Font Size: 

Climate change, although a global problem, is usually perceived by individualson a local scale, being changes in temperature patterns the most direct indicatorof global warning. Specifically, for an Atlantic climate location (as it happens inGalicia, NW-Spain), spring and fall are expected to present mild transitions betweensummer/winter, but people in this region usually perceive that this transition has disappearedin the last decade. We have considered daily temperature curves along time,being each curve attached to a certain calendar day, and therefore enabling the formulationof a regression model with functional covariate (the temperature curve) andcircular response (the calendar day). Such a model is fitted with observations from theearly 21st century. For observations from recent years (for which the correspondingcalendar day is known), the fitted model is used to predict the calendar day, observinga remarkable misalignment between observed and predicted days for a certaincurve. Apart from the interesting practical results, the methodological proposal isstudied, and the asymptotic bias and variance of a Nadaraya-Watson type estimator,jointly with its asymptotic distribution, is derived. Simulation results support the finite sample performance of the proposal.