Testing reanalysis datasets in Antarctica: Trends, persistence properties and trend significance

TL;DR

This study evaluates the reliability of three major Antarctic reanalysis datasets by comparing their climate trend estimates, persistence properties, and significance levels with observational data over 35 years.

Contribution

It provides a comprehensive assessment of reanalysis datasets' accuracy in capturing Antarctic climate trends and persistence, highlighting their strengths and limitations.

Findings

NCEP1 and NCEP2 show spurious warming trends.

ERA-Interim is generally reliable except at Amundsen-Scott.

Reanalysis datasets differ significantly in trend significance and persistence properties.

Abstract

The reanalysis datasets provide very important sources for investigating the climate dynamics and climate changes in Antarctica. In this paper, three major reanalysis data are compared with Antarctic station data over the last 35 years: the National Centers for Environmental Prediction and the National Center for Atmospheric Research reanalysis (NCEP1), NCEP-DOE Reanalysis 2 (NCEP2), and the European Centre for Medium-Range Weather Forecasts Interim Re-Analysis (ERA-Interim). In our assessment, we compare the linear trends, the fluctuations around the trends, the persistence properties and the significance level of warming trends in the reanalysis data with the observational ones. We find that NCEP1 and NCEP2 show spurious warming trends in all parts of Antarctica except the Peninsula, while ERA-Interim is quite reliable except at Amundsen-Scott. To investigate the persistence of the data sets, we consider the lag-1 autocorrelation $C(1)$ and the Hurst exponent. While $C(1)$ varies quite erratically in different stations, the Hurst exponent shows similar patterns all over Antarctica. Regarding the significance of the trends, NCEP1 and NCEP2 differ considerably from the observational datasets by strongly exaggerating the warming trends. In contrast, ERA-Interim gives reliable results at most stations except at Amundsen-Scott where it shows a significant cooling trend.