Final Report Summary - SENSUM (Framework to integrate Space-based and in-situ sENSing for dynamic vUlnerability and recovery Monitoring)
Executive Summary:
The Framework to integrate Space-based and in situ sENSing for dynamic vUlnerability and recovery Monitoring or SENSUM project is a multi-disciplinary project that set out to develop innovative methodologies for the exploitation of Earth Observing (EO) remote sensing and ground-based methods to better understand those factors associated with the exposure and vulnerability of the built environment to, in particular, seismic and landslide hazard.
The aims of the project required the identification of various indicators relevant to earthquake and landslide risk that could be identified via EO products, as well as those that would need in-situ observations. Such indicators were related not only to natural parameters such as precipitation or slope, but especially to the vulnerability and exposure of the human environment. A related aspect was the development of tools for change detection, which not only would be exploited for gauging changes in the built environment over time (i.e. a result of increasing urbanisation for example), but also for monitoring recovery operations during the post-event period (e.g. extent and changes in tent camps, road accessibility, etc.).
In-situ observation and analysis schemes were built upon and expanded within SENSUM. The most promising for these is the GFZ-Mobile Mapping (GFZ-MOMA) system, involving an omni-directional camera and associated interpretation software, to implement Remote Rapid Visual Screening (RRVS) activities. This system in turn exploits the sampling framework developed within the project and based on the so-called focus maps, a means by which an optimised path for gathering observations is developed based on various weighted indicators (possibly derived from remote sensing data). The result is a sampling plan where the path considers those areas with a higher statistical need/value to acquire data/information. Furthermore, information obtained from foot-surveys making use of the GEM (Global Earthquake Model) building taxonomy can be integrated, leading to a more accurate picture of an urban areas vulnerability and, subsequently, risk.
A critical (and successful) element of SENSUM involved the very intense end-user interactions in the form of not only meetings and symposia, but also workshops that centred around a scenario training exercise referred to within the project as “the Game”. The Game takes a group of end-users and, after dividing them into relevant groups (decision makers,, information providers, etc.) presents a time compressed scenario and, from the interactions, allows shortcomings to be identified on a number of levels (knowledge, awareness of available products, institutional and cultural limitations, etc.).
Project Context and Objectives:
A critical requirement for the effective mitigation of, and response to, disasters of any type, natural or man-made, is the availability of appropriate information. Such information must be up-to-date, sufficiently resolved (in time and space) and accurate, while being able to be readily understood and rapidly applied within robustly defined uncertainties. It must cover the entire disaster risk cycle (see Figure 1.1) which we consider involving mitigation (pre-disaster), response (post-disaster) and recovery, and be effectively exploited by Disaster Risk Reduction (DDR) practitioners, land-use planners, decision-makers, infrastructure operators, and the general public. This information must also be able to take into consideration the dynamic nature of the risk to human society, especially its greater exposure to natural hazards due to increasing populations (especially growing urbanisation), the greater dependency on complex technological infrastructure and supply chains, and climate and environmental change.
Adequate information is essential to assessing the three components of risk associated with natural disasters: the actual natural hazards that afflict an area, including their potential spatial, temporal, and causal relationships; the exposure of an area’s population to these hazards; and the vulnerability of the area’s built environment to these events. The latter two are in turn related via various social and economic factors, both temporal and spatial, while in some cases they may also affect the hazards themselves (e.g. growing populations require more space for living and agriculture, hence leading to the potential for increased environmental degradation).
However, relevant information is often missing or inadequate, especially in economically less developed regions such as Central Asia, which are also experiencing rapid spatio-temporal changes in their urban environments. Such a deficiency is also often the case for more developed countries such as Germany, where the required information may not be available or is so only for limited areas, or not readily accessible owing, for example, to it being dispersed amongst different municipalities or government departments.
One contribution to solving this problem is the better recognition of the potential of Earth Observation (EO) remote sensing imagery. EO products’ greatest value lies in their almost-global coverage, coupled with their capacity to introduce a time-dependent element to any resulting datasets and models. While the usefulness of such information is recognised in many quarters, there is still a concern about the level of expertise and resources that are required to adequately exploit such information (in terms of both the necessary tools, the understanding of the imagery itself and the associated efforts and expense), and what can realistically be learned from such products. These questions then call upon the need to integrate such observations with imagery (and other information) that is gained in situ and which can identify features not possible from space- or airborne sensors, and which can also consider the temporal change aspects of the problem.
The lack of information about urban vulnerability and exposure, and subsequently the investigation of the potential added value by the combined use of EO and in situ observations, are thus the fundamental problems confronted by the Framework to integrate Space-based and in situ sENSing for dynamic vUlnerability and recovery Monitoring or SENSUM project.
SENSUM, whose consortium is made up of 8 European and Central Asia partners representing education, research and commercial institutions (see Table 1.1) is a multi-disciplinary project setting out to develop innovative methods to exploit observations and data from EO remote sensing satellites (the focus in this project being exclusively on optical sensors) and ground-based methods (e.g. omni-directional camera surveys) to better understand the factors associated with exposure and vulnerability. Figure 1.2 outlines the operational scheme that is envisaged to arise from the products of SENSUM. Such a strategy must consider the specific protocols adopted by each country or region for information collection, depending upon their level of technical expertise, available resources, hazard nature, distribution of the exposed assets, and DRR institutional culture. The importance of the latter point cannot be underestimated, and considerable effort was expended in understanding this. In addition, as seen in Figure 1.2 there is the need to integrate EO images with ground-based data, allowing a higher spatial resolution, respecting also the temporal component of disaster management cycle. This in turn is exploited for the development of more effective sampling frameworks, leading to the production of products (e.g. maps) that are of benefit to end-users, all the while keeping track of how the various datasets change with an information management system.
Figure 1.3 presents the work package (WP) structure of the project, which also best describes the associated work plan. The work proceeds (usually in parallel) from WP2 which deals with the collection, processing and integration of remote sensing and in situ data for dynamic indicator analysis, to WP3, which considers data collection from the perspective of statistical sampling, and integrates it over different spatial scales, leading to what are referred to as “focus maps”, meaning maps that identify areas where statistically there is a greater need/value in acquiring data/information. The identification of the different indicators that are useful for hazard assessment (which in SENSUM are focused on landslides and earthquakes) and dynamic vulnerability assessment are the themes for WP4, while the indicators relevant to dynamic recovery monitoring are covered by WP5. Naturally, the uncertainties associated with any results are critical, leading to the systematic accuracy assessments in WP6, which also targets the compilation of a catalogue of relevant, available remote sensing sources. Knowledge about the project, its results and developed products, has been disseminated throughout the DRR community (both practitioner and research) via WP7, while the project is managed through WP1 and WP8. A critical aspect of the dissemination involves so–called policy statements, which are presentations and documents for DRR decision and policy makers.
The primary goals of the project may therefore be outlined as follows:
• Provide analysis and mapping tools based on space- and ground-based data acquisition (primarily imagery) for continuous exposure and vulnerability assessment and analysis, and for post-event recovery monitoring;
• Provide user-oriented guidelines to the DRR community which clearly defines EO and in situ-based exposure products in terms of their capabilities and limitations; in this context, specifically, the demonstration of the integrative use of EO and in situ sensing is one of the focal points of exploration;
• Apply the tools developed by SENSUM to the specific hazards of earthquakes and landslides, including their temporal and spatial relationships, to test cases (see below) that represent examples involving varying amounts of available information;
• Explore and encourage the application and integration of EO remote sensing and ground-based methods to multi-type risk analysis and disaster response and recovery, including the consideration of the time-dependent component. Such efforts require the reassurance of DRR practitioners that such actions will not require excessive levels of expertise or resources, hence the commitment of the SENSUM project of the use of Free and Open Software Source (FOSS) tools, datasets and archives;
• Enhance the scientific and technical capacities of the Central Asian partner countries for DRR. This includes science and engineering students and researchers, DDR practitioners (decision and policy makers and responders), land-use planners and infrastructure operators.
As the methodologies and products developed within SENSUM need to be tested and validated, and to ensure their relevance to as broad a range of environments as possible, the SENSUM project included several test cases, which may be classified as follows:
• “Data poor” test case, here the Isfara-Batken region between the Kyrgyz and Tajik Republics (subject to earthquakes and landslides). This area has very little information available about any aspect of risk, but especially exposure and vulnerability. Its relevance is also connected to the cross-border character of the area, a critical issue when considering more effective DRR actions.
• The “data rich” area of Cologne, Germany (subject to seismic events, flooding and windstorms). Cologne has a large amount of information dealing with all aspects of risk, however, such information may also be out of date, or inadequate to thoroughly assess the city’s risk. There is also the prospect of difficulties of actually obtaining or even knowing what information is available, given institutional limitations.
• The “data intermediate” or “data critical” area of Izmir, Turkey (earthquakes). By this term, we mean that while there is considerable information available on all aspects of risk, however, it is still inadequate given the fairly high level of hazard and exposure of this location.
The resulting SENSUM products therefore contribute to the cost-efficient, multi-scale, and multi-temporal gathering of information. They acknowledge the need for tools that do not require a too-high a level of technical expertise, while as mentioned above, the primary consideration will be towards software tools and data that are available within a FOSS context. In terms of GIS tools, this means the use of such applications as QGIS , a free and open source GIS tool which is able to be used on a range of computer platforms and is under a continuous state of development. In terms of EO datasets, there is Landsat archive which provides continuous and near global coverage extending over 40 years from a series of 7 satellites and is free and open for download. In addition, there are the current and future Sentinel EO spacecraft of the European Space Agency (ESA) which will also be providing their observations (optical and radar) within the FOSS framework.
Project Results:
The following outlines the main outcomes of the SENSUM project. It needs to be remembered that while some tools developed within the project may already be exploited, many are still in need for further development. Likewise, some of the less “tangible” outcomes, such as our discussions with DRR practitioners, require a much longer timeframe before their true fruition can be recognised. It will also be emphasised how the different outcomes are in fact closely related to each other, as outlined in Figure 2.1.
Summary of currently available and future remote sensing products
When undertaking an integrative and comprehensive risk analysis, DRR practitioners, urban planners, decision makers and infrastructure operators commonly encounter the problem of appropriate data collection. Shortcomings in available data include when data sources are too generalized, outdated, inconsistent or simply non-existent, especially for economically less developed parts of the world. Throughout the disaster risk community, remote sensing is today perceived as a promising tool for economical, up-to-date and independent data collection and has been employed in various investigations within the geo-risk context (see, for example, the International Charter ). While the use of EO data for the study of earthquake and landslide hazard spans a wide array of applications - e.g. ranging from site characterization and terrain analysis to fault mapping, monitoring the dynamics of active fault systems/volumetric mass movements and the detection of hazard-induced land surface changes - remote sensing for vulnerability-centred investigations is a less well-established field of research. In this context, SENSUM primarily aimed at the derivation of pre-event vulnerability indicators related to the physical and structural vulnerability of human assets and the built-environment at various spatial scales. A thorough review of current EO datasets and products, future missions, their technical specifications as well as their (potential) applications within the DRR context served as a starting point for SENSUM product development, encompassing the versatility of remote sensing data and techniques with regards to the analysis of earthquake and landslide risks. The resulting report (deliverable D2.1 “Present-day and future remote sensing data”) provided a non-exhaustive, but comprehensive, overview of globally available present-day and future remote sensing data and products.
For their use in disaster risk research and management, the technical specifications of remote sensing data, such as their geometric and temporal resolution, their scene size and thus, the spatial scale of analysis they can be applied to, as well as limiting factors including data cost and coverage, determine the potential for operational applications. Data from past- and present-day sensors featuring coarse geometric resolutions and large-scale aerial coverage generally enable the overall evaluation of pre- and post-event disaster situations at a low cost. On the contrary, data availability from Very High Resolution (VHR) systems, such as airborne LIDAR or VHR optical data over larger areas for in-depth analysis, is still limited due to their high costs.
Upcoming EO missions have the potential to play a key role in future research by continuing and improving the observing capabilities of current missions for earthquake- and landslide-related investigations, as well as operational emergency response service delivery in the near future. As a prime example, the ESA Sentinel missions will feature enhanced geometric and thematic observing capabilities and increased revisit capabilities at no cost to users, and will further ensure data continuity of long-lasting missions as several past- and present-day missions, such as Landsat and SPOT, reach the end of their technical lifetime. In this context, the enhanced capabilities of the described missions are believed to enable a leap forward in earthquake and landslide risk research, both for hazard- as well as vulnerability-related investigations. A selection of promising launched, planned or proposed missions is presented in Figure 2.2 (see deliverable D2.1) with a focus on their technical lifetimes. The respective project report further holds detailed information on technical sensor specifications such as geometric resolution, swath width, revisit capabilities and potential fields of applications. A stringent categorization regarding the spatial scale of analysis is applied based on the aerial coverage, i.e. the swath-width-dependent scene size, and the geometric resolution of the sensors according to the harmonized classification scheme presented by the European space programme Global Monitoring for Environment and Security (Copernicus/GMES ), envisaging the possible future enlarging of the respective product portfolio beyond European countries.
Apart from the perspective on current and future mission, international initiatives from both have produced remote sensing based geoinformation products over the past years that provide valuable input to the vulnerability-related research of SENSUM. On the one hand, space-based pre-operational emergency response services have produced a large product portfolio and significant experience in post-event mapping applications in recent years. On the other, both large-scale global and regional land cover inventories have been produced as first-level approximations of human exposure and second-level descriptors of the built environment. These information layers are essential as they present almost the only data basis for risk-related spatial investigations in data-poor regions of the world, such as Central Asia. Nevertheless, a better understanding of each dataset’s strengths and weaknesses is still required – a challenge encountered through a systematic cross-validation of multi-scale space- and in situ sensed exposure products within SENSUM (see Section 2.4).
Overall, from a technical perspective, the constantly increasing availability and accessibility of remote sensing technologies and the enhanced technical capabilities of future missions will provide new opportunities and data continuity for a wide range of vulnerability-related investigations such as SENSUM and similar future projects.
End-user elicitation
One of the aims of SENSUM was to investigate decision-makers information needs and the potential demand for SENSUM products. This was done using scenario planning with senior disaster managers in Kyrgyzstan, Tajikistan and Turkey, reported in the SENSUM deliverables D4.1 “End-users assessment” and D5.1 “Comparison of outcomes with end-users needs”. These realistic earthquake scenarios were played-out during one-day exercises that allowed disaster personnel to simulate the post-earthquake decision-making process. The findings pinpoint how information derived from multi-resolution imagery can be used in planning and assessing recovery of transportation networks, transitional shelters and the built environment.
The scenario planning game developed for SENSUM involved getting people to dynamically ‘play’ through an imagined future. The ‘Game’ collapsed real time to highlight significant issues and to focus on strategic decisions. Scenario planning was developed by Herman Kahn at the Rand Corporation in the early 1950’s (e.g. Fahey and Randall, 1997) and used extensively in military planning. In the disaster management field, the approach has been applied in ‘shakeout’ type preparedness exercises and drills aimed at raising public awareness.
