Report of the Round Table Session
Erbe, C.1*, Folegot, T.2*, Aulanier, F.3, Binnerts, B.4, Jiménez, G. 5, Risch, D.6, and Simon, L.7
1 Centre for Marine Science & Technology, Curtin University, Perth, Western Australia
2 Quiet-Oceans, France
3 Fisheries & Oceans Canada, Canada
4 TNO, The Netherlands
5 Seiche Ltd. / University of Bath
6 Scottish Association for Marine Science (SAMS-UHI), UK
7 RTSYS, France
This report can be referenced as:
Erbe, C., Folegot, T., Aulanier, F., Binnerts, B., Jiménez, J., Risch, D., and Simon, L. (2015). Report of the Mapping and Modelling Session, oceanoise2017, Vilanova i la Geltrú, Barcelona, Spain, 10-15 May. (Editors Michel André & Peter Sigray). Retrieved from https://2023.oceanoise.com
The purpose of modelling and mapping
The prediction of noise emission from anthropogenic activities, the modelling of noise footprints of individual operations, and the mapping of anthropogenic noise from single or multiple events are becoming routine in environmental impact assessments (EIA) of coastal and marine operations. Noise modelling and mapping is used as a predictive tool to show areas of potential risk of bioacoustic impacts, and to compare predicted levels to set thresholds for acoustic exposure. Noise modelling and mapping is a tool for EIAs, and hence for noise management, and is used to assess “good environmental status” (GES). Noise maps can further be used in monitoring and mitigation, including real-time prediction and monitoring. Predictive noise modelling and mapping is sometimes used to test the efficacy of mitigation methods. Also, noise maps, if overlain with habitat maps, can show areas of risk (overlap), as well as areas of opportunity (lack of overlap), and hence inform marine spatial planning and conservation management. Last but not least, noise maps serve a communication function, aiding information exchange and communication amongst stakeholders in the marine environment.
Examples of noise mapping projects
Several “large” (in terms of financial investment, level of effort, number of collaborators, geographic scale, and expected outcomes) mapping projects have been initiated in recent years. These include “Listening to the Deep-Ocean Environment” (LIDO, http://www.listentothedeep.com), NOAA’s “Cetacean and Sound Mapping” (CetSound, http://cetsound.noaa.gov), “Achieve Quieter Oceans” (AQUO, http://www.aquo.eu), “Baltic Sea Information on the Acoustic Soundscape” (BIAS, https://biasproject.wordpress.com), and “Suppression of Underwater Noise Induced by Cavitation” (SONIC, http://www.sonic-project.eu).
Noise sources that are mapped
Noise maps typically focus on anthropogenic sources such as ship traffic, seismic surveys, and pile driving. They rarely also include ambient noise levels, e.g. from wind or animal choruses. Combining abiotic, biotic and anthropogenic sources in noise maps would yield a more realistic representation of the soundscape and put anthropogenic footprints into perspective. However, this is often not the purpose of the mapping exercise. The purpose drives what is mapped and how. There are no agreed methods, let alone standards, for the production and display of noise maps, making it difficult to compare maps.
How to produce noise maps
Noise maps are produced by gridding a geographic area or habitat of concern, positioning all noise sources, and modelling sound propagation from all sources to all possible receiver locations.
The specific quantities that are mapped depend on the purpose of the map. For EIAs, the maps typically display cumulative sound exposure levels (CSEL) or root-mean-square sound pressure levels (SPLrms), as these quantities appear in many countries’ noise regulations. If the question is how often certain thresholds are exceeded, the map displays the probability of reaching specific levels.
Maps are most commonly static, in the sense that they display levels reached over a period in time at a grid of stationary receivers. Such maps apply to bioacoustic impact assessments of sessile or benthic animals, or animals confined in space. Such maps are also useful to highlight consistent problem or risk areas.
For migratory animals, or animals that move across maps within the time frame of the study, or animals that change movement in response to noise, it would be best to model the (variable) received levels at the animal, followed by a statistical approach to estimate bioacoustic impact on the population or group of animals. In this case, static maps that show long-term averages or cumulative levels in space are not useful. Rather, a whole series of maps would be needed to track received levels at the animals. Or, maps should be dynamic, as opposed to static, e.g. displayed as a movie. Agent-based models populate a habitat with animats and track received levels at individual animats, who behave and respond according to a set of behavioural rules. A lot of information about the animals’ behaviour in response to a multi-variate environment is needed in order to yield representative behavioural rules and ultimately sensible model outputs.
