Conservation Technology
Metadata
- Author: Serge A. Wich and Alex K. Piel
- Full Title: Conservation Technology
- Category: books
- Document Tags: #planet #tech
Highlights
- Anthropocene—where human activities alter the dynamics to life, including climate, ocean, and forests (Location 8739)
- What do conservation biologists need to know to support conservation management and policy? Which data, resolution, or parameters are not being captured that need to be? What existing technology can help obtain the type and quality of data needed? How do we enhance data capture? From a management perspective, can we identify not just hotspots of biological interest, but particular key resources (specific trees, water sources) rather than broader areas or general vegetation types (Location 8740)
- We often want to do this longitudinally, especially to examine trends and change over time, and also geographically, across increasingly vast landscapes. Moreover, long-term projects are faced with how to digitize and standardize data collection protocols while maintaining interobserver reliability, all the while confronted with rotating staff and often the training of volunteers/interns/students. (Location 8741)
- line transects (Yapp, 1956). In this method, researchers walk a straight line, often of random bearing, and record the perpendicular distance of all direct or indirect wildlife observed from the line. (Location 9179)
- as point counts, the same approach is sometimes used for roadside tallies of wildlife as well (reviewed in Schwarz & Seber, 1999). These types of surveys are based on the proportion of observed evidence to search effort and the assumption that change over time is attributable to population increase or decrease. (Location 9613)
- ‘geolocators’, which were attached to animals, used the time of day to calculate an animal’s position, and so were prone to large error margins as much as tens of kilometres from the animal’s actual location (Location 9617)
- The most common type of remote sensing from either satellites or drones is multispectral, whereby sensors assess the radiation (or brightness) emitted from surface areas on earth. (Location 10052)
- For example, LiDAR (light detection and ranging) on low flying aircraft, but also from space-borne satellites and drones, estimates above-ground carbon stocks and overall biomass, ecosystem structure, and broadly provides critical data for environmental management (Location 10052)
- hyperspectral remote sensing—multispectral sensors with hundreds of bands across a nearcontinuous range, including visible, infrared, ad electromagnetic spectrum—have (Location 10052)
- For conservationists, thermal imaging sensors are providing data to answer questions about species distribution, counts, and the location of fires (Location 10487)
- Satellites and drones offer views from above and thus suffer from an inability to access ground-based biodiversity if those are covered by vegetation or are too small. Two of the most common methods to capture ground-truthing data are camera traps (Chapter 4) and acoustic sensors (Location 10487)
- biologgers now reveal animal movement in three dimensions, integrating accelerometers, barometers, and gyroscopes (Location 10488)
- Environmental DNA is animal genetic material originating from the hair, skin, faeces, or urine of animals but that has degraded and can be extracted from water, soil, or sediment (Location 10924)
- eDNA collection is not constrained by weather conditions (Location 10925)
- The main barriers to the use of information provided by remote sensing are often not technological. The barriers can include a lack of clearly defined management questions and a lack of budget for training and transfer of technology, as well as a lack of providing results to decision-makers and organizations that manage the environment, wildlife, and other natural resources (Location 13983)
- research from social psychology to organizational management has convincingly shown that empirical evidence is only a minor factor in influencing human decision-making and behavioural change (Location 13984)
- Recent understandings of the science–policy interface from communication, science, and technology studies suggest that information and knowledge should not be seen as a product to be transferred from researchers to users but rather a ‘process of relating that involves negotiation of meaning among partners’ (Location 14419)
- adjust from trying to solve research-implementation gaps to creating more research-implementation spaces where researchers, decision-makers, and other partners engage, collaborate, identify, and understand how data, information, and knowledge are produced, by whom and for whom. (Location 14419)
- it is important to consider how EO applications, knowledge, and tools are perceived and filtered through existing local beliefs, traditions, values, experiences, and concerns along with the capacity and resources available to use that information (Location 14420)
- Conservation is inherently spatial and complex (Game et al., 2013). It requires using geographic data to understand the current and changing distributions of species, the extent of species’ habitats and its change, and ecosystem processes along with human activities to design and implement strategic actions that will minimize or eliminate the most important threats to biodiversity (Location 14420)