-Running the Game in Bishkek and Izmir
The scenario planning game was played in Bishkek, Kyrgyzstan and Izmir, Turkey. In Bishkek, the cross-border exercise was attended by 20 participants, including 7 people from the Ministry of Emergency Situations in Kyrgyzstan and 3 from a similar Ministry in Tajikistan, 2 from the Red Crescent, and 1 from the United Nations Development Programme (UNDP) in Bishkek. In Turkey there was a larger number of attendees (27), from AFAD Izmir and Bornova Municipality and from the central government AFAD HQ, as well as from the Turkish Red Crescent and Turkish Amateur Radio Society. The players were divided into three ‘teams’ – Events, Decisions and Information. The transcripts of the Game clearly revealed the interplay between events, decisions and information in a way that other types of enquiry, for example, user-needs surveys or interviews, would have failed to show. During the game, the discussions at each table were highly informative for the SENSUM team.
The choices of SENSUM products focused on damage assessment and early recovery. The SENSUM researchers believe that there is considerable potential for using remote sensing products beyond that communicated through these choices. However, this positive finding needs to be qualified in the light of analysing the game transcripts using Kahneman’s model of Fast and Slow thinking (Kahneman, 2011). One of the main insights was that the decision-making team in Bishkek, and to a lesser extent in Izmir, had difficulty using the information SENSUM could provide. At the time, we thought that the main reason was the lack of experience in using remotely sensed GIS data, time pressure and the inability of the information group to provide answers quickly enough. In Izmir, even when information began to flow quickly enough to influence decisions, the decision-making team seemed to use it to confirm what they already knew or had already decided, rather than inform new thinking.
On reflection, the answers seem obvious. It was unreasonable of us in hindsight to have expected these disaster managers to be able to develop strategic plans. They have been selected because they can react effectively using Kahneman’s “Fast” thinking, making decisions based on experience and established heuristics, with little recourse to information or analysis. Neither group were able or willing to think beyond Year 1, with long-term recovery planning needing a different way of thinking than relief efforts.
The obvious answer is for the authorities to set up two separate teams working independently, but with some degree of coordination. One would be the standing civil defence group of disaster managers – the people we had been working with in Bishkek and Izmir. The second team would include economists and planners responsible for developing recovery and development plans. The challenge this poses to information providers (the sorts of people SENSUM’s products are aimed at) is that two entirely different types of information are needed. To have any influence on crisis decisions, the information would have to be provided quickly in real time. Meanwhile, longer-term recovery planning by contrast requires detailed and accurate GIS-type information that can inform plan making. This is expensive and needs time, resources and highly skilled personnel that may not be available in many places prone to disasters, especially Central Asia.
-Outcome of the game exercises
The scenario planning game worked remarkably well in the two case studies, and the participants found it realistic and absorbing. They were able to imagine how the scenario might play out and use their imagination and experience to generate the detailed content. The Game had two main aims: a) to gain a deeper understanding on the real data needs of the end-users and b) to define a list of vulnerability and recovery indicators that would guide the development of appropriate SENSUM products. Both of these were achieved.
The Game also investigated what is required for remote sensing and GIS to become an integral part of the disaster risk management process. During the Game, users in both countries had difficulty in transferring information needs into requests for data products. Kahneman’s thesis of Thinking Fast Thinking Slow, suggests that the problem is more fundamental. The need to respond quickly means that disaster managers use the so-called System 1 thinking (based on intuition, experience and rules of thumb) almost exclusively and have difficulty using the kind of information SENSUM is able to provide currently.
Thinking Fast Thinking Slow also suggests that governments need to authorise two teams to respond to disasters. One, the usual civil defence team managing relief and immediate recovery who are good at System 1 thinking and the other who are good at both System 1 and System 2 (meaning evidence based) thinking, planning long-term physical, social and economic recovery (see Platt et al., 2014a; Platt et al., 2014b; Platt, 2015). Following such a scheme would therefore lead to a more comprehensive (and efficient) response to natural disasters and the enhancements of a society’s resilience.
Framework for data collection
The framework for data collection (see Figure 1.2 to see its relevance within the entire SENSUM scheme and deliverable D3.5 “Sampling framework” amongst others) may be divided into:
• What indicators are actually being sort?
• What changes have occurred and how these may be detected?
• What methodologies for in situ data collection are being employed (for example, omni-directional cameras, foot surveys).
• How does one actually design the sampling strategy?
-Identification and extraction of appropriate indicators from remote sensing
Physical vulnerability can be measured through a wide spectrum of indicators that have been presented in the literature (see Table 2.1 List of vulnerability indicators as defined after the literature review process made by EUCENTRE and refined after discussion within the consortium.). During the SENSUM project, two series of proxies related to earthquake and landslides were identified (see Figure 2.3 for the identification and selection process and deliverable D2.2 “Vulnerability indicators”) as being obtainable from remote sensing. The so-called Earth Observation Tools developed within SENSUM were also designed to provide a set of useful and easy-to-use algorithms both for GIS users and software tool developers . The former were released as a plugin for Quantum GIS while the latter are available as source codes, open to modifications and the implementation of new workflows.
Earthquakes
The selection of the adopted physical vulnerability indicators was made considering two main elements: the first reflects the differences in the seismic performance of buildings and infrastructure, while the second considers other affecting indicators that enhance or decrease the physical direct losses. According to the definition of a community’s vulnerability, the main essential values are people and their livelihood (UNISDR, 2004): buildings combine both of these key values and therefore serve as a prevalent monitoring representative.
A literature review of more than 70 papers was performed to try to match and intersect the requirements of a pure-risk-oriented approach with restrictions defined by remote and in situ sensing. The resulting list is mostly earthquake-oriented; however, some of the suggested proxies were also suitable for presenting the vulnerability of other hazards (e.g. landslides) after a discussion within the consortium.
The final proposed and agreed-upon list is summarized in Table 2.1 where the indicators are differentiated according to the type of data (defined mainly by resolution) used for their extraction. For example, the extent and age of a built-up area can be retrieved from medium-resolution Landsat-like imagery, while building footprints are derived from higher-resolution datasets. Unfortunately, a few proxies are strictly bounded to the availability of ancillary information, as explicitly reported in the table. The last column is dedicated to the method used for their extraction, varying from newly developed to available datasets like OpenStreet Map.
Landslides
Landslides represent a major threat to human life, property and constructed facilities, infrastructure and the natural environment in most mountainous and hilly regions of the world. In landslide risk assessment, data availability is a major constraint, which greatly affects the type of methods and models that can be developed. Remote sensing is a promising tool for economical and up-to-date data collection, which also could be applied to monitoring the dynamic development of risk. The spatial and temporal distribution of the risk for landslides can be assessed by monitoring hazard indicators (e.g. slope characteristics), exposure indicators (e.g. number of houses or total population) and vulnerability indicators (e.g. population density, settlement structures or indicators related to structural vulnerability).
Indicators for application in landslide risk assessment as a function of space and time should be selected based on:
• their importance to the assessment of landslide risk,
• the ability to apply remote sensing techniques in the data collection,
• their relevancy to end-users’ needs.
The selection of indicators relevant for landslide risk was based on a literature review and end-user assessment, as described in Section 2.2. For landslide hazard, indicators derivable using remote sensing techniques are listed in Table 2.2. In addition, there are important landslide hazard indicators that cannot (or only with difficulty) be directly derivable from remote sensing, such as soil types, soil depth, soil moisture and geotechnical properties.
The landslide hazard indicators could be subdivided into triggering factors (e.g. extreme precipitation) and susceptibility factors (e.g. slope). The data for the susceptibility factors were reclassified into 5 classes of susceptibility, representing very low, low, medium, high and very high susceptibility to landslides, respectively. For example, Figure 2.4 illustrates the susceptibility factor "slope angle" according to the above mentioned classification, i.e. division into 5 classes (Figure 2.4a). Similarly, Figure 2.4b. shows the indicators indicator soil moisture. This was evaluated by overlaying maps of ground permeability and mean precipitation, i.e. the mean 30-day precipitation as derived from the dataset provided by the JAXA/NASA Tropical Rainfall Measuring Mission (TRMM ).
For structural vulnerability to landslides, indicators derivable with remote sensing techniques are in Table 2.3. In addition, important vulnerability indicators that are not directly derivable from remote sensing include size and conditions of openings towards the slope. Exposure and non-structural vulnerability indicators include, e.g. built-up area, built-up density, population density, accessibility and distances and land use.
For the application of these indicators within the development of focus maps (see Section 2.3.4) it was decided to focus on landslide hazard and exposure for the following reasons:
• The end-user assessment showed a stronger need for products related to landslide hazard than for products related to landslide vulnerability.
• The capabilities of remote sensing are greater for landslide hazard assessment than for landslide vulnerability assessment, i.e. more relevant data for hazard mapping is able to be gained by remote sensing.
-Change detection
Change detection is by definition the capability to detect and highlight changes occurring in space and time. Within the project’s framework, change detection products, which focus on exposure, can be used as an input indicator for other tools produced within the SENSUM consortium. The EUCENTRE team proposed a post-classification comparison algorithm based on the capability of the other algorithms to automatically process a stack of Landsat data. In particular, the tool is capable of comparing results obtained by the object-based built-up area analysis from different years. Results are also improved thanks to the expected continuity over years. A schematic view of the post-classification process is represented in Figure 2.5. The results of the built-up extraction using the object-based methods are combined together using logic operations and the capability to track changes in time due to the repeatability of the methodology.
For the areas of interest for SENSUM in terms of vulnerability and exposure mapping, medium resolution images such as those from Landsat would provide temporal information and focus area mapping. For the other intended application, that is for post-disaster recovery monitoring, users would also use temporal analyses to identify focus areas followed by high-resolution data techniques to track specific changes for recovery and building vulnerability analyses. Based on the changes within the indicator-specific GIS objects in a high resolution image, indices are calculated and plotted over time. The SENSUM method detects change, which is identified as the absence of similarity between gradient, roughness, texture and edges. For the purposes of recovery evaluation and monitoring, the method can be used to detect changes to buildings over time and therefore track the progress of initial damage, clearing, reconstruction and recovery due to an earthquake. An example of this is shown in Figure 2.6 which shows a sequence of images demonstrating the use of the SENSUM building change detection index. The pre- and post-disaster images are clipped by a GIS objects tool and the change index provides the degree of change to the selected buildings over time, indicating possible signs of initial damage and subsequent recovery.
-In situ data collection
For rapid in-situ image data capturing, a mobile mapping system (GFZ-MOMA) has been developed and extensively tested during the project. The system is composed of a Ladybug3 omnidirectional camera from Point Grey Research Ltd., a data capturing and storage unit, a navigation unit and an external battery pack that supplies the energy for up to 6 hours of autonomous operation (Figure 2.7). This camera is made up of 6 colour Complementary Metal Oxide Semiconductor (CMOS) sensors that capture concurrent image sequences up to 15fps acquisition rate. The 6 image streams are synchronized and automatically stitched into omnidirectional (panoramic) high resolution (5700x2700px) format with JPEG compression. The camera system is operated from inside the car and is mounted on the vehicle's roof with a simple mounting system (see deliverable D2.4 “Capture of in-situ image data”).
The data capturing and storage unit has been developed with the specific focus on ease of use and ruggedness for robust outdoor applications, even under rough conditions (e.g. unpaved roads, dust). A custom software application captures, synchronizes and saves the different data streams coming from the camera, the GPS and the Inertial Measurement Unit (IMU). Location is associated with each omnidrectional image by b-spline interpolation of the GPS positioning.
The navigation unit uses QGIS as the main software component for location tracking and car navigation. Its map interface is able to combine various background maps of the study area and to display pre-calculated sample areas and routes. The position can be tracked and displayed in real time with the GPS live-tracking functionality. This allows an operator to navigate the car along pre-calculated routes, while also rescheduling the path on-the-fly to cope with unexpected environmental conditions (e. g., traffic jams or road blockages).
An analysis of the captured omnidirectional image sequences is performed through visual image interpretation. Because of its similarities to commonly used screening procedures, the difference being that it is performed remotely, this novel technique is referred to as Remote Rapid Visual Screening (RRVS, see deliverables D2.6 and D3.5). A specific RRVS system has been implemented to allow for user-friendly, rapid and standardized image analysis. The system is composed of a map-interface, an omnidirectional image viewer and a customizable data entry form that currently supports the per-building indicators that were identified as part of the SENSUM project for vulnerability assessment and allows the display of building locations along with the GPS locations of the images captured during the field survey. Taxonomies are defined in order to describe the indicators and their values in a standardized and comparable way. Once a building of interest has been located in the map-interface and the nearest captured omnidirectional image has been identified and displayed in the image viewer, the building can be screened remotely by structural engineers. Uncertainties related to the assignment of attribute values can be quantified by the operator through additional qualifier attributes. In the current implementation of the RRVS tool, the degree of belief is supported as a qualitative measure of uncertainty for each attribute.
In-situ surveys using the above mentioned systems have successfully been carried out as part of the SENSUM case studies in Cologne (Germany) and in several towns and cities in the Batken district of Kyrgyzstan. A survey in Izmir (Turkey) could not be conducted due to permission problems. However, as part of a scientific service by GFZ that evolved out of the SENSUM project, an additional survey has been performed in Istanbul (Turkey). This survey was requested by the Turkish Catastrophe Insurance Pool (TCIP ) and aimed at testing the applicability of the proposed mobile mapping system and sampling framework for operational earthquake insurance claim management.
-Sampling design
As shown in the preceding section, a simple, customized mobile mapping system can efficiently and rapidly collect large amounts of high-quality, georeferenced visual data about building stock, infrastructure (e.g. roads) and in general the features contained in the visual surrounding which relate to the assessment of expected consequences in case of an earthquake. Such a system can be easily deployed wherever it is needed, does not require particular skills to be operated, and is economically viable. The data collected through the mobile mapping system can be easily distributed and analyzed remotely, and the intensity of the analysis (aka the number of remote operators allocated to the analysis) can be easily scaled up or down according to the contingent necessities and on the business case. For instance, additional skilled operators could be seamlessly integrated into the analysis protocol in case a bigger effort is requested on a temporary basis.
By applying a consistent prioritization and optimization approach, a strong increase in the efficiency of the survey can be obtained, especially when small sample sizes are required in order to decrease the impact of the survey on the available resources. In order to realize such prioritization, the previously mentioned focus maps concept has been proposed. This implements a customized, unequal inclusion probability scheme which can be used for employing PPS (Probability-Proportional-to-Size) or STR (Stratified) sampling designs with an accompanying increase in the overall statistical efficiency.
Automatic routing optimization proved to be a simple and scalable solution to optimizing the actual data collection in urban environments. The set of sampling locations obtained by means of the computed focus map can be converted into an actual path to be followed by a mobile mapping system in order to implement the survey. The proposed optimization algorithms are FOSS-based, and can exploit official or authoritative information on the existing road networks, as well as freely available datasets from OpenStreetMap (OSM). The implementation of the routing engine server-side moreover allows for the implementation of compact, flexible and reliable IT frameworks for data collection, integration and access.
