Marine turtles are a taxon of world-wide conservation concern. Effective long-term monitoring is hampered by the fact that populations are widely dispersed except during the breeding season. Thus most monitoring programmes focus on nesting beaches, necessitating resource intensive studies, often over months, that could conceivably, be less comprehensive and focus on more parsimonious sampling. We analyse 11 years of exhaustive monitoring data for two species of Mediterranean marine turtles (Chelonia mydas and Caretta caretta). We resample using a variety of plausible sub-sampling regimens to estimate the total annual nesting population. We project our dataset into the future applying a range of population change rates to explore how adopting a monitoring programme based on sub-sampling would, for example, affect our ability to detect population decline. We show that accurate annual population estimates can be achieved with as few as 14 days of survey effort providing monitoring spans the peak of nesting. Furthermore, the modelled impact of sampling-based monitoring suggests that the duration to detect population change is not increased greatly. Our findings have implications for all marine turtle monitoring and have applicability to other animal groups. It is often considered desirable to perform exhaustive monitoring, with aversion of basing policy recommendations on partial data. However, comprehensive long-term monitoring programmes, particularly in developing nations, although presenting a number of advantages, are often impossible. Accurate total annual censuses can be achieved through a variety of sub-sampling regimens without sacrificing the ability to detect changes in the population trends over time. In this example, a solid block of at least 3 weeks sampling that encompasses the peak of the nesting season is advised. (c) 2008 Elsevier Ltd. All rights reserved.