zyp: ZYP Prewhitened Nonlinear Trend Analysis Package

In zyp: Zhang + Yue-Pilon Trends Package

ZYP Prewhitened Nonlinear Trend Analysis Package

Description

Update April 1 2022

At the time of writing, the application of the Yue Pilon (Yue et al. 2002) is actively discouraged in the hydrologic community. This is because the trend free whitening procedure (TFWP) applied in the Yue Pilon method shows very high Type-I error rates with increasing autocorrelation and hence the rate of false trend detection with this method is unacceptable (Buerger 2017; Zhang and Zwiers 2004). This version of zyp has been modified to apply Zhang et al. 2000 trend detection methods as default.

Additionally, preserve.range.for.sig.test is set to TRUE throughout to make sure significance is reinflated as per the Yue and Pilon paper exactly.

This function includes two approaches to analyze for trend, the Zhang method and Yue and Pilon method. These differ in their approach to pre-whitening to removing lag-1 autocorrelation. The magnitude of the trend is computed using the Theil-Sen approach (TSA).

In the Zhang method, trend is removed from the series if it is significant and the autocorrelation is computed. This process is continued until the dfferences in the estimates of the slope and the AR(1) in two consecutive iterations are smaller than 1 percent. The Mann-Kendall test for trend is then run on the resulting time series and TSA is used to compute the slope of the trend.

In the Yue and Pilon method, the slopes is estimated with the TSA, if almost equal to zero, then its is not necessary to conduct the trend analysis. If it differs from zero, then it is assumed to be linear and the data is detrended by the slope and the AR(1) is computed for the detrended series. This is referred to as the Trend Free Pre-whitening (TFPW) procedure. The residuals should be an independent series. The trend and residuals are then blended together. The Mann-Kendall test is the applied to the blended series to assess the significance of the trend.

One variance of this package, at least from the Yue and Pilon method, is that by default the values used to compute significance are reinflated by dividing by (1 - AR(1)). Empirically this provides better results. However, if you do not desire this behaviour, it can be controlled using the parameter 'preserve.range.for.sig.test'; setting this to 'FALSE' should give results which follow the Yue and Pilon paper exactly.

References

Buerger, G., 2017. On trend detection. Hydrological Processes 31, 4039-4042.

Wang, X.L. and Swail, V.R., 2001. Changes in extreme wave heights in northern hemisphere oceans and related atmospheric circulation regimes. Journal of Climate, 14: 2204-2221.

Yue, S., P. Pilon, B. Phinney and G. Cavadias, 2002. The influence of autocorrelation on the ability to detect trend in hydrological series. Hydrological Processes, 16: 1807-1829.

Zhang, X., Vincent, L.A., Hogg, W.D. and Niitsoo, A., 2000. Temperature and Precipitation Trends in Canada during the 20th Century. Atmosphere-Ocean 38(3): 395-429.

Zhang, X., Zwiers, F.W., 2004. Comment on "Applicability of prewhitening to eliminate the influence of serial correlation on the Mann-Kendall test" by Sheng Yue and Chun Yuan Wang. Water Resources Research 40.

See Also

zyp.trend.csv, zyp.trend.vector.

zyp documentation built on March 31, 2023, 10:04 p.m.