Workshops

Estimating Multi-decadal Trends in ENSO: A Case Study Using Darwin SLP

121
reads
Workshops

Andy Chiodi

2012-09-17
12:00:00 - 12:25:00

101 , Mathematics Research Center Building (ori. New Math. Bldg.)

The task of identifying which aspects of climate are undergoing very long term trends from anthropogenic climate change or natural processes has received increased attention in recent decades. High quality, very long climate records are few. Recent improvements to the climate observing system have increased the information available in recent decades, but because geophysical spectra tend to have considerable variance at low frequencies (they are “red”) as well as much shorter frequencies, it is unclear whether reliable estimates of long term trends can be made from multi-decadal or shorter time series.

We use the long (135yr) high quality record of sea level pressure (SLP) at Darwin, Australia, which is a very good proxy for the state of the El Nino-Southern Oscillation (ENSO), to determine the prospects for using shorter multi-decadal records to estimate longer term trends in ENSO. We have found that the distribution of 20 or 30 year sub-segments is evenly split among positive and negative trends and that the long term trend is not statistically significant. Despite this, multi-decadal (20-60 year) trends pass statistical significance tests surprisingly often when they are based on fits to autoregressive models or done via bootstrap calculations. For Darwin SLP, multi-decadal records are not long enough to reliably determine century-scale trends. The recent use of satellite-based observations and the increased sampling of the world’s oceans since the 1950s have provided a wider range of climatically relevant information in recent decades than was available previously. But we suggest caution is needed when imputing meaning to trends found in these recent multi-decadal or shorter geophysical time series. The Darwin record illustrates that a multi-decadal trend may not be a good indicator of longer term behavior.

For material related to this talk, click here.