- Geospatial technologies are a combination of three major groups of technologies involved in the collection, manipulation, storage, management, analysis, visualization, and communication of these geographic data and information. They include (1) global navigation satellite systems (GNSS), (2) GIS, and (3) remote sensing (RS) (Location 14421)
- There are two major types of RS sensors: passive and active. A digital camera mounted on a drone is an example of a passive sensor that works by detecting reflected sunlight from the object, while active sensors such as LiDAR (light detection and ranging) and SAR (synthetic aperture radar) provide their own energy to illuminate the object or scene they observe (Location 14422)
- Active and passive sensors can be divided into non-imaging (e.g. radiometers) and imaging sensors (e.g. imaging scanners) that are analogue or digital and acquired on a range of spectral and spatial resolutions (Location 14422)
- Category Subcategory Platforms Ground—Airborne—Spaceborne Sensors Active—Passive Output Non-Imaging—Imaging Image output Analogue—Digital Spectrum range Visible—Infrared—Thermal—Microwave Spatial resolution High <5 m—Medium >5 m <50 m—Low > 50 m (Location 14856)
- Maxar is the first company to deliver 30 cm native and 15 cm high definition synthetic (using image resampling algorithms) satellite imagery1 that could be collected along a swath width of 13.1 km at nadir (i.e. looking straight down), covering a total of 171.61 km2 just in one image. (Location 14856)
- A typical commercial level fixedwing drone such as an eBee can deliver imagery accuracy below 3 cm but can only fly for about 1 hour on one battery charge and cover around 220 ha (2.2 km2) while flying at 120 m asl, which is the approximate legal altitude ceiling in many countries2 (Location 14856)
- For example, the Global Land Analysis and Discovery (GLAD) lab at the University of Maryland, College Park has released and is maintaining a forest loss data set derived from 30 m Landsat imagery that has global accuracy but also has relevance at the local scale (Location 15296)
- Global Forest Watch,5 which hosts several data 5 https://www.globalforestwatch.org, GFW sources relevant for conservation. Users can delineate an area of interest and receive a notice via e-mail if forest loss or a fire is detected. Moreover, the Forest Watcher mobile application enables the offline use of Global Forest Watch’s (GFW) spatial data.6 (Location 15731)
- The GLAD team developed an alert for humid tropical forests at 30 m resolution that is available as new Landsat imagery is acquired, which means users can be updated weekly, but the prevalence of clouds in the region of interest could result in a delayed alert (Location 15733)
- ‘Open Standards for the Practice of Conservation’ or ‘Conservation Standards’ (CMP, 2013) (Location 16168)
- Conservation Standards is a science-based and collaborative planning approach that uses adaptive management to help focus conservation decisions and actions on clearly defined objectives and prioritized threats and measures success in a manner that enables adaptation and learning over time (Location 16168)
- Key questions facing stakeholders engaged in the Conservation Standards process are: ‘How species and habitats are doing and what are the major threats to their survival?’, ‘What actions are needed to minimize or eliminate those priority threats?’, and ‘Are our actions effective in minimizing or eliminating the most important threats?’ (CMP, 2007). In addressing these questions, the Conservation Standards approach is oriented around a five-step project management cycle adopted for conservation: assess, plan, implement, analyse and adapt, and share (Location 16169)
- A Conservation Standards cycle starts with defining the project’s purpose, teams, and articulating geographic scope, a vision of what the project hopes to achieve, and the conservation targets on which the plan will focus. It also includes making sense of the project’s context, including identifying threats, opportunities, and key stakeholders (Location 16170)
- The use of EO data and geospatial technologies starts with identifying, downloading, cleaning, and compiling all relevant geospatial data and RS imagery into a geodatabase to produce a series of basemaps. This geodatabase serves as the foundation for other geospatial applications and provides geographical context to the conservation planning process. (Location 16170)
- Since 2002 JGI has developed a Participatory Remote Sensing (PRS) model (Pintea, 2006) as part of its Tacare approach. It combines Participatory Rural Appraisal (PRA) with very high-resolution satellite imagery below 1-metre resolution, enabling local people to record their knowledge, values, and perspectives of their communities by mapping water sources, wildlife migrations, land tenure, land 7 https://www.protectedplanet.net/ 8 https://www.openstreetmap.org/ uses, vegetation, traditional belief sites, and others. (Location 16605)
