PARENT SESSION
Symposium #9: Uncertainty and information in ecological forecasting.

Organized by: JS Clark and C Brewer
Tuesday, August 6. 8:00 AM to 11:30 AM. Crystal Ballroom, TCC.

State space models for ecological time series.

Barber, Jarrett^*,1, Lavine, Michael^1,2,3, ¹ Institute of Statistics and Decision Sciences, Durham, NC² Nicholas School of the Environment and Earth Sciences, Durham, NC³ University Program in Ecology, Durham, NC

ABSTRACT- This talk reviews state space models for time series. These models are also known as hidden Markov models and dynamic models. State space models are attractive for the following reasons. 1. They can sometimes give a more parsimonious representation of the data than standard time series models. 2. They can account for observation error. 3. They can show explicitly how parameters change through time. 4. They can be analyzed with off-the-shelf software. These points are illustrated with simulated data and with real data from two examples. One is data from a FACE site in which the goal is to quantify how excess atmospheric CO₂ affects the amount of throughfall. The other is data from the Yadkin River basin in North Carolina where the goal is to see whether land use change is reflected in the hydrological record.

KEY WORDS: statistics, kalman filter, dynamic models