Edited by: Wen-Cheng Wang, National Taiwan Normal University, Taiwan
Reviewed by: Haavard Rue, King Abdullah University of Science and Technology, Saudi Arabia; Arliss Winship, Consolidated Safety Services, United States
†Present address: Jochen Bellebaum, Sächsische Vogelschutzwarte, Neschwitz, Germany
This article was submitted to Marine Conservation and Sustainability, a section of the journal Frontiers in Marine Science
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
The utilization of marine renewable energies such as offshore wind farming leads to globally expanding human activities in marine habitats. While knowledge on the responses to offshore wind farms and associated shipping traffic is accumulating now at a fast pace, it becomes important to assess the population impacts on species affected by those activities. In the North Sea, the protected diver species Red-throated Diver (
Worldwide, efforts are made to increase the production of renewable energy to reduce the reliance on fossil fuels and nuclear energy. The expansion of offshore wind farms is one of the main pillars to achieve decarbonisation in Europe, and as part of a concerted effort, growing numbers of offshore wind farms are also recorded in the German North Sea (
Members of the diver family (Gaviidae) are among the seabird species most susceptible to human disturbance in the North Sea (
The two diver species that most commonly occur in the study area, Red-throated Diver (
In the German North Sea, the highest spring densities of divers are found in the Eastern part of the German Bight (
For the licensing of offshore wind farms in the German North Sea Exclusive Economic Zone (EEZ), a standard concept conceived and coordinated by the BSH was developed for monitoring the environmental impacts of offshore wind farms, before, during and after construction, including regular aerial surveys of seabirds (
However, as biological systems are always a complex interplay of many factors working on different spatial-temporal levels, population trends are difficult to measure, calculate and predict. Especially in seasonal migrants, it is important to consider the entire migratory life cycle in determining the population dynamics.
In avian populations, consequences of anthropogenic activities could arise from increased mortality during the non-breeding season. Avoidance of suitable habitat could lead to reduced body condition or increased mortality if important foraging habitat is lost (
Thus, an optimal statistical framework is required to get an accurate estimate of the diver population trend during the non-breeding season integrating different data sources. We make use of a Bayesian spatio-temporal hierarchical model, applying latent Gaussian models (LGMs) with a flexible stochastic partial differential equation (SPDE) approach to model spatial structure in density and integrated nested Laplace approximation (INLA;
On this basis, our main objective was to develop a statistical framework to reliably estimate bird populations from a dataset collected using different aerial survey techniques and to assess changes in spatial distribution that might be related to offshore wind farm development. We applied this framework to estimate the diver population size in the German North Sea and the MCA over the 18-year study period and examine whether there are any changes in the population that might be related to the increase in the number of offshore wind farms since 2009.
The study area covers 28,625 km2 of the German North Sea, excluding the western-most region of the German EEZ due to poor data coverage and low diver density in that area (
Location map, showing the total prediction area (brown line) regarding the German EEZ border (gray line). The main concentration area of divers (
Available data from ship transect surveys was not included due to the disturbance effect of ships on divers (
Over the years, the data collection methodology improved, so there were two main types of aerial surveys available during the studied period: visual observer flight surveys and digital aerial surveys. Since surveys were carried out on different dates each year, there is a varying amount of effort associated with the seasons of each year. All surveys were restricted to favorable weather conditions with a sea state less than 4 (visual surveys) and 5 (digital surveys) and no clouds below flight altitude.
Observer-based visual survey flights were conducted as part of wind farm monitoring from 2001 to 2013, and followed the methods described by
Flights were performed at a height of ca. 250 ft (76 m) and a speed of around 180 km/h and observations were divided into distance bands based on their distance to the transect line using an inclinometer.
The detectability of seabirds decreases with increasing distance from the survey platform. In order not to underestimate the density of birds in the most distant transect bands with lower probability of detection, an effective strip width (ESW) was calculated, which is smaller than the total transect width (
Digital aerial surveys (
The APEM system (
The recorded footage (video or stills) was evaluated by professional ornithologists, with a separate step for random sample quality control. Flight height in digital surveys was greater than in visual survey flights, so survey aircrafts could fly above the wind turbines and disturbance to birds was minimized. Precise geographical positions of each observation were recorded using GPS technology. Although survey flights were generally only conducted during favorable weather conditions, parameters such as sea state, glare, cloud cover, air and water turbidity were recorded, and pictures of insufficient quality were excluded from analysis.