- Viability assessments of conservation targets that include species, habitats, or ecosystem process help project teams build a set of hypotheses to guide conservation and research. It begins by identifying key ecological attributes (KEAs) for each of the targets. A KEA is ‘an aspect of a target’s biology or ecology that if present, defines a healthy target and if missing or altered, would lead to the outright loss or extreme degradation of that target over time’ (Location 17914)
- The conservation standards group KEAs go into three indicator classes: • Size is a measure of the area or abundance of the conservation target’s occurrence (e.g. acres of habitat). • Condition is a measure of the biological composition, structure and biotic interactions that characterize the occurrence (e.g. presence of key species). • Landscape context is an assessment of the target’s environment, including ecological processes and regimes that maintain the target occurrence such as flooding, fire regimes, and many other kinds of natural disturbance, and connectivity such as species targets having access to habitats and resources or the ability to respond to environmental change through dispersal or migration (e.g. average distance in km between habitat patches) (Location 17915)
- Once the project teams and stakeholders define conservation targets and indicators for measuring chimpanzee population and habitat health, they need to use the available evidence to identify, map, and prioritize the most important direct threats or pressures and the actors behind those threats (Location 18789)
- For a standard list of human activities that could potentially degrade a target, see CMP’s Conservation Threat Classification (Salafsky et al., 200810). (Location 18789)
- For projects in dry environments such as in western Tanzania, Guinea, or Senegal where GLAD products do not characterize dry deciduous forests, woodlands, and scattered trees that effectively, or areas covered by persistent clouds like in the western region of the Republic of Congo, GLAD products can be complemented with other global products such as PALSAR-2 global forest/non-forest map from JAXA.11 (Location 18790)
- Not all tree cover loss is deforestation. (Location 18791)
- community mapping, ODK, Survey 123, and other geospatial technologies to inform and facilitate a participatory land-use planning process owned and driven by the local communities. (Location 19663)
- JGI leverages and uses Esri’s ArcGIS cloud as its geospatial science platform with specific tools, data workflows, and models organized under different Decision Support Systems (DSS). (Location 20102)
- The application of EO satellite data to conservation has several limitations. First of all, cloud coverage strongly limits the availability and overall quality of optical satellite imagery, especially in the tropics (Hilker et al., 2012). Very high-resolution satellite data are also expensive and conservation organizations might find it difficult to access because of the high cost (Boyle et al., 2014). The cost of hardware and software to store, manage, process, and classify the EO data could also be a major constraint for conservation practitioners and local governments (Steering Committee on Space Applications and Commercialization, 2003). (Location 20538)
- Users also need to have significant expertise in data processing and software development to fully utilize RS data (He et al., 2015). Finally, end-users increasingly rely on internet-based workflows to identify, download, and analyse EO data (Sudmanns et al., 2020). These factors mean that many parts of the world that do not have access to high-speed internet are left behind in taking advantage of the disruptive changes that are transforming the RS and GIS fields. (Location 20539)
- Policy development often lags behind advances in technology, which is also true of the new generation of highly detailed RS technologies applied in the absence of legal and policy constraints (Slonecker et al., 1998). This includes the development of new and higher resolution sensors, a shift of the responsibility for data collection and use from the public to the private sector, and expansion of actors into the international arena that further complicates issues because of the need for global policy and regulations (Location 20972)
- EO satellite data could be used to provide detailed insights into people’s lives, such as predicting the socioeconomic status of individual households (Location 20973)
- Court cases in the USA have concluded that the individual’s privacy is not protected when the subject or the property is in plain or public view. (Location 20973)
- However, in the United States, the collection of satellite imagery finer than 0.31 m (native) is restricted. Similar regulations exist in the European Union. On the other hand, these requirements do not apply to Chinese or Indian companies that are rapidly making advancements in high-resolution imaging technology (Location 20973)
- EO data could be used to directly map or predict with relatively high certainty the potential location of an endangered species (He et al., 2015; Jantz et al., 2016). Special consideration should be given that such data does not end up in the hands of poachers or illegal loggers. (Location 20974)
- spectranomics (Asner & Martin, 2009; Asner & Martin, 2016). New hyperspectral sensors, planned on spaceborne platforms by Planet and NASA, will enable researchers to monitor in detail forest canopy function and composition and connect species functional traits with biodiversity indicators directly. (Location 20975)
