An ultra-fine network for rivers

How are species richness and the properties of running water connected worldwide? A team led by IGB scientist Sami Domisch is investigating this question. Researchers have developed the highest-resolution map of the world’s river systems ever made. The map provides a basis for detailed analysis of what characterizes riverine habitats and how they are interconnected.

Rivers are the “lifelines” of all landmasses on earth. This is also shown by the map that Sami Domisch developed together with other researchers from IGB and Yale University: A finely ramified network of potential river sections stretches across all continents. The map is based on the “Hydrography90m” dataset that the researchers spent two and a half years creating on the US university’s supercomputer.

Of course, this map is not the first of its kind. Rivers and their distribution around the world have already been represented in numerous models. All these maps are based on satellite data of topographic relief: wherever there are landscape cracks with certain characteristics, there may also be a watercourse. And yet no other data set is as detailed as hydrography90m.

“We took a high-resolution elevation model of the Earth and extracted the flownet from it using open-source software. In contrast to other previous datasets, Hydrography90m also maps short and very short arms of rivers,” said Sami Domisch. The precision is in the name: the shortest unit is 90 meters long. Since small rivers make up the largest share of the global river network (around 70%), they play a particularly important role in river biodiversity.

A model calculates the probable discharge from environmental parameters

The data set includes a total of 726 million potential river stretches. The term “potential” is crucial in this context: “At first we don’t know where rivers really flow,” says Sami Domisch. The scientist and his team are currently modeling runoff to identify rivers that actually carry water, either year-round or intermittently.

To do this, they use data from 30,000 measuring stations worldwide, at which water volumes have been collected in defined river sections for years. In addition, the researchers have access to extensive data on a wide range of environmental parameters such as precipitation, temperature, land use, soil conditions and slope gradient. In the model, these parameters are related to the measured amounts of water.

“We work with machine learning. This means that with each new data set, our model recognizes better and better which parameter variables are related to which amounts of water,” says Giuseppe Amatulli, first author of the study. If the model works, it can be applied to all river stretches worldwide, even if they do not have a measuring station: in this case, from the available environmental parameters, the model calculates the likely runoff, ie the amount of water in the river for the entire area.

In order to validate the model, the researchers first “feed” it with 70 percent of the existing water volume data sets. The model trained in this way is then given the task of determining the corresponding amounts of water from the environmental parameters of the remaining 30%. If these quantities correspond sufficiently to the actually measured values, the model works correctly – if not, the model can be improved.

However, systematic model deviations can also lead to certain parameters playing an important role for which the researchers have no data, such as water withdrawal by humans. With the adapted model, the discharges of all river sections worldwide can then be determined.

“Especially in dry regions, there are probably far fewer water-bearing rivers than our data set suggests,” says Sami Domisch. This assumption is also supported by a study conducted by authors using the less detailed HydroRIVERS dataset. They estimated that only about 60% of the world’s rivers carry water intermittently or year-round.

The model provides answers to central questions: How long are rivers that carry water permanently or temporarily? Where is there high or low radiance? And what effects does this have on biodiversity? Detailed statements are also possible on questions such as these, since hydrography90m records the catchment areas of river sections on a very small scale.

Since environmental data is already available for each of these 5-hectare catchments, this data can be used to characterize the existence of species communities, eg which climate data or gradients are associated with these communities. For example, various parts of the world have a Mediterranean climate – not just the Mediterranean Basin, but also some parts of the US West Coast.

The analysis of the species composition there allows conclusions to be drawn about the biogeography of these habitats, i.e. which environmental influences contribute to the existence of certain species. And that is just the beginning:

“If we know how much water flows where, we can analyze river habitats worldwide down to the smallest tributary in detail,” says Sami Domisch. Even in parts of the world that are practically inaccessible to humans.

The study was published in Earth System Science Data.

Provided by the Leibniz Institute for Freshwater Ecology and Inland Fisheries (IGB)