Norway Comes to New Zealand: Edward Kidson, Jørgen Holmboe, and the Modernization of Australasian Meteorology

Edward Kidson, Director of the Meteorological Service of New Zealand from 1927 until his death in 1939, was an instrumental figure in modernizing Australasian meteorology. Throughout the 1920s, Kidson promoted the methods of synoptic analysis emanating from the Bergen School of Meteorology. However, it was not until the 1930s that he began in earnest to apply these methods to weather charts for the Australasian region. This development was aided by two visits he made to Bergen and by a personal correspondence he maintained with Jacob Bjerknes during the 1930s. In 1932, Kidson presented the first Norwegian-style analysis conducted for a Southern Hemisphere region, promptly following this with a more extensive study. However, these analyses were not of a sufficient standard at that stage to be adopted in forecasting practice. It was the fortuitous visit to New Zealand of Norwegian meteorologist Jørgen Holmboe, in 1934, that finally facilitated the transition. Holmboe was attached to the Lincoln Ellsworth Antarctic Expedition, but damage to their aircraft caused them to spend the winter of 1934 in New Zealand. Holmboe was engaged at the Meteorological Service in Wellington during this period, working with Kidson to apply Norwegian methods to the region. Kidson had hoped to further embed this practice by employing Tor Bergeron in New Zealand during 1938. Bergeron had accepted an offer from Kidson but canceled at the last minute after contracting rheumatic fever. Nevertheless, shortly after Holmboe’s visit, daily analyses were being conducted along Norwegian lines, bringing Australasian meteorology into the twentieth century.

Forecasting Performance Comparison by Using Power Transformation between VAR and Bayesian VAR Models

This article is devoted to examine the performance of power transformation in VAR and Bayesian VAR (BVAR) forecasts, in comparison with log-transformation. The effect of power transformation in multivariate time series model forecasts is still untouched in the literature. We examined the U.S. macroeconomic data from 1960 to 1987 and the Taiwan’s technology industrial production from 1990 to 2000. Our results showed that the power transformation provides outperforming forecasts in both VAR and BVAR models. Moreover, the non-informative prior BAVR with power transformation is the best predictive model and is recommendable to forecasting practice.