How do your professional career and/or your personal experience relate to space technologies and water? How did you first get in touch with space technologies?

Space flight is often referred to as an exciting and awe-inspiring achievement of human technology and exploration. However, the diverse impact of space technology on our society and its importance for our daily lives is often overlooked and insufficiently communicated. Having been fascinated by space flight and geophysics since my early days, I increasingly learnt about its benefits and the great potential of space applications in the context of humanitarian action and development cooperation. While the space sector has been a driving factor for scientific and technological enhancement in recent decades – leading to many spin-offs and high-end products used daily all over the world – applications such as satellite navigation and communication have shaped the way we interact on a global scale. Earth observation satellites provide data on numerous environmental parameters and thus allow quantifying and understanding global change, a fundamental prerequisite for adapting to our changing environment. On top of these few examples, for me, space flight represents a contemporary way of exploratory and pioneering spirit, the urge of humanity to delve further into the unknown. The technological challenges in combination with socially relevant applications and the spirit of curiosity shaped my decision to study aerospace engineering and to focus on environmental risks and human security.

Could you tell us about your current work, your latest project or your proudest professional moment?

My current work centres on methodological developments for rapid building damage assessment. In order to assess the impact of large-scale natural disasters as well as for planning and coordinating emergency response efforts, the humanitarian community increasingly makes use of geospatial information. Near real-time damage assessment after a disaster event is a critical component in emergency response efforts since it provides vital information regarding the condition and functionality of critical infrastructure such as buildings, roads, bridges, or airports. However, the prompt generation of such comprehensive map products poses a significant challenge, since infrastructural damage is largely assessed through on-site surveys and visual analyses of remote sensing imagery. Accordingly, the assessment process requires important and usually limited resources such as time and human capacity. As part of my PhD, I am currently working on deep learning-based methods for automatic and rapid building damage assessment. These models provide large-scale overviews of affected areas to help emergency response units with operating faster and more efficiently, aiming to ultimately reduce casualties and fatalities. Subsequently, these methods are incorporated into risk models to better anticipate the impacts of future flood events while building resilience based on adapted urban planning.

As an aerospace engineer, what was the most interesting thing you have learnt about designing satellites aimed at measuring water-related parameters?

For me, the most interesting part of the design of satellites for measuring water-related parameters is the diversity of sensors and the creativity of developing scientists. Satellite remote sensing sensors can be classified into two broad categories: active and passive sensors. Passive sensors detect natural radiation emitted or reflected by the Earth's surface or atmosphere. These sensors measure the intensity of radiation in different wavelengths – such as visible, infrared, and microwave – to infer information about the properties of the Earth's surface and atmosphere. Active sensors, on the other hand, emit radiation and measure the response of the Earth's surface or atmosphere to the emitted signal. Both passive and active sensors have their unique strengths and limitations, and the choice of sensor depends mostly on the application, but also on the required spatial and temporal resolution.

To give a few examples: one of the most common applications in the context of my work is mapping the extent of floods and permanent water bodies using radar or optical data. This is an important information for emergency response and disaster management operations. Furthermore, thermal sensors can measure the temperature of surface water bodies, which is crucial for understanding water circulation patterns and detecting anomalies such as thermal plumes from industrial facilities. Radar altimeters measure the sea surface height with high accuracy. These measurements are used to monitor sea level rise and the variability of ocean currents. Optical sensors can measure the colour of the ocean, which is related to the concentration of phytoplankton and other microscopic organisms. These measurements are used to study ocean productivity, the global carbon cycle, and the impacts of climate change on marine ecosystems. Scatterometers infer information about ocean wind speed and direction based on the geometry of waves. This helps predict storm surges, ocean currents, and wave heights. To conclude, there is an impressive variety of sensors to measure highly relevant parameters.

Could you elaborate on the relation between deformation of land surface and ground water? What instruments are used to monitor this?

The relationship between surface-level deformations and groundwater is complex, as changes in the water table can cause the land surface to deform, and deformation of the land surface can in turn affect the flow of groundwater. Differential Interferometric Synthetic Aperture Radar (DInSAR) is a remote sensing technique that uses satellite radar to detect changes in ground elevation over time. This technique is commonly used for monitoring subsidence, deformation, and other ground movements associated with natural and anthropogenic activities. By comparing radar images acquired at different times, DInSAR can detect these changes and provide information on the spatial and temporal patterns of groundwater movement. However, it is important to note that DInSAR is no direct measurement of groundwater levels. Rather, it provides information on ground deformation in the order of millimetres that can be used to infer changes in groundwater levels. To accurately interpret the radar data and distinguish between different sources of deformation, other complementary data, such as groundwater level measurements, hydrological modelling, and geologic information, may also be needed.

