Monitoring approachshare of 246
The central finding
Sensors turn people into counts and paths. Vision turns scenes into actions and interactions.
01 · The search
From a wide evidence field to a focused corpus.
Nine technology terms were paired with ten natural-environment groups across Web of Science and Scopus. Screening followed PRISMA-ScR.
0.92%
of all identified records entered the final evidence base. Every step narrows the question from technology in public space to empirical monitoring of human behaviour in nature.
WoS 11,108Scopus 15,657
IdentificationRecords identified
26,765DeduplicationAfter duplicates removed by title
15,970Preliminary eligibilityRecords screened
12,078Title screeningTitles assessed by eligibility
2,145Abstract screeningAbstracts assessed by eligibility
306IncludedArticles in the review
24602 · The field
The literature accelerated. Its centre of gravity did not.
By 2024, sensor-based studies still formed four-fifths of the corpus. Vision and hybrid methods are growing, but remain comparatively narrow streams.
Cumulative publications by monitoring approach
Primary capture technologystudies
GPS apps
121
GPS devices
62
Computer vision
39
Other technologies
24
03 · What becomes visible
Human behaviour appears in three dimensions.
Most studies examined more than one dimension. Spatial distribution is not a fourth behaviour: it is the shared analytical lens arranging presence, movement, and activity across space.
Presence is the count.
Who is there, how many people visit, and when demand peaks. Scalable sensor data make this the most established dimension.
Behavioural dimension × methodbubble = studies
COUNTVisitor volume across place and time
04 · From method to evidence
Technology becomes evidence through an analytical pipeline.
The restored three-stage Sankey follows each study from capture technology, through its algorithm family, to the behaviour it ultimately makes visible.
Technology → algorithm → behaviourflows = studies
The diagram preserves the review’s original method pipeline: GPS dominates location-based analysis, while camera and hybrid pipelines contribute most strongly to activity recognition.
05 · Begin with the question
Do not choose a tool before deciding what must become visible.
Method selection is a conceptual choice as much as a technical one. Explore how different management questions lead to different evidence strategies.
Visitor pressure
Start with counting.
Counters, passive wireless sensing, and aggregated mobility data can reveal volume and peak demand. They are efficient, but say little about activity or experience.
Best-fit evidencePresence
Typical methodsCounters · Wi-Fi · mobility data
Critical limitationContext remains thin
06 · The open frontier
A fuller view requires more than a more powerful sensor.
The review points toward integration, representativeness, outdoor-specific models, and privacy-by-design.
01 · CONNECT
Link reliable counts with continuous movement.
Fixed entrances miss less prominent routes; detailed GPS studies often remain small. Future systems need both scale and continuity.
02 · TRAIN
Build models for behaviour outdoors.
General vision datasets miss subtle actions such as landscape appreciation, group interaction, and wildlife observation.
03 · INCLUDE
Design for people missing from digital traces.
Smartphone and platform data can underrepresent older adults, children, and marginalized groups — a serious limitation for equity research.
04 · PROTECT
Treat privacy as part of the method.
Combining location and imagery can increase reidentification risk. Data minimization, local processing, anonymization, and clear disclosure are essential.
Published research
Monitoring human behaviors in natural environments using sensor and vision technologies
Mingze Chen & Keunhyun Park · Environmental Development · 2026 · Article 101546