A software developed at the WSL for collecting data on past and current natural hazards

Abstract

Since its invention, photography has been a simple and direct means of documenting landscapes. Historical or current terrestrial oblique photographs capturing natural hazards often provide detailed information and can nowadays be easily taken by anyone using various devices.

The image2world software, originally developed at WSL (wsl.ch/monoplotting) and now taken over by the company image2world GmbH (i2w.ch), allows the use of individual terrestrial or aerial oblique images to map natural hazards (and other landscape features) in a cost-effective and efficient manner. Conventional photographs taken by smartphone, drone, or helicopter are transformed into 3D maps available to professionals, researchers, or just the curious.

This presentation will show examples of the software's use in the context of natural hazards, including the mapping of floods, landslides, and rockfalls, the analysis of the effectiveness of snow bridges, as well as in the field of immediate geolocation of natural events (Rapid Mapping), which can contribute to the organization of search and rescue operations.

Due to their complexity, large-scale wildfire and flood events put immense pressure on authorities to quickly gain a clear overview of the disaster situation for an adequate operational planning. Satellite-based emergency mapping (SEM) services such as the Copernicus Emergency Management Service (CEMS) rapid mapping service provide geospatial crisis information on demand and fast in support of authorities and responders before, during or immediately following a disaster. Although the standard SEM workflow has evolved in recent years, particularly in the field of satellite image analysis, the gap between the initial warning and the SEM activation still delays product availability.

For understanding where the delays stem from, we analysed the steps taken by the actors involved in the SEM process. Service providers perform the rapid mapping upon SEM activation and publish the produced crisis information, e.g. via the CEMS. In order to produce timely and accurate map products such as burnt or flooded area maps service providers need the SEM process to be triggered as early as possible with clearly defined areas of interest (AOIs). The SEM is typically activated by end users such as civil protection authorities and emergency services. Though a number of early warning tools are available, some crucial steps until SEM activation remain user-driven (Mühlbauer et al., 2024). First, end users need to manually identify the AOI, often from multi-source data such as warnings, weather forecasts, observations, etc. In addition, they need to put effort in getting aware of the availability of satellite data to capture the AOIs once they get affected by the event. Service providers usually use acquisition planning tools where they intersect the AOI with planned satellite overpasses. Furthermore, it is unclear to end users when the generated products eventually become available.

Accordingly, our research question here revolves around technical ways of improving users’ situational awareness and hence reducing the time needed from the initial warning to satellite data acquisition to the availability of analysis results. For addressing the latter, we examined and developed a tool that automatically processes and fuses multi-source web data (e.g., public alerts, sensor observations, weather forecasts), identifies AOIs and intersects them with relevant satellite acquisitions. Our approach improves the end user's situational awareness by automatically generating decision proposals regarding EO data and product availability. Situational awareness is further improved by an interactive spatiotemporal visualization of AOIs and satellite acquisitions. The user is supported by transparency on the underlying data sources, the expected (and actual) time of satellite data acquisition, attributes of relevance and overlap of satellites for events.

The talk will focus on how Rapid Mapping works, the challenges involved and our experiences with Rapid Mapping.

Rapid Mapping is a 24/7 on-call service of the Swiss Federal Government for the timely collection and/or provision of geodata (e.g. aerial or satellite imagery) in the event of natural disasters for the purpose of event documentation and, in certain cases, event management. In cooperation with the National Emergency Operations Centre (NEOC), the Federal Office for the Environment (FOEN) coordinates the needs of federal and cantonal agencies and, if necessary, other stakeholders in the event of large-scale or significant events with a high degree of urgency for data collection. After a positive assessment, the FOEN instructs the Federal Office of Topography, swisstopo, to obtain the data.

swisstopo is Switzerland’s geoinformation center and is responsible for the provision of analysis-ready geodata before and after natural hazard events and supports various stakeholders (federal government, cantons, municipalities) in documenting natural hazard events. swisstopo makes data freely available within the framework of Open Government Data by offering newly collected data (post-disaster) and existing swisstopo geodata (pre-disaster) for comparison purposes.

Rapid Mapping offers a range of so-called base products. These include digital imagery from a variety of imaging platforms (satellites, aircraft, helicopters) and sensors, selected according to the task at hand and availability. As with all airborne or spaceborne imagery, there are limiting factors that make it impossible to guarantee rapid mapping products within a given timeframe. These include weather conditions (e.g. cloud cover), but also the general availability of specific resources (manpower, equipment, etc.) at the time of the event.

