Author: Richard Bittman
As part of the „Building Bridges between Earth Observation and Environmental Sciences“ project, I recently completed a research exchange at the University of Bonn with Prof. Anna Cord’s team. This collaboration strengthened my doctoral research on avian habitat characterization using remote sensing techniques.

With Prof. Anna Cord
Research Focus
My work addresses two key questions:
- Can continuous multispectral satellite data effectively characterize bird habitats? This approach moves beyond traditional discrete land cover classifications to capture subtle environmental gradients that influence species distribution.
- How do spatial scales and positional errors affect our ability to discriminate between different bird groups? This is particularly relevant when working with occurrence databases containing records with varying degrees of positional errors.
Research Approach
I applied multispectral satellite imagery to characterize the habitat preferences of five ecologically distinct bird groups in the Czech Republic:
- Forest species (e.g., Black Woodpecker, European Robin)
- Aquatic species (e.g., Mallard, Grey Heron)
- Synanthropic species (e.g., Common Swift, Black Redstart)
- Open landscape species (e.g., Eurasian Skylark, Yellowhammer)
- Ecotone species (e.g., European Goldfinch, Great Tit)

Illustrative example of differences in NDVI values between bird groups, analyzed environment and positional error
Using breeding season occurrence data (2014-2017) from the Czech Nature Conservation Database and Landsat 8 imagery, six complementary spectral metrics were analyzed:
- Three vegetation indices: NDVI (vegetation health), MNDWI (water detection), and SATVI (bare soil assessment)
- Three Tasseled Cap transformation components (a powerful technique in remote sensing that converts satellite imagery into composite indices highlighting specific landscape features): Brightness (total reflectance), Greenness (photosynthetically active vegetation), and Wetness (moisture content)
These metrics were analyzed across multiple spatial scales and positional error levels to capture habitat characteristics comprehensively.
Preliminary Research Findings
Smaller cell sizes provided the best overall discriminatory power (up to 400 m). The ability to differentiate between groups remained high for positional errors up to 1 km but dropped sharply beyond this threshold.
Different spectral characteristics showed varying efficacies in discrimination. The NDVI index demonstrated the highest overall discriminatory power, while others such as Tasseled Cap Greenness were less effective but with differences in selected bird groups.
Research Implications
Our preliminary findings indicate:
- The method is applicable and effective
- Positional error thresholds are identifiable
- Smaller spatial scales and lower positional errors generally provide better results
- For habitat discrimination, medium resolution satellite data is often sufficient
- Spectral characteristics should be selected based on specific taxonomic groups
Looking Forward
These preliminary results have important implications for species distribution modeling and habitat characterization, particularly for data containing varying degrees of positional error. During discussions with my Bonn collaborators, we have identified several promising future directions:
- Testing these spectral characteristics as predictors in species distribution models
- Using higher spatial resolution data (e.g., Sentinel-2) for species without distinct habitat preferences
- Validating occurrence records by comparing spectral characteristics with expected values
- Assessing the quality of median composites in remote sensing applications
The exchange experience not only advanced my current research, but also opened new collaborative pathways and boosted my confidence. This ongoing work represents an important step toward more nuanced habitat mapping and biodiversity monitoring using remote-sensing technologies.

With Prof. Anna Cord and her team
Research Exchange Experience
My time in Bonn proved invaluable in bringing my research to completion. I received critical feedback from Prof. Anna Cord and her team that shaped my methodological approach. Regular consultations with my colleagues allowed me to refine my analyses and interpretations. These preliminary results could have important implications for species distribution modeling and habitat characterization, particularly when working with data containing varying degrees of positional error. The exchange experience not only advanced my current research and helped me move toward finalizing the manuscript but also opened new collaborative pathways and boosted my confidence.