3.1
Trend detectability in climatic parameters
Elizabeth C. Weatherhead, CIRES/Univ. of Colorado, Boulder, CO; and G. C. Reinsel, G. C. Tiao, A. J. Stevermer, and J. E. Frederick
How many years of data are needed to detect a trend? This simple question is dependent on a number of issues including the accuracy of the instrumentation, the magnitude of the trend and the fundamental noise characteristics--including autocorrelation. For individual parameters, such as temperature, reasonable estimates can be made for all of these parameters. Incorporating these estimates into statistical analysis can offer probabilistic estimates of how long it will take to detect a specific trend. Work by these authors shows that the length of time is highly dependent on where one is looking for a trend. Changes in many parameters, such as temperature and total column ozone are expected to change but these changes are dependent on location. Similarly, noise characteristics are also observed to depend on location. Careful analysis can show that some places are inherently better for detecting trends in specific parameters.
The analysis involves a number of assumptions such as the statistical model used, the effects of regional averages (such as from satellites), the limitations of sampling, etc. A number of these assumptions have been tested and the results will be explained. In general, the largest uncertainty in this analysis is often due to our limitations in estimating future trends in the environment.
This type of analysis can be helpful for a number of reasons. With a number of atmospheric factors expected to change in the future, it is useful to understand which are likely to show statistically significant trends earliest. For individual parameters, choosing optimal monitoring locations will allow for earlier detection of environmental changes. Optimal monitoring will also allow for scientific questions to be answered earlier and adjustments in our understanding based on the monitoring results. Programmatically, it is useful to establish reasonable expectations for how long it will take to detect the expected trends. Changes in location and number of monitoring sites as well as calibration procedures can be justified in terms of number of years saved in the detection of change.
Session 3, Time series
Wednesday, 10 May 2000, 8:40 AM-9:58 AM
Next paper