Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Spatial Distribution Patterns in the Very Rare and Species-Rich Picea chihuahuana Tree Community (Mexico)

Spatial Distribution Patterns in the Very Rare and Species-Rich Picea chihuahuana Tree Community (Mexico)

  • Christian Wehenkel, 
  • João Marcelo Brazão-Protázio, 
  • Artemio Carrillo-Parra, 
  • José Hugo Martínez-Guerrero, 
  • Felipe Crecente-Campo
PLOS
x

Correction

23 Nov 2015: Wehenkel C, Brazão-Protázio JM, Carrillo-Parra A, Martínez-Guerrero JH, Crecente-Campo F (2015) Correction: Spatial Distribution Patterns in the Very Rare and Species-Rich Picea chihuahuana Tree Community (Mexico). PLOS ONE 10(11): e0143899. https://doi.org/10.1371/journal.pone.0143899 View correction

Abstract

The very rare Mexican Picea chihuahuana tree community covers an area of no more than 300 ha in the Sierra Madre Occidental. This special tree community has been the subject of several studies aimed at learning more about the genetic structure and ecology of the species and the potential effects of climate change. The spatial distribution of trees is a result of many ecological processes and can affect the degree of competition between neighbouring trees, tree density, variability in size and distribution, regeneration, survival, growth, mortality, crown formation and the biological diversity within forest communities. Numerous scale-dependent measures have been established in order to describe spatial forest structure. The overall aim of most of these studies has been to obtain data to help design preservation and conservation strategies. In this study, we examined the spatial distribution pattern of trees in the P. chihuahuana tree community in 12 localities, in relation to i) tree stand density, ii) diameter distribution (vertical structure), iii) tree species diversity, iv) geographical latitude and v) tree dominance at a fine scale (in 0.25 ha plots), with the aim of obtaining a better understanding of the complex ecosystem processes and biological diversity. Because of the strongly mixed nature of this tree community, which often produces low population densities of each tree species and random tree fall gaps caused by tree death, we expect aggregated patterns in individual Picea chihuahuana trees and in the P. chihuahuana tree community, repulsive Picea patterns to other tree species and repulsive patterns of young to adult trees. Each location was represented by one plot of 50 x 50 m (0.25 ha) established in the centre of the tree community. The findings demonstrate that the hypothesis of aggregated tree pattern is not applicable to the mean pattern measured by Clark-Evans index, Uniform Angle index and Mean Directional index of the uneven-aged P. chihuahuana trees and P. chihuahuana tree community and but to specific spatial scales measured by the univariate L-function. The spatial distribution pattern of P. chihuahuana trees was found to be independent of patches of other tree species measured by the bivariate L-function. The spatial distribution was not significantly related to tree density, diameter distribution or tree species diversity. The index of Clark and Evans decreased significantly from the southern to northern plots containing all tree species. Self-thinning due to intra and inter-specific competition-induced mortality is probably the main cause of the decrease in aggregation intensity during the course of population development in this tree community. We recommend the use of larger sampling plots (> 0.25 ha) in uneven-aged and species-rich forest ecosystems to detect less obvious, but important, relationships between spatial tree pattern and functioning and diversity in these forests.

Introduction

The endemic Picea chihuahuana Martínez, a relict stranded by a warming climate during the current interglacial period [1], is listed as “Endangered” on the “Red List” of threatened species in Mexico [2]. At elevations of between 2,100 and 3,000 m a.s.l., 40 populations comprising at least 42,600 individuals have been detected in three separate clusters in the Sierra Madre Occidental. The size of the populations ranges between 21 and 5,546 individuals, including trees, saplings and seedlings. Individual trees can reach up to 50 m in height and 120 cm in diameter and an age of at least 272 years [3] [1]. The lower branches are almost horizontal, starting at 2 to 5 m in height, while higher branches are extended and somewhat raised, forming a conical crown [4]. Picea chihuahuana preferentially inhabits areas of rough terrain located on hillsides and canyons in areas facing northwest or northeast, with slopes ranging from 35% to 80%, at the margins of streams and rivers [5][6]. The species is often associated with other species of the genera Pinus, Quercus, Abies, Pseudotsuga, Populus, Prunus, Juniperus and Cupressus [6], [7]. The dominant type of disturbance seems to be tree fall gaps in the canopy caused by windstorms, fungi pathogens and insects [1], [8], but not by fire [9].

