Abstract

This study analyses the effects of the construction of eight offshore wind farms within the German North Sea between 2009 and 2013 on harbour porpoises (Phocoena phocoena). It combines porpoise monitoring data from passive acoustic monitoring using Porpoise Detectors (POD data 2010-2013) and aerial surveys (2009-2013) with data on noise levels and other piling characteristics. These data were analysed in detail in connection to pile driving activities, most of which occurred with application of noise mitigation techniques in order to reduce disturbance effects.
Prior to investigating piling effects on porpoises, baseline analyses were conducted to identify the seasonal distribution of harbour porpoises in different geographic subareas. Daily POD data and aerial survey data highlighted similar seasonal patterns with higher densities in spring and summer. Highest porpoise occurrence was found next to the SAC Sylt Outer Reef in the northeast of the German Bight in early summer. Another high density area occurred near the SAC Borkum Reef Ground in the southwest almost year round, which is in line with previous findings.
In addition to porpoise monitoring data, noise measurements from the seven wind farms constructed between 2010 and 2013 were combined and noise levels extrapolated where measurements did not exist. Analyses of these measured noise levels revealed that there was high variability within each wind farm. Median noise levels during noise mitigated piling were about 10 dB lower than those measured during unmitigated piling. However, several noise levels measured during noise mitigated piling were as high as those during unmitigated piling and there was a high variability in these measurements within projects ranging over about 20 dB. This high variability probably results from differences in the combination of noise mitigation systems and how well a particular system worked at the time. It may also result from several environmental factors such as water depth, sediment and wind speed that all affect sound propagation. Probably due to these reasons, there was also no clear difference in noise levels between foundation types. The present study shows that noise mitigation systems used during this study were still under development and thus did not always work consistently well.
Establishing the relationship of noise levels to porpoise responses is crucial for environmental impact assessments based on noise prognosis for specific projects. Non-parametric analyses revealed a clear gradient in how much porpoise detections declined at different noise level classes: Compared to a baseline period 25-48 h before piling, porpoise detections declined by over 90% at noise levels above 170 dB, but only by about 25% at noise levels between 145 and 150 dB. Below 145 dB this decline was smaller than 20% and may thus not clearly be related to noise emitted by the piling process. Based on the complete POD-dataset analysed at an hourly resolution using GAM techniques and controlling for other environmental variables, we also found a clear gradient in the decline of porpoise detections during piling with noise level. While a decline in porpoise detections was found at noise levels above 143 dB SEL05, not all porpoises left the noise impacted area at that noise level.
In further analyses, distance from piling was used as a proxy for noise to analyse detailed effect ranges. This was done because it increased the sample size (noise data did not exist for each POD-position and each piling) and model fit using distance instead of noise improved. Analyses pooling all available POD-data yielded an effect range up to 17 km when analysed with General Additive Models (GAM). Non-parametric analyses revealed significant declines in porpoise detections during piling when compared to 25-48 h before in up to 20-30km, but only in up to 10-15km was this decline at least 20%. With increasing distances to the construction site, the magnitude of decline during piling clearly decreased.
When noise mitigation was considered within this GAM model, the estimated effect range of 14km during noise mitigated piling was lower compared to the complete dataset (17km) or unmitigated piling (between 20 and 34km). Caution is required when interpreting these results because of the relatively low dataset for unmitigated piling events. Nevertheless, it shows that noise mitigation effectively reduced porpoise disturbance. This reduction in disturbance may be less than would usually be expected under properly working noise mitigation (when effects may be expected to only reach up to about 5km). This is probably related to the high variability in noise level measurements due to the fact that noise mitigation systems were still under development and did not always work reliable at that time. Considerable improvement has happened since then. Our result that piling noise above 143 dB SEL05 led to disturbance effects in porpoises (even though not all porpoises were affected at these noise levels), supports earlier estimations by Nehls et al. (2016) that properly working sound mitigation, under which 160 dB are not exceeded at a 750m distance (as intended by the regulatory framework), would lead to a substantial reduction of the area in which porpoises are affected by about 90%.