A decision on the resolution of the noise maps has to be made at the beginning of the modelling and mapping exercise. It would be computationally prohibitive to model global ship noise at 100 m resolution. For EIAs of local operations, a fine resolution is required and the area to model is likely small. For large-scale marine spatial planning, a larger area is modelled and the resolution likely doesn’t have to be as fine. In particular near the source, levels change a lot over short ranges. The resolution affects the displayed values. With a grid of, e.g., 1 km, the displayed received level in a cell that contains a source might be several 10s of dB less than it would be on a 10 m grid.
Maps are inherently two-dimensional, printed images. Depth-information is mostly not displayed. Animals that dive experience different received levels as a function of depth. Maps can be produced for different depths. Or, in EIAs, in order to be conservative, received levels are commonly maximised over depth. In dynamic maps, or agent-based models, variability of the noise field with depth can more easily be included.
The temporal resolution is another a priori decision. For EIAs, noise should be mapped only for the days or seasons that animals are expected. For highly mobile animals, a fine temporal resolution is necessary (e.g. in the form of dynamic maps). For real-time mitigation and monitoring, maps need to be “quick” and updateable at small steps in time. Regulations in some countries dictate the temporal resolution, with some impact thresholds for pulsed sound set per pulse or cumulative over 24 h.
Frequency resolution is another parameter to consider. Not all animals are responsive to the same band of frequencies. Animals are more affected at certain frequencies than others.
Overall, there are five map parameters whose resolution needs to be set up front: longitude, latitude, depth, time and frequency. The purpose of the map determines the resolutions chosen. If used for EIAs, then regulatory requirements and, last but not least, the target species will affect the selection of parameters and their resolution. Ultimately, it would be sensible to display noise maps in terms of quantities and resolutions that relate to the way sound is perceived by or affects animals.
Common input parameters
The input parameters needed to produce noise maps are those that are used in common noise models. They relate to the hydroacoustic characteristics of the water column (temperature, salinity,…), geoacoustic properties of the seafloor (density, attenuation, …), bathymetry, and potentially additional parameters (sea surface roughness, ice cover, …). There are public databases for some of these environmental parameters. In addition, much more detailed information (with better temporal or spatial resolution) often exists within government or industry, not publicly accessible. Regulators sometimes require that data be made publicly available. Occasionally, data are kept proprietary for a certain number of years only, and become available afterwards. Ambient noise (e.g. wind-dependent noise) could be included in noise maps, but is often not considered. To map anthropogenic noise, input parameters related to the source type and location are needed, e.g. ship AIS data including ship type, size, draught, speed etc.
All of the input parameters and the noise model have a level of uncertainty, which determines the uncertainty of the output noise map. An error analysis should always be done. To assess the uncertainty of the noise map, a sensitivity analysis can be performed based on the variability of each parameter. Most commonly, a single realisation of the model and map are presented. It would not be difficult to only show data (e.g. noise levels in geographic cells) that have less than a threshold value in uncertainty. Or, uncertainty itself can be plotted in a map. This might highlight areas where errors could be large and where more measurements in combination with modelling need to be done. Maps that show percentiles or probabilities of certain levels rather than long-term averages may be preferable in EIA scenarios.
Noise modelling and mapping typically occurs prior to operations. The outputs are often validated in situ by recording the received sound on moored, drifting or towed hydrophones. For large-scale mapping projects, ocean observatories such as CTBTO provide an opportunity for validation of noise maps. In situ recordings can be used to improve models. E.g., the geoacoustic properties of the seafloor are often the largest source of uncertainty in noise models, and in situ recordings can be used to adjust the seafloor parameters to provide more accurate model outputs at the same location later. Validation and modelling can go hand-in-hand when real-time data feed back into ongoing modelling effort.
oceanoise2017 provided a broad overview of current and recent noise modelling and mapping projects. Maps differed largely in their spatial and temporal resolution, and in the quantities displayed.
It was recognised that the goal of the map drives the type of map. It would be helpful to derive a set of guidelines on how to produce the “best” map fit for a specific purpose, in order to reduce errors and variability, and to make maps more comparable.
Uncertainty is mostly not displayed, and guidance and agreement on error analysis and improvement of accuracy might be helpful.
Maps appear to be most commonly used for EIAs, either short-term, operation-specific EIAs, or long-term, habitat-level EIAs. How displayed noise levels link to GES, remains an active area of research. The application of noise maps to conservation management and marine spatial planning is perhaps the ultimate goa