In order to maximize the reliability of the system, auxiliary information might be collected and kept up-to-date in order to provide a consistent reference for prioritization and optimization of the survey activities (see Figure 2.8). The collected data would moreover have a significant value, since a better description of the exposure and vulnerability of the building stock could be achieved by, for example, analyzing the collected data and integrating them into a flexible geographical database. Pre-event and post-event surveys can be independently optimized by integrating multiple data (possibly including statistical inference), thus resulting in a consistent improvement in the operational capabilities in the field of data collection and related analytics.
Combining focus maps, dynamic sampling frameworks and mobile mapping for the data collection phase, a new generation of iterative and incremental data collection and integration approaches can be realized (see Figure 2.9). This would provide end users interested in vulnerability monitoring and risk assessment with a powerful and scalable tool to increase the reliability and timeliness of the resulting model, and to undertake more informed mitigation and prevention activities.
-Information life-cycle management
With an increase in data volume and complexity from a large variety of sources – reaching from satellite remote sensing and crowd-sourcing to cadastral and in-situ surveys – storage and management of geospatial information becomes an important topic, and one that was addressed in the SENSUM project (see deliverable D3.7 “Information integration”). Especially when considering the management of the information life-cycle (“from the cradle to the grave”), the integration of heterogeneous information and the distributed versioning and release of geospatial information are important topics. In the SENSUM project, a conceptual and technological framework was developed to tackle major questions regarding the information life-cycle management, the non-distributed and distributed versioning of geospatial information and its release options. A life-cycle management solution was provided based on a relational spatio-temporal database model that is coupled with Git and GeoGig repositories and workflows.
Figure 2.10 shows a schema of the information life-cycle as followed within the SENSUM project and implemented at the database level in PostgreSQL . New information enters the database workflow in the creation phase. Along with the actual object to be added to the database, several metadata attributes need to be defined. Along these are the source of the information item and its value (e.g. accuracy). The use of data entry forms for user -driven information creation can enforce these additional metadata attributes. Once new information is created, it is integrated with already existing information. In case of geospatial information, it is checked if other information items already exist in the database at the same location of the newly created item. In case of an overlap between new and old items, their values are compared and the information item with the lower value is replaced with the higher value item. The overall values of the database are than updated accordingly and the information is ready for usage for further analysis and interpretation by the user. Several customized spatio-temporal query functions and views are implemented in order to improve information accessibility during the usage phase of the life-cycle model. The database model supports transaction logging and therefore archiving of records. Once information is modified or disposed, its previous state can be archived and the database transaction can be logged. In case the modification refers to an updating by information with higher values (meaning better information became available), the archiving mechanism allows to retrieve or recover previous database states at any given transaction time. In case of the modification or disposal referring to real-world changes, this mechanism allows the archiving of the full history of an object from its creation (e.g. construction of a building), through modifications (e.g. retrofitting of a building) until its disposal (e.g. destruction of a building).
The content of the database can be released and versioned at any time in order to distribute and share it in a decentralized way. Information to be released can be the whole database or subsets of it in the form of summary views. Once information is released it enters a new workflow. For releases and distributed versioning we support the use of Git and GeoGig. A possible release workflow is schematized in the bottom part of Figure 2.10.
Framework for data validation
Despite the unclear definition of the term exposure, i.e. objects such as people, properties, infrastructure or economic units adversely affected by a potential hazard, it is widely recognized that the understanding and spatial quantification of exposed elements is of crucial importance for a comprehensive risk assessment. Due to the large-scale extent of human activities on our planet, remote sensing has been widely recognized in today’s scientific community as a valid tool to capture the elements at risk objectively and across large spatial extents. It is applicable on various spatial scales ranging from global to local to capture physical elements of the built environment of human settlements, as well as quantified indicators relating to these exposures. On a local scale, EO-based geo-information can be further refined and added by techniques of in situ sensing for an even more detailed characterization of the structural vulnerability of the built environment. In this context, SENSUM provides a set of user-friendly open source software tools to recall established EO information extraction methods for exposure mapping from space as well as workflows to derive important vulnerability information of human settlements on a per-building level using techniques of in situ imaging and ground-based sampling to add to the existing available exposure product portfolio. Together with existing global to regional EO-derived exposure layers, the proposed methods can contribute to a non-exhaustive, but comprehensive description of the vulnerability of human assets on our planet.
Nevertheless, classifications based on remotely or in situ sensed datasets are basically abstract approximations of the true physical environment of the Earth’s surface being monitored. Thus, today’s scientific and – above all – end-user community still struggles with a clear understanding and definition of the quality and accuracy of the manifold available spatial exposure products. In this regard, a systematic cross-validation framework for both SENSUM product developments (remote sensing toolset and in situ imaging workflows) and existing global and regional exposure products was produced. The respective report, deliverable D6.2 “Accuracy Assessment and Cross-Validation”, establishes clearly the value of remote and in situ sensed geo-information within the context of SENSUM. Significant criteria for data selection from a user perspective, such as accuracy, thematic detail, geometric resolution, cost and added value, are reported as product definition tables, enabling a systematic, comprehensive and user-oriented guideline for application in risk-related applications. In this context, validation efforts consisted of three main tasks based on the specific data subject to analysis. These are:
1. Systematic accuracy assessment of global exposure products addresses the validation of external EO-based exposure products, mainly global layers of urban extent which present valuable first-level approximations of human settlements on a global scale. In this, a systematic multi-scale concept has been developed to assess the accuracy of these products based on a review and selection of state-of-the-art techniques and methods for meaningful accuracy assessment. Applied techniques include absolute accuracy measures based on the error matrix, measures of relative inter-map agreement (e.g. receiver operating characteristics) as well as methods for pattern-based accuracy assessment. The developed approach respects the different spatial scales and resolutions of the layers under study, as well as pattern-based differences to assess accuracies in structurally different areas (e.g. rural vs. urban). The results (see the examples presented in Figure 2.11) show that new high-resolution products investigated within the context of SENSUM, namely the Global Urban Footprint (GUF, DLR ) and the Global Human Settlement Layer (GHSL, JRC ), hold great potential for mapping the spatial outline of settlement patterns (even smaller settlements in fragmented, low-density rural areas) in a very detailed manner and will significantly improve the geometric capabilities of the global exposure database, mapping even small-scale settlements of rural landscapes, as is the case for the test site in Central Asia. On the contrary, existing low-resolution (LR) products such as Globcover (GLOBC, ESA ) or the MODIS Map of Urban Extent (MODIS, SAGE-WISC ) map larger urban areas accurately, but are too coarse/generalized for applications in rural areas.
2. Systematic accuracy assessment of regional exposure products addresses the validation of external EO-based exposure products of regional coverage (e.g. coming from Copernicus/GMES) that present second-level structural qualifications of the built environment beyond binary formats (i.e. built-up, not built-up). In this context, several pan-European geo-information layers coming from the Copernicus/GMES programme have been assessed and validated. With regard to these layers, statements dealing with their accuracy vary greatly within the literature and across test sites, since information is heavily dependent upon different workflows of automated image processing, scene-specific quality of satellite imagery, as well as visual inspection/screening differences between interpreters. However, concerning risk-related applications for larger urban areas, the accuracy, especially in terms of the quality of the structural information content, is of great interest for approximating their structural vulnerability. The validation efforts carried out in this regard thus focused on both mapping accuracy and the presented structural information content (see the example presented in Figure 2.12) based on available, appropriate and up-to-date reference data.
3. Cross-validation and statistical testing of methodologies proposed within SENSUM for exposure data collection highlights the added value brought to the user by these SENSUM developments. These include the open-source remote sensing software developed for vulnerability indicator extraction (deliverable D2.2) and the sampling framework for in situ data collection (deliverable D3.5). The former were assessed using appropriate reference data on city and building level by the use of standard measures quantifying thematic accuracies based on the error matrix (Figure 2.13). Overall, the feasibility and general viability of all implemented workflows and algorithms provide the user with a powerful open-source tool to extract settlement information (which is the core of exposure models) from aerial and satellite imagery with a high level of both reliability and automation. Regarding the in situ data capturing, less conventional validation approaches based on sample data for Cologne were chosen due to the distinct nature of the proposed in situ data collection approaches. For example, the validation of the sampling scheme primarily aimed at the quantification of the efficiency of the selective sampling approach in terms of a comprehensive representation of the structural urban configuration (Figure 2.14). In addition, selected parameters that contain structural and demographic vulnerability information on the per-building scale obtained through visual screening employing the RRVS (see above) were evaluated with respect to the appropriate reference data. These assessments underlined the effectiveness of the approach, as well as the high potential degree of confidence that can be associated with the in situ derived information content.
Test sites – use of SENSUM products
The following will summarise some of the field activities undertaken during SENSUM and, where appropriate, outline the manner in which the SENSUM products have been exploited in each case.
-Central Asia
Kyrgyzstan
First order focus map
The first order focus maps for the Batken province, Kyrgyzstan, consist of hazard and exposure maps for landslides induced by precipitation and earthquakes. The method used is a heuristic approach largely based on the hotspots method that has been applied both globally and in drill-down studies (e.g. Nadim et al., 2006). The basis is the expert-based assignation of weights for available susceptibility and triggering factors. The susceptibility factors were terrain slope, lithology, ground moisture, and land cover. The triggering factors were 30-day precipitation and seismic intensities. Due to the lack of ground-based precipitation measurements, the maps for moisture and the precipitation triggering factor were derived using satellite-based estimates from the TRMM mission. Seismic intensities were provided by a previous study performed by Bindi et al, (2012). Exposure conditions were evaluated in terms of population and road networks. The maps were provided as an aid to field surveys performed by CAIAG during late spring and summer 2014 in the area of Suluktu, Kyrgyzstan.
A landslide hazard map for precipitation triggered events, obtained with the hotspots methodology, is shown in Figure 2.15. The importance of the different triggering and susceptibility factors can be calibrated against the information available in landslide inventories and physical processes.
Landslide inventory
Field work on landslide studies were conducted during SENSUM in the south-western part of Tien-Shan. This region is located in the Batken administrative oblast and borders with Tajikistan. In the study region, landslides are abundant on a relatively small area of 15 km radius from the town of Suluktu. Landslides here occur on the sedimentary cover of Mezo-Cenozoic and Quaternary loess loam soils (see Figure 2.16) which completely cover the bedrock in the intermountain areas. Most of landslides in the area occurred on the northern foothill slopes. Landslides located in the sub-mountain belt in the western and south-western part of the Fergana basin occurred predominantly on the back side slopes of anticline structures. In most cases, such slopes are north facing as a result of modern tectonic movements. Anthropogenic landslides also occur in this area, triggered by the exploitation of coal deposits.
An inventory study of landslides was conducted on the basis of digital geology, geomorphology, soil cover, and hydrogeology maps, resulting in landslide polygons. Distance measurements using laser instruments (TruePulse 360 ) and space images (Bing , Google and RapidEye ) were also employed. Modelling of slope parameters was carried out on the basis of the ASTER DEM with a 23 х 23 m resolution. During the course of the field work, samples of the upper landslide layer rocks were taken to identify the physical mechanical ground properties. In total, 45 landslides were surveyed, involving the consideration of 50 characteristics.
Second order focus map
The resulting landslide inventory was employed for several purposes in the drill-down analyses (i.e. in the production of 2nd order focus maps), namely serving as input for the development of an empirical run-out model and for a triggering model for the Suluktu area. The model used for the production of 2nd order focus maps for landslide hazard has three components:
1. Triggering: An infinite slope model was used. Only rainfall triggered landslides were considered.
2. Run-out: a statistical/empirical landslide run-out model was applied, that has been developed based on the data from the CAIAG landslide inventory.
3. Monte Carlo simulation: The Monte Carlo Simulation was performed by randomly varying the soil parameters and landslide dimensions according to the statistical variation observed in the CAIAG landslide inventory. Log-normal distributions were applied for the parameters.
The results of the analysis were combined with the exposure (built-up areas) and the resulting map is shown in Figure 2.17.
Tajikistan
Seismic and landslide studies were undertaken in the Jorgatal area, located in the north-east part of Central Tajikistan, on the border with Kyrgyzstan. The district is drained in an east-west direction by the Surkhob River, starting at the confluence of the Kyzylsu River (from the northeast, with its source in Kyrgyzstan) and the Muksu River (from the east, with its sources in the Pamir).
Seismic and landslide hazards investigation were estimated for Jirgatal area based on the published sources, archive data and interpretation and analysis of the existing satellite imagery and Google Earth resources like a most current sources. A map of seismic hazard, namely a probabilistic seismic hazard map (Figure 2.18) showing a 10% probability of exceedance during a 50 year period was created for Jirgatal area (in terms of peak ground acceleration (PGA) based on the existing active faults map and earthquakes catalogue (based on the Earthquake Model Central Asia catalogue, e.g. Mikhailova et al., 2015) making use of the software tool CRISIS 2007, a seismic and tsunami hazard tool .
Most attention was paid to landslides, although debris flows and snow avalanches are the most common natural hazards for study area. Moreover, efforts were not concentrated on individual landslides, but rather they took into account their probability, dividing potential landslides areas into 5 levels of possible hazard. This is a new generation of special landslide hazard maps based on the results of the interpretation of satellite images and the analysis of traditional landslide surveys, including the region’s geomorphology, lithology of the Quaternary deposits and engineering geology. The resulting landslide hazard map is shown in Figure 2.19.
In addition, we calculated the probability of the landslides under the influence of seismic loading based on three parameters: slope angle, curvature of slope and calculated PGA values for the area. The equation proposed by Uchida et al. (2004) was used, and the results are given in Figure 2.20. Finally (not shown here) a substantial database describing the settlements of the Jirgatal area was compiled, including population, gender balance, schools, health centres, types and distribution of buildings and bridges.
-Turkey
Izmir
The availability of medium and high resolution satellite imagery for the city of Izmir guaranteed the chance to test the set of released products. Figure 2.21 shows the evolution of the built-up area over the years obtained to the stack satellite and change detection (post-classification) algorithms applied to Landsat data.
Figure 2.22 and Figure 2.23 are related to indicators extracted from high resolution imagery (World-View 2). In particular, the former displays the extraction of footprints in a high-density residential area; the different colours being associated with the roof types. The latter is related to the output of the density algorithm used to highlight areas with major concentration of buildings within the area of interest.
Van
The destructive 7.1 Mw earthquake that struck eastern Turkey near the city of Van on 23 October 2011 provided an opportunity to test the SENSUM products devised for assessing and monitoring post-earthquake recovery. Many issues surround post-disaster recovery and reconstruction, usually requiring the combined effort of planners, government authorities and community interaction. However, natural disasters produce an unusual set of constraints and tensions that leads to specific needs for each case. The aim of SENSUM is therefore to provide the appropriate tools to assist in this process.
