experiencing rapid and variable rates of subsidence due to groundwater extraction. We show that the algorithm is able to detect changes in the radar timeseries around businesses in the city.

Area A - Board 134

Radar satellites can acquire imagery at night and irrespective of weather conditions (e.g. even when cloudy). Measurements from these satellites provide valuable additional input that can be used for a multitude of applications in support of disasters: determining of landslide and rock fall areas, re-defining fault tracks after an effective earthquake; mapping the geographical extent of flooded regions; monitoring the effects of droughts on soils, vegetation, crops; and tracking the extent of forest fires and oil spills. The extent of damages to buildings and infrastructure following a disaster can be assessed.
Sarpol-e Zahab earthquake is the largest event after the occurrence of 1909 earthquake around the shaking area. These two events sourced from the Zagros fold belt in the Zagros vicinity that has the most of the active faults of Iran. Zagros region is the continuation of Bitlis Suture in Turkey and it has main Zagros fault, Zagros Mountain front fault, High Zagros fault and other fundamental faults. Naturally, the area is under the effect of high seismic activity.
A strong shaking with Mw 7.3 occurred on Sunday November 12, 2017 around the Sarpol-e Zahab town in the border area between Iran and Iraq. It has a number of foreshocks and aftershocks increasing the total deformation, cumulatively. In this study, we have investigated how land and earth's surface deformed after such a strong earthquake and its scatters. Because, the ground deformation inspection in some villages are indispensable for the safety of citizens and infrastructures. The best way for monitoring of surface deformation in such a big event is the SAR technique which use microwave signals. This system could operate effectively during day and night under the different weather conditions. The Interferometric SAR (InSAR) allows accurate measurements of surface deformations with the mm resolution. There are several methods for the application of SAR techniques. One of them is Differential InSAR (DInSAR) indicating an uplift and subsidence around epicentral area precisely and we preferred to use it for sensitive vertical displacement in the target area. According to the results, epicentral area has been exposed a vertical displacement range between 90cm and -41cm in different sides of the fault.

Area A - Board 135

The high seismicity along the south-east coast of Kamchatka Peninsula due to the Pacific subduction slab and the related activities of what are the most active volcanoes in the world, attracted in the past decades the scientific interest over this remote region. While the subduction zone and mechanism had been thoroughly constrained, not much is known about the complex tectonic setting immediately north of the slab. This region had long been considered aseismic, however, in the recent past some strong events occured and acknowledged its highly potential seismic hazard.
The 29th March 2017 Mw 6.6 earthquake occured exactly north of the active Kamchatka-Aleutian trench junction and the ground motion had been registered by global and regional seismic networks. We focused our investigation on the coseismic surface displacement using Differential Interferometric SAR technique (DInSAR) and retrieved the seismic source geometric and kinematic parameters. ALOS-II ascending and descending and Sentinel-1 descending deformation maps have been jointly inverted using elastic half-space fault modeling.
Predominant uplift deformation was detected from both geometries, amounting to around 23 cm near the shoreline. The nonlinear inversion suggested a north-west-dipping oblique thrust faulting with right-lateral rupture. We constrained the maximum slip to 2.5m as described by empirical magnitude-slip scale laws. The fault geometry is consistent with that determined from seismological analysis, however there is a conspicuous disagreement in epicenter location. Finally we provided a geodynamic interpretation of the proposed fault model.

Area A - Board 136

The Virunga Volcanic Province (VVP) in the Democratic Republic of the Congo (North Kivu region) hosts two of the most active and hazardous volcanoes in Africa, Nyiragongo and Nyamulagira. Nyamulagira, which is a large shield volcano, is characterized by frequent eruptions but does not represent a major risk for the main agglomerations, in comparison to its neighbour Nyiragongo located only 13 km away. Nyiragongo is a stratovolcano hosting the largest persistent lava lake on Earth and directly threatens the city of Goma and its neighbouring populated area. In 1977 and 2002, flank eruptions of Nyiragongo caused major disasters. For many years already, urbanisation in that region undergoes sustained rapid growth, and the region counts 1 million inhabitants today, making the continuous monitoring of these volcanoes and the better understanding of the volcanic processes a top priority. In addition to continuous space-based observations (e.g., OMI, Sentinel), the recent and progressive deployment of the broadband seismic network KivuSNet since 2014, currently composed of sixteen telemetered seismic stations and three infrasound arrays within the VVP and Lake Kivu area, opens new monitoring opportunities.

