Info

trends.

JJA Geopotential Height [m] Anomalies ECHAM5, 20C/A2, run 1, zonal means

" -75" " -60" " -45" " -30 " "-15 " " 0 15 30 45 60 75

JJA Geopotential Height [m] Anomalies ECHAM5, 20C/A2, run 1, zonal means

" -75" " -60" " -45" " -30 " "-15 " " 0 15 30 45 60 75

1980 2000 2020 2040 2060

Time [years]

Fig. 2 Time evolution (20th century combined with SRES A2) of ECHAM5 seasonal mean JJA geopotential height profiles in five zonal IPCC+ regional bands (from IPCC Arctic region/ARC via the northern mid-latitudes/NHM, the tropics/TRO, and southern mid-latitudes/SHM to the IPCC Antarctic region/ANT). The time period 2001-2050, which forms the basis for the trend analysis, is highlighted with a frame

1980 2000 2020 2040 2060

Time [years]

Fig. 2 Time evolution (20th century combined with SRES A2) of ECHAM5 seasonal mean JJA geopotential height profiles in five zonal IPCC+ regional bands (from IPCC Arctic region/ARC via the northern mid-latitudes/NHM, the tropics/TRO, and southern mid-latitudes/SHM to the IPCC Antarctic region/ANT). The time period 2001-2050, which forms the basis for the trend analysis, is highlighted with a frame models is 10 hPa, yielding probably untrustworthy results there (Cordero and de Forster 2006). At higher latitudes, the year to year variations in the strength of the polar vortex cover a possible climate change signal and thus polar regions are not suitable for this kind of climate change tracing during the hemispheric winter, as shown for Antarctica (Fig. 2, bottom panel). The atmospheric parameters N, T, and q (not shown) also exhibit strong climate signals during the first part of the 21th century.

For most regions, the trends of the model runs agree quite well in magnitude and direction (Fig. 3, left, shows an example), a fact that deserves great importance for qualifying as best trend indicators (see Sect. 3). The trend signal of various runs of an individual model shows generally similar vertical peculiarities. Large discrepancies occur for trends at high latitudes in the winter hemisphere, as shown for Alaska in Fig. 3, right panel: above about 100 hPa, CCSM3 has a tendency towards pronounced negative temperature trends, ECHAM5 to rather positive

Fig. 3 Temperature trends for two selected regions (60°N-60°S and Alaska) of ECHAM5 seasonal mean DJF temperatures

DJF Temperature Trends 2001-2050 (A2) 60°S-60°N (G60) Alaska (ALA)

DJF Temperature Trends 2001-2050 (A2) 60°S-60°N (G60) Alaska (ALA)

Fig. 3 Temperature trends for two selected regions (60°N-60°S and Alaska) of ECHAM5 seasonal mean DJF temperatures

500 ECHAM5 (run 1-3) CCSM3 (run 1-5) HadCM3 (run 1)

1000

1000

500 ECHAM5 (run 1-3) CCSM3 (run 1-5) HadCM3 (run 1)

1000

1000

trends. Such discrepancies lead to an exclusion of those regions as trend indicators, due to the criteria defined.

Besides the consistent direction of trends across the ensemble of model runs, trend significance and goodness-of-fit determine the identification of best trend indicators. Figure 4 illustrates these quantities for refractivity in fall as an example.

SON Refractivity [N units]

Fig. 4 Trend significances (left) and goodness-of-fit (right) in northern Europe (NEU, top) and eastern Asia (EAS, bottom) for SON refractivity. The results of all model runs (columns in the plots) at all pressure levels (rows of the plots) are presented. In the left hand side plots, dark gray cells indicate significances greater than 99%, medium gray 95-99%, and light gray 90-95%, respectively. A positive sign of the trend is marked by a circle spot in the cell

Fig. 4 Trend significances (left) and goodness-of-fit (right) in northern Europe (NEU, top) and eastern Asia (EAS, bottom) for SON refractivity. The results of all model runs (columns in the plots) at all pressure levels (rows of the plots) are presented. In the left hand side plots, dark gray cells indicate significances greater than 99%, medium gray 95-99%, and light gray 90-95%, respectively. A positive sign of the trend is marked by a circle spot in the cell

Most regions show high significances of trends at selected height ranges. The significances plots (Fig. 4, left) give for each parameter a good overview of regions and height ranges that are particularly sensitive to climate change. However, goodness-of-fit turned out to be the limiting factor for the assessment of best trend indicators. Figure 4 (right) presents the respective values for R2. The shading is chosen in a way that dark boxes indicate better fits. Best fits agree mostly quite well with high significances of the trends, but regions which are dominated by a higher variability— which leads to a worse ratio between the total variance in the data and the variance explained by the regression and thus to a worse goodness-of-fit value—are sorted out. As shown with the direction of trends (see Fig. 3), these regions are mainly found at high latitudes of the winter hemisphere.

In the last step of the study, the results of the trend analysis were merged according to the criteria defined in Sect. 3, allowing the identification of geographical regions and height domains qualified for best trend indicators. Figure 5 presents a summary of the trend study's results. For each RO accessible parameter, one example season is shown. Within each plot, one box represents one pressure level (the rows) of one region (the columns). If the criteria defined are fulfilled, the box is colored. The gray shading indicates the significance level of the detected trends, the circle spot marks the positive trend direction (increase).

0 0

Post a comment