UAV-derived habitat predictors contribute strongly to understanding avian species–habitat relationships on the Eastern Qinghai-Tibetan Plateau

Andreas Fritz*, Li Li, Ilse Storch, Barbara Koch

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

11 Citations (Scopus)


Species–habitat relationships are of core interest in ecological studies. In this research, we explored the potential of small, unmanned aerial vehicles (UAVs) to derive a comprehensive set of habitat predictors to explain distribution variance of an alpine bird community in Jikdril county, on the eastern Qinghai-Tibetan Plateau (QTP). In 2014 and 2015, we carried out two breeding season bird surveys on 140 plots, from eight sample blocks in the Nyanpo Yutse region. In summer 2014, we conducted 39 flights over the eight sample blocks, and obtained 5500 images which covered 45 km2 of alpine grassland habitat. We used photogrammetry software to generate high-resolution orthophotos and digital surface models with an average pixel size of 13 cm. We derived both 3D- and 2D-UAV-based models as habitat predictors used to analyze explained variance in the local bird community at four different sample scales (plot radii of 50 m, 100 m, 150 m and 200 m). We performed canonical correspondence analyses (CCA), with a model-building approach based on permutation. Furthermore, the proportion of total variance explained by each predictor group was calculated, and analysis of results showed that UAV data derived predictors played a significant role in explaining the variance of the sampled bird community on the QTP. In particular, 3D-based predictors contributed strongly to the explanation of variance (50 m: 45.44%, 100 m: 44.60%, 150 m: 49.21%, 200 m: 36.22% of the total explained variance per sample size), followed by 2D-based predictors over the same four sample scales (50 m: 35.45%, 100 m: 34.57%, 150 m: 31.19%, 200 m: 39.48%). Our results indicate that a sample plot scale of 150–200 m radius leads to the most comprehensive explanation of variance (all predictors: 39.59%, P < 0.0001 and 39.13% P < 0.0001). UAV data provide us with fine-scale (10 cm resolution) continuous raster data on land cover, topographic and landscape features, which are hard to acquire with conventional habitat measurements. We therefore successfully demonstrate the performance of UAV-derived data in facilitating ecological research at different spatial scales.

Original languageEnglish
Pages (from-to)53-65
Number of pages13
JournalRemote Sensing in Ecology and Conservation
Issue number1
Publication statusPublished - Mar 2018
Externally publishedYes


  • 3D habitat features
  • Tibetan avifauna
  • UAV
  • canonical correspondence analyses
  • drones
  • remote sensing

Cite this