Using high-resolution aerial imagery to assess populations of wintering waterbirds

Full text

4th International Conference on
Progress in Marine Conservation
in Europe 2015
Proceedings of the Conference
Stralsund, Germany, 14 - 18 September 2015
Editors
Henning von Nordheim
Katrin Wollny Goerke

Cover picture: Conference poster (© M. Putze, C. Pfützke, S. Gust, F. Graner, S. Bär)
Editors’ addresses:
Prof. Dr. Henning von Nordheim Federal Agency for Nature Conservation (BfN)
Marine Nature Conservation Department, Isle of Vilm
18581 Putbus, Germany
E-Mail: vilm.marin@bfn.de
Katrin Wollny-Goerke meeresmedien
Kakenhaner Weg 170
22397 Hamburg, Germany
E-Mail: info@meeresmedien.de
Scientific Supervision:
Prof. Dr. Henning von Nordheim Department II 5 „Marine Nature Conservation“
Further information on the actual status and background of marine protected areas under the Habitats
Directive and the Birds Directive of the EU in the German Exclusive Economic Zone (EEZ) can be found on
the BfN web page www.bfn.de/meeresnaturschutz.html.
This publication is included in the literature database “DNL-online” (www.dnl-online.de)
BfN-Skripten are not available in book trade. A pdf version can be downloaded from the internet
at: http://www.bfn.de/0502_skripten.html.
Publisher: Bundesamt für Naturschutz (BfN)
Federal Agency for Nature Conservation
Konstantinstrasse 110
53179 Bonn, Germany
URL: http://www.bfn.de
The publisher takes no guarantee for correctness, details and completeness of statements and views in this
report as well as no guarantee for respecting private rights of third parties. Views expressed in this publica-
tion are those of the authors and do not necessarily represent those of the publisher.
This work with all its parts is protected by copyright. Any use beyond the strict limits of the copyright law
without the con-sent of the publisher is inadmissible and punishable.
Reprint, as well as in extracts, only with permission of Federal Agency for Nature Conservation.
Printed by the printing office of the Federal Ministry for Environment, Nature Conservation, Building and
Nuclear Safety
Printed on 100% recycled paper.
ISBN 978-3-89624-188-7
Bonn, Germany 2016

Using high-resolution aerial imagery to assess populations of
wintering waterbirds
Timothy Coppack1, Alexander Weidauer2, Axel Schulz2, Görres Grenzdörffer3
1 
Kingdom
2 Institute of Applied Ecology GmbH (IfAÖ), Germany
3 University of Rostock, Germany
1 Introduction
Birds inhabit heterogeneous environments across a wide range of temporal and spatial scales.
Migratory waterbirds in particular are highly reactive to changes in climate, food availability and

the marine environment.
The Baltic Sea with its many protected areas holds large numbers of moulting and wintering
waterbirds. While some of the species are listed as “vulnerable” under Annex I of the EU Birds
Directive, e.g. Red-throated Diver (Gavia stellata), most of the waterfowl species listed under
Annex II are currently among those species showing the strongest overall population declines
(European Commission 2015). The Long-tailed Duck (Clangula hyemalis) has recently been
uplisted to “vulnerable” because of an apparent rapid decline detected in the wintering popu-
lation in the Baltic Sea since the 1990s (BirdLife International 2015, see also BelleBauM et al.
2014).
The Natura 2000 network, comprising Special Protection Areas (SPAs) under the EU Birds
Directive and Special Areas of Conservation (SACs) under the EU Habitats Directive, now
covers over 4 % of Europe’s seas (European Commission 2015). SPAs are designated ac-
cording to criteria such as “1 % of the population of listed vulnerable species” or “wetlands of
international importance for migratory waterfowl”. These criteria are currently based on obser-
vational census data collected during ship-based and aerial transect surveys (e.g. MarKOnes
et al. 2015).
Seasonal or monthly surveys of seabirds and waterbirds represent snapshots (random
samples) of local populations that form part of unknown meta-populations. The robustness
        
