How are bear sighting maps created? Real danger zones revealed from a 'collection of points'
Hello, I am Iriyama, the GIS comedian.
▷ Click here
for MapQuest, where Iriyama works. ▷ Click here
to learn more about Iriyama, the GIS comedian. ▷ Click here for a GIS software comparison table.
1. Giving 'meaning' to a collection of points
Bear sightings are occurring one after another in Akita Prefecture. In news footage, red dots are lined up on a map, looking almost random. However, researchers are deciphering 'bear behavior patterns' from that 'collection of points.' What makes this possible is a technique called **
Clustering (Cluster Analysis)** . It sounds like technical jargon, but in short, it is a technology for 'grouping similar locations together.' It is like sorting cooking ingredients by genre; groups such as 'sightings along mountains,' 'sightings around orchards,' and 'sightings near settlements' naturally emerge.
Among these, a method called **DBSCAN (Density-Based Spatial Clustering of Applications with Noise)** is particularly suitable for animals like bears whose behavior 'changes with the natural environment.' This is a method that finds groups by focusing on 'density,' automatically detecting areas where sightings are concentrated as a single unit, while excluding sporadic sightings (so-called 'stray bears').
2. Why 'density-based': The shape of bear behavior
When bear behavior is represented by points on a map, mysterious characteristics emerge. For example, points may line up along mountain ridges or be distributed in long, narrow strips along rivers. This is evidence that bears are traveling through valleys to find food or migrating seasonally through forests rich in acorns.
Such 'long and narrow,' 'distorted,' and 'irregular' distributions cannot be well represented by the conventional k-means method. k-means decides in advance 'how many clusters to divide into' and tries to create round-shaped groups. However, bear behavior is not round. It shows band-like distributions along terrain, such as along valleys and mountain ridges. In contrast, DBSCAN 'naturally groups only high-density areas,' so it can reproduce shapes close to the reality of bear behavior.
3. The meaning of treating a single bear as 'noise'
Among sighting information, there are many reports of 'only once' or 'only one head.' While such sporadic sightings are important for countermeasures, they
do not indicate behavioral patterns. DBSCAN automatically treats points where a certain level of density cannot be confirmed as 'noise' (outliers) and excludes them. Through this processing, it is possible to
highlight only the 'truly dangerous areas.'
This is extremely important for government and researchers. To 'accurately grasp danger zones,' it is necessary to look at the 'trends' of behavior without being misled by sporadic sightings. DBSCAN can be said to be a method perfectly suited for that purpose.
4. Clusters are determined 'automatically'
When you hear clustering, you might imagine some difficult settings, but DBSCAN is simple. You only set two things—
'how close they must be to be in the same group (eps)' and 'how many items must be gathered to be considered a group (minpoints)'. Just by deciding these two, the number of clusters is determined automatically. Since the range of bear behavior changes with the seasons, this 'flexibility' is extremely important.
In fact, when we analyzed bear sighting data in northern Akita Prefecture,
a dense cluster in autumn in the Yoneshiro River basin (near Kitaakita City, latitude 40.0086, longitude 140.3679)
a sighting cluster centered on orchard areas in the southern part of Noshiro City (latitude 40.1744, longitude 140.0381)
a movement route type cluster in the mountainous area between Takanosu and Kamikoani (latitude 39.9465, longitude 140.2375)
emerged clearly.
5. Grasping danger zones as 'areas' with ConvexHull
The technology of enclosing a collection of points with lines to make them an 'area' is called **ConvexHull**. When you enclose bear sighting clusters with ConvexHull, 'danger zones' appear visually on the map. This can be utilized when the government designates 'this range as a priority alert' or when deciding patrol routes.
For example, when you enclose the clusters in the Yoneshiro River basin with a Convex Hull, a 'fan-shaped danger zone' is drawn from the mountain foothills to the rural areas.
This matches locations where encounters with people actually occur frequently, and it is an analysis result that aligns with both scientific data and field experience.
6. Matching bear ecology with cluster distribution
The shapes of the clusters extracted by DBSCAN match the bears' seasonal behavior surprisingly well.
Season | Main Cluster Distribution | Behavior/Food Resources | Spring | Along valleys and streams | Butterbur, budding plants | Summer | Mountains to orchards | Berries, insects | Autumn | Foothills and rural areas | Acorns, persimmons, chestnuts
In other words, the 'density' of the clusters visualizes the 'vector of the bears' appetite'.
The distribution of points becomes a map of behavior, and that map becomes the foundation for safety measures.
7. How even beginners can 'read clusters'
You don't have to be an expert to have tips for reading cluster maps.
Points are dense = The surrounding area is rich in 'food' or 'water sources'
Long and narrow shape = Bear movement paths (valleys, streams, forest roads)
Large area = Seasonal 'concentrated activity area'
The map generated by DBSCAN is not just academic data, but a tool that allows you to intuitively understand 'where and why bears appear'.If you use GIS software (such as QGIS), anyone can try this analysis on a map.
8. Summary: Data leads to 'coexistence with bears'
The purpose of clustering is not to eliminate bears.
Knowing 'where and why they appear' helps prevent damage before it happens—that is the role of data analysis.
The simple combination of DBSCAN and Convex Hull accurately depicts the 'contact points' between bear behavior and human living areas.
For future disaster prevention and coexistence, it is important for the government, researchers, and citizens to look at the same map and share the same reality.And the power to visualize that 'reality' is the greatest value of clustering.
#BearSighting #DBSCAN #ConvexHull #GIS #AkitaPrefecture #BrownBear #ClusterAnalysis #OpenData #DisasterPrevention #Coexistence
▷ Click here
for MapQuest where Iriyama works. ▷ Click here
to learn more about GIS performer Iriyama. ▷ Click here for the GIS software comparison table.