- Another clear trend is the emergence of big EO data in the geospatial cloud platforms such as Esri ArcGIS, Microsoft AI for Earth, Maxar GBDX, and Google Earth Engine. These new computing and data storage platforms enable researchers to store, access, process, analyse, collaborate, and share large amounts of EO data at an unprecedented level. Integration of EO RS data with social sensing, mobile, and other sensors and application combined with deep learning algorithms will also improve the accuracy of land cover and landuse change classes (Location 20975)
- Even though satellite imagery provides researchers with increasingly higher resolution and more frequent images globally, there can be challenges with using satellite data (see Chapter 2). First, in the humid tropics, cloud cover can hamper the collection of images, especially in certain seasons. Second, in artic areas, the almost continuous darkness throughout parts of the year hampers data collection. Third, some analyses require higher resolution data than satellites offer at present (Anderson & Gaston, 2013). Fourth, where higher resolution data are available, their costs can be prohibitively expensive. (Location 23596)
- require specialist geographic information system (GIS) and/or remote sensing software. (Location 24031)
- Hunting for meat is widespread worldwide and is generally decimating wildlife (Benítez-López et al., 2019; Ripple et al., 2019). Hunting for body parts focuses on a subset of species such as rhinos and elephants and has led to large reductions in their numbers.1 Hunting for both meat and body parts appears difficult to reduce. Current methods to reduce poaching of such highly prized species, and poaching in general, include improving law enforcement, promoting alternative livelihoods, improving antipoaching efforts, and so forth. Still, their effectiveness is limited (Location 24031)
- Drones can be split into three categories: fixed-wing, multirotor, and hybrid VTOL (vertical take-off and landing) (Location 24032)
- Irrespective of the drone system, there are three additional important components: the power source, flight controller, and a ground control station (GCS). (Location 24032)
- most of the systems flown by conservationists are powered by rechargeable lithium-ion polymer (LiPo) batteries with a small number using lithiumion (Li-ion). (Location 24032)
- There are several proprietary and open-source flight controllers that vary in their level of complexity. Many of these are highly advanced and integrate several sensors such as global navigation satellite system (GNSS) and inertial measurement unit (IMU), and allow the drone to follow pre-programmed coordinates during a mission. (Location 24033)
- The GCS has a data link with a drone and can be used for various applications such as sending commands to the drone, obtaining data on the location of the drone, its battery status, and seeing images that a camera on a drone is taking. The GCS can be a laptop, tablet, or mobile phone, depending on the specific drone and software. (Location 24033)
- Table 3.1 Drone categories Multirotor Fixed wing Hybrid VTOL Launch area Small large Small Flight duration Short (<1 hr) Long (>1 hr) Intermediate (30–60 minutes) Payload Heavy (few kilograms) Light (<1 kg) Light (<1 kg) Pilot experience Minimal (although larger systems require more training) Substantial training Intermediate (Location 24467)
- The location is important with respect to flight regulations, privacy issues related to data collection, weather (particularly wind speed in relation to the wind speed that the drone can operate to), presence of fine particles (e.g. sand), telemetry connectivity issues due to vegetation, take-off-andlanding options, and so forth (Duffy et al., 2017). (Location 24469)
- The data that need to be collected will impact the sensor choice (e.g. standard visual spectrum, thermal infrared, multispectral, hyperspectral), the flight length of which the drone needs to be capable, payload capacity, etc. In addition, there are several other aspects to consider, such as whether batteries can be transported internationally, whether in case of an accident or malfunction there are spare parts that can be purchased locally or whether these need to be brought on the trip, whether there is someone who can carry out repairs, and so on. (Location 24469)
- At present, there are three main uses of drones in conservation (mapping, animal counts, and poacher detection) which require two largely different analytical pipelines. For mapping, it is important to obtain full coverage of the area of interest. Hence, flights are often conducted in a grid pattern during which photos are being collected with a high percentage of overlap. The high percentage of overlap is important so that the software that applies the Structure-from-Motion process (SfM) to yield a map will work well (Westoby et al., 2012; Wolf et al., 2014). There are a number of commercially available packages (e.g. Pix4Dmapper,7 Agisoft Metashape8) that have user-friendly workflows to process individual images from grid flights into an orthomosaic. Alternatively, open-source packages such as VisualSfM/CMVS9 and Microsoft ICE10 can be used to process images into one large map of the area of interest. These are less user friendly and have fewer features, but they can provide good results as well. (Location 24470)
