Climatic classification of the Tibet Autonomous Region using multivariate statistical methods

Abstract

A climatic regionalization of the Tibet (Xizang) Autonomous Region (TAR) is developed based on a multivariate analysis of temperature and precipitation records of two data sets from the periods 1971–1980 and 1980–1989. For these two decades 15 selected variables of the 29 meteorological stations — in total more than 50,000, partly handwritten original data — were computed. Three different statistical approaches, Cluster Analysis (CA), Factor Analysis (FA) and Discriminant Analysis (DA) were followed up. The vast expanse (1.2 million km²) and the pronounced relief of the TAR (between 500 m and 8,848 m in elevation), the low density and unequal distribution of meteorological stations, as well as uncertainties inherent in the selection of statistical methods, restricted and hampered the separation of coherent climatic subregions using multivariate statistics. However, a solution comprising of nine clusters, representing the different climatic conditions of Tibet, was found.

The results of the multivariate statistical analyses were compared with traditional classification schemes published by the Chinese Academy of Sciences (CAS). The Chinese classifications from the years 1982 (CAS 82) and 1984 (CAS 84) are here presented to the scientific community of the western countries for the first time (Fig 2 and 3). The main aim of these Chinese climatic classifications in Tibet was the delineation of areas with the best climatic conditions for agricultural land-use. In comparison, the aim of this paper was to establish a new classification of Tibet's climate, mainly based on measurements of climatic elements analyzed with multivariate statistical procedures following up the approach of McGregor (1993). The Factor Analysis (FA) included the temperature factor, which explains 46% of the total variance, whereas 32% is explained by the winter moisture and 12% by the moisture factor.

For temperatur zonation and, in particular, for the distribution of precipitation, the high mountain topography of Tibet is interpreted as the determining factor. The new climatic classification derived from multivariate statistics allows a more dynamic interpretation of the climatic cluster pattern. First, it shows the channelized influence of the summer monsoon along the southeast-northwest- and east-west-trending broad valleys north of the Gulf of Bengal. This is also obvious in the pattern of the temperature and moisture factors. Second, it can be supposed that a general south-trending influence of the winter monsoon, in combination with a northwest-trending influence from the Bay of Bengal leads to more irregular relief- and wind-dependent pattern, which can be estimated from the distribution of the winter moisture factor.