If you have been reading previous blogs, I am sure that by now, you are familiar with the fact that ecologists like to count things, especially animals! And wildlife managers like doing that even more than we do, maybe because they need that data to manage their area and populations efficiently.

It is said that there are many ways to skin a cat, and so there are for counting animals! Hence our main objective in our paper recently published in HYSTRIX was to compare several methods that could be used to estimate the giraffe population from waterhole monitoring on Ongava Game Reserve. We used an image dataset from 61 camera traps deployed for 20 days during the 2016 dry season.

The first step was, therefore, to establish which individuals were present in each of the 30 913 images containing giraffes… A daunting task that took some time to complete…You can read more about the methods used here. Then we establish a capture history for each individual, summarising when it was seen and where. From there, the statistical work could start! This was performed by our colleague Christophe Bonenfant from the LBBE in France.

But first things first: during these 21 days, we identified a total of 101 individual giraffes (58 adult females and 41 adult males, as well as two juveniles of unknown sex), and the vast majority of them (66%) showed a high site fidelity coming to drink at only one waterhole (giraffes were observed at 10 out of 12 waterholes, but each of these individuals only used one). On average, males visited waterholes every five days and females every four days.

We compared population estimates generated by capture-recapture (CR) models that require individual identification (and are considered a benchmark) to those from rarefaction curves and N-mixture models, which do not need individuals to be uniquely identified. CR models in which the detection probability was allowed to vary between sex and individuals estimated a population of 104 giraffes living on Ongava. In comparison, rarefaction curves produced an estimate of 117 individuals, and N-mixture models yielded very contrasting results of 87 to 215 giraffes depending on the model fitted. We therefore recommend using CR methods considering variability in detection rates to avoid overestimating the population size.

During the same period, human observers were also collecting data at the same waterholes for 72h only. They recorded only the date, time, number of individuals, age, and sex. We obtained a total of 51 giraffe observations for a total of 77 individuals, but some of them were likely the same. And since giraffes come to drink on average every 4-5 days, the observers have missed many individuals! It would require observers to stay much longer at waterholes and take pictures of the giraffes for us to generate reliable population size estimates from this data. While this would be costly in terms of human logistics, it would result in far fewer images, as once observers would have taken a picture of both sides of each giraffe coming to drink, they would not need to take a photo every 30 seconds like the camera trap do. And this is the kind of trade-off that field biologist face daily: which method is the most adapted and cost-effective for the questions that need to be answered?

But whichever method is chosen (overserves or camera traps), make sure to harness the power of Artificial Intelligence to save time in the individual identification process.

If you want to learn more about giraffes on Ongava, head here.You can read the full article here: http://www.italian-journal-of-mammalogy.it/Counting-giraffes-A-comparison-of-abundance-estimators-on-the-Ongava-Game-Reserve,159422,0,2.html