One of the main physical indicators used to assess disaster recovery is the erection and disbandment of camps. The identification of open spaces to first site camps, the number of tents needed and present during the relief phase, and finally the duration these camps are present, all give an indication of the post-disaster needs of a community and the speed of recovery. The number of tents can be used as a proxy to estimate the number of displaced persons needing permanent housing. For SENSUM, as part of a suite of tools to assist with monitoring recovery, the team developed an automated method of camp detention using mathematic morphology (see deliverable D2.6 “Supervised/unsupervised change detection”). This method is especially useful for objects characterised by clear shapes, sizes and contrasts. It uses morphological erosion and morphological dilation to identify targeted objects such as tents.
Using a series of images acquired for Van (one before and two at different times after the earthquake), out of a total number of 209 tents at a particular camp, only 33 were unclassified, giving a false negative error of less than 16%. Only 6 out of 209 tents were counted twice, giving a false positive error of 3%. Therefore, overall the number of tents correctly identified was 170, giving an accuracy of around 81% accuracy (Wang et al., 2015). This is shown in Figure 2.24. The set of imagery acquired for Van were also used to test SENSUM products on open space, roads and buildings identification and change detection.
-Germany
Cologne
The municipality of Cologne has a large amount of information dealing with all aspects of risk, however, such information is potentially out of date, or inadequate to thoroughly assess the city’s current risk. With respect to flood risk for example, detailed information about the location and height of buildings was available to the project in the form of cadastral data and a 3D city model. Moreover, a flood hazard map could be used for the study site. However, detailed information about the building stock with respect to particular indicators related to flood risk, such as the number and location entrances, occupancy of the first floor, number and location of windows, etc., is not available. Therefore, the SENSUM project in Cologne focused largely on a prioritization of an in-situ survey that aimed at identifying and quantifying these specific flood-relevant building indicators.
The focus map and resulting sample path for the Cologne in-situ survey are presented in Figure 2.25. The focus map was produced by combining information on the spatial distribution of buildings with a 100 year return period flood scenario. Equal weights were assigned to the building density and the inundation scenario in order to balance the relative importance of these factors for the sampling. A multiplicative pooling operator was chosen to prioritize the data collection based on the areas that are exposed to both inundation and characterized by a higher density of buildings, therefore filtering out most of the urban environment in Cologne.
The routing operation was performed on the basis of OpenStreetMap data with cost-constraints being fixed to the travel length. Therefore, the sample path is optimized in the sense of providing a shortest path through all the sample points identified on the basis of the focus map. The complete path´s length amounted to ~100 km. The collected geo-referenced imagery was then analysed off-line using the RRVS. A sample of 800 residential buildings has been selected and several attributes of the buildings have been collected, including attributes of particular interest for the estimation of the vulnerability to floods.
Figure 2.26 shows a summary of the flood relevant attributes for the 800 sample buildings that were captured by the GFZ-MOMA system and analysed by a human operator through the RRVS tool and procedure. Data capturing was performed in one day, whereas image interpretation took about 3 days. It can be seen from Figure 2.26 that the clear majority of buildings have a single door entrance with more than half of the samples having an elevated ground floor entrance (Hochparterre). Out of 800 buildings, 163 buildings have some sort of entrance below street level that makes the building particularly vulnerable to flooding. The main construction material of the observed buildings is masonry without any reinforcement, whereas the occupancy is dominated by residential.
The outcomes of the survey were presented to the stakeholders in Cologne during a technical meeting and the benefits and characteristics of the proposed approach have been discussed during a subsequent policy-briefing meeting.
Potential Impact:
The activities in this section are subdivided into the following:
• The dissemination of the project’s goals and results to as wide an audience as possible, although focusing on the DRR community (both research and practitioner).
• How the products developed during the project may be exploited following the completion of the project, in terms of their commercial application, utilization by the DRR community and ongoing development.
• What would be the impact of the SENSUM project on the DRR community (again, in terms of research objectives and practice).
Dissemination and communication
The communication of the goals, results and products of any project, especially one like SENSUM which seeks to directly contribute to DRR activities, is critical to its lasting legacy. In general, the activities undertaken during the project (generally within WP7, see Figure 1.3) are subdivided as follows:
• Maintenance of a project website and geoinformation portal.
• Workshops, symposia, and training courses to present and demonstrate the SENSUM products.
• Academic and technical conferences and meetings.
• Publications in the relevant literature, both professional and academic.
• Policy briefings to civil protection and other authorities.
• Printed material.
• “The SENSUM Game”.
The SENSUM game was outlined in the previous section and so will not be covered here. Likewise, most information about publications and conference presentations is presented in Section A. The following will therefore provide details about the other activities undertaken with the categories listed above, with further details outlined in deliverables D7.3 “Dissemination plan” and D7.4 “Scientific technical communication”. It should, however, be mentioned that although the project has ended, opportunities for the exploitation of its products and the development of new proposal concepts to expand upon already undertaken work are continuously being undertaken.
-Project website
The project website , which is often the first point of contact for many interested parties, was continuously developed during the course of the project. In addition to general information about the project’s goals, activities, and the consortium partners, the various tools, technologies and methodologies developed are presented. The project deliverables are also available from the website, along with printed material, including a brochure outlining the project as a whole, and information sheets about specific products.
-Workshops, symposia and training courses
• An outreach event, “Symposium on earthquake and landslide risk in Central Asia and Caucasus: exploiting remote sensing and geo-spatial information management”, was held in Bishkek, Kyrgyzstan, in January, 2014 (see Figure 3.1). The main goal of this event (see deliverable D7.5 “Outreach Event”) was for the SENSUM consortium members to engage with regional DRR practitioners, decision-makers and other stakeholders in presenting the goals and early results of SENSUM. Of particular importance were efforts to enhance the awareness of the capabilities and opportunities offered by remote sensing for DRR activities. Around 25 participants (in addition to the SENSUM members) attended, representing a range of educational, DRR and development institutions (e.g. Kyrgyz Ministry of Emergency Situations, UNDP, World Bank). From this meeting, a proceedings of 15 research papers was produced, and is available in both hard copy and via the project website (see deliverable D7.6 “Proceedings Volume”).
• During the Understanding Risk Forum , held in London, England, a session with an emphasis on issues relevant to SENSUM activities was organised by members of the SENSUM consortium (UCAM and CAR) entitled “Where Sky meets ground: integration of remote sensing and ground based information for multi-hazard vulnerability monitoring and recovery assessment”. As part of this, a simplified session of the SENSUM “Game” was presented (see D7.4).
• An international summer school “Data Fusion of Risk-related Remotely Sensed and Geospatial Data” (part of a series run by the University of Pavia, Italy) was organised by the SENSUM consortium and held at the University of Pavia. This aim of the school was to provide the attendees (12 graduate students) with basic knowledge on the problems related to mapping the time-dependent hazard and vulnerability of seismic and landslide areas, the presentation of newly developed open-source tools to address the above issues by exploiting Earth Observation (EO) data as developed by SENSUM, and the introduction of the QGIS tool as an effective open-source environment for research and operational use of these products. The event proved to be most successful, with the expectations of the students generally met in terms of organisation of the course, quality of the lectures, interactions with the lecturers, and the opportunities for discussion and the sharing of opinions. Details about this event are presented in deliverable D7.7 “Summer School”.
• CAIAG conducted a training course in November, 2014, on “Monitoring and creation of a landslide database” in Osh, Kyrgyz Republic. This was attended by 25 employees of the Monitoring Department of the Ministry of Emergency Situations of the Kyrgyz Republic. The main purpose of the training course was to provide general information about the causes of landslides and the current state of natural disaster monitoring, including the presentation of ground-based and remote sensing methods for landslide studies. The course consisted of two parts: the first was the theoretical component where the CAIAG staff provided the seminar participants with the general information about landslides and landslide research methods, including database development; the second part of the training was dedicated to training in using tools during field studies of natural disasters (further details may be found in deliverable D7.10 “Policy briefings”).
-Policy briefings
As discussed elsewhere in this report, one of the key objectives of the SENSUM project was to encourage the use of Earth Observing (EO) imagery in DRR activities. As part of this, the project’s members engaged in what we refer to as “policy briefings” (see deliverable D7.10) which involved the presentation of the SENSUM project’s goals and outcomes to higher-level (decision and policy makers) members of the DRR community. Such activities included:
• The production of a so-called SENSUM policy brief document (in English, German, Turkish and Russian) which outlines how the SENSUM consortium sees the role of EO observation in DRR activities.
• On November 6th-7th 2014, the SENSUM project was presented to representatives of the Turkish DRR community at the Izmir University of Economics, Izmir, Turkey. These included members of the Disaster and Emergency Management Presidency of Turkey (AFAD) and representatives from various offices of the Izmir Municipality and academic researchers. After welcoming presentations, which included an overview of AFAD activities, the SENSUM project was presented, covering the project (and remote sensing in general) as a whole and especially the developed tools and products, along with some examples. This was followed by a panel discussion made up of AFAD, municipality, research and SENSUM representatives.
It became clear during the panel discussion that although the intention was to explore what current and future efforts in the area of remote sensing and in situ surveys would be most helpful to DRR actions, the issue of communication between the various end-users, or rather, deficiencies in communication, dominated the forum. The problem of a lack of resources, funding and expertise were also highlighted, specifically the situation (which was discussed) where although the appropriate resources (including personal) may have been available in the past, there is frequently a lack of will or capacity to maintain any infrastructure developed, for example, information collections (e.g. GIS data bases). This point will be returned to in the discussion on impact below.
• Policy briefings were also held in Germany with representatives of the German Civil Protection or Bundesamt für Bevölkerungsschutz und Katastrophenhilfe (BBK) and the Cologne City Council. The meeting with BBK (19th November, 2014) was informal, and in addition to describing the SENSUM project, as well as gaining a better understanding of the role of BBK within the context of German DRR activities, a draft of the policy briefing was discussed. It was found that the concepts put forward by the SENSUM consortium were consistent with those of BBK, hence the document remained largely unchanged following this meeting.
The meeting with the Cologne Municipality (20th November 2014) involved a SENSUM group taking part in the Sitzung des GeoBeirates, a gathering where geo-information practitioners meet to present their ongoing projects and available products. An informal discussion with municipality representatives following this gathering indicated their interest in the project’s outcome, and the consortium is optimistic that future collaborations will arise, especially given that Cologne is frequently a test for European DRR research projects.
• A meeting was organized by the SENSUM partner IGEES with officers and employees of the Committee of Emergency Situations and Civil Defence (CoES) of the Republic of Tajikistan (6th December, 2014). The meeting was organized by Colonel Jamshed Kamolov - Head of the Department of protection of population and territories (CoES). The speakers reported on the main results of the SENSUM project, and highlighted some general issues of seismology and the development of seismology in Tajikistan, as well as the results of the work within the framework of the SENSUM project in the Jirgatol district (northern Tajikistan, bordering with Kyrgyzstan, discussed in the previous section). The meeting was considered a success, with the audience strongly encouraging such work to be continued in other parts of Tajikistan. It was acknowledged that these projects have a positive effect on the reduction of risk arising from hazards like landslides, and contribute to protecting the civilian population and the country’s material resources.
-Printed material
Several printed items were developed during the project and are available from the project website. These include information sheets on some of the technological developments (e.g. the GFZ-REM or Rapid Environmental Mapping for geo-risk modelling system), a high-quality project brochure which is available in English, Russian, Turkish and German, and the SENSUM policy statements. Furthermore, most project deliverables are also available on the website.
Exploitation
The exploitation of the SENSUM results and products may be considered as including the ongoing development of the outcomes (recognising that advances in technology etc. are never ending), generally within other research projects, their consideration as part of other projects, although their further development may not be an end in itself, and in their use by DRR practitioners and other assorted end-users, which includes their commercial usage. When considering this, it was soon understood during the course of the project that the value of the SENSUM products lies in their ability to provide timely (rapid), relevant (focused) and cost-effective information on exposure and vulnerability indicators and post-event recovery progress within a multi-hazard context. The fact that the tools are free and open-source, and rely largely on free and globally available data sources, will allow them to, at the very least, be considered for use by a range of end users.-However while the value of remote sensing and in situ data collection technologies is recognized, end-users often lack the expertise necessary to strategically place the developed technologies within their in-house mechanisms.
The technologies developed within SENSUM that are considered exploitable may be divided into those associated with the EO tools, prioritization tools for sampling and the rapid environmental mapping methodologies, and user elicitation tools, such as “the game”. This therefore leads to an exploitation framework divided between services and training, which would also encompass capacity building activities. A number of exploitation activities are in fact already underway covering actions ranging from other EC research projects (e.g. RASOR - Rapid Analysis and Spatialization of Risk), programs funded by bodies such as the World Bank (risk assessment, capacity building), and data collection activities for various end-users (e.g. IGDAŞ, Istanbul Gas Distribution Industry and Trade Incorporated Company ), the latter being the sorts of activities the individual partners may engage in as part of short-term income-generating actions.
However, the primary vision of the consortium is focused on the capacity building of policy-makers, experts and stakeholders who work in risk reduction and management, and emergency response-related research and activities. Therefore, training activities/courses are seen as both a crucial, and lucrative, avenue to pursue. Potential customers and clients include intra-governmental agencies (e.g. Red Cross, World Bank), government ministries involved in DRR, development agencies (e.g. USAID) and other actors concerned with the exploitation of remote sensing for DRR actions (e.g. International Charter Space and Major Disaster). Such activities would include the SENSUM “Game” (especially useful for governmental agencies) to elicit information, and providing courses on DRR relevant to the SENUSM products, such as the use of remote sensing, integrating earth observations and in situ data, exposure and vulnerability monitoring, as well as more fundamental topics, such as general hazard and risk assessment. As an example of the sort of income such courses may generate, a 5 day course recently provided to the World Bank by CAR (Cambridge Architectural Research , subcontracted by a consortium partner) generated US$200,000 (it involved 20 days course reparation, 15 person/days delivering the course, and 15 days preparing a distance learning package. i.e. $4,000/day).
Impact
One impact from SENSUM is how the project partners will interact with potential end users when developing future projects. It was found from the various discussions, especially those during the policy briefings, that:
• The major concern of end-users was the lack of resources (e.g. geospatial expertise, IT infrastructure, ability to cover data costs, scientific collaboration links, etc.);
• Communication within and between the various end user organisations is a major issue, an outcome already identified in previous projects (e.g. the MATRIX project, Komendantova et al., 2014);
• The user community still struggles with a clear definition of the issues related to the access and cost of EO derived geo-information; in this context, a clear understanding on the strengths and weakness of the EO derived product portfolio, in particular the added value of the combined use of EO and in situ sensing is on demand (and has been partly tackled by validation efforts within SENSUM).
• There are still difficulties in conveying scientific information to DRR practitioners.
