Trend Detection in Regional-Mean Temperature Series: Maximum, Minimum, Mean, Diurnal Range, and SST

Abstract

Regional climate trends are of interest both for understanding natural climate processes and as tests of anthropogenic climate change signatures. Relative to global trends, however, their detection is hampered by smaller datasets and the influence of regional climate processes such as the Southern Oscillation. Regional trends are often presented by authors without demonstration of statistical significance. In this paper, regional-average annual series of air temperature and sea surface temperature for the New Zealand region are analyzed using a systematic statistical approach that selects the optimum statistical model (with respect to serial correlation, linearity, etc.), explicitly models the interannual variability associated with observable regional climate processes, and tests significance. It is found that the residuals are stationary and are a red noise process [ARMA(1,0)] for all the series examined. The SOI and a meridional circulation anomaly index (both high-pass filtered) are ?explanatory variables? for interannual variability. For national-average air temperature (AT), linear and exponential trend models are equally valid but for simplicity the linear model is preferred. Failure to model the serial correlation in AT would result in an estimated confidence interval for trend that is too small by 36%. The use of the explanatory variables tightens the confidence interval by 15%. Significant trends were detected. The trend in AT for 1896?1994 is 0.11 ± 0.035°C decade?1 (95% confidence interval). This is about double the trend reported for global data, which may be due to the relative absence of sulfate aerosols in the South Pacific region. The trends in maximum and minimum temperature over this period are not statistically different. However, for the later period of 1951?90, the trend in maximum temperature reduces to an insignificant value, while the trend in minimum temperature remains high, resulting in a significant downward trend in diurnal range of 0.10°C decade?1. Similar diurnal range behavior in other regions has been tentatively attributed to increasing cloudiness. The trend in a regional SST series for 1928?94, 0.07°C decade?1, is about half the trend in AT for the same period. The trend in the difference, SST?AT, ?0.06°C decade?1, is statistically significant. This implies the existence of an atmospheric warming source for the additional air temperature trend, and may mean that the heat fluxes between the atmosphere and ocean in the New Zealand region are subject to a large trend, with the direction of flux change being toward the ocean. The results of the study are consistent with the IPCC predictions of climate change.