During your internship at UNOOSA / UN-SPIDER you developed a flood mapping and drought monitoring tool, can you tell us more about it?

The mandate of the United Nations Platform for Space-based Information for Disaster Management and Emergency Response (UN-SPIDER) is to enable developing countries the use of all types of space-based information in all phases of the disaster management cycle. By doing so, UN-SPIDER bridges a very important and interesting gap between the technical community and end-users. Therefore, the driving factor in the development of both tools was to address the needs of end-users, i.e. disaster management organizations. One major constraint in many countries is insufficient internet connectivity and computing power. Therefore, cloud computing can be an excellent way to leverage advanced computing resources, improve access to technology, and drive innovation. We found a good solution by using so-called Jupyter Notebooks hosted in cloud environments. These are interactive platforms which combine software code, text and visualisation. It is a great framework for running and following step-by-step procedures. In this way, end-users can generate their products while building capacity and an understanding of different workflows and processing techniques. Both tools cover the full processing chain from data query and download up to the export of the final data products – either radar-based flood extent maps or the results of a multi-temporal analysis of spectral vegetation indices to support drought monitoring and early warning.

Recently you have been studying and mapping floods using traditional approaches but also machine learning. Can you elaborate on the state-of-the-art approaches and models related to flood detection and early warning? Where do you see the greatest potential and biggest challenges in this field of research?

Flood detection and early warning systems are critical for minimizing the negative impact of floods on society and the environment. In recent years, there has been a growing interest in using machine learning techniques in this context. Machine learning models have the potential to enhance the accuracy and speed of flood detection and prediction, thereby improving early warning systems. However, there are also significant challenges to address.
In general, various methods and perspectives/disciplines that deal with floods exist. To give a few examples; many approaches for flood detection and early warning rely on physical models based on rainfall data, river flow data, and other hydrological variables. Sensors placed in rivers and other water bodies can detect changes in water levels and provide real-time data. Another promising field for flood detection and early warning is the use of social media and crowd-sourced data. By analysing these data sources, researchers can quickly identify flood locations, estimate flood severity, and inform the public about potential dangers.

Remote sensing technologies provide another valuable data source, which allows for the monitoring of large areas. These technologies can detect changes in water levels and track flood progression, enabling timely response and mitigation of potential floods. In this domain, machine learning-based approaches have become a dominant processing method in recent years. These algorithms learn from training examples – labelled data sets of satellite imagery with their corresponding flood extents. During training, the models learn to identify patterns in the satellite imagery corresponding to flooded areas and can then be applied to new satellite data to automatically identify the flood extent. On the other hand, rule-based approaches involve predefined rules or thresholds to identify flooded areas in satellite imagery. For example, a rule-based approach may use a threshold value for water reflectance to identify flooded areas.

In comparison, machine-learning approaches have the advantage of being more flexible and adaptive than rule-based approaches. Machine learning algorithms can learn from a larger and more diverse data set and can capture more complex relationships between satellite imagery and flood extents. However, there are also significant challenges such as the lack of quality data for training and testing. In some cases, data may be limited, outdated, or of poor quality, making it challenging to develop accurate models. Furthermore, machine learning algorithms can become overly complex and learn to recognize specific patterns in the training data rather than generalizing to new data, which can result in poor performance on new data sets – a phenomenon also known as overfitting. 
Overall, the use of machine learning for flood detection and early warning is an exciting area of research with great potential for improvement. However, it is important to address the challenges associated with data quality and interpretability to ensure the effectiveness and trustworthiness of these models. Personally, I remain sceptical because purely data-driven approaches such as machine learning rely solely on patterns and correlations in the data, without necessarily understanding the underlying physics or causal relationships – something I believe to be fundamental in science.

In your Master’s thesis, you researched approaches to map flooding and the quantification of uncertainties. Can you elaborate on how mapping uncertainties influence decision-making in humanitarian operations?