Project-specific models yielded large differences in effect ranges as well as effect magnitude. Declines in detection rates during piling in 0-5km distance were smallest at the wind farm DT with 51% and largest at BARD with 83%. This also applies to effect ranges, which for DT were estimated to be 6km based on GAM models and 0-5km based on non-parametric statistics. During all other projects significant declines by at least 20 % were found in at least 5-10km but occurred in up to 20-30km distance. Such differences between projects cannot be explained by differences in noise levels alone as DT was not significantly quieter than several other projects. Instead it may be linked to a relatively high quality of feeding habitat and a lower motivation of porpoises to leave the noise impacted area, but exact reasons are currently not known.
From aerial survey data there was an indication for porpoise densities to be increased during and up to 12 h after piling at distances above 20km. This effect could not be confirmed by POD-data, which could be related to the smaller spatial coverage of the latter. Elevated densities at distances above 20 km rely on little data, however, and need to be interpreted cautiously.
Effect duration after piling was about 20-31 h at the close vicinity of the construction site (up to 2km) and decreased with increasing distance. Project-specific estimates ranged between 16 and 46 h (when defined as the first local maximum after an initial increase in detection rates), with the exception of DT where effect duration was difficult to define (no local maximum reached).
In all wind farm projects, we observed significant decreases in porpoise detections already prior to piling at distances of up to 10km. This was independent of piling or deterrence measures. The most likely explanation for this are effects by increased shipping activity during preparation works in combination with increased sound propagation at low wind speed. It was found that deterrence effects prior to as well as during piling reached further at lower wind speed indicating that the effects of wind and sea state on sound propagation may be underestimated.
There was no indication for the presence of temporal cumulative effects. Only at BWII we found some indication for potential habituation of porpoises to piling. Neither analyses of hourly nor daily POD-data revealed any further indication for habituation. However, without any knowledge of porpoise residency patterns within the German Bight and individual responses to disturbance, this topic remains difficult to address.
From analyses of daily POD-data there was some indication for piling effects on porpoise detections to differ between seasons: Piling effects were longer lasting during winter and autumn than during spring and summer. As porpoise density tends to be lower in autumn and winter this effect may be related to longer lasting effects at lower porpoise densities. However, this could not be confirmed when looking at area-specific piling effects. Piling effects were not generally longer lasting in areas of lower porpoise densities.
Using results from aerial survey data and POD-data analyses, the PCoD model was applied to estimate disturbance consequences of wind farm construction on the population level. After exploration of the interim PCoD model, several limitations of the model were pointed out that may be improved before providing a realistic estimation for porpoise population trends as a result of disturbance. Applying the PCoD model using conservative input parameters for construction effects arising from the present study (increasing the chances for the model to predict a population decline), the risk of a decline of 1% of the population in the German Bight is estimated to be below 30%. The predicted median decline is below the 1% generally considered as critical for all chosen time periods and varies between 0.9% for the piling period and 0.2% for twelve years after piling had finished.
There were no indications for such a population decline of harbour porpoises over the five year study period arising from analyses of daily POD data and aerial survey data at a larger scale. Despite extensive construction activities over the study period and an increase in these over time, there was no negative trend in acoustic porpoise detections or densities within any of the subareas studied. In some areas, POD-data even detected a positive trend from 2010 to 2013.
On a regional scale, porpoise distribution patterns, as found by aerial survey data, differed between years. These regional changes could partly be related to wind farm construction sites but only within a radius of 20km around piling. However, there was no evidence for an overall change in distribution patterns at a larger scale within the German Bight over the 5-year study period.
Even though clear negative short-term effects (1-2 days in duration) of offshore wind farm construction were found on acoustic porpoise detections and densities, there is currently no indication that harbour porpoises within the German Bight are presently negatively affected by wind farm construction at the population level. This is even though sound mitigation techniques were still under development and further improved after this study period.