Search results

This report highlights the high potential of earth observation for consistent and objective monitoring of elements at risk at various spatial scales. The product portfolio of EO derived geo-products ranges from global low resolution land cover datasets to high resolution spatially accurate building inventories on a local scale. Despite the importance of adequate data for a comprehensive risk analysis as a critical factor affecting the constraints and requirements for the scientific community, end-users, stakeholders and policy makers, an immense discrepancy exists between data-rich countries of the developed world where extensive geospatial information is available, and less-developed data-poor countries. While data-poor countries mainly rely on international efforts to provide low resolution land cover / use maps of global coverage, significant international efforts to provide geo-products of medium to high resolution on a regional scale have only been undertaken in data-rich countries of the developed world such as Europe in the past. With regard to user-oriented product generation in project SENSUM, a multi-scale and multi-source reference database has been set up to systematically screen available products with regard to data availability for the three project test sites of strongly differing data availability: Cologne (data-rich), Izmir (intermediate), Isfara/Batken (data-poor). At a later stage, these data will serve as a reference to evaluate and document the capabilities and limitations of the proposed products and range them with regard to the current GMES product portfolio. From the final database content, it becomes clear that data-poor countries of Central Asia mainly rely on coarse resolution products of global coverage which, however, provide multi-categorical thematic detail. In contrast, medium and high resolution datasets are spatially restricted to the European test sites due to trans-European mapping efforts initiated there. However, two currently developed global products – namely DLR's Global Urban Footprint as well as JRC's Global Human Settlement Layer – will be a major leap forward regarding the derivation of high resolution and accurate reference data for human exposures on a global scale – as they will provide consistent and geometrically detailed land cover information at unprecedented spatial resolutions. Furthermore, a viable option for future research and applications is presented, namely Volunteered Geographical Information (VGI) by crowd-sourcing of extensive mapping communities such as the Open StreetMap project. However, as remote sensing methods alone cannot provide all information needed for a comprehensive vulnerability and risk estimation, especially when political or socio-economical vulnerability is considered, the call for future research is on the integration of EO and in-situ data. Furthermore, the higher-ranking goal of activities in project SENSUM should address potentials for integration of the proposed products and methodologies in the GMES service offer, particularly envisaging the future expansion of the existing GMES service and product offer to further non-European countries.

In terms of an integrative and comprehensive risk analysis, the issue of an appropriate data collection is widely recognized. In this context, remote sensing is generally perceived as a promising tool for an economical, up-to-date and independent data collection and has been employed in various investigations in the geo-risk context. A review of past applications reveals the versatility of remote sensing data and techniques with regard to the analysis of earthquake and landslide risks. The study of earthquake hazard components reach from geologic site characterisations and fault mapping, detection of pre-event surface temperature anomalies for early warning, kinematics and dynamics of active fault systems, to the detection of earthquake-induced changes to the land surface. For landslides, remote sensing techniques and applications contribute to basic terrain analysis including monitoring of slope stability and land cover, the detection and mapping of pre-event landslide-related activities such as slow mass movements as well as the volumetric quantification of landslide related changes to land cover and form. Remote sensing for vulnerability-centred investigations is a less long established field of research. Nevertheless, a vast amount of research aims at the derivation of pre-event vulnerability indicators – related to physical, demographic and socioeconomic aspects of vulnerability–. Also work on post-event emergency response and damage assessment employing especially very high resolution data has been carried out. In the context of geo-risk analysis, both for pre- and post-event as well as hazard- and vulnerability-centred investigations, past- and present-day remote sensing missions have proven useful. Data with a coarser geometric resolution and larger spatial coverage per scene are able to contribute to overall evaluation of pre- and post-event situations. In contrast, it is yet both difficult and expensive to obtain high geometric resolution data such as airborne LIDAR or VHR optical data over larger areas affected, specifically in case of an earthquake event. In addition to that, the geometric resolution of the sensors and the associated scene size and thus, the spatial scale of analysis, need always be selected in consideration of the context of the objects to be analysed. Future earth observation missions have the potential to play a key role in earthquake- and landslide-related investigations and to continue or even improve geoinformation products. For example, the ESA Sentinel missions will feature enhanced geometric and thematic capabilities and increased revisit capabilities at low cost. Other future missions such as ALOS-2, the RADARSAT constellation, DESDynI, CARTOSAT-3, ALOS-3, TerraSAR-X 2, TanDEM-L, WorldView-3, or the hyperspectral sensors EnMAP and HyspIRI still need to be assessed in terms of cost and applicability with respect to the particular research questions addressed. For example, TanDEM-L will open up opportunities to better understand the earth's dynamic surface processes enabling continuous monitoring of earth surface deformation e.g. due to seismic movements, volcanic eruptions, landslides, subsidence or uplift in the range of centimetres. 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. Apart from the future perspective, international research groups from both government and academia have produced remote sensing based geo-products in past years that will provide valuable input to vulnerability-related research in the context of this project. These include on the one hand large area global or regional land cover datasets that can be used as a first approximation of human and physical exposure as well as for the disaggregation of population census data. However, with regard to these products, a better understanding of each data set's strength and weakness is still on demand. On the other hand, space-based pre-operational emergency response services have produced a large product portfolio and significant experiences in post-event mapping applications in recent years. Overall, from a technical perspective, the constantly increasing availability and accessibil-ity of modern-day remote sensing technologies and the enhanced technical capabilities of future missions will provide new opportunities and data continuity for a wide range of geo-risk investigations. However, still today and in the future, data cost will be the main limiting factor.

Abstract. In this paper, we anticipate geospatial population distributions to quantify the future number of people living in earthquake-prone and tsunami-prone areas of Lima and Callao, Peru. We capitalize upon existing gridded population time series data sets, which are provided on an open-source basis globally, and implement machine learning models tailored for time series analysis, i.e., based on long short-term memory (LSTM) networks, for prediction of future time steps. Specifically, we harvest WorldPop population data and teach LSTM and convolutional LSTM models equipped with both unidirectional and bidirectional learning mechanisms, which are derived from different feature sets, i.e., driving factors. To gain insights regarding the competitive performance of LSTM-based models in this application context, we also implement multilinear regression and random forest models for comparison. The results clearly underline the value of the LSTM-based models for forecasting gridded population data; the most accurate prediction obtained with an LSTM equipped with a bidirectional learning scheme features a root-mean-squared error of 3.63 people per 100 × 100 m grid cell while maintaining an excellent model fit (R2= 0.995). We deploy this model for anticipation of population along a 3-year interval until the year 2035. Especially in areas of high peak ground acceleration of 207–210 cm s−2, the population is anticipated to experience growth of almost 30 % over the forecasted time span, which simultaneously corresponds to 70 % of the predicted additional inhabitants of Lima. The population in the tsunami inundation area is anticipated to grow by 61 % until 2035, which is substantially more than the average growth of 35 % for the city. Uncovering those relations can help urban planners and policymakers to develop effective risk mitigation strategies.