technique, the design and level of spatial coverage, and on the timing of the surveying effort re-
lative to the phenology of a given species. Observational methods introduce further uncertainty
to the resulting population estimates. Ships (sChWeMMer      
(KuleMeyer et al. 2011) disturb sensitive bird species and thus negatively affect detection rates.
Furthermore, theoretical models correcting for distance-related observer-bias generally assu-
me random distribution of individuals. This is evidently not the case in benthivorous sea ducks
that aggregate in response to the accessibility of their invertebrate food. A biased detection in

estimates with far-reaching consequences for conservation policy.
Recent developments in digital aerial imagery allow a less invasive and safer census of marine
156

wildlife, thereby solving major problems with previous survey methods (BuCKlanD et al. 2012,
Taylor et al. 2014, COPPaCK et al. 2015). Survey data resulting from orthogonal digital images
no longer need to be corrected for distance-related detection bias. The recent switch from ana-

acquisition, data processing and archiving require expensive equipment and experienced staff,
which currently sets limits to the affordable number of surveys per annual cycle. Thus, there is
an urgent need to consider trade-offs between image quality (resolution, signal-to-noise ratio)
-

In this pilot study, we carried out experimental trials based on gapless, vertical imagery of a

sub-sampling to determine the minimum coverage required for quantifying aggregations of
waterbird species relevant to marine conservation and spatial planning, i.e. Common Scoter
(Melanitta nigra), Common Eider (Somateria mollissima), and Long-tailed Duck.
2 Methods
The study area of 46.75 km² was situated in the Bay of Wismar, western Baltic Sea (SPA, UM
M-V 2006), and covered a water-depth gradient ranging from 3 m in the south to about 20 m
in the north. On March 12 2014, a complete survey of the area was performed in 4 h with a
  
included 33 parallel transects of 8.5 km length each oriented in north to south direction and
covering the entire study area. Digital images were collected and stored with a PhaseOne
-
dicularly into the hatch of the aircraft. The aircraft’s position was continuously logged by GPS
(Leica GPS1200), which automatically synchronized the release mechanism of the camera.
Each photo contained 10328 by 7760 pixels and covered an image footprint of 200 x 150 m
at 2 cm ground sample distance (GSD), i.e. each pixel represents an area of 2 cm × 2 cm at
   -
ween consecutive photos. Image overlap between neighbouring transects lay at around 20 %.
The survey took place under suitable weather conditions (wind speed < 5 m / s, sea state < 3,
visibility > 5 km).

GIS environment for further editing, taking the overlapping areas and areas affected by glare
into account (cf. steFFen 2014, COPPaCK et al. 2015). The remaining image strips were visually
screened with a purpose-programmed viewer software by a single trained person. Each detec-

The region of interest was grouped into cells with 38 rows and 49 columns in an east-to-west



statistical parameters over the variation of sampling experiments with a given effort (for ex-
ample 50 %, 33 %, 25 %, 12 %, etc.). An effort of 25 % (1/4), for example, provides 4 variants
for sampling the region of interest and yields 4 average density values, an effort of 12,5 % (1/8)
yields 8 values, and so forth. These spectra of densities over effort were compared with the
157

accurate density values determined at 100 % coverage (i.e. for Common Scoter 93 indiv./km²,
for Common Eider 80 indiv./km², for Long-tailed Duck 43 indiv./km², cf. Figure 1).
In this experiment, we only varied coverage and did not test the outcomes of different sampling

3 Results
Figure 1 shows the relationship between simulated survey effort and estimated densities for
the three duck species Common Scoter, Common Eider, and Long-tailed Duck. A stepwise
reduction of sampling effort (spatial coverage) led to an increase of the variation of calculated
densities. The increase of variation was evident in all species below 50 % (1/2) of total covera-
ge. This effect of sampling effort on the survey outcome was independent of the orientation of

Figure 1:
Top: Grid maps of three sea duck species (Common Eider, Long-tailed Duck, Common Scoter) based on gapless
aerial photos taken on March 12 2014 in the German Baltic Sea (Bay of Wismar).
-
on), which simulated different proportions of spatial coverage from 100 % (1/1) to 10 % (1/10). The dotted horizon-
tal lines show the average densities found at 100 % coverage. The graphs show that sampling efforts below 25 %
may result in systematic under- and over-estimations of population density by factors > 2.
158