List of Websites:
http://www.sensum-project.eu/de(öffnet in neuem Fenster)
Massimiliano Pittore (GFZ) pittore@gfz-potsdam.de Project Coordinator
Kevin Fleming (GFZ) kevin@gfz-potsdam.de Project Manager
The Framework to integrate Space-based and in situ sENSing for dynamic vUlnerability and recovery Monitoring or SENSUM project is a multi-disciplinary project that set out to develop innovative methodologies for the exploitation of Earth Observing (EO) remote sensing and ground-based methods to better understand those factors associated with the exposure and vulnerability of the built environment to, in particular, seismic and landslide hazard.
The aims of the project required the identification of various indicators relevant to earthquake and landslide risk that could be identified via EO products, as well as those that would need in-situ observations. Such indicators were related not only to natural parameters such as precipitation or slope, but especially to the vulnerability and exposure of the human environment. A related aspect was the development of tools for change detection, which not only would be exploited for gauging changes in the built environment over time (i.e. a result of increasing urbanisation for example), but also for monitoring recovery operations during the post-event period (e.g. extent and changes in tent camps, road accessibility, etc.).
In-situ observation and analysis schemes were built upon and expanded within SENSUM. The most promising for these is the GFZ-Mobile Mapping (GFZ-MOMA) system, involving an omni-directional camera and associated interpretation software, to implement Remote Rapid Visual Screening (RRVS) activities. This system in turn exploits the sampling framework developed within the project and based on the so-called focus maps, a means by which an optimised path for gathering observations is developed based on various weighted indicators (possibly derived from remote sensing data). The result is a sampling plan where the path considers those areas with a higher statistical need/value to acquire data/information. Furthermore, information obtained from foot-surveys making use of the GEM (Global Earthquake Model) building taxonomy can be integrated, leading to a more accurate picture of an urban areas vulnerability and, subsequently, risk.
A critical (and successful) element of SENSUM involved the very intense end-user interactions in the form of not only meetings and symposia, but also workshops that centred around a scenario training exercise referred to within the project as “the Game”. The Game takes a group of end-users and, after dividing them into relevant groups (decision makers,, information providers, etc.) presents a time compressed scenario and, from the interactions, allows shortcomings to be identified on a number of levels (knowledge, awareness of available products, institutional and cultural limitations, etc.).
Project Context and Objectives:
A critical requirement for the effective mitigation of, and response to, disasters of any type, natural or man-made, is the availability of appropriate information. Such information must be up-to-date, sufficiently resolved (in time and space) and accurate, while being able to be readily understood and rapidly applied within robustly defined uncertainties. It must cover the entire disaster risk cycle (see Figure 1.1) which we consider involving mitigation (pre-disaster), response (post-disaster) and recovery, and be effectively exploited by Disaster Risk Reduction (DDR) practitioners, land-use planners, decision-makers, infrastructure operators, and the general public. This information must also be able to take into consideration the dynamic nature of the risk to human society, especially its greater exposure to natural hazards due to increasing populations (especially growing urbanisation), the greater dependency on complex technological infrastructure and supply chains, and climate and environmental change.
Adequate information is essential to assessing the three components of risk associated with natural disasters: the actual natural hazards that afflict an area, including their potential spatial, temporal, and causal relationships; the exposure of an area’s population to these hazards; and the vulnerability of the area’s built environment to these events. The latter two are in turn related via various social and economic factors, both temporal and spatial, while in some cases they may also affect the hazards themselves (e.g. growing populations require more space for living and agriculture, hence leading to the potential for increased environmental degradation).
However, relevant information is often missing or inadequate, especially in economically less developed regions such as Central Asia, which are also experiencing rapid spatio-temporal changes in their urban environments. Such a deficiency is also often the case for more developed countries such as Germany, where the required information may not be available or is so only for limited areas, or not readily accessible owing, for example, to it being dispersed amongst different municipalities or government departments.
One contribution to solving this problem is the better recognition of the potential of Earth Observation (EO) remote sensing imagery. EO products’ greatest value lies in their almost-global coverage, coupled with their capacity to introduce a time-dependent element to any resulting datasets and models. While the usefulness of such information is recognised in many quarters, there is still a concern about the level of expertise and resources that are required to adequately exploit such information (in terms of both the necessary tools, the understanding of the imagery itself and the associated efforts and expense), and what can realistically be learned from such products. These questions then call upon the need to integrate such observations with imagery (and other information) that is gained in situ and which can identify features not possible from space- or airborne sensors, and which can also consider the temporal change aspects of the problem.
The lack of information about urban vulnerability and exposure, and subsequently the investigation of the potential added value by the combined use of EO and in situ observations, are thus the fundamental problems confronted by the Framework to integrate Space-based and in situ sENSing for dynamic vUlnerability and recovery Monitoring or SENSUM project.
SENSUM, whose consortium is made up of 8 European and Central Asia partners representing education, research and commercial institutions (see Table 1.1) is a multi-disciplinary project setting out to develop innovative methods to exploit observations and data from EO remote sensing satellites (the focus in this project being exclusively on optical sensors) and ground-based methods (e.g. omni-directional camera surveys) to better understand the factors associated with exposure and vulnerability. Figure 1.2 outlines the operational scheme that is envisaged to arise from the products of SENSUM. Such a strategy must consider the specific protocols adopted by each country or region for information collection, depending upon their level of technical expertise, available resources, hazard nature, distribution of the exposed assets, and DRR institutional culture. The importance of the latter point cannot be underestimated, and considerable effort was expended in understanding this. In addition, as seen in Figure 1.2 there is the need to integrate EO images with ground-based data, allowing a higher spatial resolution, respecting also the temporal component of disaster management cycle. This in turn is exploited for the development of more effective sampling frameworks, leading to the production of products (e.g. maps) that are of benefit to end-users, all the while keeping track of how the various datasets change with an information management system.
Figure 1.3 presents the work package (WP) structure of the project, which also best describes the associated work plan. The work proceeds (usually in parallel) from WP2 which deals with the collection, processing and integration of remote sensing and in situ data for dynamic indicator analysis, to WP3, which considers data collection from the perspective of statistical sampling, and integrates it over different spatial scales, leading to what are referred to as “focus maps”, meaning maps that identify areas where statistically there is a greater need/value in acquiring data/information. The identification of the different indicators that are useful for hazard assessment (which in SENSUM are focused on landslides and earthquakes) and dynamic vulnerability assessment are the themes for WP4, while the indicators relevant to dynamic recovery monitoring are covered by WP5. Naturally, the uncertainties associated with any results are critical, leading to the systematic accuracy assessments in WP6, which also targets the compilation of a catalogue of relevant, available remote sensing sources. Knowledge about the project, its results and developed products, has been disseminated throughout the DRR community (both practitioner and research) via WP7, while the project is managed through WP1 and WP8. A critical aspect of the dissemination involves so–called policy statements, which are presentations and documents for DRR decision and policy makers.
The primary goals of the project may therefore be outlined as follows:
• Provide analysis and mapping tools based on space- and ground-based data acquisition (primarily imagery) for continuous exposure and vulnerability assessment and analysis, and for post-event recovery monitoring;
• Provide user-oriented guidelines to the DRR community which clearly defines EO and in situ-based exposure products in terms of their capabilities and limitations; in this context, specifically, the demonstration of the integrative use of EO and in situ sensing is one of the focal points of exploration;
• Apply the tools developed by SENSUM to the specific hazards of earthquakes and landslides, including their temporal and spatial relationships, to test cases (see below) that represent examples involving varying amounts of available information;
• Explore and encourage the application and integration of EO remote sensing and ground-based methods to multi-type risk analysis and disaster response and recovery, including the consideration of the time-dependent component. Such efforts require the reassurance of DRR practitioners that such actions will not require excessive levels of expertise or resources, hence the commitment of the SENSUM project of the use of Free and Open Software Source (FOSS) tools, datasets and archives;
• Enhance the scientific and technical capacities of the Central Asian partner countries for DRR. This includes science and engineering students and researchers, DDR practitioners (decision and policy makers and responders), land-use planners and infrastructure operators.
As the methodologies and products developed within SENSUM need to be tested and validated, and to ensure their relevance to as broad a range of environments as possible, the SENSUM project included several test cases, which may be classified as follows:
• “Data poor” test case, here the Isfara-Batken region between the Kyrgyz and Tajik Republics (subject to earthquakes and landslides). This area has very little information available about any aspect of risk, but especially exposure and vulnerability. Its relevance is also connected to the cross-border character of the area, a critical issue when considering more effective DRR actions.
• The “data rich” area of Cologne, Germany (subject to seismic events, flooding and windstorms). Cologne has a large amount of information dealing with all aspects of risk, however, such information may also be out of date, or inadequate to thoroughly assess the city’s risk. There is also the prospect of difficulties of actually obtaining or even knowing what information is available, given institutional limitations.
• The “data intermediate” or “data critical” area of Izmir, Turkey (earthquakes). By this term, we mean that while there is considerable information available on all aspects of risk, however, it is still inadequate given the fairly high level of hazard and exposure of this location.
The resulting SENSUM products therefore contribute to the cost-efficient, multi-scale, and multi-temporal gathering of information. They acknowledge the need for tools that do not require a too-high a level of technical expertise, while as mentioned above, the primary consideration will be towards software tools and data that are available within a FOSS context. In terms of GIS tools, this means the use of such applications as QGIS , a free and open source GIS tool which is able to be used on a range of computer platforms and is under a continuous state of development. In terms of EO datasets, there is Landsat archive which provides continuous and near global coverage extending over 40 years from a series of 7 satellites and is free and open for download. In addition, there are the current and future Sentinel EO spacecraft of the European Space Agency (ESA) which will also be providing their observations (optical and radar) within the FOSS framework.
Project Results:
The following outlines the main outcomes of the SENSUM project. It needs to be remembered that while some tools developed within the project may already be exploited, many are still in need for further development. Likewise, some of the less “tangible” outcomes, such as our discussions with DRR practitioners, require a much longer timeframe before their true fruition can be recognised. It will also be emphasised how the different outcomes are in fact closely related to each other, as outlined in Figure 2.1.
Summary of currently available and future remote sensing products
When undertaking an integrative and comprehensive risk analysis, DRR practitioners, urban planners, decision makers and infrastructure operators commonly encounter the problem of appropriate data collection. Shortcomings in available data include when data sources are too generalized, outdated, inconsistent or simply non-existent, especially for economically less developed parts of the world. Throughout the disaster risk community, remote sensing is today perceived as a promising tool for economical, up-to-date and independent data collection and has been employed in various investigations within the geo-risk context (see, for example, the International Charter ). While the use of EO data for the study of earthquake and landslide hazard spans a wide array of applications - e.g. ranging from site characterization and terrain analysis to fault mapping, monitoring the dynamics of active fault systems/volumetric mass movements and the detection of hazard-induced land surface changes - remote sensing for vulnerability-centred investigations is a less well-established field of research. In this context, SENSUM primarily aimed at the derivation of pre-event vulnerability indicators related to the physical and structural vulnerability of human assets and the built-environment at various spatial scales. A thorough review of current EO datasets and products, future missions, their technical specifications as well as their (potential) applications within the DRR context served as a starting point for SENSUM product development, encompassing the versatility of remote sensing data and techniques with regards to the analysis of earthquake and landslide risks. The resulting report (deliverable D2.1 “Present-day and future remote sensing data”) provided a non-exhaustive, but comprehensive, overview of globally available present-day and future remote sensing data and products.
For their use in disaster risk research and management, the technical specifications of remote sensing data, such as their geometric and temporal resolution, their scene size and thus, the spatial scale of analysis they can be applied to, as well as limiting factors including data cost and coverage, determine the potential for operational applications. Data from past- and present-day sensors featuring coarse geometric resolutions and large-scale aerial coverage generally enable the overall evaluation of pre- and post-event disaster situations at a low cost. On the contrary, data availability from Very High Resolution (VHR) systems, such as airborne LIDAR or VHR optical data over larger areas for in-depth analysis, is still limited due to their high costs.
Upcoming EO missions have the potential to play a key role in future research by continuing and improving the observing capabilities of current missions for earthquake- and landslide-related investigations, as well as operational emergency response service delivery in the near future. As a prime example, the ESA Sentinel missions will feature enhanced geometric and thematic observing capabilities and increased revisit capabilities at no cost to users, and will further ensure data continuity of long-lasting missions as several past- and present-day missions, such as Landsat and SPOT, reach the end of their technical lifetime. In this context, the enhanced capabilities of the described missions are believed to enable a leap forward in earthquake and landslide risk research, both for hazard- as well as vulnerability-related investigations. A selection of promising launched, planned or proposed missions is presented in Figure 2.2 (see deliverable D2.1) with a focus on their technical lifetimes. The respective project report further holds detailed information on technical sensor specifications such as geometric resolution, swath width, revisit capabilities and potential fields of applications. A stringent categorization regarding the spatial scale of analysis is applied based on the aerial coverage, i.e. the swath-width-dependent scene size, and the geometric resolution of the sensors according to the harmonized classification scheme presented by the European space programme Global Monitoring for Environment and Security (Copernicus/GMES ), envisaging the possible future enlarging of the respective product portfolio beyond European countries.
Apart from the perspective on current and future mission, international initiatives from both have produced remote sensing based geoinformation products over the past years that provide valuable input to the vulnerability-related research of SENSUM. On the one hand, space-based pre-operational emergency response services have produced a large product portfolio and significant experience in post-event mapping applications in recent years. On the other, both large-scale global and regional land cover inventories have been produced as first-level approximations of human exposure and second-level descriptors of the built environment. These information layers are essential as they present almost the only data basis for risk-related spatial investigations in data-poor regions of the world, such as Central Asia. Nevertheless, a better understanding of each dataset’s strengths and weaknesses is still required – a challenge encountered through a systematic cross-validation of multi-scale space- and in situ sensed exposure products within SENSUM (see Section 2.4).
Overall, from a technical perspective, the constantly increasing availability and accessibility of remote sensing technologies and the enhanced technical capabilities of future missions will provide new opportunities and data continuity for a wide range of vulnerability-related investigations such as SENSUM and similar future projects.
End-user elicitation
One of the aims of SENSUM was to investigate decision-makers information needs and the potential demand for SENSUM products. This was done using scenario planning with senior disaster managers in Kyrgyzstan, Tajikistan and Turkey, reported in the SENSUM deliverables D4.1 “End-users assessment” and D5.1 “Comparison of outcomes with end-users needs”. These realistic earthquake scenarios were played-out during one-day exercises that allowed disaster personnel to simulate the post-earthquake decision-making process. The findings pinpoint how information derived from multi-resolution imagery can be used in planning and assessing recovery of transportation networks, transitional shelters and the built environment.
The scenario planning game developed for SENSUM involved getting people to dynamically ‘play’ through an imagined future. The ‘Game’ collapsed real time to highlight significant issues and to focus on strategic decisions. Scenario planning was developed by Herman Kahn at the Rand Corporation in the early 1950’s (e.g. Fahey and Randall, 1997) and used extensively in military planning. In the disaster management field, the approach has been applied in ‘shakeout’ type preparedness exercises and drills aimed at raising public awareness.
