Analyzing multi-sectoral drought impacts: a mixed-methods approach combining knowledge- and data-driven insights

Droughts have far-reaching impacts on society across social, economic, and ecological di­mensions. In Europe, recent droughts caused severe damage in sectors such as forestry and agriculture. Understanding how droughts impact societies across different sectors is required for managing and adapting to these events. Specifically, it is imperative to assess multi-sectoral impact patterns that emerge from cascading and compounding impacts (CCI). This dissertation investigates multi-sectoral drought impact patterns, particularly the rela­tionships between impacts stemming from CCI. Specifically, it aims to address two challenges: missing multi-sectoral drought impact datasets and an empirical understanding of the rela­tionships between impacted sectors. Therefore, I developed a mixed-methods approach that combines a data-driven approach using novel text mining and a knowledge-driven approach based on participatory modeling. For assessing multi-sectoral drought impacts, this dissertation developed a text mining­-based approach to extract different drought impact types from newspaper articles. Natural language processing techniques served to infer information on the reported impact type and its location from over 50,000 newspaper articles. The resulting dataset describes the magnitude of reported impacts in Germany from 2000 to 2022 at a district level. Validation showed that the dataset reflects indicators such as precipitation deficit or crop yield losses. Two case studies demonstrate the applicability of this data for understanding multi-sectoral drought. Impact patterns at large spatial-temporal scales. The first case study compared the impacts of the 2018-2022 multi-year drought in Germany to single-year droughts. The study shows that multi-year droughts hold legacy and cascading effects that change the impacts experienced in regions throughout multiple years, making them different from single-year droughts. The second case study modeled the occurrence of flash droughts in Germany, demonstrating that reported drought impacts can complement the identification of flash droughts. A knowledge-driven approach using participatory modeling complements the data-driven insights in this dissertation on CCI patterns. Combining the individual knowledge of stakeholders affected by drought impacts in the German federal state of Thuringia, this study demonstrates how this knowledge-driven approach complements the data-driven approach by providing a more fine-grained representation and understanding of CCI. More specifically, it reveals that combining the stakeholder knowledge makes complex features of CCI explicit (such as feedback loops) that the data-driven approach could not detect. In summary, this dissertation innovatively combines text mining and participatory modeling to provide a robust understanding of the complexity of multi-sectoral CCI of drought. Within such methodological pluralism, combining both approaches enhances an understanding of large-scale spatio-temporal patterns and more detailed mechanisms underlying the CCI patterns. The data-driven approach leverages multi-sectoral impact data from text mining to detect patterns on an extensive spatial and temporal scale. Using stakeholders’ expert knowledge via participatory modeling allows for a complementing, more fine-grained investigation of the underlying CCI. This approach can help practitioners monitor drought impacts using text mining and understand regional differences in impact patterns.