- Often the resulting orthomosaics are combined with contextual layers in GIS software packages such as ArcGIS, QGIS, and Global Mapper. Some of these programmes have started to integrate the SfM processing into their GIS so that such contextual information can be easily added to the orthomosaics. The maps produced through the SfM process can be further processed for land-cover classification and land-cover change analyses. Some of those analyses can be conducted within the GIS platforms such as ArcGIS and QGIS. (Location 24471)
- ERDAS Imagine or use cloud options such as Google Earth Engine on which orthomosaics can be uploaded and analysed. (Location 24904)
- For animal counting and poacher detection, the requirement is that objects of interest (e.g. rhinos, poachers, cars) are detected in the individual images. Generally, detection and classification are conducted manually. This is costly and timeconsuming and, at present, represents a major challenge for the development of end-to-end workflows where data are collected and analysed in a cost- and time-efficient way. To alleviate this issue, researchers are working on automating the detection and classification of objects (Lhoest et al., 2015; Gonzalez et al., 2016; Longmore et al., 2017). There are a large number of different approaches to automating the detection of animals on images, such as spectral thresholding (Chabot and Bird, 2012), blob detection (Ward et al., 2016), and using local maxima and isolines (Lhoest et al., 2015). A popular approach is to apply neural network approaches in which computer algorithms are trained with a subset of the data, validated, and then used with test data. This approach has led to some promising results for detecting animals and poachers (Maire et al., 2015; Bondi et al., 2018; Bondi et al., 2019). (Location 24905)
- There are many options for object detection and classification using machine learning for experienced programmers (Lamba et al., 2019). However, to the best of the authors’ knowledge, there are no off-the-shelf open-source programs that allow non-programmers to label images, train a number of machine learning algorithms, validate these, and then use a test data set to determine their accuracy in detecting and classifying the objects of interest. There are commercial options available through Google11 and Microsoft12 that are relatively user-friendly but might still deter users due to the steep learning curves and high costs involved when processing large numbers of images. Nevertheless, companies such as Microsoft and Google provide 11 https://cloud.google.com/ml-engine/ 12 https://azure.microsoft.com/en-us/services/machinelearning-studio/ computational and web-based services to facilitate the uptake of machine learning at a large scale… (Location 24906)
- These maps can provide a manager with details on which land-cover types occur in an area, their spatial extent, and any changes if a series of maps are produced through time. Because of the large spatial extent of many conservation projects, satellites are most commonly used for land-cover mapping. However, sometimes higher resolution is required for specific areas and drones can be particularly useful for such cases. Such information can then be used to… (Location 24907)
- Because conservation researchers often need to map relatively large areas, they often use fixed-wing drones instead of multirotor drones for… (Location 24908)
- For complex environments, such as tropical forests, 90% sidelap is recommended by Pix4Dmapper to process an accurate orthomosaic. In less complex areas, 60–70% sidelap might be sufficient to produce an accurate orthomosaic. Due to the shorter flight duration of multirotor drones compared with fixedwing and hybrid VTOL systems, the former maps much smaller areas. (Location 25342)
- By far the most common sensor is one that captures data in the visual spectrum. The orthomosaics produced with such images often provide very high-resolution information to the user. They can be used to classify land-cover types or detect specific features such as a particular tree species using the orthomosaic and/or the point cloud associated with it (Location 25342)
- Such cameras can, for instance, have four bands (green, red, red edge, and near-infrared) as well as a standard visual spectrum sensor. This provides more options for land-cover classification but also to determine indices of vegetation health (e.g. Normalized Difference Vegetation Index (NDVI), Green NDVI (GNDVI), etc. (Michez et al., 2016; Assmann et al., 2019)). (Location 25343)
- hyperspectral cameras, which have a much higher number of bands than multispectral cameras. However, the costs of these sensors are high (e.g. €40,000, at the time of writing, for a Rikola camera14) and as a result they are infrequently used (Location 25344)
- Determining the distribution and density of animals and deriving abundances from those data is a key aspect of conservation science. (Location 25344)
- Drones equipped with visual spectrum cameras have been the most commonly used system for animal detection, but a growing number of studies have been using thermal sensors to detect animals (Location 25345)
