Kosciuszko National Park 2024 wild horse population survey

The NSW Government undertakes regular surveys of the wild horse population in Kosciuszko National Park, using internationally recognised scientific methods. This work is essential to inform wild horse management in the park in support of meeting the legal requirement for the wild horse population to be reduced to 3,000 by 30 June 2027.

Reducing horse numbers is essential to protect biodiversity and ensure the park's resilience against climate change and other environmental pressures. This approach seeks to balance cultural heritage with the need to protect the park's diverse and fragile sub-alpine and alpine ecosystems.

The 2024 population survey shows the NSW Government is on track to meet the 2027 target, with early signs of ecological recovery already visible in impacted areas.

Based on feedback from scientists and wild horse advocates, mark recapture distance sampling (MRDS) and thermal camera technology survey methods were trialled alongside standard distance sampling (SDS) which was used in the 2020, 2022 and 2023 population surveys.

The final survey report includes results from standard distance sampling and mark recapture distance sampling methods, while findings from the thermal camera technology trial will be reported once finalised. In the survey report, SDS generally refers to multiple covariate distance sampling (MCDS). More information is provided on this in the section below ('What methods were used for the 2024 survey?')

A new contractor was engaged to design, analyse and report on the 2024 survey. They provided sophisticated data recording hardware and expertise in the use of thermal camera technology. 

The survey blocks cover 39% of the park and were the same as those used in the 2020, 2022 and 2023 surveys (see survey area map). When the Kosciuszko National Park Wild Horse Heritage Management Plan was adopted in 2021 it was estimated that horses occupied 53% of the park.

For the first time, the 2024 survey provides an estimate of the population in 3 of the park's 4 wild horse retention areas, where 3,000 horses are to be retained by June 2027. The Tom Groggin retention area was excluded due to its small size, making an accurate estimate impossible.

Population estimates for survey blocks

Statisticians and scientists use a confidence interval to express how reliable an estimate is. A 95% confidence interval is the accepted convention in biological sciences.

• The 2024 survey using standard distance sampling estimates with 95% confidence that the population in the survey blocks is between 1,579 to 4,007 horses.
• The 2024 survey using mark recapture distance sampling estimates with 95% confidence that the population in the survey blocks is between 2,131 to 5,639 horses.

That compares to the 2023 standard distance sampling survey which estimated with a 95% confidence that the population was between 12,797 to 21,760 horses.

Population estimates for retention areas

• The standard distance sampling method estimates with 95% confidence that the population in the retention areas is 1,766 to 4,050 horses.
• The mark recapture distance sampling method estimates with 95% confidence that the population in the retention areas is 2,373 to 5,717 horses.

The population estimate for the retention areas is higher than in the survey blocks because the southern retention area includes land that was not included in the survey blocks, due to steepness, but is known to contain horses. This means the southern retention area is larger in size than the southern survey block, making the population estimate higher.

Different population estimates show that there can be a lot of variation when sampling, which is why relying on a single population estimate number is avoided. The various estimates suggest that the population is close to or above the goal set by the Kosciuszko National Park Wild Horse Heritage Management Plan, which aims to keep 3,000 horses in specific areas by 30 June 2027.

Next steps

NSW National Parks and Wildlife Service will continue working with experts in the field to ensure the 2025 survey uses the best available methods.

Copies of the raw data sheets will be made available on request. Please email your request to [email protected]

How was the 2024 survey effort different?

The contractor analysed the amount of survey effort required as part of the 2024 survey design, resulting in an increase in the total length of transect lines surveyed. This adjustment aimed to maximise precision and minimise the 95% confidence interval, given the population of horses was expected to be significantly less than in previous surveys.

This additional survey effort meant transect lines (transects) had to be closer together. In the 2023 survey, the transects for the northern and southern survey blocks were spaced 1.5 km apart. Based on previous literature, this had been assessed on a precautionary principle to be the minimum distance between transects before a reasonable risk of double counting, which could occur when horses counted on one transect move to the adjacent one and are re-counted.

