Are Global Tide Gauge Data Stationary in Variance?

Tide gauges distributed all over the world provide valuable information for monitoring mean sea level changes. The statistical models used in estimating sea level change from the tide gauge data assume implicitly that the random model components are stationary in variance. We show that for a large number of global tide gauge data this is not the case for the seasonal part using a variate-differencing algorithm. This finding is important for assessing the reliability of the present estimates of mean sea level changes because nonstationarity of the data may have marked impact on the sea level rate estimates, especially, for the data from short records.     

The state‐of‐the‐art in marine navigation

Tide gauges distributed all over the world provide valuable information for monitoring mean sea level changes. The statistical models used in estimating sea level change from the tide gauge data assume implicitly that the random model components are stationary in variance. We show that for a large number of global tide gauge data this is not the case for the seasonal part using a variate-differencing algorithm. This finding is important for assessing the reliability of the present estimates of mean sea level changes because nonstationarity of the data may have marked impact on the sea level rate estimates, especially, for the data from short records.