Here, we present an overview of two main scientific results focusing on the volcanic activity at Nyamulagira and Nyiragongo over four years between early 2014 to late 2017. In order to improve our ability to interpret signs of volcanic unrest, we discuss the potential of multidisciplinary approaches using remote sensing observations (space-based SO2 emission estimates, high-resolution satellite SAR) and ground-based data (seismic, infrasound).

First, the seismic dataset is combined with SO2 emission for a time period spanning April 2014 to March 2017 (see also Theys et al., same session). Since magma movements at active volcanoes are commonly associated with modifications in SO2 degassing levels as well as changes in seismicity (e.g., long-period -LP- events and/or continuous tremor due to fluid-driven resonance processes), we develop a multidisciplinary continuous tracking methodology for magma migration over these 3 years in the VVP. Ambient seismic noise interferometry techniques are used to identify the variations of tremor source zones, combined with the detection and location of LP seismic events using a cross-correlation based approach. For time periods of limited data availability, a template matching technique provides further insights, even with only two available seismic stations. These data are combined both with the time series of SO2 emission variations and its spatial variations, the latter in time periods of favourable wind conditions allowing for the separation of the contribution of each volcano to the total emissions. We find that this methodology allows for a robust discrimination of magma migration in and out of the shallow plumbing system at Nyamulagira volcano, improving our ability to interpret signs of volcanic unrest on a daily time scale. On the other hand, Nyiragongo features a nearly constant background seismic tremor source and low SO2 degassing associated with the dynamics of its lava lake.

Then, we present results focusing on Nyiragongo volcano by combining high-resolution satellite SAR amplitude information with seismic and acoustic records. Since its last effusive eruption in 2002, Nyiragongo has been an open-vent volcano characterized by the world's largest persistent lava lake. This lava lake provides a unique opportunity to detect pressure change in the magmatic system by analyzing its level fluctuations. We demonstrate that this information is contained in the seismic and infrasound signals generated by the lava lake’s activity. The continuous seismo-acoustic monitoring permits quantification of lava lake dynamics, which is analysed retrospectively to identify periods of volcanic unrest. Synchronous, high-resolution satellite SAR images are used to constrain the lava lake level by measuring the length of the SAR shadow cast by the rim of the pit crater where the lava lake is located. Seventy-two estimations of the lava lake level were obtained with this technique between August 2016 and November 2017. These sporadic measurements allow for a better interpretation of the continuous infrasound and seismic data recorded at the closest station (~6 km from the crater). Jointly analysed seismo-acoustic and SAR data reveal that slight changes in the spectral properties of the continuous cross-correlated low-frequency seismo-acoustic records can be used to track fluctuations of the lava lake level on a daily and hourly basis. We notably observe that a major drop of the lava lake and the appearance of significant LP “lava lake” seismic events in November 2016 are a probable consequence of deep magma lateral intrusion (inferred from the increase of deep seismicity detected by the network). This study thus highlights the use of SAR measurements for calibrating continuous seismo-acoustic records with the lava lake level and consequently, the potential to continuously monitor Nyiragongo’s lava lake activity and subsequent information about pressure changes within the magmatic system.

Area A - Board 137

Mozambique is considered the most risk prone country for floods in Southern Africa. In February 2000, Mozambique suffered its worst flooding in almost 50 years: Over 800 people died and hundreds of thousands were displaced. On 11 February the Limpopo River, north of Maputo, broke its banks, contaminating the water supply and bringing dysentery to the local population. The worst came on 22 February when Cyclone Eline hit the Mozambique coast near Beira, with winds of 260 km/h and torrential rains. Eline worked its way inland, dropping huge quantities of water on the Limpopo River catchment. That water followed its way down the Limpopo River valley, and several days later flash floods occurred in the Gaza Province of Mozambique, arriving suddenly and burying the low-lying farmlands in the Ch´okwe and Xai Xai Districts under 4–8 m of water. Although all the provinces in the country were affected by floods the most affected province was Chokwe. The frequent floodings call for cost-effective method for flood mapping and damage assessment.

