Strandings can provide important information on the composition of populations, their wellbeing and the health of their environment as explained in an earlier post. Discovering and analysing stranding events can prove challenging though, especially in remote areas. Thus, a group of scientists from British Antarctic Survey and four Chilean research institutes have suggested and tested a possible approach through satellite imagery to help in the task of detecting and quantifying stranding events of large cetaceans.
They reanalysed the biggest reported baleen mass stranding event of 343 reported sei whales in the Chilean Central Patagonia in 2015 to test their hypothesis with Very High Resolution (VHR) remote sensing technology. At present, the satellite images provide a spatial resolution of 31 – 50 cm grayscale and 1.2 – 2 m in visible colour and infrared. Two different approaches were conducted: The initial analysis consisted of a manual scan and count of whale-like objects. The close match between prior conducted aerial survey locations and whales identified through satellite imagery showed that VHR images can be used to identify and evaluate stranding events of large cetaceans. This method proofed to be more reliable – discovering even more whales than the aerial survey had - though challenges arise in overlapping places and time consumption.
Next, automatic evaluation was tested with a Spectral Angle Mapper. Herein a target-finding algorithm is used to detect whale spectral signatures that were chosen manually beforehand to discover other whale-like objects in the imagery. The outcome of the automatic approach proofed variable, depending on weather conditions such as cloud cover and tide height, thereby altering the quality of the picture and reliability of the programme. Due to the “variety in shapes and heterogenous spectra it is unlikely that a totally automated solution can be constructed using the current resolution imagery”. However, a semi-automated approach might offer a solution by manually evaluating areas that have been highlighted by automation as ‘most likely’. Further, recent availability of higher resolution might improve automatic processes due to better shape discrimination and larger numbers of pixels.
Analysing stranding events through satellite imagery has the potential to provide a quick and relatively reliable view of whale strandings, therefore posing a possible alternative or addition to regular evaluations through surveys and detections by chance. Some aspects still prove challenging though, presenting the need for further investigation and research in the area. For one, satellite imagery cannot be accessed and evaluated immediately which limits the urgent response. Since it is still faster than the average current time of response though, it poses a powerful tool for the monitoring of remote areas. “With careful planning, a combination of manual and semi-automated searching is likely to reduce image searching times and make large-scale satellite imagery interrogation more feasible”. The detection and evaluation through VHR would be accompanied by regular stranding responses, facilitating identification of species, physical response and sampling for further investigation.
The minimum size at which an animal can be identified is not yet understood and is likely to depend on the image quality and environment. Nevertheless, it has shown effective for large, baleen species, whose mass strandings are much rarer than those of smaller odontocetes and should receive particularly acute attention as they can signal unusual environmental conditions.
The analysis of satellite imagery on the sei whale stranding in Chile confirmed to be an effective method to calculate stranded whale numbers and distribution without being limited by geographical remoteness. The “ease of access to satellite imagery, low potential cost, large spatial coverage and relatively simple logistics” compared to more traditional survey techniques means that VHR satellite imagery could become a valuable tool for monitoring mass mortality and other stranding events in remote areas. By providing regular snapshots of remote areas where whale strandings are relatively likely, stranding patterns could be monitored and events could be caught closer to the time they happen. Future work should focus on the investigation of machine learning algorithms and automated analysis with the use of higher resolution satellite imagery to improve capacity to search the imagery.
The full paper can be accessed through https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0222498
コメント