Results

A selection of screenshots showing different aspects of the data using the derived parameters described in Sect. 3.2 (trends/10 years and SNR/10 years, respectively) is presented in Figs. 2 and 3 for the ECHAM5 model and in Figs. 4 and 5 for the ERA-40 reanalysis data, respectively. The color intensity depicts the density of the value in the whole data set. Features selected in several views are depicted in red, features selected only in the current view are shown in blue, and context information in black.

The distribution of the signal-to-noise ratio for each time-step for temperature and refractivity is shown in Fig. 2a and 2d, respectively. Selected are values of the | SNR | > 1. This selection is carried out directly in the scatterplot using the brushing mechanism described in Sect. 3.1.1. Any change in this selection (or in any of the subsequently described ones) is immediately propagated into all other views. In Fig. 2b and 2e , the variation of the linear trend (relative trend for the refractivity) over time is shown using the curve view. The additional constraint of selecting high SNR values reflects in the red colored curves. The inversely proportional behavior of temperature and refractivity is clearly observed. Note that the negative temperature (positive refractivity) trend with high significance (high SNR) is prominent over the whole time period, compared to the positive temperature (negative refractivity) trend with high significance only visible in later time periods.

The trend distribution with latitude in Fig. 2c and 2f shows, among other things, that points with high trend values are often also associated with comparably low SNR values (in blue).

Fig. 2 ECHAM5 A2, temperature (in K) in (a)-(c) and refractivity (in N-units) in (d)-(f): Selection of high values of SNR in (a)/(d), the corresponding pattern in the trend variation over time in (b)/(e), and the trend distribution vs. latitude in (c)/(f) are depicted. |SNR| > 1 shown in red

Fig. 3 ECHAM5 A2, temperature (in K) represented in (a)-(c) and refractivity (in N-units) in (d)-(f): Selection of one specific feature ((a)/(d), denoted by rectangle box) and the corresponding trend distribution versus latitude in (b)/(e); overview latitude vs. pressure level plot, again without rectangle box selection in (c)/(f), |SNR| > 1 shown in red

Fig. 3 ECHAM5 A2, temperature (in K) represented in (a)-(c) and refractivity (in N-units) in (d)-(f): Selection of one specific feature ((a)/(d), denoted by rectangle box) and the corresponding trend distribution versus latitude in (b)/(e); overview latitude vs. pressure level plot, again without rectangle box selection in (c)/(f), |SNR| > 1 shown in red

Fig. 4 ERA-40, geopotential height trend (in gpm/10 years): Exclusion of southern high latitudes in black, high SNR in red in (a); resulting trend variation over time, selected latitudes in red in (b); additional selection of |SNR| > 1 in (c) and (d); additionally restriction to post-1979 time period in (e) and (f)

Fig. 4 ERA-40, geopotential height trend (in gpm/10 years): Exclusion of southern high latitudes in black, high SNR in red in (a); resulting trend variation over time, selected latitudes in red in (b); additional selection of |SNR| > 1 in (c) and (d); additionally restriction to post-1979 time period in (e) and (f)

The height distribution in Fig. 3a and 3d shows that positive temperature (negative refractivity) trends stem from the troposphere, and that the negative temperature (positive refractivity) trends stem from the upper pressure levels corresponding to the lower stratosphere region, as expected. The SNR-selected data points are marked in red again. Specific selection (rectangle box) is now imposed for negative temperature (positive refractivity) trends at the 150 hPa and 200 hPa levels. With this new selection applied, the red-colored trend distribution with latitude changes from Fig. 2c/2f to what is shown in Fig. 3b/3e. It can be observed that the significant trend features stem only from northern high latitudes, where they are well inside the stratosphere, as expected.

An overview of the spatial distribution of data points with |SNR| > 1 (without rectangle box selection) is gained in the plots 3c and f, where pressure levels versus latitude are plotted. Values with high SNR are in particular found at the topmost levels (20-30 hPa) in the ECHAM5 model run, corresponding to robust negative temperature (positive refractivity) trends in the lower stratosphere observable over the whole investigated time period (see Fig. 2b/2e). Note that at low levels with weaker |SNR| > 1 presence, the red color dots are effectively "buried" in the predominant blue color dots (visible on the screen, though).

For the ERA-40 reanalysis the distribution of the geopotential height trend with latitude in Fig. 4a (again |SNR| > 1 in red) shows high trend variation in southern

Fig. 5 ERA-40, temperature trend (in K/10 years); selection of a confined trend curve ensemble in red in (a); corresponding latitude vs. trend distribution in (b); exclusion of southern high latitudes and |SNR| > 1 values in (c); restrict additionally to post-1979 in (d)-(f)

Fig. 5 ERA-40, temperature trend (in K/10 years); selection of a confined trend curve ensemble in red in (a); corresponding latitude vs. trend distribution in (b); exclusion of southern high latitudes and |SNR| > 1 values in (c); restrict additionally to post-1979 in (d)-(f)

high latitudes, which is a spurious feature according to Santer et al. (2004). The selection was restricted to data over 90°N to 60° S accordingly. The resulting time variation of the geopotential height trend shown in Fig. 4b (only the latitude selection applies here) now has most outliers removed, with the remaining data points in red. Applying the selection of |SNR| > 1 values again further improves the result (Fig. 4c), with some outliers still present in the early years of the data set. The same selections apply for the plot in Fig. 4d, still showing a somewhat ambiguous distribution with pressure levels. Restricting to the post-1979 era when satellite data were assimilated (Uppala et al. 2004) displays a better-contrasted height distribution in Fig. 4e. The pressure levels versus latitude plot (Fig. 4f) is based on all selections described above and presents best SNR for the lower stratosphere at northern high latitudes (negative trends, see Fig. 4e) and the lower tropical troposphere (positive trends, partly up to the lower stratosphere), respectively.

High variation of the ERA-40 temperature trends in time is observed in Fig. 5a. Some of the temperature trend curves showing the highest variation over time are selected in the curve view (red) and analyzed further. The corresponding latitude distribution of the trend in Fig. 5b reveals that it is a spurious feature since it stems from the southern high latitudes. Restricting to 90°N to 60°S and selecting |SNR| > 1 (deselecting the curve ensemble before) leads to the trend variation in Fig. 5c. Limiting the time period to the post-1979 era further removes highly variable trends and yields the spatial distributions depicted in Fig. 5d-f. As seen in Fig. 5f, high

SNR (in red) appears in the lower stratosphere region at northern high latitudes for negative trends and in the tropical upper troposphere for positive trends (again, the latter feature is better visible on the screen).

0 0

Post a comment