Covering large area at regular revisits, satellite remote sensing has been playing an important role in disaster management such as floods, droughts, cyclones, etc., especially for preparedness, warnings and response stages. Several studies have been conducted in Mozambique to analyze the risks of floods and mitigate the impact. However, Mozambique has lack of resources and often uses wide range satellite imagery with lower resolution. In General there is no near-real-time monitoring of floods using satellite due to the lake of professionals in the field and financial resources which leads to no access to higher resolution satellite data. With the launches of Sentinel-1 and Sentinel-2, free and open data with global coverage and high temporal resolution became routinely available. The objective of this research is to investigate multitemporal Sentinel-1 SAR and Sentinel-2 MSI for near real-time flood mapping and monitoring in Mozambique. All available Sentinel-1 SAR and Sentinel-2 MSI data during the 2017 and 2018 floods mentioned above will be collected. The methods is based on the analysis of the Senitnel-1 SAR backscatter changes in the flooded areas. The idea is to use the acquired Sentinel-1 time series before the event to characterize the backscatter variation in the area and then use the post event image/images to detect the flooded pixels and monitoring the event. Then pre-flood Sentinel-2 image(s) is used for land cover mapping. Finally the flood map from Sentinel-1 is overlaid on the land cover map derived from Sentinel-2 for damage assessment. The preliminary results show that the combination of Sentinel-1 SAR and Sentinel-2 MSI data is promising for near real-time flood mapping and damage assessment.

Area A - Board 138

The threat of water-borne diseases increases dramatically under flood conditions. Therefore, a first requirement in flood relief work and disease control activities is to map the extent of floods. In August of 2018, regions around the Vembanad Lake in the state of Kerala, India, suffered from extensive flooding, leading to displacement of over a million people, and causing some 400 deaths. Damage to homes, public infrastructure and the environment was unprecedented, placing a huge economic burden on the state. Authorities declared it the worst flood in one hundred years. During the crisis, newspapers presented maps that high-level officials of the Government of Kerala were using to direct relief work during the emergency, and which were generated from ESA Sentinel 1 and 2, as well as NASA Landsat satellites, showing the extent of flooding. Scientists and volunteers tested lake water, well water and drinking water for contamination in flooded areas. Here, we study the extent and locations of inundated areas in the region, based on maps produced from Sentinel-2 data acquired before, during and after the floods, and compare the results derived from different algorithms. We examine the steps required to make such information available in near-real time, on an operational basis, during similar emergencies. An integrated multi-faceted approach is required to minimise outbreaks of waterborne diseases in the aftermath of flooding, with scientists, government officials, medical officers, the media and the general public working together effectively. Satellite-based observations have a crucial part to play in this scheme.

Area A - Board 139

Key-words : Disaster risk ; Land cover ; Erosion ; Time series high resolution images ; High spatial resolution images ; Supervised learning ; Haïti