-Running the Game in Bishkek and Izmir
The scenario planning game was played in Bishkek, Kyrgyzstan and Izmir, Turkey. In Bishkek, the cross-border exercise was attended by 20 participants, including 7 people from the Ministry of Emergency Situations in Kyrgyzstan and 3 from a similar Ministry in Tajikistan, 2 from the Red Crescent, and 1 from the United Nations Development Programme (UNDP) in Bishkek. In Turkey there was a larger number of attendees (27), from AFAD Izmir and Bornova Municipality and from the central government AFAD HQ, as well as from the Turkish Red Crescent and Turkish Amateur Radio Society. The players were divided into three ‘teams’ – Events, Decisions and Information. The transcripts of the Game clearly revealed the interplay between events, decisions and information in a way that other types of enquiry, for example, user-needs surveys or interviews, would have failed to show. During the game, the discussions at each table were highly informative for the SENSUM team.
The choices of SENSUM products focused on damage assessment and early recovery. The SENSUM researchers believe that there is considerable potential for using remote sensing products beyond that communicated through these choices. However, this positive finding needs to be qualified in the light of analysing the game transcripts using Kahneman’s model of Fast and Slow thinking (Kahneman, 2011). One of the main insights was that the decision-making team in Bishkek, and to a lesser extent in Izmir, had difficulty using the information SENSUM could provide. At the time, we thought that the main reason was the lack of experience in using remotely sensed GIS data, time pressure and the inability of the information group to provide answers quickly enough. In Izmir, even when information began to flow quickly enough to influence decisions, the decision-making team seemed to use it to confirm what they already knew or had already decided, rather than inform new thinking.
On reflection, the answers seem obvious. It was unreasonable of us in hindsight to have expected these disaster managers to be able to develop strategic plans. They have been selected because they can react effectively using Kahneman’s “Fast” thinking, making decisions based on experience and established heuristics, with little recourse to information or analysis. Neither group were able or willing to think beyond Year 1, with long-term recovery planning needing a different way of thinking than relief efforts.
The obvious answer is for the authorities to set up two separate teams working independently, but with some degree of coordination. One would be the standing civil defence group of disaster managers – the people we had been working with in Bishkek and Izmir. The second team would include economists and planners responsible for developing recovery and development plans. The challenge this poses to information providers (the sorts of people SENSUM’s products are aimed at) is that two entirely different types of information are needed. To have any influence on crisis decisions, the information would have to be provided quickly in real time. Meanwhile, longer-term recovery planning by contrast requires detailed and accurate GIS-type information that can inform plan making. This is expensive and needs time, resources and highly skilled personnel that may not be available in many places prone to disasters, especially Central Asia.
-Outcome of the game exercises
The scenario planning game worked remarkably well in the two case studies, and the participants found it realistic and absorbing. They were able to imagine how the scenario might play out and use their imagination and experience to generate the detailed content. The Game had two main aims: a) to gain a deeper understanding on the real data needs of the end-users and b) to define a list of vulnerability and recovery indicators that would guide the development of appropriate SENSUM products. Both of these were achieved.
The Game also investigated what is required for remote sensing and GIS to become an integral part of the disaster risk management process. During the Game, users in both countries had difficulty in transferring information needs into requests for data products. Kahneman’s thesis of Thinking Fast Thinking Slow, suggests that the problem is more fundamental. The need to respond quickly means that disaster managers use the so-called System 1 thinking (based on intuition, experience and rules of thumb) almost exclusively and have difficulty using the kind of information SENSUM is able to provide currently.
Thinking Fast Thinking Slow also suggests that governments need to authorise two teams to respond to disasters. One, the usual civil defence team managing relief and immediate recovery who are good at System 1 thinking and the other who are good at both System 1 and System 2 (meaning evidence based) thinking, planning long-term physical, social and economic recovery (see Platt et al., 2014a; Platt et al., 2014b; Platt, 2015). Following such a scheme would therefore lead to a more comprehensive (and efficient) response to natural disasters and the enhancements of a society’s resilience.
Framework for data collection
The framework for data collection (see Figure 1.2 to see its relevance within the entire SENSUM scheme and deliverable D3.5 “Sampling framework” amongst others) may be divided into:
• What indicators are actually being sort?
• What changes have occurred and how these may be detected?
• What methodologies for in situ data collection are being employed (for example, omni-directional cameras, foot surveys).
• How does one actually design the sampling strategy?
-Identification and extraction of appropriate indicators from remote sensing
Physical vulnerability can be measured through a wide spectrum of indicators that have been presented in the literature (see Table 2.1 List of vulnerability indicators as defined after the literature review process made by EUCENTRE and refined after discussion within the consortium.). During the SENSUM project, two series of proxies related to earthquake and landslides were identified (see Figure 2.3 for the identification and selection process and deliverable D2.2 “Vulnerability indicators”) as being obtainable from remote sensing. The so-called Earth Observation Tools developed within SENSUM were also designed to provide a set of useful and easy-to-use algorithms both for GIS users and software tool developers . The former were released as a plugin for Quantum GIS while the latter are available as source codes, open to modifications and the implementation of new workflows.
Earthquakes
The selection of the adopted physical vulnerability indicators was made considering two main elements: the first reflects the differences in the seismic performance of buildings and infrastructure, while the second considers other affecting indicators that enhance or decrease the physical direct losses. According to the definition of a community’s vulnerability, the main essential values are people and their livelihood (UNISDR, 2004): buildings combine both of these key values and therefore serve as a prevalent monitoring representative.
A literature review of more than 70 papers was performed to try to match and intersect the requirements of a pure-risk-oriented approach with restrictions defined by remote and in situ sensing. The resulting list is mostly earthquake-oriented; however, some of the suggested proxies were also suitable for presenting the vulnerability of other hazards (e.g. landslides) after a discussion within the consortium.
The final proposed and agreed-upon list is summarized in Table 2.1 where the indicators are differentiated according to the type of data (defined mainly by resolution) used for their extraction. For example, the extent and age of a built-up area can be retrieved from medium-resolution Landsat-like imagery, while building footprints are derived from higher-resolution datasets. Unfortunately, a few proxies are strictly bounded to the availability of ancillary information, as explicitly reported in the table. The last column is dedicated to the method used for their extraction, varying from newly developed to available datasets like OpenStreet Map.
Landslides
Landslides represent a major threat to human life, property and constructed facilities, infrastructure and the natural environment in most mountainous and hilly regions of the world. In landslide risk assessment, data availability is a major constraint, which greatly affects the type of methods and models that can be developed. Remote sensing is a promising tool for economical and up-to-date data collection, which also could be applied to monitoring the dynamic development of risk. The spatial and temporal distribution of the risk for landslides can be assessed by monitoring hazard indicators (e.g. slope characteristics), exposure indicators (e.g. number of houses or total population) and vulnerability indicators (e.g. population density, settlement structures or indicators related to structural vulnerability).
Indicators for application in landslide risk assessment as a function of space and time should be selected based on:
• their importance to the assessment of landslide risk,
• the ability to apply remote sensing techniques in the data collection,
• their relevancy to end-users’ needs.
The selection of indicators relevant for landslide risk was based on a literature review and end-user assessment, as described in Section 2.2. For landslide hazard, indicators derivable using remote sensing techniques are listed in Table 2.2. In addition, there are important landslide hazard indicators that cannot (or only with difficulty) be directly derivable from remote sensing, such as soil types, soil depth, soil moisture and geotechnical properties.
The landslide hazard indicators could be subdivided into triggering factors (e.g. extreme precipitation) and susceptibility factors (e.g. slope). The data for the susceptibility factors were reclassified into 5 classes of susceptibility, representing very low, low, medium, high and very high susceptibility to landslides, respectively. For example, Figure 2.4 illustrates the susceptibility factor "slope angle" according to the above mentioned classification, i.e. division into 5 classes (Figure 2.4a). Similarly, Figure 2.4b. shows the indicators indicator soil moisture. This was evaluated by overlaying maps of ground permeability and mean precipitation, i.e. the mean 30-day precipitation as derived from the dataset provided by the JAXA/NASA Tropical Rainfall Measuring Mission (TRMM ).
For structural vulnerability to landslides, indicators derivable with remote sensing techniques are in Table 2.3. In addition, important vulnerability indicators that are not directly derivable from remote sensing include size and conditions of openings towards the slope. Exposure and non-structural vulnerability indicators include, e.g. built-up area, built-up density, population density, accessibility and distances and land use.
For the application of these indicators within the development of focus maps (see Section 2.3.4) it was decided to focus on landslide hazard and exposure for the following reasons:
• The end-user assessment showed a stronger need for products related to landslide hazard than for products related to landslide vulnerability.
• The capabilities of remote sensing are greater for landslide hazard assessment than for landslide vulnerability assessment, i.e. more relevant data for hazard mapping is able to be gained by remote sensing.
-Change detection
Change detection is by definition the capability to detect and highlight changes occurring in space and time. Within the project’s framework, change detection products, which focus on exposure, can be used as an input indicator for other tools produced within the SENSUM consortium. The EUCENTRE team proposed a post-classification comparison algorithm based on the capability of the other algorithms to automatically process a stack of Landsat data. In particular, the tool is capable of comparing results obtained by the object-based built-up area analysis from different years. Results are also improved thanks to the expected continuity over years. A schematic view of the post-classification process is represented in Figure 2.5. The results of the built-up extraction using the object-based methods are combined together using logic operations and the capability to track changes in time due to the repeatability of the methodology.
For the areas of interest for SENSUM in terms of vulnerability and exposure mapping, medium resolution images such as those from Landsat would provide temporal information and focus area mapping. For the other intended application, that is for post-disaster recovery monitoring, users would also use temporal analyses to identify focus areas followed by high-resolution data techniques to track specific changes for recovery and building vulnerability analyses. Based on the changes within the indicator-specific GIS objects in a high resolution image, indices are calculated and plotted over time. The SENSUM method detects change, which is identified as the absence of similarity between gradient, roughness, texture and edges. For the purposes of recovery evaluation and monitoring, the method can be used to detect changes to buildings over time and therefore track the progress of initial damage, clearing, reconstruction and recovery due to an earthquake. An example of this is shown in Figure 2.6 which shows a sequence of images demonstrating the use of the SENSUM building change detection index. The pre- and post-disaster images are clipped by a GIS objects tool and the change index provides the degree of change to the selected buildings over time, indicating possible signs of initial damage and subsequent recovery.
-In situ data collection
For rapid in-situ image data capturing, a mobile mapping system (GFZ-MOMA) has been developed and extensively tested during the project. The system is composed of a Ladybug3 omnidirectional camera from Point Grey Research Ltd., a data capturing and storage unit, a navigation unit and an external battery pack that supplies the energy for up to 6 hours of autonomous operation (Figure 2.7). This camera is made up of 6 colour Complementary Metal Oxide Semiconductor (CMOS) sensors that capture concurrent image sequences up to 15fps acquisition rate. The 6 image streams are synchronized and automatically stitched into omnidirectional (panoramic) high resolution (5700x2700px) format with JPEG compression. The camera system is operated from inside the car and is mounted on the vehicle's roof with a simple mounting system (see deliverable D2.4 “Capture of in-situ image data”).
The data capturing and storage unit has been developed with the specific focus on ease of use and ruggedness for robust outdoor applications, even under rough conditions (e.g. unpaved roads, dust). A custom software application captures, synchronizes and saves the different data streams coming from the camera, the GPS and the Inertial Measurement Unit (IMU). Location is associated with each omnidrectional image by b-spline interpolation of the GPS positioning.
The navigation unit uses QGIS as the main software component for location tracking and car navigation. Its map interface is able to combine various background maps of the study area and to display pre-calculated sample areas and routes. The position can be tracked and displayed in real time with the GPS live-tracking functionality. This allows an operator to navigate the car along pre-calculated routes, while also rescheduling the path on-the-fly to cope with unexpected environmental conditions (e. g., traffic jams or road blockages).
An analysis of the captured omnidirectional image sequences is performed through visual image interpretation. Because of its similarities to commonly used screening procedures, the difference being that it is performed remotely, this novel technique is referred to as Remote Rapid Visual Screening (RRVS, see deliverables D2.6 and D3.5). A specific RRVS system has been implemented to allow for user-friendly, rapid and standardized image analysis. The system is composed of a map-interface, an omnidirectional image viewer and a customizable data entry form that currently supports the per-building indicators that were identified as part of the SENSUM project for vulnerability assessment and allows the display of building locations along with the GPS locations of the images captured during the field survey. Taxonomies are defined in order to describe the indicators and their values in a standardized and comparable way. Once a building of interest has been located in the map-interface and the nearest captured omnidirectional image has been identified and displayed in the image viewer, the building can be screened remotely by structural engineers. Uncertainties related to the assignment of attribute values can be quantified by the operator through additional qualifier attributes. In the current implementation of the RRVS tool, the degree of belief is supported as a qualitative measure of uncertainty for each attribute.
In-situ surveys using the above mentioned systems have successfully been carried out as part of the SENSUM case studies in Cologne (Germany) and in several towns and cities in the Batken district of Kyrgyzstan. A survey in Izmir (Turkey) could not be conducted due to permission problems. However, as part of a scientific service by GFZ that evolved out of the SENSUM project, an additional survey has been performed in Istanbul (Turkey). This survey was requested by the Turkish Catastrophe Insurance Pool (TCIP ) and aimed at testing the applicability of the proposed mobile mapping system and sampling framework for operational earthquake insurance claim management.
-Sampling design
As shown in the preceding section, a simple, customized mobile mapping system can efficiently and rapidly collect large amounts of high-quality, georeferenced visual data about building stock, infrastructure (e.g. roads) and in general the features contained in the visual surrounding which relate to the assessment of expected consequences in case of an earthquake. Such a system can be easily deployed wherever it is needed, does not require particular skills to be operated, and is economically viable. The data collected through the mobile mapping system can be easily distributed and analyzed remotely, and the intensity of the analysis (aka the number of remote operators allocated to the analysis) can be easily scaled up or down according to the contingent necessities and on the business case. For instance, additional skilled operators could be seamlessly integrated into the analysis protocol in case a bigger effort is requested on a temporary basis.
By applying a consistent prioritization and optimization approach, a strong increase in the efficiency of the survey can be obtained, especially when small sample sizes are required in order to decrease the impact of the survey on the available resources. In order to realize such prioritization, the previously mentioned focus maps concept has been proposed. This implements a customized, unequal inclusion probability scheme which can be used for employing PPS (Probability-Proportional-to-Size) or STR (Stratified) sampling designs with an accompanying increase in the overall statistical efficiency.
Automatic routing optimization proved to be a simple and scalable solution to optimizing the actual data collection in urban environments. The set of sampling locations obtained by means of the computed focus map can be converted into an actual path to be followed by a mobile mapping system in order to implement the survey. The proposed optimization algorithms are FOSS-based, and can exploit official or authoritative information on the existing road networks, as well as freely available datasets from OpenStreetMap (OSM). The implementation of the routing engine server-side moreover allows for the implementation of compact, flexible and reliable IT frameworks for data collection, integration and access.
