Research of RG Shao
The research group of Prof. Yaping Shao is primarily concerned with the modelling and parametrisation of the interactions between the components of the Earth system and the related dynamical processes. Integrated models are developed to provide a framework for multidisciplinary knowledge integration and to become a powerful tool to uncharted areas in Earth system studies. The present research focuses of the group are:
.
Parametrisation of extremely heterogeneous land-surface processes
Land-surface processes have a significant influence on the exchange of energy, mass, and momentum between the atmosphere, biosphere, and lithosphere. We have a relatively good understanding of these processes in areas characterised by homogeneous terrain, but in extremely heterogeneous regions which exhibit numerous different soil types and/or complex orographic features like mountains or steep slopes, our understanding of the land-surface processes is still insufficient and the capacity to parametrise them is limited. State-of-the-art weather models commonly use the Monin-Obukhov similarity theory (MOST) for surface flux parametrisations, which are not suitable for complex terrain. To address these limitations, our research uses an alternative approach for calculating surface fluxes that does not rely on MOST. Using the WRF model, we conduct several simulations with different settings and domains to evaluate this new approach against MOST and compare the results.
.
Soil inorganic carbon (SIC)
Soil inorganic carbon (SIC) is a major component of the global soil carbon pool, with high-stock regions concentrated in arid and semi-arid zones that broadly coincide with major dust source areas. Our research investigates how natural dust activity drives the long-range redistribution of SIC by combining high-resolution soil datasets with regional atmospheric modelling. We are compiling a high-resolution soil texture–SIC database for dust source regions, developing a dust–carbon module for WRF-Chem, and constructing an East Asia–scale SIC budget to quantify dust-driven stock changes in source and deposition areas over the past ~40 years. Building on these efforts, we will use multi-scenario simulations to assess how ecological restoration and climate change alter dust–SIC fluxes and to identify potential secondary ecological risks linked to reduced material inputs.
.
Human and Earth System COupled Research (HESCOR)
The core of our contribution to HESCOR lies within the Our-Way Framework (OWF) of models: We examine how environmental variability influenced migration routes and habitat suitability of genetically modern humans by integrating the Human Existence Potential (HEP) Model with a Human Dispersal Model (HDM). By combining palaeoclimate simulations with archaeological site data within this statistical modelling framework we can simulate spatio-temporal density patterns of possible human presence and potential migration corridors between hotspots of high suitability for survival. For localised simulations we are further developing the constrained agent-based model (CABM) to capture the interactions of individuals rather than densities of populations.
We develop convolutional neural network approaches that learn landscape-scale climatic suitability for early hunter-gatherers and farmers from gridded environmental predictors and their spatial context, enabling suitability mapping at larger spatial extents while keeping the modelling consistent with the broader HEP framework.
One application of the OWF focuses on modelling the climate-driven dispersal of Homo sapiens into East Asia during the Late Pleistocene (80–30 ka). This work provides a quantitative assessment of the role of climate variability in shaping early human expansion into East Asia.
Other applications include early neolithic farming societies in Europe (~5000 BC), the reoccupation of Europe after the Last Glacial Maximum (LGM) by hunter-gatherers, and structured populations ins southern Africa during the middle stone age.
Within the HESCOR initiative at the University of Cologne, a CKAN-based data portal is being developed and operated to support interdisciplinary data discovery, documentation, and reuse. The work includes maintaining the underlying infrastructure and supporting workflows that keep metadata, datasets, and search functionality consistent across the platform.
.
Vegetation Reconstruction
We work on data-driven vegetation modelling for paleoclimate applications, linking bioclimatic and orbital forcing variables to vegetation type and seasonal phenology. Building on earlier machine-learning models for predicting dominant vegetation types, recent work focuses on hybrid ML-parameterised seasonality models to reconstruct large-scale paleovegetation seasonality patterns (“green-wave” dynamics) from modern Earth observation constraints and paleoclimate reconstructions, providing inputs for coupled human–Earth system modelling.
.
Uncertainty quantification
Uncertainty quantification has received increasing attention in atmospheric sciences over the last years. Despite its high attention and importance, accurate estimation of uncertainties in atmospheric data remains a challenge because of their complexity and difficulty of validation.
One approach is a statistical estimation by comparing different datasets over a long period. This approach enables estimation of statistical uncertainties of all datasets and thus is particularly suited for applications where many uncertain datasets are available. In the framework of the HESCOR project, this statistical approach is further developed and applied to paleo-vegetation estimates from different vegetation models and pollen-based records.
In contrast, dynamical approaches can be used to quantify case-dependent uncertainties of dynamical forecast models by exploring their underlying physical principles. While the standard ensemble approach requires several model executions, the parametric approach explicitly propagates the main uncertainty parameters within a single execution of the extended forecast model. The approach was developed as forecast step of the parametric Kalman Filter (PKF) in the context of data assimilation and has recently been applied to quantify complex case-specific uncertainties during a strong wildfire event in eastern Canada in summer 2023 (in collaboration with Environment and Climate Change Canada: Vogel et al., 2025, under review).