Plant and Soil 220: 139–150, 2000. © 2000 Kluwer Academic Publishers. Printed in the Netherlands. 139 Small-scale spatial soil-plant relationship in semi-arid gypsum environments Agust ´ in Rubio 1,∗ and Adri´ an Escudero 2 1 Departamento de Silvopascicultura. Escuela T´ ecnica Superior de Ingenieros de Montes, Universidad Polit´ ecnica de Madrid, Madrid, E-28040, Spain and 2 Departamento de Biolog ´ ia Vegetal. Escuela Universitaria de Ingenier ´ ia T´ ecnica Agr ´ icola, Universidad Polit´ ecnica de Madrid, Madrid, E-28040, Spain Received 12 April 1999. Accepted in revised form 6 January 2000 Key words: gypsum environment, plant distribution, soil heterogeneity, soil patterning, spatial pattern, vegetation bands Abstract Studies on soil patterning on a small scale in arid and semi-arid regions have rarely been conducted. Many papers implicitly assume that plant distribution is controlled by some soil variables acting at small scales. We have directly tackled the relationships between soil and some biotic variables including plant distribution at small scales in an Iberian semi-arid gypsum environment. This has been carried out by means of Canonical Correspondence Analysis as a hypothesis-testing tool. CCA models show that the spatial data matrix is able to explain a relevant fraction of the soil data set (P< 0.001). The most important variable, as firstly selected in the CCA stepwise selection procedure, suggests the existence of a vegetation-elevation gradient in relation to soil physical properties; the rest of selected variables indicates the existence of other spatial trends which may be related to certain microgeomor- phological features. On the other hand, only the cover of annuals and the cover of litter are selected in the case of the biotic data set as constraining matrix, but not the cover of any perennial plant. Partial CCA models indicated that the remaining information explained by the spatial data set after adjusting the biotic set as covariables is also significant (p< 0.001). This variability is not related to the existence of vegetation bands as shown by the two selected variables in the case of the partial CCA models. The primary source of spatial soil variation is related to the existence of three community bands and these differences are able to explain even the change of plant life forms in vegetated band. The soil parameters controlling the changes are mainly related to texture and surface features. However, we detected other sources of spatial soil variation out of this primary model. This hierarchical spatial pattern seems to be related to some geomorphological traits of the landscape, such as soil crust strength, presence of gypsum crystals or bare zones, and not to the presence of mature gypsophytes (at least the five most frequent) which might ameliorate the soil environment. Furthermore, the biotic data set is not able to explain any new fraction of soil variability out of that already explained by the spatial data set. Abbreviations: CCA – canonical correspondence analysis; TVE – total variation explained Introduction It has been long accepted that interactive soil–plant re- lationships induce spatial patterning in soil properties, and that individual plant performance and even plant communities may respond to soil heterogeneity caused ∗ FAX No: +3413366377. E-mail: arubio@montes.upm.es by abiotic, but also by some biotic sources (Jackson et al., 1990). It is also known that soil heterogeneity is a basic element for competitive and/or facilitative interactions between plants, mainly in stressed envir- onments (Chapin et al., 1994) such as semiarid hab- itats (Fowler, 1986) and consequently may determine patterns in plant and community distribution.