High spatial resolution remote sensing makes it possible to characterize natural disasters impacts and to map exposure and vulnerability in order to improve society and land resilience to extreme weather events. Combining time series of high spatial resolution products obtained from various sensors allows to monitor various processes such as erosion, runoff and land cover evolution on large scale. In Fall 2016, Matthew hurricane severely impacted the Cayes agricultural plain located between the mountainous Macaya National Park and the Caribbean Sea in Haïti. Floods and violent winds have led to high erosion rates and landslides in the headwater catchments located in the north while important vegetation and soil losses were observed in the plain agricultural fields. This 280 km² plain had also suffered two droughts periods in 2014 and 2015, and the country was impacted in 2017 by Irma hurricane. This work aims at monitoring climatic disasters impacts on ground deformation and land cover changes from 2014 to 2018 in the Cayes plain. To this end, different monitoring chains have been applied to various satelite products provided by Recovery Observatory of Haïti (Pléiades, SPOT6/7), PEPS – CNES (Sentinel 2 L2A) and USGS (Landsat 8). Land cover maps (2014 and 2018) have been performed using the Moringa processing chain (THEIA – CES OSO ; developped by CIRAD taking advantage of the GEOGLAM-JECAM site experiments). Moringa made it possible to produce a segmentation from very high spatial resolution image (ortho-image, 25 cm, 2014 ; Pléiades, 50 cm, 2018) and temporal analyzes from very high temporal resolution image (Landsat 8, 30m – 16 days, 2014 -2018 ; Sentinel 2, 10m, 5 days, 2018). To generate classification, Moringa has performed machine learning (Random forest method) with ground truth data. Ground deformation mapping evolution in plain (from 2014 to 2018) were performed using radar interferometry calculating the phase difference between Sentinel-1 time series products. Erosion and landslides evolution in mountainous areas were identified using a very high spatial resolution DEM (Lidar, 150cm, 2014) combined to a very high spatial resolution ortho-image (25 cm, 2014) and high spatial resolution SPOT 6/7 image (150 cm, 2017). Eventually, maps produced in this work will be used to define a management plan of this agricultural territory that faces numerous climatic hazards. The goal is to enhance the local resilience to natural hazards to reach a sustainable development in a context of climate change.

Area A - Board 140

Seasonal floods are part of a normal hydrological cycle in the Amazon region, where there is a network of large and extensive rivers bordering forest, rural areas and several urban settlements historically located on the riverbanks seeking the easy transportation to other regions. Traditionally, urban facilities have respected the limits of floods. However, increasing urbanisation has pressured the occupation of many of these areas, which has been a challenge for public agencies that need to develop actions to control risks and eventually remove populations from areas potentially flooded. As public organisations in the Amazon lack the material and technological resources an adequate supervision of these flood-prone urban areas, it is justified to investigate decision support tools that use medium resolution images to monitor the water bodies near to urban areas. The increasing free availability and accessibility of medium spatial resolution (10-30 m) satellite imagery as Sentinel-2 (S-2) are helpful to map flood areas and create conditions for continuous monitoring programs. Several flood mapping approaches have been developed to extract the extent of inundated areas, but few studies have investigated this issue together with urban settlements in the large Amazon Basin, at a 10m spatial resolution, particularly regarding the changes during the low- and high- water seasons. The main advantages of using remote sensing to support risk management of flooded areas are: (i) the low cost; (ii) the reliability of the source of information enabling multiple observations and; (iii) analyses throughout a flood cycle. In this study, we examine the performance of the S-2 multispectral instrument (MSI) imagery for delineating inundated areas along the river that borders the city of Marabá, Eastern Brazilian Amazon. Based on spectral information )extracted from S-2 MSI, we compare different water indices (NDWI, Modified NDWI and Water Ratio Index) to better detect the thresholds values of change able to separate flood and non-flood areas for the years from 2016 to 2018. Water indices are ranked regarding their efficacy and sensitivity to detect built-up land and produce enhanced water information for the study area. Moreover, comparative results using Google Earth Engine Platform were performed to evaluate a proposal of an automated framework to integrate urban maps and water indices.

Area A - Board 141

Understanding disaster risk in all its dimensions of hazard characteristics, exposure and vulnerability of assets and population is the first priority action of the Sendai Framework for Disaster Risk Reduction (2015-2030).