As there were surveys from different digital methods and a visual survey technique performed simultaneously in the same area, it was possible to include the differences in detection rate associated with each technique within the model approach, assuming a perfect detection rate for the HiDef system (
For divers, identification to species level is rather difficult using aerial surveys. For both methods (digital and visual) a significant part of all individuals was only identified as diver sp. Analyses were therefore conducted including all individuals observed. However, from previous studies (e.g.,
In total, 56 surveys were performed using conventional (visual) methods, and 84 surveys using digital methods. During the period 2001–2018, 16 years of data were analyzed since no data were available during the spring season for the years 2006 and 2007 (
Survey effort by data collection method used for analyses between 2001 and 2018 during the spring season. Conventional and digital flight tracks are shown in green and blue, respectively, offshore wind farms are depicted in red and the main concentration area of divers with yellow border and dark gray background. As reference, the German EEZ contour is shown as a black line.
Data coverage was rather low for the first years (2001–2005), but nevertheless, the north-eastern part with the highest density of divers was covered in all years except 2005 and 2009. From 2008 onward, coverage was high in most years, and in several years the study area was almost completely covered (>90%).
To capture the general population trend and for computational convenience, a constrained refined Delaunay triangulation spatial mesh (
Spatial mesh used for the spatial-temporal model. Main diver concentration area (gray area) and SPA “Eastern German Bight” (green line) are depicted.
An explicit spatio-temporal Bayesian hierarchical model was used to incorporate all these dependencies and to assess changes in bird distribution, where dependences were incorporated as Gaussian Random Markov Fields (GRMF) (
where
different at each
Where
For the Mátern model, Penalized Complexity priors (PC-priors;
Finally, prediction points were masked according to the desired prediction area (German EEZ and MCA) and the total estimated abundance was calculated by year and prediction area together with the 95% credible intervals.
Mean abundances and credible intervals within each of the time periods considered (e.g., 2001–2005, 2002–2012, and 2013–2018) were calculated using independent models. For this, effort and observations within each period were aggregated and the temporal component (e.g., year) was removed to obtain reliable mean abundances and margins of error for each period.
To evaluate the model’s predictive performance, mesh nodes were randomly split into two subsets: a training dataset including 80% of the nodes, and a validation dataset containing the remaining 20% of the nodes. The model was run on the training dataset and its predictive accuracy for each year was assessed on the validation dataset. This calibration-validation procedure was repeated 20 times, and for each run the performance of the model was assessed using the Pearson’s correlation coefficient (r) between the observed abundances and predicted posterior means at the testing locations.
All model calculations were performed using the R statistical software (
Compared to the reference method HiDef, DAISI showed no significant differences, while visual surveys (after applying distance sampling) showed significantly lower detections, and APEM showed the highest variability (
The overall accuracy of the model (r) was 0.71 (
Model cross-validation results. Dashed horizontal red line indicates average
No clear pattern was detected in the posterior estimates of the annual diver abundances for the study area (
Posterior estimation of annual diver abundance and total number of wind farms built every year during spring:
The overall pattern of interannual abundance was closely related to the pattern of abundance observed in the MCA (
No apparent relationship was found between diver abundance and the number of OWFs constructed in the MCA, although six OWFs had been constructed in this area between 2012 and 2018. Excluding the years prior to 2005, which showed high uncertainty in the posterior abundance estimates given the low spatial coverage sampled, the years with the highest estimated abundances in the MCA were 2016 [posterior mean = 13,149 individuals; 95% CI = (11,594–15,524)], 2014 [posterior mean = 12,068 individuals; 95% CI = (10,330–13,848)] and 2018 [posterior mean = 11,881 individuals; 95% CI = (10,132–13,607)]. See
Densities predicted from the spatio-temporal model (see
Mean posterior spatial distribution for divers during the spring season for year 2010, before the development of the offshore wind farms in the northern area (upper panel) and year 2018, under the current development of offshore wind farms (lower panel). Red polygons indicate offshore wind farms already constructed at the beginning of each period. Diver main concentration area (
Species monitoring programmes that run for years result in point-referenced spatio-temporal datasets that tend to be correlated spatially and temporally, but the subsequent ecological processes are almost never fully explained by the environmental variables collected.
