Climate Change: A New Metric to Measure Changes in the Frequency of Extreme Temperatures using Record Data
- Electronic Supplementary Material –
Lalith Munasinghe
Tackseung Jun ()
David H. Rind
Here we investigate the mean temperature and RED estimates from 1930 to 1969 in the Northern Hemisphere (NH) using DelCCR2 and HadCRUT3v (on landmass). Note that the NH mean temperature anomaly in the period 1930-1950, measured as deviation from the mean temperature between 1960-1991, is higher in DelCCR2 than inHadCRUT3v (See Fig. A1 c): the mean deviations are -0.03°C and 0.09°C in HadCRUT3v and DelCCR2, respectively. These differences disappear after the early 1950s. The extensive use of interpolation in DelCCR2 to fill out the missing temperature observations especially during the early time periods may be responsible for the earlier differences in mean temperature. Further investigation of this difference is beyond the scope of this paper.
However, it is important to examine the impact of the differences in mean temperature on RED estimates. Here we show that the differences in mean temperature do not necessarily correspond to similar differences in the RED estimates. In Fig. A1 a-b, we compared the RED estimates in the NH using DelCCR2 and HadCRUT3v. From 1930 to 1950, one might expect RED-H Mx (RED-L Mn) estimates using DelCCR2 to be higher (lower) than RED-H Mx (RED-L Mn) using HadCRUT3v since the mean temperature is higher in DelCCR2 during this time period. On the contrary, RED-H Mx using DelCCR2 are lower than RED-H Mx using HadCRUT3v (See Fig. A1 a), and RED-L Mn using DelCCR2 are higher than the RED-L Mn using HadCRUT3v (See Fig. A1 b). Therefore, although the mean temperature in the time period 1930-1950 is higher in DelCCR2, the frequency of extreme high temperatures is lower and frequency of extreme low temperatures is higher in DelCCR2. Moreover, the higher mean temperature prior to 1950 in DelCCR2 does not systematically affect the RED estimates between 1950-1969: The RED-H Mx and RED-L Mn using DelCCR2 and HadCRUT3v are very close to each other if the RED for both data sets in 1950 are normalized to one (Between 1969-2008, The RED-H Mx and the RED-L Mn using DelCCR2 and HadCRUT3v are consistent with the RED estimates shown in Fig. 1 c-d where the RED for both data sets in 1969 are normalized to one).
<Insert Fig. A1. Annual estimates of RED and five-year moving average of mean temperatures in the Northern Hemisphere. >
1