The CGSL currently focuses on three types of modelling.

Soil-landscape evolution modelling

Soil-landscape evolution models (SLEMs) simulate how soils and landscapes change, or evolve, over timescales of decades to millennia, due to different soil forming processes and geomorphological processes. SLEMs are increasingly equipped to simulate soil and landscape evolution under changing climatic conditions or under changing land use. By simulating soil and geomorphological processes at the same time, these models can reveal complex interactions or co-evolution in the landscape, that help us to better understand the evolution of soils, or even help us to design sustainable soil and land management strategies.

The CGSL is involved in the development of a soil-landscape evolution model named Lorica. Lorica is a reduced-complexity model, that simulates a set of natural and anthropogenic soil and geomorphological processes. Lorica is a raster-based model, where the relief of the landscape is represented by a digital elevation model. The soils are represented by multiple vertically stacked layers beneath each raster cell. The model simulates the redistribution of sediments of multiple grain sizes and different types of organic matter through space and time.

Recent developments of Lorica focus on additional sets of processes, which we’re previously not included in the model. These extensions are named after the type of processes that were added. For example, HydroLorica includes a detailed simulation of the soil water balance that drives different processes in the model, while ChronoLorica integrates SLEMs and geochronological methods by tracing of geochronological markers through the landscape.

Literature:

• Lorica: Temme, A.J.A.M., Vanwalleghem, T., 2016. LORICA – A new model for linking landscape and soil profile evolution: development and sensitivity analysis. Computers & Geosciences. https://doi.org/10.1016/j.cageo.2015.08.004
• HydroLorica: Van der Meij, W.M., Temme, A.J.A.M., Wallinga, J., Sommer, M., 2020. Modeling soil and landscape evolution – the effect of rainfall and land use change on soil and landscape patterns. SOIL, https://doi.org/10.5194/soil-6-337-2020

Geochronological age modelling

To determine rates and phases of landscape change, soils and sediments can be dated with various methods (for example with luminescence dating). Through this dating, age-depth profiles can be generated. These profiles show how the age of soil or sediment changes with depth below the surface. These age-depth profiles can contain different types of uncertainties. For example, there can be uncertainty due to the measurements in the lab, heterogeneities in the samples that are dated, or mixing processes that disturb the age-depth profiles after deposition of the sediments.

These uncertainties can be reduced using various age modelling methods. For example, the stratigraphical order of the samples can be used to constrain the ages of the different samples, as samples that are located closer to the surface should be younger by definition, because they were deposited later. With other techniques, the age-depth profiles can be corrected for the effect of soil mixing after deposition. The CGSL can help developing and applying statistical and numerical modelling techniques to reduce uncertainty in measured sediment ages and age-depth profiles.

Literature:

• Van der Meij, W.M., Reimann, T., Vornehm, V., Temme, A.J.A.M., Wallinga, J., van Beek, R., Sommer, M., 2019. Reconstructing rates and patterns of colluvial soil redistribution in agrarian (hummocky) landscapes. Earth Surface processes and Landforms, https://doi.org/10.1002/esp.4671

Semi-automated geomorphological mapping

Geomorphological mapping is time-consuming and often difficult to reproduce, because there is a lot of manual labour involved. The CGSL explores novel, semi-automated techniques that can aid in geomorphological mapping exercises. Especially, the use of deep learning shows promising results for the semi-automated development of geomorphological maps.

In the reference below, we explored the use of Convolutional Neural Networks (CNNs) for geomorphological mapping. CNNs are deep learning algorithms that learn to recognize patterns or objects using a series of mathematical operations. CNNs are the driving force behind many image recognition programs, such as automatic face recognition in different social media. We used CNNs to recognize typical landforms on digital elevation models (DEMs). The first results show that CNNs can be a useful aid in geomorphological mapping, by identifying the locations of landforms. However, there is still a lot of development required before CNNs can produce entire semi-automated geomorphological maps.

Literature:

• Van der Meij, W.M., Meijles, E.W., Marcos, D., Harkema, T.T.L., Candel, J.H.J., Maas, G.J., 2022. Comparing geomorphological maps made manually and by deep learning. Earth Surface Processes and Landforms. https://doi.org/10.1002/esp.5305