Context . Distance sampling is widely used to estimate the size of wildlife populations, including kangaroos. However, the performance of distance-sampling abundance estimates has seldom been evaluated for wild mammal populations of known size.

Aims . We evaluated the precision, accuracy, bias and interval coverage of abundance estimates from walked line-transect sampling, a commonly used distance-sampling method, for a marked free-ranging population of eastern grey kangaroos (Macropus giganteus) at Yanakie Isthmus, Wilsons Promontory National Park, south-eastern Australia.

Methods . In each of two study periods (November 2012 and May 2013) we first determined the true size of the uniquely marked kangaroo population by conducting 10 intensive searches of the study area. We then conducted distance sampling along six systematically spaced line transects. We walked each transect four times in November 2012 and seven times in May 2013. Data were analysed using Program DISTANCE.

Key results . Our intensive searches revealed that 141 and 124 collared kangaroos were present in the study area in November 2012 and May 2013, respectively. When transects were walked four or more times (i.e. ≥400 observations), maximum precision (coefficient of variation; CV of ∼13%) was achieved in both survey periods. Walking transects twice (i.e. ∼200 observations) produced abundance estimates with CVs of <20% in each study period. The accuracy (root mean square error) of abundance estimates varied from 1 to 13 (November 2012) and from 3 to 28 (May 2013). Bias ranged from −9% to 23%, but stabilised at between −1% and −9% when transects were walked four or more times in each study period. The 95% confidence intervals for the abundance estimates always included the true population size.

Conclusions . Our results indicated that walked line-transect distance sampling is a precise and accurate method for estimating eastern grey kangaroo abundance. The small negative biases that occurred when sample sizes were large were likely to be due to some animals moving outside the study area.

Implications . Provided that the key design elements and assumptions are met, estimates of kangaroo abundance from walked line-transect distance sampling should have good precision (CV < 20%) and minimal (<10%) bias.