This very rare pine-spruce-cedar community (hereafter referred to as the P. chihuahuana tree community) covers an area no more than 300 ha. It remains largely untouched by humans because of its isolated location in very rugged mountainous areas [1], [1], [10]. The P. chihuahuana tree community has been the subject of several studies aimed at learning more about the genetic structure [3], [7], [10], [11], [12], [13], [14] and ecology of the species [1], [6], [15] and about potential effects of climate change [16], [17], [18]. The overall aim of most of these studies has been to obtain data to help design preservation and conservation strategies [7]. However, the structure (specifically the spatial tree pattern) of the Mexican Picea chihuahuana tree community has not yet been analyzed.

Forest structure is both a product and factor of ecosystem processes and biological diversity. Understanding forest structure can therefore help in understanding the history, function and future of a forest ecosystem [19]. Moreover, information about forest structure provides an essential basis for the analysis of forest ecosystem disturbance [20].

Forest structure refers to the patterns and relationships between attributes, including structural type, size, shape and spatial distribution (vertical and horizontal) and also the characteristics of components such as tree crown, foliage, tree bark, tree bole, wood tissue, standing dead trees, fallen trees, roots, pit and mound topography, landscape structure, soil structure, shrub, herb and moss layers, and forest floor and organic layers. Many of these components are fundamental to the functioning and diversity of ecosystems. For example, forest canopies, which differ both vertically and horizontally, are important for intercepting radiation, controlling microclimate and determining habitats [19], [21].

The spatial distribution of trees is a result of many ecological processes and can, for example, affect the degree of competition between neighbouring trees [22], [23], [24], [25], tree density [26], the size distribution and variability, regeneration, survival, growth, mortality and crown formation of forest trees [22], [27], [28], [29] as well as the biological diversity within a forest community [30].

Numerous scale-dependent measures have been established to describe spatial tree structure, such as the Clark-Evans index [31]), Diggle’s F and G-functions [32], Ripley’s K-function [33], [34], [35], the uniform angle index [36], [37], [38] and the mean directional index [39], [40].

In this study, we examined the spatial distribution pattern of trees in the P. chihuahuana tree community in 12 localities, in relation to i) tree stand density, ii) diameter distribution (vertical structure), iii) tree species diversity, iv) geographical latitude and v) tree dominance at a fine scale (in 0.25 ha plots), with the aim of obtaining a better understanding of the complex ecosystem processes and biological diversity [21]. Because of the strongly mixed nature of this tree community, which often produces low population densities of each tree species and random tree fall gaps caused by tree death, we expect aggregated patterns in individual P. chihuahuana trees and in the P. chihuahuana tree community [26], repulsive Picea patterns to other tree species and repulsive patterns of young to adult trees [41]. We also assumed no differences between i) the spatial distribution of the northern and southern populations of the P. chihuahuana tree community, and ii) the spatial distribution of suppressed and dominant trees, because of similar degrees of competition-induced mortality [42].

Material and Methods

We confirm that the field studies provide the specific location of study (Fig 1, S1 Dataset). No vertebrate studies were carried out. Field permit was granted by SEMARNAT, Mexico (http://www.semarnat.gob.mx/).

thumbnail
Fig 1. Location of the study area on the Sierra Madre Occidental, Durango (Mexico).

Map of the 12 locations of the Picea chihuahuana Martínez tree community under study in the States of Durango and Chihuahua (Mexico): 1) La Tinaja (TN), 2) El Ranchito (RC), 3) El Cuervo (CV), 4) Talayote (TY), 5) Las Trojas (TR), 6) El Venado (VN), 7) La Quebrada (LQ), 8) Paraje Piedra Rayada (PPR), 9) Quebrada de los Durán (Arroyo del Indio Ignacio) (QD), 10) Cebollitas (CB), 11) San José de las Causas (SJ), and 12) La Pista (Arroyo de La Pista) (LP).Data sources: Own compilation based on freely-accessible digital maps from INEGI, Mexico (http://www.inegi.org.mx/geo/contenidos/mapadigital/).