For the 2024 survey, the distance between transects was decreased to 1.23 km in the southern block and 1.37 km in the northern block. To minimise the risk of double counting, every second transect line was surveyed on one day, and on a separate day the alternate lines were surveyed. This method of surveying was used in previous surveys for the Cabramurra and Snowy Plains blocks and was used again for these blocks in 2024.

What methods were used for the 2024 survey?

Consistent with the 2020, 2022 and 2023 surveys, standard distance sampling was the primary survey method used for the 2024 survey. Aerial surveys using distance sampling are regarded scientifically as international best practice for sampling large animals across broad landscapes.

NSW National Parks and Wildlife Service are open to trialling alternative survey methods to improve the accuracy of population estimates. One such method trialled in the 2024 survey, and also suggested by horse advocates, is mark recapture distance sampling.

Standard distance sampling

This method involves 2 specialist observers seated on either side of a helicopter counting groups of horses in distance categories within a 300 metre wide transect (i.e. 150 metres on either side of the transect centreline). Animals get more difficult to see the further they are from the transect, so the observer must record the estimated distance of the animal from the transect centreline.

Not all of the horses within the 300-metre-wide transects are seen by observers. This is a proven concept in peer-reviewed scientific papers and has been verified in published studies performed on animal populations of known size. Hence, the results are analysed in a software package known as Distance, which adjusts the counts made by the observers based on the reduction in sightability, also known as the detection function (or g(x) in the report). That is, an adjustment is made by calculating the proportion of horses that were actually seen by the observers.

An analysis using only the detection function is known as conventional distance sampling (CDS). However, other variables, such as vegetation cover, north or south side of aircraft, time since sunrise/to sunset and the seating position of each observer may also influence sightability and can be used to adjust the detection function. This is known as multiple covariate distance sampling (MCDS). Conventional distance sampling and multiple covariate distance sampling are the 2 methods of standard distance sampling (SDS).

Data was analysed using both conventional distance sampling and multiple covariate distance sampling. Unsurprisingly, multiple covariate distance sampling provided more robust results with vegetation cover ('habitat') as the covariate most affecting sightability.

Further information on distance sampling can be obtained at Distance project.

Mark recapture distance sampling

This method uses the same approach as standard distance sampling (SDS) with the addition of a third observer in the front left-hand seat of the helicopter (the pilot sits on the right).

This method has been avoided in the past as our standard operating procedures for this type of low-level flying operation require a dedicated navigator to sit in this position. Through a risk assessment and the inclusion of additional safety mitigations this method was approved as a trial for this operation.

In mark recapture distance sampling, 2 observers sit on the left side of the aircraft to view the same horses – the front observer 'marks' a particular animal (by counting it), and the rear observer may or may not independently 'recapture' it (by recounting it), hence the name.

The main difference between mark recapture distance sampling and standard distance sampling is that standard distance sampling assumes the observers see 100% of animals on the centreline of the transect (i.e. at zero distance from the observer in the helicopter). Mark recapture distance sampling does not make the same assumption and analyses the counts from the observers at the front and rear of the left-hand side of the aircraft to determine the proportion of horses that are seen on the centreline of the transect. This perception bias is the probability of detecting a horse on the transect centreline. As a result of this additional analysis, mark recapture distance sampling is regarded as providing a more accurate population estimate.

The probability of detecting horses on the transect centreline can then be used to adjust the estimates for multiple covariate distance sampling, which are presented in the final report as mark recapture distance sampling estimates.

Horses and foals

Observers distinguish adult horses from foals, which allows horses and foals to be analysed separately, and combined.

In the report, the population estimate for mark recapture distance sampling (MRDS) is slightly higher for adult horses (3,949 within the survey area) than for adult horses and foals combined (3,885). While this seems counterintuitive, it is due to the detection function being impacted by the low numbers of recorded foals, which in turn has negatively skewed the density estimate.