On the twenty-first-century wet season projections over the Southeastern United States
As shown in earlier studies, diurnal variations in the summer climate in the SEUS are significant (Carbone and Tuttle 2008; Misra et al. 2011a, b; Stefanova et al. 2012), yet notoriously difficult for coupled global climate models to resolve (Dai 2006). In Fig. 3 we show the fraction of diurnal variations that explain the seasonal variations in summer rainfall from observations (NCEP Stage IV) and RSM20th. This fraction (F) is computed as the ratio of
where A is the seasonal mean diurnal amplitude, P is the seasonal mean of the total rainfall, overbars indicate climatological mean, and N is the number of years (8 years for NCEP Stage IV observations and 30 years for RSM20th). RSM20th (Fig. 3b) clearly underestimates the observed fractional variance of diurnal variability (Fig. 3a). However, the observed fraction (F) is computed over an 8-year period, compared to the 30-year period of RSM20th, which could explain some of the differences in Fig. 3. It is quite apparent that in the observations the variability of diurnal rainfall explains a large fraction of the summer seasonal rainfall variability. The observed fraction ranges from 15 to 40 % in the SEUS (Fig. 3a), while in the RSM20th it ranges between 15 and 25 % (Fig. 3b), especially in the eastern half of the domain. We were unable to compute the same fraction (F) from CCSM3-20th because the data are not available at intervals shorter than 6 h.
Similarly, we have compared the diurnal phase of rainfall between NCEP Stage IV and RSM20th (Fig. 4). This figure shows the climatological hour of maximum rainfall in the summer season for observations (Fig. 4a) and for RSM20th (Fig. 4b). RSM20th can simulate this feature reasonably well over most of the SEUS, peaking between 24 and 02 GMT (*18-20 LST). However, there are some apparent disparities over the Florida panhandle and along the Carolina coasts, where the RSM20th summer rainfall tends to peak later than observations by a couple of hours.
Dai (2006), analyzing CCSM2 (which he claimed was similar in its diurnal behavior to CCSM3), showed that the timing of maximum rainfall in the summer season over SEUS was around 12-14 LST, which is roughly four to six hours earlier than observed. Similarly, Bukovsky and Karoly (2008) showed that in CCSM3 the diurnal peak of rainfall over the SEUS in the twentieth-century simulation of CCSM3 is *1300LST (or 1900UTC). The RSM has consistently been shown to have significantly high fidelity in simulating the diurnal variability of the rainfall over the SEUS (Misra et al. 2011a, b; Stefanova et al. 2012). We believe the improvement of the diurnal variations in RSM20th is largely a result of the behavior of the parameterization (especially convection) scheme in RSM at the given resolution of 10 km. Additional improvements are enhanced resolution of the orography, vegetation, and the coastlines.
|Fig. 3 a The diurnal fraction of variance of rainfall calculated from NCEP STAGE IV rainfall observations. b Similar to a, but calculated from RSM20th. Dimensions plotted here are fractional [larger image]|
|Fig. 4 a The observed timing of summer maximum rainfalls from NCEP STAGE IV rainfall data. b Timing of maximum rainfall from RSM20th. Both panels are in GMT [larger image]|
Projected summer seasonal climate
Figure 5a and b shows the seasonal rainfall from CCSM3- 20th and CCSM3-21st, respectively; Fig. 5c shows their differences. It is apparent from Fig. 5c that CCSM3 projects a dipole-like pattern in the twenty-first-century summer season rainfall anomaly relative to the corresponding twentieth-century seasonal rainfall: Peninsular Florida becomes drier, and the rest of the SEUS becomes wetter. This dipole pattern has also been observed by Rauscher et al. (2011), who attributed this pattern of change in the summer rainfall to the atmospheric response from remote impacts of warming in the tropical western Pacific Ocean. Figure 6a and b shows similar plots for the summer rainfall from RSM20th and RSM21st, respectively; Fig. 6c shows their difference. The pattern of rainfall anomaly (Fig. 6c) projected by RSM21st is different from that of the CCSM3, showing a universal drying of the SEUS in the late twenty-first century relative to the corresponding summer season climatology in the twentieth century. The reason for this universal drying of the summer rainfall in RSM21st is explained further in the following subsection.
|Fig. 5 (left) June, July, and August 31-year mean rainfall from the CCSM3. Units are in mm/day. a The period 1969-1999. b The period 2040-2070. The bottom c is the difference in the two periods [larger image]||Fig. 6 (right) June, July, and August 31-year mean rainfall from the RSM. Units are in mm/day. a The period 1969-1999. b The period 2040-2070. The bottom c is the difference in the two periods [larger image]|
Projected diurnal variations in the summer climate
|Fig. 7 a and b are RSM20th and RSM21st summer seasonal maximum rainfall, respectively. c A difference in the two. Units are in mm/day [larger image]|