Of particular importance is the fact that swisstopo's flight service has been upgraded to the highest priority level for rapid mapping missions within the framework of national airspace usage priorities, thus recognising the benefits of this service and facilitating its use in crisis management. While swisstopo operates its own flight service in collaboration with the Swiss Air Force to produce aerial image data, satellite data is acquired and managed through the National Point of Contact for Satellite Images (NPOC, 2024), which is managed by swisstopo. Through this contact point, various additional satellite data - both free and commercial - can be obtained, such as products from the Copernicus Sentinel portfolio or very high resolution imagery.

Access to analysis-ready rapid mapping products is provided through the Federal Geodata Infrastructure. The products are freely available in read-only formats at map.geo.admin.ch. Both pre- and post-disaster data are presented via this geodata infrastructure, together with a functionality that facilitates image comparison (Fig. 1).

The summer of 2024 was particularly challenging for the Rapid Mapping Service. Due to recurrent heavy rainfall, the service was activated four times in less than two weeks in different regions of Switzerland. Thanks to the various collection platforms, useful data was made available to the relevant authorities within the required timeframe in all four cases, making an important contribution to the management and documentation of the events and even to the situation.

Disasters such as floods cause severe damage and affect millions of people every year. To respond quickly and effectively, emergency services need up-to-date, comprehensive and accurate information on the extent of the hazard, exposed assets and damage. In recent years, the rapidly growing number of satellites in orbit and the data they provide have also made it possible to prepare for specific events or to monitor vulnerable regions of the world on an ongoing basis.

The recent floods in Germany, for example, demonstrated the importance of continuous monitoring and close cooperation with humanitarian actors. The heavy rainfall events and subsequent widespread flooding in southern Germany in June 2024 were preceded by official warnings from the German Weather Service, and emergency services were able to take preparatory measures before the onset of the flooding. In addition to the activities by the Copernicus Emergency Management Service (CEMS) and national entities, the Center for Satellite-based Crisis Information (ZKI) of the German Aerospace Center (DLR) provided first aid responders with updated web-based crisis information on the evolving flood situation. To improve the situational awareness of the event, the web application included not only aerial images and the analyzed flood extent from different satellite sensors but also datasets on population, buildings, land use and critical infrastructure (ZKI, 2024).

As part of the Indicator Monitoring for Early Acquisition of Innovative Satellite Sensors in Natural Disasters (IFAS) project, the DLR is working to improve satellite-based emergency mapping by initiating the process chain in anticipatory action. A study by CEMS has shown the timely benefit of early tasking satellite imagery based on hydrological forecasts (Wania et al., 2021). In regards to this finding and to improve the response time for flood disasters, the DLR is collaborating closely with European Space Imaging (EUSI), an industrial partner providing very high-resolution satellite imagery, as well as with first aid responders. By automating components of the rapid mapping process, this project aims to minimize the time delay in the availability of satellite data and the provision of crisis information to anticipate the development of a crisis at an early stage.

The project therefore touches on various aspects of different phases in disaster management: In the monitoring and preparedness phase, heterogeneous data sources including official alerts and forecasting information are collected and matched with possible satellite overpasses to initiate a timely acquisition of (very high-resolution) satellite data for a potential crisis event. The aggregation of data sources can further facilitate the pre-definition of areas of interest for satellite data acquisition, which is then automated using EUSI's Tasking Archive API. Once satellite imagery has been delivered, the data is automatically downloaded, rapidly analyzed, and integrated into a web-based crisis information product and shared with dedicated civil protection actors to support response activities. Their feedback is then used to continuously improve the visualization of crisis information products.

Detecting buildings from remote sensing data typically requires high-resolution satellite images, which are costly and limited with respect to accessibility. Although open-access satellite data offer global coverage with frequent revisits, low resolution poses a significant challenge for accurate building detection. By constructing multiple detection models tailored to specific building location characteristics, such as building density and size, we develop an approach that enhances building detection accuracy. First, we fine-tune a super-resolution model to increase the resolution of Sentinel-2 images from 10 m to 2.5 m, thereby improving the visibility of relatively small buildings. We then classify the study area using a logistic regression model based on population, Normalized Difference Vegetation Index, nighttime light intensity, and distance to the coastline. This classification facilitates the grouping of areas into distinct categories based on the building location characteristics. Subsequently, we develop separate building detection models for each classified area and evaluate their performance against a general detection model trained on all data. Our results demonstrate that models optimized for specific building location characteristics significantly outperform the general model, particularly for areas with large buildings. This study highlights the importance of considering building location characteristics when constructing detection models and provides a framework for improving the accuracy of building detection using low-resolution, open-access satellite data.