International Financial Institutions (IFIs) play a significant role as facilitators of funding in developing countries, in direct cooperation with national mandated disaster authorities to prevent and mitigate the adverse effects of natural disasters and foster sustainable development.
In this sense, the recently launched ESA EO4SD Disaster Risk Reduction project aims to promote the adoption of Earth Observation-based products and services mainstreamed into the working processes of IFIs funded projects that seek to prevent or mitigate the adverse impacts of natural disasters in developing countries.
The EO services proposed in this project comprises a set of mature and operational services of the European industry that needs to be tailored and customized to the specific DRM needs of the IFIs and Client states. The main thematic components of the service portfolio will be presented including hazard modelling of major natural hazards like geological (landslides, terrain subsidence, earthquakes or tsunamis), hydro-meteorological hazards (flooding, coastal and storm surge events) and climatological hazards like droughts or wildfires. All of these analysis are supported by multi-purpose EO services (Reference mapping, Land use & Land cover, topographic information or settlement characterization). Hazard modelling results will be linked to the exposure and vulnerability components to allow quantitative assessments of the overall risk scenarios.
The exposure is defined as the population and assets present in (potential) hazard areas that are thereby subject to the potential adverse impact or damage/losses. Vulnerability as the characteristics and specific context of any given exposed population and asset (eg: building construction characteristics, population age, incomes). Both components in its physical and socioeconomic dimensions are related to the available local, national or international census data and statistics at a disaggregated level to properly fuse this ancillary data with EO data in the quantitative assessments. Inherent constraints and limitations of final EO-derived products and services for end-users will be also reviewed.

Primary focus of the project is to use free and open satellite data from Sentinels compared to commercial Very High Resolution satellite imagery to overcome the barrier of a wide adoption in the operational workflows in IFIs and national agencies of the Client States with a mandate in DRM operations.
Some examples will be presented with the benefits provided by EO-services in IFIs funded projects , either in the design or implementation phases.

Area A - Board 142

METEOR (Modelling Exposure Through Earth Observation Routines) is a three-year ODA project funded by the UK Space Agency International Partnership Programme to develop innovative application of Earth Observation (EO) technologies to improve understanding of exposure with a specific focus on the countries of Nepal and Tanzania. http://meteor-project.org/

The escalating impacts of natural hazards are caused mostly by increasing exposure of populations and assets. A major challenge when making Disaster Risk Management (DRM) decisions is poor understanding of the distribution and character of exposure in ODA countries. Exposure needs to be mapped, monitored and modelled by Governments, NGOs, affected communities and businesses, seeking to bolster resilience and growth. Robust, quantitative methods are required to justify resilience decisions and risk mitigation. Projects have aimed to map exposure using Earth Observation (EO) using a range of approaches, though the application of these in DRM has been greatly limited by the fact that many have been poorly calibrated, for instance by being based solely upon readily available data, or were designed only for a particular setting.

METEOR takes a step-change in the application of EO exposure data by developing and delivering rigorous and open routines (protocols) and standards to allow quantitative assessment of exposure, with explicit uncertainties. These protocols and standards will be co-developed for broad application to ODA countries and will be tested and validated in two contexts (Nepal and Tanzania) to ensure they are fit-for-purpose. Geohazard footprints will also be developed for those two countries. The process of building capacity and co-delivering new consistent data will promote welfare and economic development in these countries, and demonstrate the applicability of the techniques elsewhere. METEOR will deliver country-wide openly-available exposure data for the 47 least developed ODA countries. Better-informed DRM decisions that meet the demands of international drivers (e.g. SDGs, Sendai Framework) will be underpinned by our national-scale data.

The project is led by the British Geological Survey with partners from Disaster Management Department of the Prime Minister's Office, Tanzania; the National Society for Earthquake Technology (NSET), Nepal; Fathom Global; Global Earthquake Model Foundation; Humanitarian OpenStreetMap Team; ImageCat Inc; and Oxford Policy Management Limited.