Map products based on remote sensing data have been used to support near real-time emergency response to the onset of flood events for a long time. A map provides a synoptic overview of the situation and supports more targeted planning and distribution of limited resources in response to local needs. Currently, flood maps illustrate a single possible interface boundary between water and non-water classes. However, geospatial products contain inherent uncertainties from different sources, such as input training data quality and model inaccuracies. Improvements to uncertainty quantification can provide additional information for informed decision-making, especially in response to non-trivial problems where wrong decisions can result in adverse consequences. Therefore, rather than returning a single possible interface boundary between classes, an area that represents the range of possible interface boundaries can be generated to visually capture the extent of segmentation uncertainty. In the deep learning domain, this can be achieved by introducing stochastic components to the model such as probability distributions which account for aggregated uncertainties.
A survey with official authorities and non-governmental organizations revealed that the humanitarian community is interested in the inclusion of uncertainty information in geospatial products. The additional information can help decision-makers to better understand the potential risks and uncertainties associated with different scenarios. For example, if a flood map indicates a high level of uncertainty in the predicted flood extent, decision-makers may choose to take a more conservative approach to planning and response activities. On the other hand, if a flood map indicates a low level of uncertainty, decision-makers can have confidence in more targeted decisions to mobilize resources and respond to the crisis. Overall, the inclusion of uncertainty information in flood maps can help decision-makers make more informed and effective decisions in humanitarian operations, ultimately leading to better outcomes for the affected communities.

Since you have studied flood management in West Africa within the context of a Regional Academy on the United Nations project, can you expand on your research findings? 

Despite floodplains supplying many of the key natural resources for the world’s population, floods can pose a threat to lives and infrastructure. Especially in cities, climate change alongside increasing urbanization and land-use change further exacerbate flood risks. In African cities, there is a rising trend of flood risk and population growth. In the context of managing floods in Africa, interviews conducted with experts from organizations such as the African Union Commission, World Bank, UNESCO, NADMO Ghana, and NASRDA Nigeria highlighted the benefits and challenges of using satellite remote sensing. The interviews revealed that space-based technologies offer valuable information for flood mapping, forecasting, and risk modelling. However, they also face limitations due to low awareness, limited local capacities, and restricted data accessibility. To address these challenges and maximize the benefits of satellite remote sensing in flood management, several key points should be considered.

Firstly, expanding intra-national mechanisms to transfer knowledge and skills from national agencies to local municipalities is crucial. This ensures that frontline responders have the necessary expertise to effectively utilize space-borne products. Secondly, capacity-building and awareness-raising efforts need to be prioritized, requiring adequate resources and funding. It is essential to establish frequent communication and coordination among relevant institutions to achieve synergies and avoid duplicating efforts. Furthermore, collaborations with privately held companies might offer potential for the future. Implementing data-sharing policies would enable disaster management agencies to access remote sensing data with higher resolution from private companies. Cloud computing-based resources can also be utilized to facilitate rapid mapping during humanitarian emergencies. Lastly, a strong focus on prevention is vital for sustainable long-term flood risk management. Space-based technologies allow forecasting and identifying areas that are exposed to an increased flood risk due to their topological and geographic characteristics before disasters strike. These analyses provide a quantitative basis for land-use policies and building codes. Hence, the results support resilient urban planning, mitigate displacement and harm to society, and reduce damage to infrastructure and financial loss. So ultimately, it is not just a question of technological development, but also of political will and the implementation of measures. By addressing these insights gained from a variety of subject matter experts, satellite remote sensing could potentially be leveraged more effectively in managing urban floods in Africa, leading to improved preparedness, response, and long-term resilience.

What do you need to innovate and what would your ideal working environment look like?

In my opinion, the most important factor for an ideal working environment is team dynamics; a healthy balance between interpersonal understanding, a shared ambition for quality work, and a fault-tolerant culture that encourages creativity, exploration, collaboration, and exchange of ideas. Innovation often comes in multidisciplinary settings, so a diverse team with a broad range of skills and expertise can be beneficial. Additionally, an environment that fosters continuous learning and professional development can help individuals stay up-to-date with the latest technologies and trends. Other important factors that can contribute to an ideal working environment include clear communication channels, a strong sense of purpose and direction, a culture of accountability and transparency, and opportunities for feedback and recognition.

What is your favourite aggregate state of water?

My favourite aggregate state of water is liquid because it accommodates a beautiful and mysterious world of its own.