A0385
Title: A Bayes space point of view on climate warming
Authors: Christine Thomas-Agnan - CNRS (France) [presenting]
Abstract: The impact of temperature warming is often analyzed with temperature temporal summaries, which can lead to a loss of valuable information. To address this issue, it is possible to take temperature into account as a functional parameter (function of time). An alternative and complementary point of view is put forward, which considers temperature distributions over time periods. Using the Bayes space formalism, the yearly distributions of maximum (or minimum) daily temperatures are analyzed across Vietnam's provinces over a 30-year period (1987-2016). The daily maximum temperatures are preprocessed using maximum penalized likelihood, resulting in density samples expressed on a B-spline basis. First, the presence of outlying densities is investigated both spatially and temporally with the ICS method (invariant component selection) adapted for density objects. Regional effects of the temperature density relative changes of these distributions between the initial period 1987-1989 and the final period 2014-2016 are examined using a Bayes-space version of functional analysis of variance. Lastly, the impact of climate warming is assessed on rice yield production using a scalar on density regression model.