In order to maximize the reliability of the system, auxiliary information might be collected and kept up-to-date in order to provide a consistent reference for prioritization and optimization of the survey activities (see Figure 2.8). The collected data would moreover have a significant value, since a better description of the exposure and vulnerability of the building stock could be achieved by, for example, analyzing the collected data and integrating them into a flexible geographical database. Pre-event and post-event surveys can be independently optimized by integrating multiple data (possibly including statistical inference), thus resulting in a consistent improvement in the operational capabilities in the field of data collection and related analytics.
Combining focus maps, dynamic sampling frameworks and mobile mapping for the data collection phase, a new generation of iterative and incremental data collection and integration approaches can be realized (see Figure 2.9). This would provide end users interested in vulnerability monitoring and risk assessment with a powerful and scalable tool to increase the reliability and timeliness of the resulting model, and to undertake more informed mitigation and prevention activities.
-Information life-cycle management
With an increase in data volume and complexity from a large variety of sources – reaching from satellite remote sensing and crowd-sourcing to cadastral and in-situ surveys – storage and management of geospatial information becomes an important topic, and one that was addressed in the SENSUM project (see deliverable D3.7 “Information integration”). Especially when considering the management of the information life-cycle (“from the cradle to the grave”), the integration of heterogeneous information and the distributed versioning and release of geospatial information are important topics. In the SENSUM project, a conceptual and technological framework was developed to tackle major questions regarding the information life-cycle management, the non-distributed and distributed versioning of geospatial information and its release options. A life-cycle management solution was provided based on a relational spatio-temporal database model that is coupled with Git and GeoGig repositories and workflows.
Figure 2.10 shows a schema of the information life-cycle as followed within the SENSUM project and implemented at the database level in PostgreSQL . New information enters the database workflow in the creation phase. Along with the actual object to be added to the database, several metadata attributes need to be defined. Along these are the source of the information item and its value (e.g. accuracy). The use of data entry forms for user -driven information creation can enforce these additional metadata attributes. Once new information is created, it is integrated with already existing information. In case of geospatial information, it is checked if other information items already exist in the database at the same location of the newly created item. In case of an overlap between new and old items, their values are compared and the information item with the lower value is replaced with the higher value item. The overall values of the database are than updated accordingly and the information is ready for usage for further analysis and interpretation by the user. Several customized spatio-temporal query functions and views are implemented in order to improve information accessibility during the usage phase of the life-cycle model. The database model supports transaction logging and therefore archiving of records. Once information is modified or disposed, its previous state can be archived and the database transaction can be logged. In case the modification refers to an updating by information with higher values (meaning better information became available), the archiving mechanism allows to retrieve or recover previous database states at any given transaction time. In case of the modification or disposal referring to real-world changes, this mechanism allows the archiving of the full history of an object from its creation (e.g. construction of a building), through modifications (e.g. retrofitting of a building) until its disposal (e.g. destruction of a building).
The content of the database can be released and versioned at any time in order to distribute and share it in a decentralized way. Information to be released can be the whole database or subsets of it in the form of summary views. Once information is released it enters a new workflow. For releases and distributed versioning we support the use of Git and GeoGig. A possible release workflow is schematized in the bottom part of Figure 2.10.
Framework for data validation
Despite the unclear definition of the term exposure, i.e. objects such as people, properties, infrastructure or economic units adversely affected by a potential hazard, it is widely recognized that the understanding and spatial quantification of exposed elements is of crucial importance for a comprehensive risk assessment. Due to the large-scale extent of human activities on our planet, remote sensing has been widely recognized in today’s scientific community as a valid tool to capture the elements at risk objectively and across large spatial extents. It is applicable on various spatial scales ranging from global to local to capture physical elements of the built environment of human settlements, as well as quantified indicators relating to these exposures. On a local scale, EO-based geo-information can be further refined and added by techniques of in situ sensing for an even more detailed characterization of the structural vulnerability of the built environment. In this context, SENSUM provides a set of user-friendly open source software tools to recall established EO information extraction methods for exposure mapping from space as well as workflows to derive important vulnerability information of human settlements on a per-building level using techniques of in situ imaging and ground-based sampling to add to the existing available exposure product portfolio. Together with existing global to regional EO-derived exposure layers, the proposed methods can contribute to a non-exhaustive, but comprehensive description of the vulnerability of human assets on our planet.
Nevertheless, classifications based on remotely or in situ sensed datasets are basically abstract approximations of the true physical environment of the Earth’s surface being monitored. Thus, today’s scientific and – above all – end-user community still struggles with a clear understanding and definition of the quality and accuracy of the manifold available spatial exposure products. In this regard, a systematic cross-validation framework for both SENSUM product developments (remote sensing toolset and in situ imaging workflows) and existing global and regional exposure products was produced. The respective report, deliverable D6.2 “Accuracy Assessment and Cross-Validation”, establishes clearly the value of remote and in situ sensed geo-information within the context of SENSUM. Significant criteria for data selection from a user perspective, such as accuracy, thematic detail, geometric resolution, cost and added value, are reported as product definition tables, enabling a systematic, comprehensive and user-oriented guideline for application in risk-related applications. In this context, validation efforts consisted of three main tasks based on the specific data subject to analysis. These are:
1. Systematic accuracy assessment of global exposure products addresses the validation of external EO-based exposure products, mainly global layers of urban extent which present valuable first-level approximations of human settlements on a global scale. In this, a systematic multi-scale concept has been developed to assess the accuracy of these products based on a review and selection of state-of-the-art techniques and methods for meaningful accuracy assessment. Applied techniques include absolute accuracy measures based on the error matrix, measures of relative inter-map agreement (e.g. receiver operating characteristics) as well as methods for pattern-based accuracy assessment. The developed approach respects the different spatial scales and resolutions of the layers under study, as well as pattern-based differences to assess accuracies in structurally different areas (e.g. rural vs. urban). The results (see the examples presented in Figure 2.11) show that new high-resolution products investigated within the context of SENSUM, namely the Global Urban Footprint (GUF, DLR ) and the Global Human Settlement Layer (GHSL, JRC ), hold great potential for mapping the spatial outline of settlement patterns (even smaller settlements in fragmented, low-density rural areas) in a very detailed manner and will significantly improve the geometric capabilities of the global exposure database, mapping even small-scale settlements of rural landscapes, as is the case for the test site in Central Asia. On the contrary, existing low-resolution (LR) products such as Globcover (GLOBC, ESA ) or the MODIS Map of Urban Extent (MODIS, SAGE-WISC ) map larger urban areas accurately, but are too coarse/generalized for applications in rural areas.
2. Systematic accuracy assessment of regional exposure products addresses the validation of external EO-based exposure products of regional coverage (e.g. coming from Copernicus/GMES) that present second-level structural qualifications of the built environment beyond binary formats (i.e. built-up, not built-up). In this context, several pan-European geo-information layers coming from the Copernicus/GMES programme have been assessed and validated. With regard to these layers, statements dealing with their accuracy vary greatly within the literature and across test sites, since information is heavily dependent upon different workflows of automated image processing, scene-specific quality of satellite imagery, as well as visual inspection/screening differences between interpreters. However, concerning risk-related applications for larger urban areas, the accuracy, especially in terms of the quality of the structural information content, is of great interest for approximating their structural vulnerability. The validation efforts carried out in this regard thus focused on both mapping accuracy and the presented structural information content (see the example presented in Figure 2.12) based on available, appropriate and up-to-date reference data.
3. Cross-validation and statistical testing of methodologies proposed within SENSUM for exposure data collection highlights the added value brought to the user by these SENSUM developments. These include the open-source remote sensing software developed for vulnerability indicator extraction (deliverable D2.2) and the sampling framework for in situ data collection (deliverable D3.5). The former were assessed using appropriate reference data on city and building level by the use of standard measures quantifying thematic accuracies based on the error matrix (Figure 2.13). Overall, the feasibility and general viability of all implemented workflows and algorithms provide the user with a powerful open-source tool to extract settlement information (which is the core of exposure models) from aerial and satellite imagery with a high level of both reliability and automation. Regarding the in situ data capturing, less conventional validation approaches based on sample data for Cologne were chosen due to the distinct nature of the proposed in situ data collection approaches. For example, the validation of the sampling scheme primarily aimed at the quantification of the efficiency of the selective sampling approach in terms of a comprehensive representation of the structural urban configuration (Figure 2.14). In addition, selected parameters that contain structural and demographic vulnerability information on the per-building scale obtained through visual screening employing the RRVS (see above) were evaluated with respect to the appropriate reference data. These assessments underlined the effectiveness of the approach, as well as the high potential degree of confidence that can be associated with the in situ derived information content.
Test sites – use of SENSUM products
The following will summarise some of the field activities undertaken during SENSUM and, where appropriate, outline the manner in which the SENSUM products have been exploited in each case.
-Central Asia
Kyrgyzstan
First order focus map
The first order focus maps for the Batken province, Kyrgyzstan, consist of hazard and exposure maps for landslides induced by precipitation and earthquakes. The method used is a heuristic approach largely based on the hotspots method that has been applied both globally and in drill-down studies (e.g. Nadim et al., 2006). The basis is the expert-based assignation of weights for available susceptibility and triggering factors. The susceptibility factors were terrain slope, lithology, ground moisture, and land cover. The triggering factors were 30-day precipitation and seismic intensities. Due to the lack of ground-based precipitation measurements, the maps for moisture and the precipitation triggering factor were derived using satellite-based estimates from the TRMM mission. Seismic intensities were provided by a previous study performed by Bindi et al, (2012). Exposure conditions were evaluated in terms of population and road networks. The maps were provided as an aid to field surveys performed by CAIAG during late spring and summer 2014 in the area of Suluktu, Kyrgyzstan.
A landslide hazard map for precipitation triggered events, obtained with the hotspots methodology, is shown in Figure 2.15. The importance of the different triggering and susceptibility factors can be calibrated against the information available in landslide inventories and physical processes.
Landslide inventory
Field work on landslide studies were conducted during SENSUM in the south-western part of Tien-Shan. This region is located in the Batken administrative oblast and borders with Tajikistan. In the study region, landslides are abundant on a relatively small area of 15 km radius from the town of Suluktu. Landslides here occur on the sedimentary cover of Mezo-Cenozoic and Quaternary loess loam soils (see Figure 2.16) which completely cover the bedrock in the intermountain areas. Most of landslides in the area occurred on the northern foothill slopes. Landslides located in the sub-mountain belt in the western and south-western part of the Fergana basin occurred predominantly on the back side slopes of anticline structures. In most cases, such slopes are north facing as a result of modern tectonic movements. Anthropogenic landslides also occur in this area, triggered by the exploitation of coal deposits.
An inventory study of landslides was conducted on the basis of digital geology, geomorphology, soil cover, and hydrogeology maps, resulting in landslide polygons. Distance measurements using laser instruments (TruePulse 360 ) and space images (Bing , Google and RapidEye ) were also employed. Modelling of slope parameters was carried out on the basis of the ASTER DEM with a 23 х 23 m resolution. During the course of the field work, samples of the upper landslide layer rocks were taken to identify the physical mechanical ground properties. In total, 45 landslides were surveyed, involving the consideration of 50 characteristics.
Second order focus map
The resulting landslide inventory was employed for several purposes in the drill-down analyses (i.e. in the production of 2nd order focus maps), namely serving as input for the development of an empirical run-out model and for a triggering model for the Suluktu area. The model used for the production of 2nd order focus maps for landslide hazard has three components:
1. Triggering: An infinite slope model was used. Only rainfall triggered landslides were considered.
2. Run-out: a statistical/empirical landslide run-out model was applied, that has been developed based on the data from the CAIAG landslide inventory.
3. Monte Carlo simulation: The Monte Carlo Simulation was performed by randomly varying the soil parameters and landslide dimensions according to the statistical variation observed in the CAIAG landslide inventory. Log-normal distributions were applied for the parameters.
The results of the analysis were combined with the exposure (built-up areas) and the resulting map is shown in Figure 2.17.
Tajikistan
Seismic and landslide studies were undertaken in the Jorgatal area, located in the north-east part of Central Tajikistan, on the border with Kyrgyzstan. The district is drained in an east-west direction by the Surkhob River, starting at the confluence of the Kyzylsu River (from the northeast, with its source in Kyrgyzstan) and the Muksu River (from the east, with its sources in the Pamir).
Seismic and landslide hazards investigation were estimated for Jirgatal area based on the published sources, archive data and interpretation and analysis of the existing satellite imagery and Google Earth resources like a most current sources. A map of seismic hazard, namely a probabilistic seismic hazard map (Figure 2.18) showing a 10% probability of exceedance during a 50 year period was created for Jirgatal area (in terms of peak ground acceleration (PGA) based on the existing active faults map and earthquakes catalogue (based on the Earthquake Model Central Asia catalogue, e.g. Mikhailova et al., 2015) making use of the software tool CRISIS 2007, a seismic and tsunami hazard tool .
Most attention was paid to landslides, although debris flows and snow avalanches are the most common natural hazards for study area. Moreover, efforts were not concentrated on individual landslides, but rather they took into account their probability, dividing potential landslides areas into 5 levels of possible hazard. This is a new generation of special landslide hazard maps based on the results of the interpretation of satellite images and the analysis of traditional landslide surveys, including the region’s geomorphology, lithology of the Quaternary deposits and engineering geology. The resulting landslide hazard map is shown in Figure 2.19.
In addition, we calculated the probability of the landslides under the influence of seismic loading based on three parameters: slope angle, curvature of slope and calculated PGA values for the area. The equation proposed by Uchida et al. (2004) was used, and the results are given in Figure 2.20. Finally (not shown here) a substantial database describing the settlements of the Jirgatal area was compiled, including population, gender balance, schools, health centres, types and distribution of buildings and bridges.
-Turkey
Izmir
The availability of medium and high resolution satellite imagery for the city of Izmir guaranteed the chance to test the set of released products. Figure 2.21 shows the evolution of the built-up area over the years obtained to the stack satellite and change detection (post-classification) algorithms applied to Landsat data.
Figure 2.22 and Figure 2.23 are related to indicators extracted from high resolution imagery (World-View 2). In particular, the former displays the extraction of footprints in a high-density residential area; the different colours being associated with the roof types. The latter is related to the output of the density algorithm used to highlight areas with major concentration of buildings within the area of interest.
Van
The destructive 7.1 Mw earthquake that struck eastern Turkey near the city of Van on 23 October 2011 provided an opportunity to test the SENSUM products devised for assessing and monitoring post-earthquake recovery. Many issues surround post-disaster recovery and reconstruction, usually requiring the combined effort of planners, government authorities and community interaction. However, natural disasters produce an unusual set of constraints and tensions that leads to specific needs for each case. The aim of SENSUM is therefore to provide the appropriate tools to assist in this process.
