QUANTIFYING THE INCREASED PROBABILITY AND FREQUENCY OF RECORD HIGH TEMPERATURES DURING GLOBAL WARMING

Extreme values in long-term instrumental records of temperature are not as widely studied as the trends in mean temperature, but are very important. For example, incidence of heat-related illnesses and fatalities is highest during times of extreme temperatures, and utilities must be prepared for higher demand during extreme weather. Global warming provides not only an increase in the average temperature but also an increase in the frequency and severity of temperature extremes. Thus, record-breaking temperatures provide an important, easily understood way to increase public awareness of global warming.

An analysis of the statistics of extreme temperatures shows that they are dependent on a number of factors. In times of constant temperature, record-breaking high and low temperatures should become less and less frequent as the length of temperature records increases. Changes in standard deviation of temperature values from year to year changes the frequency of record-breaking temperatures. The limited precision of measurement and resulting rounding of temperature records to integer values reduces frequency of record-breaking temperatures.

Under a global warming scenario, the incidence of record-breaking temperatures approaches a constant non-zero probability over time. With a possible predicted warming of three degrees centigrade over the next century, simulations show that the probability of record-breaking warm temperatures exceeds 7% in any given year. This means that on average, two days per month will be the warmest on record, and one month per year will be the warmest on record. By contrast, a no-warming scenario gives only a 1% annual probability of a record temperature in a 100 year time series and 0.5% in a 200 year time series.

Analysis of instrumental temperature records over the last 100 years shows significantly more frequent record highs and less frequent record lows than predicted by simple random variation. Both observations are due to warming over this period. Similar analysis of preindustrial temperature records shows no abnormal incidence of high temperature extremes and thus no evidence of warming.