Ignoring these dependencies, as most widely popular generalized linear and additive models (GLMs and GAMs) do, not only reduces their predictive ability, but can also lead to incorrect results (
In this study, we performed a Bayesian spatio-temporal hierarchical model with a progressive spatio-temporal distribution structure (
Other key advantages of this modeling framework are:
Bayesian models can incorporate our knowledge of the unknown parameters governing species behavior, expressed through probability distributions, rather than just fixed estimates, as in frequentist approaches (
The Delaunay triangulation allows more information to be collected in areas with more observations, which contributes to more accurate predictions. It is also less computationally demanding and takes into account the boundary effect (
Both, space and time are treated continuously, incorporating intrinsic components to account for spatial and temporal autocorrelation (
The application of this statistical framework to this long-term dataset, offers the opportunity to uncover changes in the diver population that could be related to the offshore wind energy development in these waters over the last ten years, and five years of offshore wind farm expansion within the MCA.
Despite these major changes in the environment, our results showed that the number of divers staging in the German North Sea during spring fluctuated between years but on average remained on the same level throughout the study period.
The NW-European flyway population of Red-throated Divers is estimated at 150,000–450,000 (
The mean spring population for the German North Sea was estimated by the present study at 16,330 divers (95% CI: 15,088–17,912) during 2013–2018 when most of the wind farms were built, and 15,942 divers (95% CI: 14,836–17,176) during 2002–2012 (excluding the years 2006 and 2007) with few or no wind farms built. Our estimates for the early years are thus in the range of the study by
Seabird species vary strongly in their sensitivity to OWFs (
Assuming a gradual avoidance for divers of about 10 km around wind farms (
It is also noteworthy that diver numbers in the MCA stayed relatively constant during the study period, confirming the importance of the 7,000 km2 area which was identified by
Even with an optimal statistical framework, the causes of a population change are usually hard to investigate, especially in migratory seabirds, and impacts might occur during wintering, migration or on the breeding grounds (e.g.,
Since the number of divers has not decreased but is distributed over a smaller area, as they avoid the wind farms and their immediate surroundings, the density within the unaffected areas has increased accordingly. It might be argued that a carrying capacity limit could be reached in future (
Only if alternative staging areas are unavailable, carry-over effects leading to reduced survival or breeding success might be expected (e.g.,
In this context, it must be considered that individuals are not stationary, but moving over large distances during winter and spring. Recently, a telemetry study (
This study highlights the advantages of using spatio-temporal approaches to assess population trends and bird distributions and their application on environmental impact studies. Standardized monitoring programmes need to be maintained to assess the long-term trend of avian populations. Ain the case of the diver population in the German North Sea, although the population shows no signs of decline so far, the long-term effects of disturbance by offshore wind farms on divers are still unknown.
To get more insight into the importance of the main concentration area for divers, staging duration and turn-over rates should be included in future analyses. Monitoring data from adjacent areas (e.g., Denmark) could further improve population estimates and the detection of population changes.
The datasets presented in this study can be found in the Marine Data Archive repository at
All authors designed the study. LS, AF, AB, JB, BB, and RV processed the visual and digital aerial survey data. FB and RV performed the model analysis. CB, AD, and RV led the writing of the manuscript. All authors contributed substantially to the drafts and gave final approval for publication.
RV, CB, FB, AD, and GN were employed by the company BioConsult SH GmbH & Co., KG. LS, AF, and AB were employed by the company IBL Umweltplanung GmbH. JB, BB, and WP were employed by the company Institut für Angewandte Ökosystemforschung GmbH.
We would like to thank the Federal Agency for Nature Conservation (BfN) at Vilm, Landesbetrieb für Küstenschutz, Nationalpark und Meeresschutz Schleswig-Holstein (
The Supplementary Material for this article can be found online at: