Counting Kruger’s rhinos
Photo by Cathy Greaver, Savanna and Grassland Research Unit, SANParks
Numbers of animals are often of interest and importance for conservationists and the public. Trends in numbers are key indicators of conservation effectiveness. There are numerous ways to determine how many individuals of a species live in a park at a specific time, including aerial surveys, call-up surveys, dung counts and camera trapping. In smaller areas, individual identification allows registration studies like wild dog and cheetah photo censuses. These require an identity register of every animal (similar to ID registration of citizens at Home Affairs), which can be difficult for certain species that are not easy to photograph.
In some instances, mark-recapture methods use the re-sighting probability of marked individuals to determine what proportion of the population observers missed and thereby helps derive total population size.
Another approach is a total count, which is like a human population census – observers work through the entire park and count every individual they see. Applying total counting methods in large areas is easier if surveyors use a helicopter, fixed-wing aircraft, or even drones.
Aerial approaches for counting animals should be easy, right? To check if this is true, take 30 seconds to count the number of elephants you see in the aerial picture below.
Now compare your count to the number of red dots on the same picture below – there are 10 to be precise! Your count was most likely less than 10 as the number of elephants are hard to see because they are under trees. Your count suffered from one type of bias, namely availability bias!
There is also a good chance that someone else counts a different number of elephants on the picture than you do. You most likely have different observation skills – which leads to observer bias. You have just experienced two of the factors that influence surveyors every year when they count rhinos!
Both total counts and surveys that count animals on a sample of blocks suffer from these kinds of problems. They include detectability, availability and observer biases that affect the accuracy of a count as well as sampling and survey errors that influence the precision of any estimate of population size. All of these affect our ability to detect changes (see infographic).
Infographic by Corli Wigley-Coetsee, Savanna and Grassland Research Unit, SANParks
Given all these biases and sample errors, which affect the accuracy and precision of our estimates, we use statistical confidence intervals – that is the estimated range with a 95% chance of including the actual number of rhinos. We report our actual estimate within the 95% confidence interval as well as the range around this when we report rhino population sizes. The narrower the confidence interval, the better the precision of the estimate.
What works best for rhinos?
This depends on the size of the area as well as the size of the rhino population. Generally, larger areas with more rhinos lend themselves to the use of aerial count techniques. However, dense vegetation affects the use of aerial surveys, as these are not good for elusive animals living in dense forests, such as the Javan rhino.
In Kruger, we estimate the population size for white and black rhino through a sample block counting method. We survey 50% of the total area (Kruger is almost 2 million hectares) using 3 km x 3 km blocks randomly placed across 35 different landscapes and then extrapolate for the whole park.
So, why don’t we just survey the total area?
Research, including our own study, shows that the precision of a population estimate improves as an aerial survey covers more and more area until the coverage gets to about 40% of the total area. At this point, the precision of a population estimate levels off so that more counting does not give a better population estimate. We, therefore, settled on counting 50% of the total area to get meaningful and efficient estimates of rhino numbers in terms of time and money. Bear in mind that a typical rhino survey in Kruger takes a month and costs almost R 2 million in aircraft costs alone.
So how do we get from the total number counted on the blocks to a population estimate for the park?
We use a well-known estimation process that calculates the density of rhinos on the blocks surveyed within a landscape type. By multiplying that density by the total area of that landscape type in Kruger; correcting for availability, observer and detectability biases; and using the range of density estimates on blocks within the same landscape, the method provides an estimate and the 95% confidence interval for the number of rhinos in that specific landscape. Adding the total numbers estimated for each of the 35 landscapes then gives the rhino estimate for Kruger.
These influences of biases and sample error are very prevalent even when authorities do total counts. You may hear a single, clean figure being presented with total count estimates, but as you have experienced with counting elephants, this cannot be the case, especially when counting animals over very large areas.
Thus, counting rhinos (or other large animals across large landscapes) using aerial observations is not as easy or predictable as counting the money in your wallet. Scientists take into account the many influences affecting population estimates to explain the study caveats and must present the uncertainty in the data. This is not a result of poor study design, but in recognition of the limitations of methods used and the inherent variability in the biological systems we attempt to measure. Nevertheless, repeated population estimates, taking into account the various biases and errors, show trends over time. These are important for understanding how populations respond to threats and management actions designed to address those threats.
