Members of the “Data-intensive Systems and Visualization Group” at the Institute of Applied Computer Science of the TU Ilmenau under the leadership of Prof. Mäder develop and run the “Flora Incognita” application, a mobile app that can identify more than 16.000 plant species based on AI algorithms, also created and trained in-house. Plant identifications by citizen scientists create a new source of species occurrence data that enables biodiversity research.

A new paper, published by the Flora Incognita Research Group at the MPI for Biogeochemistry in Jena and the TU Ilmenau, shows that opportunistic plant observations, collected by thousands of plant enthusiasts via the plant identification app Flora Incognita (Mäder et al. 2021) can support phenological monitoring initiatives. Instead of being recorded on purpose, opportunistic plant observations are created when a plant catches a person’s eye and is captured in this instant.
Why is that important? Understanding plant phenology is helping scientists assess (for example) how nature reacts to climate change. Traditional species-specific monitoring of pheno phases (such as bud break, leaf-out, beginning of flowering, and leaf senescence) is done by trained volunteers – and those numbers are steadily declining.


There are many ways to collect phenological data, from remote sensing via satellites to drones and PhenoCams, but also by cameras within the canopy of forests or human observations. (from: Katal et al. 2022)

Can Flora Incognita be used for phenology monitoring?

In Germany, the “official” plant phenology monitoring is mainly conducted by the German Weather Service (Deutscher Wetterdienst, DWD). Here, every observer is given a specific "station," which corresponds to a dedicated spot for the observation of one tree, shrub or herbaceous plant. The number of these stations varies depending on the observed species. This means that some species have more stations than others. During the growing season, observers need to check on the plants they are studying at least two times each week and record the day when specific pheno phases begin. But now there’s a new way to generate onset-of-flowering data of individual plants: The new study’s lead authors Negin Katal and Michael Rzanny found a way to process Flora Incognita data in a way that they resemble DWD stations:

They identified the locations of the DWD stations, and for each species, created a 5 km circle around them. Within this circle, they collected all the Flora Incognita observations for that species within a certain altitude range. At least 35 of those opportunistic records were needed to create a “Flora Incognita station”. If those could not be reached, even within an additional buffer (1 km at a time, up to 55 km), no Flora Incognita station was established at that location.


Process diagram, showing how plant observation data are converted into onset of flowering dates and related to DWD (German meteorological service) observation stations for one examplary species. (from: Katal & Rzanny et al. 2023)

As a next step, for each of those Flora Incognita stations, the onset of flowering in 2020 and 2021 was calculated by fitting the number of observations against their Day-of-Year by using a special statistical method ( a parametric bootstrapping approach based on the Weibull distribution). The percentile closest to the median of the DWD interpolation in 2020 was then chosen as the species-specific onset of flowering for both years. The resulting interpolation maps show similar results to those provided by manual documentation done by the DWD for most species.

Spatially interpolated maps based on the DWD and Flora Incognita stations for the onset of flowering of Sambucus nigra and Taraxacum officinale in 2020 and 2021. The color scale indicates the day of the year of the onset of flowering in each grid cell (from: Katal & Rzanny et al. 2023).

Take-away message
The main finding of the study is that the onset of flowering can be derived from opportunistic plant observations, at least for annual herbaceous species or shrubs with a conspicuous flowering stage. This way, this new source of species-occurring data is able to complement traditionally collected phenology data, especially since those are often based on the observation of trees, which prove not to be suitable for analysis when captured opportunistically. By harnessing unstructured and opportunistic data in this manner, we can make a valuable contribution to quantifying phenological shifts associated with ongoing climatic changes.

We want to thank the many users of Flora Incognita who contribute to this new source of data. It is your curiosity that enables research like this.

The publication is now freely available:
Katal, N., Rzanny, M., Mäder, P., Römermann, C., Wittich, H. C., Boho, D., Musavi, T. & Wäldchen, J. (2023). Bridging the gap: How to adopt opportunistic plant observations for phenology monitoring
Frontiers in Plant Science14.doi: 10.3389/fpls.2023.1150956

Visit the Flora Incognita website for more information on the plant identification app.

Further References:
Katal, N., Rzanny, M., Mäder, P., & Wäldchen, J. (2022). Deep learning in plant phenological research: A systematic literature review. 
Frontiers in Plant Science13

Mäder, P., Boho, D., Rzanny, M., Seeland, M., Wittich, H. C., Deggelmann, A., & Wäldchen, J. (2021). The flora incognita app–interactive plant species identification. Methods in Ecology and Evolution. 12: 1335– 1342.


Prof. Dr. Patrick Mäder (JP)
Department of Computer Science and Automation
Data-intensive Systems and Visualization Group