National Geographic Relief Shading

Cartographers add a shaded relief image to their maps, because landforms are easy and quick to interpret with shaded relief. While standard algorithms create shaded relief in a fraction of a second, the resulting images do not meet the quality of hand-drawn high-quality shaded relief. Such standard digital shadings are not capable of clearly showing the structure and shapes of landforms and they lack aesthetic expressiveness. In previous research, we have addressed these shortcomings by replicating manually shaded relief with machine learning. We used manual shaded relief imagery created by expert cartographers at the Swiss national mapping agency, whose maps are considered the gold standard for terrain representation. The combination of these shaded relief images with machine learning resulted in the award-winning Eduard app, which enables cartographers to create expressive high-quality relief shadings of alpine mountain ridges as found in Swiss topography (https://eduard.earth). Unfortunately, Eduard is less successful in creating shaded relief for terrain shapes that cannot be found in Switzerland, such as volcanoes, mesas, canyons, sand dunes, repeating rolling hills, or large flat plains. The reason is that these terrain shapes are not depicted in the Swiss relief images that we used for machine learning. The goal of this research project is to develop improved machine learning methods with shaded relief images of these terrain shapes. We aim at inventing machine learning methodology for creating high-quality shaded relief for any typical landform found on Earth. We also aim at extending relief shading with machine learning to multiple map scales for zoomable web maps.