The high-frequency measurements of valve activity in bivalves (e.g., valvometry) over a long period of time and in various environmental conditions allow a very accurate study of their behaviors as well as a global analysis of possible perturbations due to the environment. Valvometry uses the bivalve's ability to close its shell when exposed to a contaminant or other abnormal environmental conditions as an alarm to indicate possible perturbations in the environment. The modeling of such high-frequency serial valvometry data is statistically challenging, and here, a nonparametric approach based on kernel estimation is proposed. This method has the advantage of summarizing complex data into a simple density profile obtained from each animal at every 24-h period to ultimately make inference about time effect and external conditions on this profile. The statistical properties of the estimator are presented. Through an application to a sample of 16 oysters living in the Bay of Arcachon (France), we demonstrate that this method can be used to first estimate the normal biological rhythms of permanently immersed oysters and second to detect perturbations of these rhythms due to changes in their environment. We anticipate that this approach could have an important contribution to the survey of aquatic systems.