One of the main physical indicators used to assess disaster recovery is the erection and disbandment of camps. The identification of open spaces to first site camps, the number of tents needed and present during the relief phase, and finally the duration these camps are present, all give an indication of the post-disaster needs of a community and the speed of recovery. The number of tents can be used as a proxy to estimate the number of displaced persons needing permanent housing. For SENSUM, as part of a suite of tools to assist with monitoring recovery, the team developed an automated method of camp detention using mathematic morphology (see deliverable D2.6 “Supervised/unsupervised change detection”). This method is especially useful for objects characterised by clear shapes, sizes and contrasts. It uses morphological erosion and morphological dilation to identify targeted objects such as tents.
Using a series of images acquired for Van (one before and two at different times after the earthquake), out of a total number of 209 tents at a particular camp, only 33 were unclassified, giving a false negative error of less than 16%. Only 6 out of 209 tents were counted twice, giving a false positive error of 3%. Therefore, overall the number of tents correctly identified was 170, giving an accuracy of around 81% accuracy (Wang et al., 2015). This is shown in Figure 2.24. The set of imagery acquired for Van were also used to test SENSUM products on open space, roads and buildings identification and change detection.
-Germany
Cologne
The municipality of Cologne has a large amount of information dealing with all aspects of risk, however, such information is potentially out of date, or inadequate to thoroughly assess the city’s current risk. With respect to flood risk for example, detailed information about the location and height of buildings was available to the project in the form of cadastral data and a 3D city model. Moreover, a flood hazard map could be used for the study site. However, detailed information about the building stock with respect to particular indicators related to flood risk, such as the number and location entrances, occupancy of the first floor, number and location of windows, etc., is not available. Therefore, the SENSUM project in Cologne focused largely on a prioritization of an in-situ survey that aimed at identifying and quantifying these specific flood-relevant building indicators.
The focus map and resulting sample path for the Cologne in-situ survey are presented in Figure 2.25. The focus map was produced by combining information on the spatial distribution of buildings with a 100 year return period flood scenario. Equal weights were assigned to the building density and the inundation scenario in order to balance the relative importance of these factors for the sampling. A multiplicative pooling operator was chosen to prioritize the data collection based on the areas that are exposed to both inundation and characterized by a higher density of buildings, therefore filtering out most of the urban environment in Cologne.
The routing operation was performed on the basis of OpenStreetMap data with cost-constraints being fixed to the travel length. Therefore, the sample path is optimized in the sense of providing a shortest path through all the sample points identified on the basis of the focus map. The complete path´s length amounted to ~100 km. The collected geo-referenced imagery was then analysed off-line using the RRVS. A sample of 800 residential buildings has been selected and several attributes of the buildings have been collected, including attributes of particular interest for the estimation of the vulnerability to floods.
Figure 2.26 shows a summary of the flood relevant attributes for the 800 sample buildings that were captured by the GFZ-MOMA system and analysed by a human operator through the RRVS tool and procedure. Data capturing was performed in one day, whereas image interpretation took about 3 days. It can be seen from Figure 2.26 that the clear majority of buildings have a single door entrance with more than half of the samples having an elevated ground floor entrance (Hochparterre). Out of 800 buildings, 163 buildings have some sort of entrance below street level that makes the building particularly vulnerable to flooding. The main construction material of the observed buildings is masonry without any reinforcement, whereas the occupancy is dominated by residential.
The outcomes of the survey were presented to the stakeholders in Cologne during a technical meeting and the benefits and characteristics of the proposed approach have been discussed during a subsequent policy-briefing meeting.
Potential Impact:
The activities in this section are subdivided into the following:
• The dissemination of the project’s goals and results to as wide an audience as possible, although focusing on the DRR community (both research and practitioner).
• How the products developed during the project may be exploited following the completion of the project, in terms of their commercial application, utilization by the DRR community and ongoing development.
• What would be the impact of the SENSUM project on the DRR community (again, in terms of research objectives and practice).
Dissemination and communication
The communication of the goals, results and products of any project, especially one like SENSUM which seeks to directly contribute to DRR activities, is critical to its lasting legacy. In general, the activities undertaken during the project (generally within WP7, see Figure 1.3) are subdivided as follows:
• Maintenance of a project website and geoinformation portal.
• Workshops, symposia, and training courses to present and demonstrate the SENSUM products.
• Academic and technical conferences and meetings.
• Publications in the relevant literature, both professional and academic.
• Policy briefings to civil protection and other authorities.
• Printed material.
• “The SENSUM Game”.
The SENSUM game was outlined in the previous section and so will not be covered here. Likewise, most information about publications and conference presentations is presented in Section A. The following will therefore provide details about the other activities undertaken with the categories listed above, with further details outlined in deliverables D7.3 “Dissemination plan” and D7.4 “Scientific technical communication”. It should, however, be mentioned that although the project has ended, opportunities for the exploitation of its products and the development of new proposal concepts to expand upon already undertaken work are continuously being undertaken.
-Project website
The project website , which is often the first point of contact for many interested parties, was continuously developed during the course of the project. In addition to general information about the project’s goals, activities, and the consortium partners, the various tools, technologies and methodologies developed are presented. The project deliverables are also available from the website, along with printed material, including a brochure outlining the project as a whole, and information sheets about specific products.
-Workshops, symposia and training courses
• An outreach event, “Symposium on earthquake and landslide risk in Central Asia and Caucasus: exploiting remote sensing and geo-spatial information management”, was held in Bishkek, Kyrgyzstan, in January, 2014 (see Figure 3.1). The main goal of this event (see deliverable D7.5 “Outreach Event”) was for the SENSUM consortium members to engage with regional DRR practitioners, decision-makers and other stakeholders in presenting the goals and early results of SENSUM. Of particular importance were efforts to enhance the awareness of the capabilities and opportunities offered by remote sensing for DRR activities. Around 25 participants (in addition to the SENSUM members) attended, representing a range of educational, DRR and development institutions (e.g. Kyrgyz Ministry of Emergency Situations, UNDP, World Bank). From this meeting, a proceedings of 15 research papers was produced, and is available in both hard copy and via the project website (see deliverable D7.6 “Proceedings Volume”).
• During the Understanding Risk Forum , held in London, England, a session with an emphasis on issues relevant to SENSUM activities was organised by members of the SENSUM consortium (UCAM and CAR) entitled “Where Sky meets ground: integration of remote sensing and ground based information for multi-hazard vulnerability monitoring and recovery assessment”. As part of this, a simplified session of the SENSUM “Game” was presented (see D7.4).
• An international summer school “Data Fusion of Risk-related Remotely Sensed and Geospatial Data” (part of a series run by the University of Pavia, Italy) was organised by the SENSUM consortium and held at the University of Pavia. This aim of the school was to provide the attendees (12 graduate students) with basic knowledge on the problems related to mapping the time-dependent hazard and vulnerability of seismic and landslide areas, the presentation of newly developed open-source tools to address the above issues by exploiting Earth Observation (EO) data as developed by SENSUM, and the introduction of the QGIS tool as an effective open-source environment for research and operational use of these products. The event proved to be most successful, with the expectations of the students generally met in terms of organisation of the course, quality of the lectures, interactions with the lecturers, and the opportunities for discussion and the sharing of opinions. Details about this event are presented in deliverable D7.7 “Summer School”.
• CAIAG conducted a training course in November, 2014, on “Monitoring and creation of a landslide database” in Osh, Kyrgyz Republic. This was attended by 25 employees of the Monitoring Department of the Ministry of Emergency Situations of the Kyrgyz Republic. The main purpose of the training course was to provide general information about the causes of landslides and the current state of natural disaster monitoring, including the presentation of ground-based and remote sensing methods for landslide studies. The course consisted of two parts: the first was the theoretical component where the CAIAG staff provided the seminar participants with the general information about landslides and landslide research methods, including database development; the second part of the training was dedicated to training in using tools during field studies of natural disasters (further details may be found in deliverable D7.10 “Policy briefings”).
-Policy briefings
As discussed elsewhere in this report, one of the key objectives of the SENSUM project was to encourage the use of Earth Observing (EO) imagery in DRR activities. As part of this, the project’s members engaged in what we refer to as “policy briefings” (see deliverable D7.10) which involved the presentation of the SENSUM project’s goals and outcomes to higher-level (decision and policy makers) members of the DRR community. Such activities included:
• The production of a so-called SENSUM policy brief document (in English, German, Turkish and Russian) which outlines how the SENSUM consortium sees the role of EO observation in DRR activities.
• On November 6th-7th 2014, the SENSUM project was presented to representatives of the Turkish DRR community at the Izmir University of Economics, Izmir, Turkey. These included members of the Disaster and Emergency Management Presidency of Turkey (AFAD) and representatives from various offices of the Izmir Municipality and academic researchers. After welcoming presentations, which included an overview of AFAD activities, the SENSUM project was presented, covering the project (and remote sensing in general) as a whole and especially the developed tools and products, along with some examples. This was followed by a panel discussion made up of AFAD, municipality, research and SENSUM representatives.
It became clear during the panel discussion that although the intention was to explore what current and future efforts in the area of remote sensing and in situ surveys would be most helpful to DRR actions, the issue of communication between the various end-users, or rather, deficiencies in communication, dominated the forum. The problem of a lack of resources, funding and expertise were also highlighted, specifically the situation (which was discussed) where although the appropriate resources (including personal) may have been available in the past, there is frequently a lack of will or capacity to maintain any infrastructure developed, for example, information collections (e.g. GIS data bases). This point will be returned to in the discussion on impact below.
• Policy briefings were also held in Germany with representatives of the German Civil Protection or Bundesamt für Bevölkerungsschutz und Katastrophenhilfe (BBK) and the Cologne City Council. The meeting with BBK (19th November, 2014) was informal, and in addition to describing the SENSUM project, as well as gaining a better understanding of the role of BBK within the context of German DRR activities, a draft of the policy briefing was discussed. It was found that the concepts put forward by the SENSUM consortium were consistent with those of BBK, hence the document remained largely unchanged following this meeting.
The meeting with the Cologne Municipality (20th November 2014) involved a SENSUM group taking part in the Sitzung des GeoBeirates, a gathering where geo-information practitioners meet to present their ongoing projects and available products. An informal discussion with municipality representatives following this gathering indicated their interest in the project’s outcome, and the consortium is optimistic that future collaborations will arise, especially given that Cologne is frequently a test for European DRR research projects.
• A meeting was organized by the SENSUM partner IGEES with officers and employees of the Committee of Emergency Situations and Civil Defence (CoES) of the Republic of Tajikistan (6th December, 2014). The meeting was organized by Colonel Jamshed Kamolov - Head of the Department of protection of population and territories (CoES). The speakers reported on the main results of the SENSUM project, and highlighted some general issues of seismology and the development of seismology in Tajikistan, as well as the results of the work within the framework of the SENSUM project in the Jirgatol district (northern Tajikistan, bordering with Kyrgyzstan, discussed in the previous section). The meeting was considered a success, with the audience strongly encouraging such work to be continued in other parts of Tajikistan. It was acknowledged that these projects have a positive effect on the reduction of risk arising from hazards like landslides, and contribute to protecting the civilian population and the country’s material resources.
-Printed material
Several printed items were developed during the project and are available from the project website. These include information sheets on some of the technological developments (e.g. the GFZ-REM or Rapid Environmental Mapping for geo-risk modelling system), a high-quality project brochure which is available in English, Russian, Turkish and German, and the SENSUM policy statements. Furthermore, most project deliverables are also available on the website.
Exploitation
The exploitation of the SENSUM results and products may be considered as including the ongoing development of the outcomes (recognising that advances in technology etc. are never ending), generally within other research projects, their consideration as part of other projects, although their further development may not be an end in itself, and in their use by DRR practitioners and other assorted end-users, which includes their commercial usage. When considering this, it was soon understood during the course of the project that the value of the SENSUM products lies in their ability to provide timely (rapid), relevant (focused) and cost-effective information on exposure and vulnerability indicators and post-event recovery progress within a multi-hazard context. The fact that the tools are free and open-source, and rely largely on free and globally available data sources, will allow them to, at the very least, be considered for use by a range of end users.-However while the value of remote sensing and in situ data collection technologies is recognized, end-users often lack the expertise necessary to strategically place the developed technologies within their in-house mechanisms.
The technologies developed within SENSUM that are considered exploitable may be divided into those associated with the EO tools, prioritization tools for sampling and the rapid environmental mapping methodologies, and user elicitation tools, such as “the game”. This therefore leads to an exploitation framework divided between services and training, which would also encompass capacity building activities. A number of exploitation activities are in fact already underway covering actions ranging from other EC research projects (e.g. RASOR - Rapid Analysis and Spatialization of Risk), programs funded by bodies such as the World Bank (risk assessment, capacity building), and data collection activities for various end-users (e.g. IGDAŞ, Istanbul Gas Distribution Industry and Trade Incorporated Company ), the latter being the sorts of activities the individual partners may engage in as part of short-term income-generating actions.
However, the primary vision of the consortium is focused on the capacity building of policy-makers, experts and stakeholders who work in risk reduction and management, and emergency response-related research and activities. Therefore, training activities/courses are seen as both a crucial, and lucrative, avenue to pursue. Potential customers and clients include intra-governmental agencies (e.g. Red Cross, World Bank), government ministries involved in DRR, development agencies (e.g. USAID) and other actors concerned with the exploitation of remote sensing for DRR actions (e.g. International Charter Space and Major Disaster). Such activities would include the SENSUM “Game” (especially useful for governmental agencies) to elicit information, and providing courses on DRR relevant to the SENUSM products, such as the use of remote sensing, integrating earth observations and in situ data, exposure and vulnerability monitoring, as well as more fundamental topics, such as general hazard and risk assessment. As an example of the sort of income such courses may generate, a 5 day course recently provided to the World Bank by CAR (Cambridge Architectural Research , subcontracted by a consortium partner) generated US$200,000 (it involved 20 days course reparation, 15 person/days delivering the course, and 15 days preparing a distance learning package. i.e. $4,000/day).
Impact
One impact from SENSUM is how the project partners will interact with potential end users when developing future projects. It was found from the various discussions, especially those during the policy briefings, that:
• The major concern of end-users was the lack of resources (e.g. geospatial expertise, IT infrastructure, ability to cover data costs, scientific collaboration links, etc.);
• Communication within and between the various end user organisations is a major issue, an outcome already identified in previous projects (e.g. the MATRIX project, Komendantova et al., 2014);
• The user community still struggles with a clear definition of the issues related to the access and cost of EO derived geo-information; in this context, a clear understanding on the strengths and weakness of the EO derived product portfolio, in particular the added value of the combined use of EO and in situ sensing is on demand (and has been partly tackled by validation efforts within SENSUM).
• There are still difficulties in conveying scientific information to DRR practitioners.
List of Websites:
http://www.sensum-project.eu/de(öffnet in neuem Fenster)
Massimiliano Pittore (GFZ) pittore@gfz-potsdam.de Project Coordinator
Kevin Fleming (GFZ) kevin@gfz-potsdam.de Project Manager