干旱区生态环境知识资源中心

The development of rapid and cheap methodologies to gather basic information relevant to management planning is a critical task to be addressed in many semi-arid zones of Africa, Asia and Latin America. We present in this paper a case study of a vegetation survey in two areas of the Mexican Chihuahuan Desert using a rapid and cheap methodology which combines the intensive use of available multivariate techniques and cartographic information and that can be easily applied in other arid lands. The first of the two studied areas (La Laguna) is the focus of a development project for agriculture and animal production in marginal rural communities. The second area (Mapimi Bolson) is a conservation area with the status of Biosphere Reserve in the framework of the MAB Programme of UNESCO. Vegetation-environment relationships were studied on a sample of sites using a divisive polythetic classificatory procedure and principal component analysis of the floristic table followed by analyses of variance of axis scores according to environmental factors. Discriminant functions to predict vegetation types on the basis of environmental data that can be easily gathered from cartographic information were then calculated. Validation of the discriminant functions was made with another set of sampled sites. The analysis of both datasets showed that landform is the variable that best explains the variability between vegetation types. The addition of other variables influencing soil water availability such as topography, surface rockiness and soil origin confers a high degree of predictability to the mathematical models describing the relationship between vegetation types and environmental variables. This fact, in rum, allows the construction of discriminant functions that correctly identify nearly 90% of new sites. This level of predictability is obtained using only the first two dimensions of the discriminant space, i.e. most of the variability can be explained with only two discriminant functions. This combination of available techniques of analysis allowing a quick and cheap vegetation cartography could be afforded in many development projects having a modest supply of economic resources and, eventually, i.e. when sufficient resources and expertise are available, the discriminant functions could be integrated in geographical information systems to produce the final cartography. (C) 1996 Academic Press Limited