Skip Ribbon Commands
Skip to main content

Institute of Soil Landscape Research

Institute of Land Use Systems

Institute of Landscape Biogeochemistry

Institute of Landscape Systems Analysis

Institute of Landscape Hydrology

Institute of Socio-Economics

Breadcrumb Navigation

Decrypting patterns in nature with BIG DATA

Hauptinhalt der Seite

​Researchers have spent 20 years gathering an enormous wealth of data in a study area in Brandenburg covering approximately 160 km²: sensors measure data in soil, groundwater and the air, but also observations of animals and studies of plants can help us better understand interactions in the environment. Previously, linking and correctly interpreting the data proved extremely difficult due to the enormous complexity as well as the sheer quantity of data – statements about natural correlations remained vague. An excursion into theoretical physics led a team of researchers to a new methodology and then to a discovery: patterns in the sea of data.

Professor Gunnar Lischeid, Head of the Institute of Landscape Hydrology at the Leibniz Centre for Agricultural Landscape Research (ZALF), did not at first suspect that he would uncover a major error. He is interested in all kinds of water: groundwater, soil water, rivers, streams, ponds and kettle holes as well as everything that is transported in them, primarily fertilisers and nutrients. And so he was astonished to discover that many of the streams in the Uckermark region of Brandenburg were becoming cleaner over the years. The general assumption: »Our farmers are using less fertiliser.« A success for environmental protection? Prof. Lischeid and his team started to investigate.


Big Data in the Uckermark

The data which Prof. Lischeid can draw upon originates from an area studied by ZALF, which is located 90 km north of Berlin in the catchment area of the River Quillow in the Uckermark. Everything which is important for the environment has been measured there since the end of the 1990s: Weather and soil structures, tick and mosquito infestations, weeds and birds, groundwater levels and soil moisture. »Ornithologists, water experts, biogeochemists, landscape researchers and other scientists are gathering data and analysing processes here from entirely different perspectives«, says Prof. Lischeid. Due to the size of the research area, results can be transferred to other regions in Germany. In this way, important statements on interactions in the countryside arise, which are quintessential for environmental protection and nature conservation. »However: we scientists often look only at the data through our own specialist glasses. In order to illustrate correlations and interactions in nature, we have to work with extremely large quantities of data and link them together so that new reliable information emerges.« What had been lacking until now was a way to identify certain recurring patterns, not only in a small data set, but in a very large quantity of data. These patterns could represent the basis of an essential link in nature, which is as yet unknown to us or for which the evidence is missing. 

»Of course we are aware of many correlations in nature. But there are interactions that we have previously been unable to explain«, says Lischeid. Plants that actually require dry soil, are suddenly also growing in moorland. Other plants evaporate almost exactly the same amount of water every year, even though rainfall and average temperatures differ significantly from year to year. Developments are the result of cause, effect and adaptation. If this principle is valid, then there must be patterns for comparable processes which repeat themselves. Certain correlations catch your eye immediately. For others, you require very large quantities of data with as many different measurement variables as possible. Taking a closer look then no longer suffices, neither to simply identify all the links and interactions nor to demonstrate them. »I was therefore looking for an approach to track down the creative forces of nature in the flood of data.« A coincidence came to his aid in his search for clues.​


​Breaking new ground in the field of environmental research

»A former colleague of mine is a theoretical physicist. He gave me an insight into modern methods of analysis of large data sets, which have been used in physics for a long time.« Big Data is the key word: The term summarizes data which is too complex, too big or too fast moving to process with conventional computer systems and methods. For this, new methods and technologies are required. In business, for example, the data of hundreds of millions of Internet users is recorded simultaneously, linked and condensed into personal profiles within seconds. A complex algorithm calculates which advertising is shown to us during our visits to web pages, for example, where we might plan our next holiday. Thousands of pieces of measurement data feed mathematical models for day-today weather forecasts, help biologists to understand the construction of a cell and are used in physics for the representation of atoms. Fascinated by these approaches, the team led by Prof. Lischeid begins to employ these methods, thereby breaking new ground in environmental research. A first test case is already on his table: the changes in ground water quality in the Uckermark. 

