With the rapid onset of anthropogenic (Gries et al., 2019; Orsenigo & Vercellis, 2018) environmental degradation and biodiversity loss, developing new technologies to salvage ecosystems is a matter of extinction (Lahoz-Monfort et al., 2019). Due to a lack of profit motive, conservation research has been late to adopt or develop emerging technologies at a competitive rate compared to analogous industries (Lahoz-Monfort et al., 2019). Industry, which is based around the commodification and satisfaction of human needs, not the preservation of nature for nature’s sake. Thus, we only pay attention to environmental degradation when it negatively impacts us. (Hatcher, 2004; Lahoz-Monfort et al., 2019; Plater, 2014). We’re forced to pay attention now because the byproducts of our economic development must be remediated if we are to continue pushing back the carrying capacity of our environment (Kim et al., 2017; Abdouli et al., 2018; Wright & Kelly, 2017; Reijnders, 2014). Even the USA department of defense is now adapting its strategies to match the outcome predicted if we fail to mitigate the effects of climate change (Brosig et al., 2019). Thus, we have to act. But how to do so affordably? Artificial intelligence, machine learning and remotely operated vehicles are prime candidates for adoption in conservation as they are cheap, reduce labor costs, increase sample size, ease of acquisition (see whale snot sampling above) and overall increase the efficiency of conservation research. (Lahoz-Monfort et al., 2019)
In the meta-analysis by Dujon et al., (2019) 213 publications utilizing remotely operated vehicles in 256 different locations were compared and 42% of these papers used machine learning to identify organisms and map habitats. Dujon et al., (2019) found that use of ROVs allowed researchers to collect data from remote locations faster and more efficiently than manned expeditions and when machine learning was used, it saved the researchers money and labor by processing large datasets in a matter of minutes. In the same study it was found that while ROVs were used in all habitats, machine learning was most often used in heterogeneous habitats to detect clusters of organisms with larger body sizes or to count nests/burrows, most frequently in countries with advanced economies. This is problematic as developing countries have higher pollution rates (Abdouli et al., 2018) and are more actively degrading habitat than developed countries (Tilker et al., 2019). In (Oliver et al., 2018) autonomous acoustic recorders were used to collect soundscape data at four different locations around the Toolik Field Station in Alaska over a 30 day period. Because the volume of data was large, signal processing and machine learning were used to cluster the data into 100 potential source types which were manually reduced down to 20 sources and the arrival date of the migratory birds was calculated. Oliver et al., (2018) concluded that while machine learning was not presently effective at identifying species, it effectively determined the arrival times and call frequencies for migratory bird species and could be adapted to track the spawning times and population densities of insect and amphibian populations if trained with an appropriate vocal activity index.
Generally, machine learning has been used to identify large incidences of change (Oliver et al., 2018) or to parse out populations of large organisms from an overall mosaic (Dujon et al., 2019). In (Tabak et al., 2019) it was used to identify species recorded by camera traps. Tabak et al., (2019) used a convolutional neural network with ResNet18 architecture trained with a data set of 3,367,383 images. This AI was then able to classify north American ungulate species with a 98% accuracy rate at a speed of 2000 images per minute. It was also trained to identify whether an image contained an animal with a 94% accuracy on out of sample image sets from Tanzania, allowing researchers to cull through images at a record pace. In closing, remote operated vehicles and remote monitoring systems, when used in consort with machine learning and artificial intelligence allow researchers to collect more data over larger ranges for longer periods of time and when processed, to target areas of high diversity or importance for conservation or resource extraction with surgical precision (Dujon et al., 2019). It also allows researchers to more efficiently process large data sets and identify areas of focus for manual review (Dujon et al., 2019; Lahoz-Monfort et al., 2019; Oliver et al., 2018; Tabak et al., 2019) which lowers the cost of operations and labor needed for environmental research, making this technology a prime candidate for adoption in countries where funding and labor are scarce (Lahoz-Monfort et al., 2019).
Literature Cited Abdouli, M., Kamoun, O., and Hamdi, B. (2018) The impact of economic growth, population density, and FDI inflows on CO2 emissions in BRICTS countries: Does the Kuznets curve exist? Empir. Econ. 4, 1717–1742. Brosig, M., Frawley, P., Hill, A., Jahn, M., Marsicek, M., Paris, A., Rose, M., Shambaljamts, A., and Thomas, N. (2019) Implications of climate change for the U.S. army. U.S. Army War College. Dujon, A., and Schofield, G. (2019). Importance of machine learning for enhancing ecological studies using information-rich imagery. Endang. Species. Res. 39, 91–104. Gries, T., Redlin, M., and Ugarte, J. (2019) Human-induced climate change: The impact of land- use change. Theo. and App. Climatol. 135, 1031–1044. Hatcher, T. (2004) Environmental ethics as an alternative for evaluation theory in for-profit business contexts. Eval. & Progrm. Plann. 27, 357–363. Kim, Y., Manley, J., and Radoias, V. (2017) Medium and long-term consequences of pollution on labor supply: evidence from Indonesia. IZA Journ. of Lab. Econ. 6, 5–20 * Lahoz-Monfort, J., Chades, I., Davies, A., Fegraus, E., Game, E., Guillera-Arroita, G., Harcourt, R., Indraswari, K., McGowan, J., Oliver, J., Refisch, J., Rhodes, J., Roe, P., Rogers, A., Ward, A., Watson, D., Watson, J., and Wintle, B., and Joppa, L. (2019). A call for international leadership and coordination to realize the potential of conservation technology. BioSci. 69, 1–10. Oliver, R., Ellis, D., Chmura, H., Krause, J., Perez, J., Sweet, S., Gough, L., Wingfield, J., and Boelman, N. (2018) Eavesdropping on the Arctic: Automated bioacoustics reveal dynamics in songbird breeding phenology. Sci. Adv. 4, eaaq1084. Orsenigo, C., and Vercellis, C. (2018) Anthropogenic influence on global warming for effective cost-benefit analysis: A machine learning perspective. Econ. e. Polit. Indust. 45, 425–442. Plater, Z. (2014) Human-centered environmental values versus nature-centric environmental values: Is this the question? Michigan Journ. of Env. & Admin. Law. 3. 2, 273–290. Reijnders, L. (2014) Phosphorus resources, their depletion and conservation, a review. Resour. Conserv. and Recycl. 93, 32–49. Tabak, M., Norouzzadeh, M., Wolfson, D., Sweeney, S., Vercauteren, K., Snow, N., Halseth, J., DiSalvo, P., Lewis, J., White, M., Teton, B., Beasley, J., Schlichting, P., Boughton, R., Wight, B., Newkirk, E., Ivan, J., Odell, E., Brook, R., Lukacs, P., Moeller, A., Mandeville, E., Clune, J., and Miller. (2019) Machine learning to classify animal species in camera trap images: Applications in ecology. Methods. Ecol. Evol. 10, 585–590. Tilker, A., Abrams, J., Mohamed, A., Nguyen, A., Wong, S., Sollmann, R., Niedballa, J., Bhagwat, T., Gray, T., Rawson, B., Guegan, F., Kissing, J., Wegmann, M. and Wilting, A. (2019) Habitat degradation and indiscriminate hunting differentially impact faunal communities in southeast Asian tropical biodiversity hotspot. Comm. Bio. 2, 396 Wright, S., and Kelly, F. (2017) Plastic and human health: A micro issue? Enviorn. Sci. Technol. 51, 6634–6647.