Figure1: Linear regression of average yearly temperatures vs. height
Abstract
The climate of the upper Po basin is examined on the basis of temperature (1926-1986) and precipitation (1921-1986) observations from 50 stations. Linear regression of average yearly temperatures vs. height allows one to calculate the regional average temperature gradient (5.2ºC per 1000 metres), while an examination of the positive and negative residuals makes it possible to pinpoint specific local deviations linked to aspect conditions, and therefore to geomorphology.
Figure1: Linear regression of average yearly temperatures vs. height
In order to achieve a more detailed picture of regional climate patterns, data are filtered through factor analysis and a taxonomy of climate patterns is thereby obtained by means of cluster analysis.
The application of cluster analysis, however, is fraught with problems of choice of the relevant technique. Hierarchical iterative agglomerative techniques are regarded as the most appropriate, in view of the need to be able not only to cluster similar meteo stations together, but also to find out the level of intercluster similarity, while at the same time minimising intra cluster variability. The theoretical underpinnings and procedures of various techniques are examined (single linkage, complete linkage, average linkage, centroid sorting, median, flexible strategy and Ward). The Ward technique is found to be the one which better satisfies the requirement of minimising intra cluster variability, thereby aiming, at clusters as much as possible homogenous. The results of the analysis are presented in the following table (in which clusters are coded in such a way as to show the similarities with one another) and map.
| CL. | M.P. | M.T. | |
| I,1 | 635.5 | 9.7 | Rather dry, precipitation peak in the autumn, little precipitation concentration, cool to temperate |
| 1,2,1 | 853.0 | 12.6 | Moderate precipitations, dual spring/autumn peak, little concentration, temperate |
| I,2,2 | 880.5 | 10.3 | Moderate precipitations, autumn peak, more concentration, cool to temperate |
| 1,2,3 | 1244.0 | 9.1 | Considerable precipitations, autumn peak, sizeable concentration, cool to temperate |
| II,1,1 | 2021.0 | 8.9 | Very strong precipitations, spring or autumn peak, strong. concentration, cool to temperate ZP |
| II,1,2 | 1402.5 | 10.5 | Strong precipitations, spring peak, strong concentration, rather cool to temperate |
| II,2,1,1 | 993.0 | 1 1.7 | Moderate precipitations, spring peak, noticeable concentration, rather temperate |
| II,2,1,2 | 721.0 | 12.3 | Very moderate precipitations, spring, peak, no considerable concentration, temperate |
| II,2,2,1 | 796.5 | 6.4 | Moderate precipitations, spring peak, moderate concentration, considerably cold |
| II,2,2,2 | 1044.5 | 4.1 | Considerable precipitations, spring peak, noticeable concentration, cold to very cold |
In order to assess whether the reliability of the results thus obtained is impaired by the presence of incomplete data sets, the procedure is repeated excluding all stations having temperature records shorter than 30 years or precipitation records shorter than 35 years. The new analysis succeeds in identifying the same clusters, thereby providing confirmation of the taxonomy previously obtained, as shown by a comparison between the relevant dendograms.
Nethertheless, no matter how sophisticated statistical techniques are employed, a measure of subjectivity inevitably lingers on. It is a necessary reminder of the fact that science itself is a cultural construct.
Figure2: Map of study area with meteo stations classified by means of cluster analysis
Figure3: Dendogram of 40 stations cluster analysis
Figure4: Dendogram of 50 stations cluster analysis