Prof. Lischeid’s search for patterns begins. He and his team collect 2449 water samples from the streams, small natural ponds, so-called kettle holes, and from the groundwater in the Quillow area. A total of 96 different water bodies are examined for twelve measurement variables: the pH value, electrical conductivity, sulphate, nitrogen, chloride, phosphate, sodium, potassium, magnesium, calcium, ammonia content and organic carbon concentration: an entire ocean of figures. »Although environmental processes are very complex, they are often dominated by a small number of key processes. In this case I discovered that essentially the same five processes take place in these 96 measurement sites, albeit to varying degrees and with different time frames.« His resourceful companion: the computer. However the computer has one handicap – it is no great help in recognising patterns. This fascinating ability of our brain remains thus far largely closed to the computer. 

Prof. Lischeid therefore familiarised himself with »Big Data« approaches to data analysis and came across the »SOM-SM method«. This combines the processing power of a computer with the human capacity to recognize patterns. Each of the 2449 water samples is represented by a point on a diagram. Water samples with very similar values for all 12 measurement variables plot very close to each other. Those that are very different are placed far away from each other. A cloud of points of differing density emerges. Even at first glance patterns are recognizable: The location of points in the figure reveals a lot about which samples are similar, which ones display a typical pattern and which groups of measurement sites are different. The decisive factor is that these patterns are determined by all twelve measurement variables. Using different colors in the same figure, the values of individual measurement variables or individual measuring sites can be shown and changes to individual measurement variables over time can be seen. For the human brain it is very helpful that the position of the points in the figure always remains the same, only the colors of the points change. »In this way even very large data sets can be investigated very efficiently with the aid of computers« says Lischeid.


Amazing Evidence

Using these »Big Data« approaches, the alleged success story of the clean rivers and streams was ultimately revised: the researchers were able to show a gradual change in water quality in the streams, however not in the groundwater and so they became suspicious. A second feature in the clouds of points brought them onto the right track: the groundwater levels had dropped in comparison to the previous years. The reason for the improvement in the quality of water was not a reduced use of fertilisers, but the weather. In recent years which had been warm and dry, only little heavily contaminated water from the arable land had found its way into the streams. During this time they were fed with less polluted groundwater from a greater depth. »It had nothing to do with the fact that farmers were working differently, but simply with the different water levels. Using our procedures, we were finally able to scientifically prove this.« 

But this was not the only error Prof. Lischeid was able to rectify using the new models. »In various kettle holes in Brandenburg and in Mecklenburg- Western Pomerania, plant protection products were found which farmers in the catchment areas we investigated claimed they had never used. Suspicions arose immediately: Are the farmers cheating? »Our analyzes were able to absolve them: the kettle holes are in contact with the groundwater, plant protection products are often transported for kilometres under the earth.« Misjudgements such as this in the field of environmental protection can have devastating consequences – for farmers, for nature, for human beings. »Using ›Big Data‹ approaches, we can investigate the effects of interventions in the environment, e.g. through nature conservation, and explore and identify the causes of changes in a much more differentiated way. Today we have the technical capacity to change the material flows and habitats of plants and animals to a degree that threatens the foundations of human life. Environmental research can use these new approaches to uncover correlations and draw attention to developments before they become irreversible.»​


Prof. Dr. Gunnar Lischeid
researches water – in all its facets and interactions. He is Head of the Institute of Landscape Hydrology at ZALF and holds an eponymous professorship at the University of Potsdam. Following studies of Agriculture and Geology at the Universities of Bonn and Göttingen, he obtained his doctorate in Forestry Sciences and qualified as a professor in the Department of Hydrology.​


More i​nformation:



Fusszeile der Seite
© Leibniz-Zentrum für Agrarlandschaftsforschung (ZALF) e. V. Müncheberg