4 Discussion

determined and spatially distinct population of wintering waterbirds. A simulated reduction of


such that population densities were either overestimated or underestimated with decreasing
survey effort. This result emphasises the importance of adjusting areal coverage to the expec-
ted frequency distribution of birds before commissioning dedicated aerial surveys. The effect
of areal coverage on the quantitative outcome of a survey is predicted to be especially strong
when the species of interest is non-randomly distributed, like in our case.
Our study has major implications for the future design and implementation of aerial digital
surveys for assessing populations of wintering waterbirds in protected areas. Reducing spatial
coverage for economic reasons and following a traditional transect design (continuous series
of images collected along widely spaced trajectories) increases the chance that a SPA or ag-
gregation of birds is chronically undersampled. Through concentrating and equally spacing
 
values and overall probability densities into account, the relative quantity of images, survey
time and costs could be reduced, while the statistical power and biological meaning of the
surveys would increase.
The minimum technical and methodological requirements for carrying out digital aerial surveys
are subject to ongoing basic research. A sound conceptual framework based on further empiri-
cal trials, e.g. by comparing simultaneous ship-based and aerial observations, will be decisive
for the calibration of observational and camera-based survey techniques in order to evaluate
the backlog of existing data and population estimations.
References
BirDliFe internatiOnal (2015). European Red List of Birds.
BuCKlanD, s. t., Burt, M. l., rexstaD, e. a., MellOr, M., WilliaMs, a. e., WOODWarD, r. (2012).
Aerial surveys of seabirds: the advent of digital methods. Journal of Applied Ecology 49:
960-967.
BelleBauM, J., KuBe, J., sChulZ, a., sKOv, h., WenDeln, h. (2014). Decline of Long-tailed Duck
Clangula hyemalis numbers in the Pomeranian Bay revealed by two different survey me-
thods. Ornis Fennica 91: 129-137.
COPPaCK, t., WeiDauer, a. & KeMPer, g. (2015). Erfassung von Seevogel-und Meeressäugerbe-
-
formatik 1: 358-367.
eurOPean COMMissiOn (2015). The State of Nature in the European Union - Report on the status
of and trends for habitat types and species covered by the Birds and Habitats Directives
for the 2007-2012 period as required under Article 17 of the Habitats Directive and Article
12 of the Birds Directive. Report from the Commission to the Council and the European
159

Parliament, Brussels.
KuleMeyer, C., sChulZ, a., WeiDauer, a., röhrBein, v., sChleiCher, K., FOy, t., grenZDörFFer, g.,
COPPCK, t.   

MarKOnes n., guse, n., BOrKenhagen, K., sChWeMMer h., garthe s. (2015). Seevogel-Moni-
toring 2014 in der deutschen AWZ von Nord- und Ostsee. Bericht für das Bundesamt für
Naturschutz, Vilm.
sChWeMMer, P., MenDel, B., sOnntag, n., DiersChKe, v., garthe, s. 
on seabirds in offshore waters: implications for marine conservation and spatial planning.
Ecological Applications 21: 1851–1860.
steFFen, u. (2014). Entwicklung alternativer Sampling Designs bei der luftbildgestützten See-
vogelzählung unter Berücksichtigung einer GIS-basierten Modellierung am Beispiel von
Meeresenten. Masterarbeit, Universität Rostock.
taylOr, J. K. D., Kenney, r. D., lerOi, D. J., Kraus, s. D. (2014). Automated vertical photogra-
phy for detecting pelagic species in multitaxon aerial surveys. Marine Technology Society
Journal, 48: 36-48.
UM M-V (Umweltministerium Mecklenburg-Vorpommern) (2006). Managementplan für das
FFH-Gebiet DE 1934-302 Wismarbucht (gleichzeitig teilweise Vogelschutzgebiet DE 2034-
401 gemäß Vogelschutz-Richtlinie). Referat Landschaftsplanung, Management der Natura
2000 Gebiete, Schwerin.
160