Stochastic Modeling and Analysis of Temperature Data From Hot Springs in Yellowstone Caldera, Wyoming, USA

摘要

We monitored temperatures in three geothermal springs located in Lower Geyser Basin of the Yellowstone caldera, Yellowstone National Park, Wyoming, USA. The observed temperatures were analyzed using autoregressive moving average (ARMA) models to represent the underlying generating processes responsible for the observed temperature distributions. An attempt was made to attribute physical significance to the ARMA model parameters by comparing the time lags of the model coefficients with characteristic times observed for the same springs in a related study, but this effort was unsuccessful. In spite of the empirical nature of ARMA models, and their apparent lack of relationship with readily observed physical drivers, the fact that they provide minimum variance, optimal forecasts of time-correlated univariate variables, their ability to detect changes in time-series structure, and their usefulness in bootstrapping parameter estimates give them a high degree of utility, which has been underutilized in the geosciences.