With machine learning having become a typical application along with GIS, one area of focus has been habitat protection. Habitat managers and conservation specialists have struggled to find ways in which to protect wildlife threatened by a variety of mostly-human induced factors. Machine learning and GIS have proven one way in which new ideas and scenarios can be tested before any plan is carried out, saving time, money, and possibly avoiding making crucial habitat errors in plans implemented.
A recent example of using GIS and machine learning for habitat protection has been applied on the black-necked crane. This type of bird is very particular with where it can breed and relatively little is known about it. Furthermore, the remote locations in which it lives makes intensive survey and long-term monitoring of the crane difficult. A species distribution model (SDM) was built on available data, where the model would then predict regions where the crane would breed. The results could then be applied to inform on where conservation efforts should focus to protect the crane.
Knowing how species’ habitats change is also important for conservation and long-term planning. In Wisconsin, the Muskellunge breeds in different lakes in the northern parts of the state. However, in recent years these fish have been declining in numbers. In this case, a maximum entropy model was created to show several key variables are important to where the Muskellunge could best breed and survive. This included places that had moderately sheltered fetches, moderate, small areas of shallow flats, places that were away from outflowing streams, and shorelines facing east or west were a moderate factor. Classification and regression tree modeling is often the most common machine learning techniques applied. This has been used for estimating ranges of the endangered red panda, where the species, similar to the crane discussed earlier, that lives along the Hindu-Kush Himalaya region along the borders of India, Pakistan, and China.
What has made machine learning techniques so useful is how, with relatively little data, very informative models can be created to inform on habitat. This is the case with understanding the great bustard, another bird that has been difficult to track. Simply using data on its droppings has allowed scientists to utilize the TreeNet, Random Forest, and CART machine learning models to project information about its habitat. In this case, the best model fit showed that distance to residential area, water pools, and farmland had a major impact on areas where the bird preferred to be. These birds preferred to be near water pools and live in or near small farms, while they preferred to be at least some distance from residential areas (at least 400 m). Furthermore, these tools were utilized with ArcGIS, where this commercial software has been among the more typically utilized GIS platforms with machine learning.
With impacts, such as climate change, becoming more of a threat to different species, increasingly we are seeing the application of machine learning and GIS techniques for understanding how this will affect habitat losses. Plant species are often important for economies of states. For instance, in Canada, maple syrup, a leading export, is under potential threat as habitats change. Projection maps have been created to determine areas where maple syrup may best survive changes and potentially new areas where maple syrup production could concentrate in the future.
There are still challenges in using machine learning and GIS in conservation and habitat studies. For instance, in the oceans, where the seabed is poorly understood, has not led to highly accurate models that can classify sedimentation and patterns of substrata. This can affect habitat as different seabed fauna are sensitive to the type of sediments and strata that might be found. The challenge is to capture better data so that future maps can be better at predicting how the seabed is constituted.
With gains in computational power and accessibility to off the shelf models, often used in commercial or open source software, applying machine learning techniques has become far easier for conservation and habitat specialists. Increasingly, we are seeing future projections and attempts to better understand habitat changes as the planet rapidly changes and the need for conservation and projection of future change becomes increasingly important.
 For more on a recent GIS and machine learning application focused on the black-necked crane, see here: Han, X., Guo, Y., Mi, C., Huettmann, F., et al. (2017) Machine Learning Model Analysis of Breeding Habitats for the Black-necked Crane in Central Asian Uplands under Anthropogenic Pressures. Scientific Reports. [Online] 7 (1).
 For more on the Muskellunge, see: Nohner, J.K. & Diana, J.S. (2015) Muskellunge Spawning Site Selection in Northern Wisconsin Lakes and a GIS-Based Predictive Habitat Model. North American Journal of Fisheries Management. [Online] 35 (1), 141–157.
 For more on the red panda and use of machine learning techniques, see: Kandel, K., Huettmann, F., Suwal, M.K., Ram Regmi, G., et al. (2015) Rapid multi-nation distribution assessment of a charismatic conservation species using open access ensemble model GIS predictions: Red panda (Ailurus fulgens) in the Hindu-Kush Himalaya region. Biological Conservation. [Online] 181, 150–161.
 For more on using such machine learning modeling on understanding the great bustards habitat, see: Mi, C., Huettmann, F. & Guo, Y. (2014) Obtaining the best possible predictions of habitat selection for wintering Great Bustards in Cangzhou, Hebei Province with rapid machine learning analysis. Chinese Science Bulletin. [Online] 59 (32), 4323–4331.
 For more on this study on maple syrup, see: Brown, L.J., Lamhonwah, D. & Murphy, B.L. (2015) Projecting a spatial shift of Ontario’s sugar maple habitat in response to climate change: A GIS approach. The Canadian Geographer / Le Géographe canadien. [Online] 59 (3), 369–381.
 For more on the seabed study, see: Diesing, M., Green, S.L., Stephens, D., Lark, R.M., et al. (2014) Mapping seabed sediments: Comparison of manual, geostatistical, object-based image analysis and machine learning approaches. Continental Shelf Research. [Online] 84, 107–119.