Area A - Board 143

Cyclone Matthew struck southwest Haiti as a Category 4 storm on October 4th, the first Category 4 hurricane to strike Haiti since Hurricane Cleo in 1964. With upwards of 1,300 lives lost across the Caribbean, and more than 1,000 lives lost in Haiti, the storm is the deadliest hurricane to strike in the Caribbean since Jeanne in 2004. The impact of Matthew will be lasting. While flooding caused significant damage and loss of life, the main impact was felt from the wind, which in some regions has destroyed more than 95% of buildings and has completely destroyed trees and agriculture. In addition, widespread environmental damage occurred.
In this context, the Recovery Observatory project, set up by the Committee on Earth Observation by Satellites (CEOS), and co-directed by the Center National d'Etudes Spatiales (CNES), is supporting the post-Matthew recovery phase since the beginning of 2017. Among the objectives of the Recovery Observatory, there is the production of an up-to-date land-use map of Haiti based on High-Resolution optical satellite data (i.e. Sentinel-2), as well as open source data and processing tools such as the iota2 processing chain developed by the Centre d'Études Spatiales de la BIOshphère (CESBIO) and OpenStreetMap data.
The purpose of this study is to identify the potential of iota2 with limited baseline data. iota2, designed for the territory of metropolitan France, had never been tested elsewhere in a country. The iota2 land classification chain was run with the 2017 time-series of Sentinel-2 images over the whole Haitian territory, with reference data mainly sourced from OpenStreetMap. The pre-treatment of the reference data is a major step, given that the large volume of databases and their quality in terms of cleanliness of their content. The main results showed that the performance of the classification process increase both with the selection of relevant reference data and target classes. iota2 has been operational in Haiti with productions in line with reference data. Meanwhile, the potential of OpenStreetMap data in the production of land-use maps by large-scale classification remains questionable with regard to the heterogeneity of their availability over any places in the world.
The overall accuracy of the map produced was 96.4% and the Kappa index was 88.7%, Haitians geodata experts evaluated this study as globally accurate. Nevertheless, the visual evaluation of the result is not so satisfactory. Confusion between classes and over-classification, particularly of the woodland class, highlights the acute need for relevant reference data for bare soil and grassland classes.
Results of this work being considered encouraging, and Haitian partners have expressed their interest and motivation to produce nor provide the missing reference data but also to being trained to use the iota2 processing chain. Further studies are currently conducted in order to implement a bio-climatic stratification during the classification model training phase. Furthermore, a re-design of the final nomenclature taking better into account the consistency between both the input image and reference data is also in progress.

Area A - Board 145

Nowadays, underground mining exploitation of various raw materials is connected with a comprehensive assessment of possible impacts on the surface area. In most cases, the introduction of an appropriate management policy leads to the minimization of the causes and thus the effects of such exploitation. However, some phenomena, such as mining-induced earthquakes, seem to be a cause that can not be entirely eliminated. The nature of these earthquakes is very dynamic and stochastic, which creates the greatest problems in the prevention of this type of phenomena. In spite of this, today's modern tools create real opportunities to determine the impact of earthquakes, determining their exact area and size of changes. Such a solution is possible thanks to the use of satellite radar interferometry - DInSAR (Differential Interferometric Synthetic Aperthure Radar). With data on the location and date of the occurrence of mining-induces earthquakes, it is easy to use specific satellite images. This allows to determine the vertical ground movements of the terrain surface at a specific interval. The research area was limited to European countries where seismic disturbances were made. These earthquakes occurred in the area of underground mining of mining raw materials such as hard coal, copper ores and natural gas. It was decided to present several interpretations of earthquakes due to the different geological structure of the rock mass and the extraction of different raw materials. This was contrasted with the occurrence of quite high magnitude of earthquakes, the values of which were obtained on the basis of EMSC service (European Mediterranean Seismological Center) measurements. An additional factor that influenced the selection of specific earthquakes was the date of their occurrence. It was decided to analyze the earthquakes that took place in the early spring or winter period, due to the low vegetation rate. This indicator plays a very important role in the analysis of interferometric data, because its high value can affect the data disturbance, and hence the incorrect values of vertical ground movements of the surface area. Satellite images used for research were downloaded from the open and free platform: Copernicus Open Access Hub using data from Sentinel satellite missions. Interferometric data was processed in SNAP and SNAPHU software. In order to obtain the final values of vertical movements in the form of lowering basins, geospatial software was used for this. Comprehensive analyzes performed in these programs allowed to determine the area of the boundaries of the impact of mining-induced earthquakes and final values of vertical ground movements. Additionally, a comparison of the above results with the current orthophotomap or topographic map gave the possibility of obtaining information whether the impact of mining-induced earthquakes included developed or undeveloped areas.

Area A - Board 132