Christopher Schiller, a graduate of the Karlsruhe Institute of Technology (KIT) and now a PhD student at the Freie Universität Berlin, has been named one of two winners of the 2022 Young Researchers Award.
An expert jury has recognized Schiller for developing a novel demonstration of the potential for combining big data from professional and citizen science with machine-learning models to automate global-scale assessments of plant functional diversity. Nominated by the German delegation to GBIF, Schiller is the first German national to receive the award.
Plant functional traits are essential to understanding and assessing biodiversity and ecosystem processes. However, the demands and difficulty of directly observing and measuring them severely limits the available data. In hope of filling this gap, Schiller proposed to explore the potential of applying deep learning-based pattern recognition to hundreds of thousands of photographs taken by citizen scientists to predict and reveal the expression of traits hidden in plain sight.
Schiller started by using species names to link plant images and coordinates from iNaturalist Research-grade Observations via GBIF.org to expert trait observations from the TRY Plant Trait Database. With these connections established, the pattern-recognition models could start learning plant features visible (or hidden) in the photos to generate predictions for traits like leaf area, growth height, seed mass and leaf nitrogen content.
The prediction of several key traits only based on image features produced global trait maps that reflect general global macroecological patterns across growth forms and biomes. The introduction of ensemble models and prior knowledge on trait plasticity and climate even enhanced the results.
"To our surprise, just as a trained ecologist could likely approximate these traits when looking at photographs of plants and their organs, the initial models produced remarkably accurate predictions," said Schiller. "The results reveal another way in which information gathered by citizen scientists can scale up Big Data sources that improve our scientific understanding of biodiversity."
"Christopher's work reveals another area where citizen scientists can contribute to macroecological and biogeographical research," said Sebastian Schmidtlein, professor and head of the vegetation research group in the Institute of Geography and Geoecology, KIT. "I think it provides a promising perspective and excellent example of what the data gathered by GBIF and the new possibilities brought by Deep Learning can do for science."
"Christopher pursued a demanding research programme that dealt with heterogeneous, multidimensional data, complex ecological relationships and exacting state-of-the-art methods of data science," said Teja Kattenborn, a researcher in the Universität Leipzig's Remote Sensing Centre for Earth System Research who served as Schiller's thesis advisor while previously at KIT. "It's a sign of his achievement that this groundbreaking study culminated in a peer-reviewed publication in Nature Scientific Reports just months after he submitted his thesis."
After earning a Master's in Geoecology with distinction from KIT, Schilling has now started PhD research in Remote Sensing and Geoinformatics at the Institute of Geographical Sciences of the Freie Universität Berlin.
About the Award
About the Karlsruhe Institute of Technology
KIT creates and imparts knowledge for the society and the environment as “The Research University in the Helmholtz Association”, where more than 5,500 scientists cooperate in a broad range of disciplines in natural and engineering sciences, economics, the humanities, and social sciences. Learn more.
About Freie Universität Berlin
Freie Universität Berlin is a leading research institution whose 16 academic departments and central institutes maintain more than 100 international partnerships and offer over 150 degree programs across a wide range of subjects. Learn more.