108
Miguel A. Zavala
regimes and water availability. Nevertheless, the specific environmental factors to which holm oak and Aleppo pine respond and the population-level
mechanisms that control changes in species dominance over space and time
have not been adequately described.
8.3 Top-Down Models of Stand Composition
in Mixed Holm Oak-Aleppo Pine Forests
The investigation of statistical methods to interpret changes in community
composition in terms of species' responses to environmental gradients has
been central to the development of vegetation science (Whittaker 1975; Ter
Braak and Prentice 1988). Gradient analyses have provided important insights into the factors controlling plant community composition in different
biomes (e.g. Whittaker 1956; Peet and Loucks 1977). In the Mediterranean
region, however, these analyses have often produced unsatisfactory results,
apparently due to anthropogenic influence on vegetation structure (Romane
1987; Floret et al. 1989; Retana et al. 1996).
In order to detect the main sources of correlation between forest structure
and environmental factors in this region and to account for the specific patterns of variability of these stands, a series of statistical methods was developed. The mathematical derivation of these methods is beyond the scope of
this chapter and only some relevant features are reported here. The approach
used follows the general definition of statistical inference and likelihood (see
Edwards 1972) and it does not make any a priori assumption regarding the
structure of the data. Therefore, it allows us to explore the effects of different
model formulations not only on the mean but also on any other aspects of
variability in the statistical distribution that underlies our observations. In
the example shown here, I describe stand composition in holm oak-Aleppo
pine forests in relation to drought length (DL) (Walter and Lieth 1960). This
index indicates the number of months in which Turc's potential evapotranspiration (Turc 1958) exceeds rainfall for a given plot and integrates the effect of local climate and exposure on vegetation. Information about forest
structure was obtained from more than 1000 plots of the Forest and Ecological Inventory of Catalonia (Gracia 1992) and climatic data were computed
through multiple regression models that estimate local measurements as a
function of altitudinal, latitudinal and longitudinal deviations from the closest meteorological station.
According to conventional wisdom, the mean basal area of holm oak and
Aleppo pine covaried inversely along the DL gradient according to a sigmoid
function. However, the residual variation in species basal area along the
aridity gradient followed a characteristic V-shaped pattern that suggests
strong patterns of interspecific segregation between these two species
(Fig. 8.1). This figure considers the frequency distribution of holm oak pro-
Miguel A. Zavala
regimes and water availability. Nevertheless, the specific environmental factors to which holm oak and Aleppo pine respond and the population-level
mechanisms that control changes in species dominance over space and time
have not been adequately described.
8.3 Top-Down Models of Stand Composition
in Mixed Holm Oak-Aleppo Pine Forests
The investigation of statistical methods to interpret changes in community
composition in terms of species' responses to environmental gradients has
been central to the development of vegetation science (Whittaker 1975; Ter
Braak and Prentice 1988). Gradient analyses have provided important insights into the factors controlling plant community composition in different
biomes (e.g. Whittaker 1956; Peet and Loucks 1977). In the Mediterranean
region, however, these analyses have often produced unsatisfactory results,
apparently due to anthropogenic influence on vegetation structure (Romane
1987; Floret et al. 1989; Retana et al. 1996).
In order to detect the main sources of correlation between forest structure
and environmental factors in this region and to account for the specific patterns of variability of these stands, a series of statistical methods was developed. The mathematical derivation of these methods is beyond the scope of
this chapter and only some relevant features are reported here. The approach
used follows the general definition of statistical inference and likelihood (see
Edwards 1972) and it does not make any a priori assumption regarding the
structure of the data. Therefore, it allows us to explore the effects of different
model formulations not only on the mean but also on any other aspects of
variability in the statistical distribution that underlies our observations. In
the example shown here, I describe stand composition in holm oak-Aleppo
pine forests in relation to drought length (DL) (Walter and Lieth 1960). This
index indicates the number of months in which Turc's potential evapotranspiration (Turc 1958) exceeds rainfall for a given plot and integrates the effect of local climate and exposure on vegetation. Information about forest
structure was obtained from more than 1000 plots of the Forest and Ecological Inventory of Catalonia (Gracia 1992) and climatic data were computed
through multiple regression models that estimate local measurements as a
function of altitudinal, latitudinal and longitudinal deviations from the closest meteorological station.
According to conventional wisdom, the mean basal area of holm oak and
Aleppo pine covaried inversely along the DL gradient according to a sigmoid
function. However, the residual variation in species basal area along the
aridity gradient followed a characteristic V-shaped pattern that suggests
strong patterns of interspecific segregation between these two species
(Fig. 8.1). This figure considers the frequency distribution of holm oak pro-