ANALYSIS OF CHANGES OF MEAN IN IRREGULARLY SPACED AUTOCORRELATED DATA OR - DOES MATIMBA POWER-STATION CAUSE A HEAT-ISLAND EFFECT

With the planning of South Africa's first totally dry-cooled power station (Matimba, 3990 MW total generating capacity), the question arose as to whether the facility would create a heat island effect. This was to be tested by examining the air temperature both before and after the building of the power station, and the commissioning of the turbines. Factors affecting air temperature, such as altitude, season, time of day and atmospheric conditions, are taken into account. Prevailing wind speed and direction are also taken into account, as these will affect the presence of a heat island effect. Problems arise in the statistical analysis in that the data are autocorrelared, so that the standard rests are nor applicable. Further taking wind speed and direction into account results in data series that are measured at highly irregular intervals, so that standard time series analyses could not be used. This article discusses various possible solutions to the problem, concentrating on the analysis of the data using models for unequally spaced longitudinal data with serial autocorrelation. These techniques are applied to the Matimba data and show that there is no statistically significant increase in the temperature as a result of heat released into the atmosphere by the dry-cooled power station, up to the time of commissioning of the fourth of the six turbines.