by Khalil Teber, PhD candidate at the university of Leipzig in the RSC4Earth group
The 23rd and 24th of February, I organized an online workshop about disaster impact data, and how they could be jointly used with climate data to further improve our understanding of extreme events. Participants from different backgrounds took part in the discussion, and we tried together to highlight the current challenges scientists are facing when dealing with this topic. In the following, you will find more information about the workshop, and a summary of the main points of the discussion.
1 – Context
Climate and weather extreme events are natural hazards that result from unusually severe meteorological or climatological conditions, such as heatwaves, droughts or storms. When they coincide with fragile societal conditions, they turn into disasters, inflicting human and material damages. Often times, multiple disasters can happen simultaneously, or in cascades, which increases their destructiveness. In these cases, we talk of compound events.
Disaster impacts databases are the most valuable observations available for studying climate and weather related extremes and compound events. Only through these records can we study their effects on human well-being and livelihood. These databases are not only important for the detection of the events, but also to deepen our understanding of their underlying mechanisms on two different levels: on one hand, from the compounding interactions on the meteorological and climatological level, and on the other hand, the socioeconomic context that determines vulnerability to them.
The idea of this workshop sparked in a breakout discussion during the third general assembly of the DAMOCLES project on the 27th of October 2021. Given their sparse nature, disaster impact databases have many gaps, unlike meteorological and climatological databases that have continuous spatial and temporal coverage. In addition, they entail many biases depending on the reporting entities and the objectives of the curators of the data, that may range from research and humanitarian work to the insurance sector.
At this point, a survey to collect the different available impact databases (https://damocles.compoundevents.org/report.php?id=1 ) had been published, and many recommendations collected, and work to synthesize this data was already underway. However, this initial database collection and categorisation left important questions unanswered: How could the scientific know-how needed to make the best use of the data brought in? How reach more accessible to fellow researchers and to the general public?
For these reasons, the workshop tried to bring together researchers and practitioners from different fields, and frame the discussion over relevant axes for all users of the data, which are namely:
1. Present overview of different databases and data limitations that describe the diverse sectors affected by extreme and compound events.
2. Review approaches used to couple disaster data with climatological databases and discussion of their limitations and potential ways of improvements.
3. Defining data requirements for climate extremes impacts research.
2 – Description of the workshop
The workshop took place the 23rd and 24th of February 2022. Due to the pandemic, it was held online via Zoom. 21 participants attended the event (List of participants). The majority were climate scientists, except one social scientist, one participant from the insurance sector, and one NGO representative. Given the small number of participants, there were no parallel sessions or breakout rooms, and the whole group was involved in the discussions. The different sessions consisted of a series of talks, followed by group discussions.
On the first day, a sectoral overview of climate extremes using databases that have different spatial and temporal scales were presented. The presentations covered a wide range of weather and climate related disasters, namely:
- Wildfire and tropical cyclone impact models (Samuel Lüthi)
- Climate extremes and impacts in the agricultural sector (Andrea Toreti)
- Marine ecosystem and fisheries’ related disasters (Natcha Legrix)
- Hydro-geomorphological events in Portugal (Alexandre Ramos)
- Climate change and insurance disaster losses (Lucia Bevere)
- Assessing impacts of disasters on armed conflicts (Elisabeth Rosvold)
The second day focused on data collection methods and analysis of their impacts with the following presentations:
- Text mining of natural disaster impacts (Mariana de Brito)
- Approaches for coupling climate and disaster data (Jakob Zscheichler and Fabian Gans)
The keynote lecture “Merging and harmonising climate and disaster data: technical and conceptual obstacles to overcome” by Debarati Guha-Sapir at the end of the second day emphasized the technical and conceptual obstacles one needs to overcome when merging and harmonizing climate and disaster data.
You can find the detailed schedule here (Workshop schedule)
3 – Outcome of the workshop
The presentations of the first day gave a broad overview of the sectors relevant for the analysis of disasters, as well as practical examples of the challenges posed by the used databases for such analysis. In fact, many databases are available, each having its specific challenges in terms of biases, data gaps, etc..
However, the main challenges could be grouped in the following points:
- The historical records of global databases have important discrepancies when compared to local databases. This gap which is common to all global databases was caused mainly by their strict reliance on international sources of information. As penalizing as it might be for historic records, this has improved for the last 3 decades. However, local databases exist only for a small number of countries, and often for a specific class of events.
- The databases have different data entry thresholds: some databases are more restrictive than others and have specific requirements on the damages incurred to include the event. This makes them heterogeneous and comparing their records very difficult. In addition, they rely on different data sources (for example state institutions, NGOs or the media sources)… which increases the uncertainties and biases of their records.
- Geocoding has not been a priority for a long time, which makes recent initiatives in this direction extremely valuable.
From the discussion, it was clear that the limitations of the databases need to be acknowledged and assessed. All freely available databases need to be evaluated in terms of their analysis readiness from a climatological perspective.
In the second part of the workshop, the discussion was axed on different approaches used for analysis in combination with climate data:
- The aggregation of climatological data to the administrative boundaries of disaster location is the most intuitive approach. But this comes at a cost: the aggregation makes certain hazards difficult to detect, such as heatwaves and droughts for example. In addition, statistical aggregates do not always give clear signals, which makes finding overlaps between processes in the biosphere and society difficult.
- Some new approaches exist to collect disaster data with text mining using machine learning and natural language processing, but these approaches are still developing, and it remains unclear when they can be applied in a generalized way. In addition, these approaches cannot solve endemic problems of under-reporting in developing countries. Nonetheless, when possible, they offer exciting new research avenues and applications, e.g., the study of impact chains in real time by extracting locations, quantifying impacts, monitoring the evolution of the reporting.
The most difficult challenge in all impact databases remains missing data, especially the data relative to financial losses and damages. This data is hard to collect in many regions of the world where people are not insured or where there is no comprehensive insurance coverage, and often there is no such data available at all. In general, the impacts of climate and weather related disasters are complex to assess as their effects tend to continue for longer times even during and after humanitarian interventions. This is even more complicated for slow onset disasters such as droughts.
For the time being, a useful contribution to overcome these challenges is to collate the metadata of all freely available impact databases to facilitate the access, comparison and usability.
Another interesting undertaking would be the evaluation of the databases according to the FAIR principles (databases should be findable, accessible, interoperable and reusable). This might not only be useful to the users, but also to the curators of the data.
4 – Next steps
A scientific publication that reviews the main freely available databases, will be produced.
A metadata database of the freely available disaster impact data will be published online soon on the DAMOCLES website.