Detection of local climate change

It is well known that climate is the accumulation of seasonal weather events over long period of time and over a particular place. The response of anthropogenic changes in climate forcing factors occurs against the natural internal and external forcing climate variability. Climate variability which is not forced by external factors is internal climate variability and occurs in all times from weeks to centuries. External forced climate variations may be due to changes in natural forcing factors, such as solar radiation or volcanic eruption, as well as increasing concentration of greenhouse gases. Climate change detection is the procedure to demonstrate that an observed change is significantly different from the explained by natural variability. Detecting a climate change is a statistical problem (Barnett, et al., 1999).

Climate change can be detected by studying the behavior of climate indicators. The Intergovernmental Panel on Climate Change (IPCC) classified the indicators as follows: concentration indicators (CO2, O3), weather indicators (air temperature, rainfall), biological and physical indicators (sea level, artic ice) and economical indicators. A climate indicator is a sequence of observations that have been collected for long time. The climate in a particular region of the world can be represented by a stationary process, which is characterized by having a natural variability and its joint probability density function is invariant with time. However, if the probability distribution changes with time it implies that external factors cause that the mean and the covariance functions changed with time. Therefore to detect a climate change is equivalent to determine when the process changes from being stationary to nonstationary.

Usually the optimal fingerprint method is applied to detect a climate change. However, in this study a sequential hypothesis testing method is introduced to identify the time and the magnitude of the climate change. This technique is introduced here because is very simple and effective to identify climate changes. The sequential testing process is known in statistics as a control chart. The implemented control chart is the exponential weighted moving average (EWMA) method. This technique plots a 95% confidence interval for the weighted average of the process at every point in time. A significant change is signal by locating a point outside of the confidence interval. Control chart can be properly applied when the elements of the climate indicator sequence are statistically independent. However, when a climate indicator is a highly autocorrelated process then a special tests should be applied to detect the climate change (Ramirez and Sastri, 1997).

Fifty six weather stations located in Puerto Rico were studied during the period 1955 to 2004 and it was found that the minimum air temperature is a climate indicator that reveals the inherent changes. The minimum air temperature is an appropriate indicator because this temperature is almost not affected by cloud dynamics during the night. It was found that in 1998 occurs a significant increment in minimum air temperature in Puerto Rico. Figure 1 shows that a significant increment of minimum air temperature has been identified in 1998. Also the total number of tropical storm that occurs in the North Atlantic basin was studied during the period 1960 to 2004. A significant increment of hurricane activity was identified in 2000 and Figure 2 shows the increment of hurricane activity during the last 10 years. It should be noted that the detected local climate change is in harmony with global climate change. The IPCC (2001) reported that the Northern Hemisphere has recorded an increment in temperature in the 20th century and is likely to have been the largest of any century during the past 1000 years. It is also likely that the Northern Hemisphere, the 1990s was the warmest decade and 1998 the warmest year.