Careful observation and accurate reporting are essential for managing land and wildlife and for supporting effective conservation. This presentation proposes a practical tracking approach that could first be tested here in Florida. In much of the African wild, animal locations are currently identified using spotters, trail cameras, drones with cameras, and GPS collars. While these methods are valuable, they are also limited—they are expensive, labor‑intensive, and often incomplete. Data can be distorted by under‑ or over‑counting, double counting, darkness, vegetation cover, distance, or deliberate evasion by animals. Current methods also tend to record animals in isolation, without showing how specific individuals interact with one another or with nearby environmental conditions. This presentation outlines an alternative, complementary approach.

This is part 6/6 in a series of presentations drawing together a thoughtful conclusion from a recent safari to Africa that included parts of Kenya and Tanzania. References and links to the previous posts in this series are listed below.

• Part 1 Glaciers and Plates Influencing Survival
• Part 2 Adaptation in Shared Habitats
• Part 3 Understanding Specialization Extinction
• Part 4 Survival Scenarios
• Part 5 Climate and adaptability

Across Africa’s protected preserves, safari vehicles crisscross the landscape every day. Each vehicle carries tourists and typically two staff members: a driver and a guide. These professionals are highly trained and experienced spotters, able to identify a wide variety of species at a distance and in challenging conditions. They know their local areas intimately and, after years in the field, are often deeply familiar with the habits, territories, and seasonal movements of the animals they see.

PROBLEM STATEMENT:

There are insufficient field data on African cheetah.¹ At the same time, tourism activities create a powerful, largely untapped opportunity. Could the guide/driver teams already in the field provide real‑time information that can be collected and used to estimate species population counts, locations, and times of observation? This presentation offers a proposal for further investigation using scientific method and tools provided by current technology and AI.

The core responsibility of the driver/guide team is to bring tourists to places where wildlife is most likely to be seen. They drive specific routes at predictable times, often starting before sunrise and ending at or after sunset, covering large areas and multiple habitats over the course of a day.

The drivers and guides routinely communicate with one another to share the location, species, and approximate numbers of animals they are observing. They may also describe size or age class and mention nearby species in the same area. These conversations occur over two‑way radios using public broadcast wavelengths within a defined range. Drivers and guides may speak in a number of local languages, but often include English. The result is a continuous, real‑time stream of observations across a network of safari vehicles operating in the same region. Since these communications occur on citizen band radio, they can be received by anyone tuned to the same frequencies. This shared audio stream could be collected and selectively recorded whenever key observational phrases are used. Those key phrases could then be translated into a common base language and digitized. With each broadcast time‑stamped, the extracted data could be linked to other information such as location, weather, hydrology, vegetation cover, aridification, human population density, land use, disease, migration patterns, and availability of forage. The resulting dataset could then be examined statistically to describe the real‑time status of observed animals and plants across the full area of radio reception. Accuracy and usefulness of this method could be evaluated by comparison with existing monitoring approaches.

Null hypothesis:

The collection and analysis of information from the existing telecommunication network is less accurate and less detailed than the sum of existing wildlife monitoring methods.

Method:

Securing the cooperation of drivers and guides in the field is essential to this project. It is in their long‑term interest to support a comprehensive system that helps sustain the wildlife on which their livelihoods depend. In addition to direct observations, there are multiple data sources that must be correlated and cross‑referenced. As critical observers, each participating driver/guide would be identified by call sign, GPS location, and radio frequency. Training for observers should be simple and standardized. It could be delivered via a short YouTube video and an accompanying qualifying test. This process would also help the LLM (Large Language Model) become familiar with the voices and phrasing of participating speakers. All participants should be uniquely identified to support later quality‑assurance and compliance checks.

The resources that could be converted into usable numeric data include:

The method should be tested progressively. Begin with a small pilot group of about 10 participating observers. If early findings are promising, refine the protocol to improve data quality and repeat. Once the method performs reliably according to the revision plan, expand the observer population. Start with English for language interpretation, then add additional languages in order of their frequency of use. Coordinate expansion of observer participation with development of the LLM. Throughout, compare information collected from these broadcasts with results from existing monitoring methods.

Here is a sample outline of the information to be collected and integrated.

Obsrv#

Guide

GPS

Time

Teroir

Cover

Geography

Meterology

Life form

Common name

Count

Behavior

Maturation

Unik Mark

Satellite telemetry location data would be matched to each broadcast GPS coordinate. Multiple environmental factors that influence animal behavior could then be associated with each observation. These include time of day, current weather, geography, vegetation cover, relative hydration/relative humidity, barometric pressure, wind conditions, short‑term weather forecasts, moon phase, fires, and human activities such as migration, construction, or land conversion.

