Count data are often used to assess relative population size and population trends with sufficient power and confidence for wildlife population studies, including those for nesting sea turtles. Although access to sea turtles while nesting is relatively simple compared to many other migratory marine animals, optimal surveys tagging every individual through the nesting season are often not feasible due to time, financial and other logistic constraints. Partial survey counts can then be used to estimate population abundance. Several models have previously been published describing the seasonal shape in abundance for nesting turtles, but none have compared different model fits using a numerical approach and all have limited general application as they describe only 1 location or 1 species. We compared 22 non-parametric and parametric modelling approaches for 9 populations of sea turtles comprising 3 different species: green sea turtles Chelonia mydas, loggerhead sea turtles Caretta caretta and leatherback sea turtles Dermochelys coriacea. Although models showed marked differences in the shape of their fit, all models provided reasonable estimates of annual nesting abundance, with mean errors less than 8% for 50% data coverage and mostly 8 to 10% for 20% random coverage. Of the 3 models that produced significantly lower mean absolute error, we recommend using generalized additive models to estimate annual abundance due to their ease of fitting, flexibility across populations and seasonal shapes and their good predictive ability. � Inter-Research 2014.