Pastoral livelihoods in northern Kenya face severe consequences of weather extremes that intensify with climate change. Efforts to enhance their resilience include financial mechanisms that compensate households when satellite data trigger forage scarcity warnings. However, flaws exist in such triggers that partially relate to insufficient understanding of the precise link between satellite indicators and household effects. Innovative combination of new-generation satellite series and field data collection efforts can address such flaws. With CGIAR’s ‘Livestock and Climate’ Initiative, the Senior Expert will help develop Earth observation solutions to obtain greater insight into the complex dynamics related to climate, rangeland, and livestock.

Geospatial machine learning (ML) models are widely used in natural and engineering science (NES). These models and the methods to develop them rapidly evolve, making it challenging to keep up with them and reap their benefits. Besides this, many NES researchers do not have the required geospatial knowledge to develop, apply and (re)use these models because “spatial is special”, and they do not know how to document their creative process, making model (re)use unnecessarily hard. To address these issues, we propose developing training modules that increase geospatial ML literacy and geospatial ML models (re)usability.

Het herstel van aangetaste ecosystemen en de bescherming van intacte ecosystemen zijn cruciale acties om de gevolgen van klimaatverandering en het verlies aan biodiversiteit tegen te gaan. Aardobservatie door satellieten kan helpen om de impact van herstel- en instandhoudingsinterventies te evalueren en om beperkte middelen efficiënter te besteden. Helaas ontbreekt het vaak aan praktische kennis van aardobservatiedata en -toepassingen bij betrokkenen in natuurbehoud en -herstel, en ontbreken momenteel gebruikersgerichte toepassingen die de verwerking van aardobservatiegegevens en impact-evaluatie integreren. Wij stellen hier de ontwikkeling van dergelijke, ‘open’ software voor.

Humanitarian organisations are relying more and more on geodata, geographical information science and (geo)artificial intelligence (jointly geo-information workflows). This is not only aiding with the decision making during aid operations after a disaster, but also increasingly for data-informed disaster risk reduction management. Obviously these workflows and the generated information itself also brings risks with it, among others the incorrect or biased classification of people or groups. How to tackle such ethical issues is the central focus of this program.

Many researchers use virtual research environments, such as JupyterLab, where substantial data produced during the whole research lifecycle. However, data publishing and sharing typically happen only at the end of the research and shared data often lack important metadata, mainly due to the need of manual inputs. This project aims to develop and operationalize a tool (JupyterFAIR) for one-click and seamless integration of research environments and data repositories, including metadata transfer and data quality checks. The tool will significantly decrease manual intervention needed to archive research data and promote more frequent data sharing in line with FAIR principles.