LLM training: AI partner in data conversion

Use an Artificial Intelligence, Large Language Model (AI)(LLM) to review the recorded conversations, identify key observational phrases, and perform language conversion into the chosen common base language.

Not all drivers and guides may wish to participate, for reasons that could include:

• Information may be pirated and used for illegal poaching. This may already be occurring.
• Information could concentrate too many trackers in one location, potentially disrupting normal animal behavior. This effect is already seen in some popular viewing areas.
• Important information may be omitted, reducing the validity and completeness of observations.
• Drivers/guides may decline to track animals due to dangerous human behavior, poor road conditions, business season, or adverse weather.
• Seasonal patterns in tourism may reduce the amount of tracking information available because there are fewer observers in the field.
• Guides and drivers may not be consistently employed from year to year, affecting continuity of observations.

Materials and Data Storage:

Contract with a reliable data‑storage service provider. Storage is required both for the audio recordings and for the structured information in the database.

Radio communication over large areas may require long‑distance signal transmission or repeater stations to relay calls to a central recording site. To manage error, there must be a probabilistic approach to ruling out duplicate identification of the same animal. Over‑ or under‑counting can occur when multiple observers report similar animals in the same group. This possibility must be actively addressed. Two main options exist to reduce the chance of counting the same individual twice.

There are several opportunities to reduce overcounting. Skilled observers can describe distinctive markings, injuries, or behaviors that distinguish individuals, even within the same species. Statistically, it is highly improbable for two truly identical animals to occupy exactly the same territory at the same time, so spatial and temporal information can help separate records. Movement patterns for groups of the same species migrating across a defined territory during a marked period can also be modeled. Finally, this method should be embedded within a broader multi‑observer framework that may include aerial surveys or sample‑based density estimates (count per unit area).

Statistics:

A trained biostatistician will be a critical member of the team, responsible for designing the database, overseeing data collection, facilitating analysis, comparing results between and among datasets, and managing corrections for technical errors. This role also includes correlating field observations with satellite information on ground cover and with weather data for aridity, temperature, cloud cover, atmospheric pressure, fronts, and fire events.

The observation period should extend for at least three consecutive years, long enough to encompass maturation of at least one generation of cheetahs. Real‑time data assessment over this period may reveal progressive changes in the populations of cheetahs and of correlated species. Potential keystone species may emerge from these patterns and could then be investigated in greater depth.

Compare data from this method with other monitoring approaches and with historical records. Paired comparison with current data will allow formal testing of the null hypothesis.

Results:

Direct benefit: As with all field methods, this approach has limitations. Darkness, severe weather, and natural disasters affect these observations just as they do existing techniques. Seasons with reduced tourism will naturally yield fewer observations and may skew results. However, when used alongside traditional methods, this approach could substantially augment current data. Scarce resources for difficult fieldwork could then be targeted more efficiently, guided by patterns revealed through this integrated system.

Added benefits: Instead of tracking only cheetah populations, this system would capture data on all species observed by participating drivers and guides. That greatly expands the value of the project by documenting the density, location, and interactions of many species across the landscape. Over time, the data could reveal competition, mutualism, parasitic relationships, and saprophytic behaviors among species.

NEXT STEP:

Here in SW Florida, a trial program could be launched using a readily visible species such as the Osprey. The same basic steps could be followed with a small team of volunteer observers on organized spotting outings. Several locations could be included, with cell phones used to record and store observations. This is an opportunity for you to become citizen scientists and directly support ongoing research. By testing and refining the method in this pilot study, we could then submit a proven approach to existing research programs in Africa.

Email your interest to john@evergladesark.com

#cheetah, #data collection, #field observation, #population #AI #LLM, #research

References:

1. https://www.wildlifeact.com/about-wildlife-act/wildlife-species/cheetah-acinonyx-jubatus

While well‑visited protected national parks and managed fenced reserves have highly detailed data, roughly 77% of Africa’s estimated 7,100 wild cheetahs live outside protected areas on private, commercial, or communal lands. Because cheetahs are naturally elusive, range over massive territories (up to 1,500 km²), and exist at very low densities, gathering empirical field data on these “free‑roaming” populations is incredibly difficult and has historically relied heavily on informed guesswork.

Don’t forget to like us!

If you wish to receive notifications of new postings, please enter your email address and click on subscribe.

If you want to make a comment, please use the comment box.