- Although the thermal imaging data are promising, there are still challenges in classifying the various species that exist in an area from thermal data. Additionally, individuals detected from the ground and under vegetation are likely to be missed on thermal images (Kays et al., 2019). Kays et al. (2019) suggest that combining flash photography or IR illumination with thermal sensors might reduce such challenges. (Location 25345)
- Several marine mammals can be detected with drones (Hodgson et al., 2013; Koski et al., 2015), sea turtles in clear waters up to a relatively narrow depth (Rees et al., 2018), similarly for sharks (Kiszka et al., 2016; Rieucau et al., 2018; Colefax et al., 2019) and fish species such as salmon (Groves et al., 2016). (Location 25779)
- This indicates that a relatively large percentage are missed from the air due to nests being inside the tree canopy. (Location 25782)
- A 1-hour flight with a still taken every 2 seconds will lead to 1800 images. (Location 26217)
- If a trained image analyst needed 1 minute to examine each image to detect an animal (or multiple animals), it would still take 1800 minutes (30 hours) to process the images. (Location 26218)
- At present, the usage of promising machine learning methods such as deep learning is still difficult for non-programmers and requires quite substantial investments in hardware. Large companies such as Google and Microsoft are making these methods more accessible through their cloud services, and we expect that there will be an enormous amount of progress made in the next 5 years or so. (Location 26219)
- As referred to in this chapter, poaching refers to poaching for bushmeat and animal parts such as rhino horn, elephant tusks, and pangolin scales. Poaching occurs inside and outside of protected areas and rangers are usually insufficiently resourced to reduce or halt it. Besides, there is considerable risk for conservation staff during anti-poaching missions (Location 26219)
- It thus seems that poachers can be detected using thermal sensor data obtained with drones. This is promising, but much work remains to be conducted on how poachers can be differentiated from other objects that might look similar—such as similar size animals and rocks that have been warmed up by the sun (Burke et al., 2018). Also, it is important to determine the false-negative rate in such data, that is, whether there are instances in which poachers were present but were not detected on the thermal drone footage. (Location 26653)
- One of these challenges is the lack of data on the durability of the drone systems used by conservation workers. There are few or no details available on, for instance, how often a motor or electronic speed controller or other part of a drone should be replaced, or how often parts should undergo maintenance. Nor are there clear data on, for example, how many belly landings a fixed-wing drone can undergo before the fuselage needs to be replaced or undergo standard maintenance. This impacts how well we can predict the number of flights or hours of flight a drone might be capable of, which has safety implications and hampers a proper comparison of costs between drones and alternative data collection methods. (Location 27527)
- potential cost reductions are an important reason for conservation organizations to invest in drones, there is a strong need to improve our understanding of drone system durability. This is also important so that operators can develop a maintenance plan for their drones, (Location 27528)
- including whether they need to build maintenance capacity themselves or whether to outsource this to third parties, and whether bespoke maintenance is required for specific drones in their fleet. (Location 27963)
- there are very little data available providing cost comparisons for the use of drones versus alternative methods. Although there are studies such as those by Vermeulen et al. (2013) that provide some comparison between the survey costs with drones and those of a crewed aircraft, these generally do not cover all the costs involved such as training, computers, analyses time, and so forth. (Location 27963)
- Another challenge is that the areas that conservationists would like to cover with drone surveys are normally much larger than the areas off-the-shelf and affordable drones can currently cover. Even though method tests and development are being conducted on smaller areas (less than 10km2), there is a need to transition to covering larger areas once methods have matured sufficiently. With the increasing capacity of batteries and the potential of combining batteries with energy from solar cells on the wings of fixed-wing systems, there are promising developments that indicate that much longer flight durations will become possible at costs within the realm of conservation projects.17 (Location 27964)
- Training can also pose challenges, particularly related to fixed-wing systems. Although multirotors are relatively easy to fly in modes where GNSS is controlling flight altitude and position, remote pilots must also be able to fly the drones in modes 17 https://sunbirds.aero/ that do not have GNSS enabled to ensure safe operation in case of GNSS failure. Similarly, many fixed-wing systems are becoming easier to fly with largely automated take-offs and landings. However, when GNSS fails or automated flying is interrupted, it would take considerable training to be able to continue to fly and safely land the drone. (Location 27965)
