DATE: Friday, October 26, 2018 TIME: 2:30 pm, refreshments at 2:15 pm PLACE: Marine Science building, the Alampi room 71 Dudley Road, New Brunswick, NJ
Michael Stein
University of Chicago, Department of Statistics
Statistical Modeling of Environmental Time Series
When modeling environmental time series, proper accounting of periodic phenomena is often critical. Seasonal, diurnal effects and their interactions are the most common, but other periodicities may also be present, such as those related to characteristics of the Earthâs orbital cycles. Climate change may affect seasonal and diurnal cycles, further complicating their modeling and estimation. This talk will explore these issues through two examples. The first is estimating quantiles of daily temperatures as they evolve over seasons and years based on large initial condition ensembles of climate model runs. The resulting quantile maps can be used to transform observational data to provide simulations of future climate. The second example reports on some preliminary work on coastal water levels, which are impacted by numerous periodic tidal phenomena related to the relative positions of the Earth, Moon and Sun. Current efforts to address this problem include the use of what is known as âskew surgeâ, which is the difference between the observed maximum water level and the predicted high water level for each tidal cycle. Recent research has argued that skew surge and predicted tides are close to independent, which, if true, should enable sharper estimates of extreme water levels than by analyzing the water levels themselves. Unfortunately, using the official NOAA predicted tides, this assumption of independence is clearly untrue at some locations. Fortunately, it is possible to explain much of the variation in the skew surges using statistical models with periodic components, which should, in principle, lead to better estimates of extreme water levels.