Introduction and Objectives Since the 1950s in France, there has been a continuous increase in tourism, and in recent years, a significant growth in rural and medium mountain areas, notably linked to the presence of natural sites classified as 'must-visit'. In these sites, alongside the direct rise of tourist activity, often comes that of excursionists and locals, which has seen a strong increase in recent years, emphasized by the attractiveness of these territories, as wells as post-pandemic lifestyle changes and the rise of teleworking. These visitor flows in these protected natural sites poses challenges both in terms of economic and social implications (valorisation) and environmental impacts (protection). Consequently, monitoring visitors of natural sites is a crucial element of management policies, aiming to meet the dual objective of preserving environments (managing tourist pressure) and developing the tourism economy, acting as a lever for local development in terms of jobs and income (Vlès). However, the monitoring of these sites, often geographically extensive, requires the mobilization of heterogeneous data sets that are complex to collect, to combine and to process. Thus, a basic stake for all actors of management policies is to quantify and qualify these visitor flows, and associated economic flows. Furthermore, in order to evaluate the trends and/or the effects of present policies, it is necessary to measure the evolution of these flows, and then to develop replicable methods allowing year-on-year comparisons of indicators related to economic (revenues, employment...), social (nuisances...), and environmental (pressure on environments, habitats...) impacts. Moreover, continuity in visitor monitoring relies as much as possible on the implementation of reliable but neither too cumbersome nor too costly devices, for regular and easy replication over time. The objective of this presentation is to provide some insights to help developing a methodological framework for monitoring visitors in natural sites, based on the approach used to estimate visitor numbers in a French natural site, and some experiments conducted to assess alternative methods. The scope is the collection of visitation data: number and type of visitors, types, places and dates of practice. Further possible objectives of monitoring (analysis of motivations, evaluation of economic impact) are object of other publications (in progress). Methodology Within the framework of the VALO-SITES project (Methods for evaluating the socio-economic effects of management and valorisation policies of natural sites, 2019-2022), a series of experiments on monitoring visitors of natural sites was conducted with the objective of proposing reproducible methods for estimating the total number of visitors of a site, for conducting visitor surveys, and evaluating the economic impact of tourism (Marsat, 2022). These experiments were primarily conducted in Chaîne des Puys – Faille de Limagne, a natural site, part of Auvergne Volcanoes Regional Nature Park, since its establishment in 1977, which has been subject to different management policies, resulting in its inscription on UNESCO's World Heritage List in 2018. This study area clearly illustrates the complexity of these sites, notably due to their openness (multiple entrances, transit, numerous accesses, often interconnected paths...), or the diversity of visitors and practices (hiking, tourist trains, paid sites, paragliding, swimming/boating...). These factors make monitoring visitors even more challenging, while the exercise is already particularly tricky due to various problems in counting visitors (diffuse visitation, double counting...), due to often heterogeneous, incomplete, and not always reliable data, as well as to the absence of any benchmark to assess our findings (Spenceley, 2021). Results First, we present the different data sources used, such as: eco-counters (pyroelectric sensors mounted on trails leading to main summits), tickets sold at main attraction sites, and visual counting in tourist office. We then propose some methods to exploit available data sources, notably to clean eco-counters’ series and infer missing data, and to extrapolate visitor numbers in unequipped areas. Alternative methods were also tested, such as Flux Vision by Orange (Berlingué & Contet, 2017), a service providing indicators about visitors based on Orange mobile phone network data, or vehicle detection in parking areas using video and road counting. These experiments allowed for comparing these different approaches in terms of feasibility, effectiveness, and reliability. Secondly, we propose an approach to produce a comprehensive estimate of the total number of visitors in 2019, and to reconstruct daily series of visitor numbers. Based on the data collected from an extensive on-site questionnaire-based survey we conducted in 2019-2020 among more than 1,000 visitors, several parameters (proportion of tourists, recurrence, proportion of multiple visits) were estimated, enabling the conversion of a number of visits into a number of visitors, and distinguishing tourists from excursionists. Thirdly, we consider different combinations of counting and survey methods, which can be relevant, depending on their advantages and limits, and depending on the specificities of each site, to take place in the frame of long-term strategies. We comment a two steps process, which starts with an initial inventory (as complete as possible) followed by cheaper regular updates. Discussion and Conclusions Finally, we will discuss the potential of this mix of methods, and discuss their limitations, in order to propose recommendations to help decision-makers in site management policies. Although no method is universally applicable to all natural sites, this approach can provide useful insights to process and combine similar set of heterogeneous and fragmentary data, to conduct this kind of global estimation of visitor numbers in a natural site, and to implement a visitor monitoring system.