- requirement for safe storage of the large quantities of data generated from highresolution stills and videos. Once the data is being used to process orthomosaics, the data volume increases even further. Even though the cost of storage has decreased enormously, the data volumes associated with a high number of flights might still mean that this is a considerable cost, particularly when back-ups are needed. (Location 27966)
- also important to consider the costs of data analyses in terms of hardware, software, and costs for the time people put in collecting and analysing the data. Particularly high costs can be associated with the hardware (e.g. >€2000 for a desktop computer) and software (e.g. >€3000 for one of the main software packages18) to process orthomosaics depending on the software used and the hardware requirements of that software. (Location 27966)
- use of drones can lead to disturbance (due to e.g. noise, shape, flight pattern) of animals (Location 27967)
- In relation specifically to drones, there are various reviews of the literature on the disturbance of drones on animal behaviour (Mulero-Pázmány et al., 2017; Wich & Koh, 2018), indicating that disturbance is not consistent. Some studies did not report any visual evidence of disturbance in animals’ behaviour, whereas others found that animals fled (in some cases only temporarily) or produced alarm calls (Location 28400)
- Particularly in situations where the usage of drones is within a law enforcement context, it has been suggested that caution is required as drones might be perceived as being part of a conservation approach that supports protected areas as ‘fortresses’ from which local communities are excluded (Location 28403)
- It might, therefore, be more beneficial if all these technologies were considered in a larger framework with respect to social and privacy implications (Location 28403)
- An important next step is to integrate these various sensors into one system in which the various sensors feed into the same decision-making process that will further facilitate conservation management (Location 28837)
- In such a system, drones can be used to collect data with their own onboard sensors but can also be data mules to wirelessly relay data from other sensors such as camera traps or acoustic sensors. (Location 28837)
- passive acoustic monitoring (PAM) (Location 31459)
- Tags: definition
- acoustic sensors can be deployed for long periods in the field (months or years) and simultaneously at multiple locations and, because of their large detection range, can monitor places that may be difficult to access by researchers. Moreover, they are non-invasive and require low researcher-hours for deployment compared to alternative devices that demand frequent visits or maintenance. Sensors can also be used in adverse weather conditions (Location 31459)
- PAM has been the primary means of revealing cetacean habitat use and behaviour. It is beyond the scope of this chapter to comprehensively review this body of work. However, PAM has been used to reveal seasonal and daily temporal movement patterns of harbour porpoises (Phocoena phocoena), common (Delphinus delphis) and bottlenose dolphins (Tursiops truncatus) (Dede et al., 2014), or night-time movements of Yangtze finless porpoises (Neophocaena phocaenoides asiaeorientalis) (Akamatsu et al., 2008), seasonal migrations of mink whales (Balaenoptera acutorostrata) (Risch et al., 2014), and simultaneous ranging behaviour of seven adjacent killer whale (Orcinus orca) pods (Yurk et al., 2010). It has also been used to study right whale (Eubalaena glacialis) use of corridors in urbanized coastal regions (Morano et al., 2012). The study of marine mammal behaviour using PAM is usually based on either stationary underwater hydrophones or towing multiple hydrophone arrays behind sea vessels. (Location 31897)
- There is far less, but nonetheless growing application of PAM to terrestrial species. These studies involve attaching acoustic sensors to stationary locations (e.g. trees usually) and setting recording schedules (i.e. duty cycling, e.g. continuous, 10 minutes every hour, to save battery life and memory). (Location 31898)
- Acoustic triangulation uses the time difference of arrival of sounds to multiple sensors to estimate the sound origin. Besides these behaviour purposes, localization also improves algorithms to automatically detect a species by separating animal sounds from background noise and estimate density (see next), and quantify sound amplitude or directionality (Location 32334)
- An acoustic sensor is composed of any sound recorder and a microphone/hydrophone. The choice of the acoustic sensor is made as a function of the different parameters necessary to monitor, such as the species studied and the frequency range of its vocalizations, the environment (marine or terrestrial), the design of the study (study length, area covered, etc.), and the budget available, among others. (Location 34517)
Relevant:
Wireless Sensors Could Be Less Effective in Muddy Soil