https://doi.org/10.1371/journal.pone.0140442.g001

Study area

Chihuahua spruce grows in areas characterized by an average temperature of between 9 and 12°C [1], precipitation ranging from 600 mm to 1,300 mm [43] and soil pH of 5.3–6.3 [4]. In order to determine the spatial tree structure of the P. chihuahuana tree community, 12 locations where the community occurs in the State of Durango and Chihuahua (north-western Mexico) were considered (Fig 1). Each location was represented by one plot of 50 x 50 m (0.25 ha) established in the centre of the tree community. Trees of all species of diameter at breast height (DBH) ≥ 7 cm were fully scored. The DBH, height and x, y coordinates were also recorded. The stem number per hectare (N), stand basal area (G), quadratic mean diameter (dg), mean breast height diameter (d), mean total height (h), maximum diameter (dmax), and maximum height (dmax) of all tree species (total) and Picea chihuahuana M. (Pch) were computed (Table 1). The total numbers of tree species within each of the populations in the P. chihuahuana tree community have been reported by Quiñones-Pérez et al.[44].The DBH structures (as parameters of vertical structure) in the 12 plots considering all tree species showed a reverse J-shaped form (Fig 2) typical of uneven-aged forests. Fig 2 also demonstrates that the minimum balanced structure area of this tree community is very small (< 3ha) [45]. In total, 15 tree species were foundin the 12 plots: Abies durangensis Martínez, Cupressus lindleyi Klotzsch ex Endl., Juniperus deppeana Steud., P. chihuahuana, Pinus arizonica Engelm., Pinus strobiformis Engelm., Pinus cooperi Martínez, Pinus durangensis Martínez, Pinus leiophylla Schl. & Cham., Pinus teocote Schiede ex Schltdl. & Cham., Populus tremuloides Michx., Prunus serotina Ehrh., Pseudotsuga menziesii (Mirb.) Franco, Quercus sideroxyla Humb.andQuercus crassifolia Humb. In each plot, P. chihuahuana grew along with three to eight other tree species [44]. To represent the diversity profile (vsp,a) of the tree species, we selected the described diversity for each location. Thus, each location of the P. chihuahuana tree community was characterized by the total number of tree species (species richness, (υsp,0)), effective number of tree species (Simpson index, (υsp,2)) and the number of prevalent tree species (υsp,), as Hill numbers [46] in each plot. The diversity values were taken from [13].

thumbnail
Fig 2. Diameter distribution in the plots representing the 12 locations of the Picea chihuahuana Martínez tree community.

Diameter distribution in the plots representing the 12 locations of the Picea chihuahuana Martínez tree community under study in the States of Durango and Chihuahua (Mexico),considering all tree species.1) La Tinaja (TN), 2) El Ranchito (RC), 3) El Cuervo (CV), 4) Talayote (TY), 5) Las Trojas (TR), 6) El Venado (VN), 7) La Quebrada (LQ), 8) Paraje Piedra Rayada (PPR), 9) Quebrada de los Durán (Arroyo del Indio Ignacio) (QD), 10) Cebollitas (CB), 11) San José de las Causas (SJ), and 12) La Pista (Arroyo de La Pista) (LP).

https://doi.org/10.1371/journal.pone.0140442.g002

thumbnail
Table 1. Summary of important stand parameters calculated from the tree data: stem number per hectare (N), stand basal area (G), quadratic mean diameter (dg), mean breast height diameter (d), mean total height (h), maximum diameter (dmax), and maximum height (dmax) of the all tree species (total) and Picea chihuahuana M. (Pch) in the 50 x 50 m plotsin the 12 study locations and minimum (min), mean and maximum (max) parameter values for the stands.

1) La Tinaja (TN), 2) El Ranchito (RC), 3) El Cuervo (CV), 4) Talayote (TY), 5) Las Trojas (TR), 6) El Venado (VN), 7) La Quebrada (LQ), 8) Paraje Piedra Rayada (PPR), 9) Quebrada de los Durán (Arroyo del Indio Ignacio) (QD), 10) Cebollitas (CB), 11) San José de las Causas (SJ), and 12) La Pista (Arroyo de La Pista) (LP).

https://doi.org/10.1371/journal.pone.0140442.t001

Spatial Structural Analysis

Clark-Evans index (CE), Uniform Angle index () and Mean Directional index ().

The Clark-Evans index (CE) [31], the Uniform Angle index () [36] and the Mean Directional index () [47] were used to describe the spatial distribution of the trees in each study plot, on the basis of the spatial distribution of the n trees nearest to a reference tree i. The CE was estimated using one neighbour (n = 1), while and were calculated using four neighbours (n = 4) [48], [49], [40]. A Poisson distribution pattern was characterised by a CE value of 1, cluster tendency by CE < 1 and a tendency of regular distribution of trees by CE > 1, with a maximum of 2.1491 for a hexagonal arrangement of trees.

For calculation of , Wi must first be calculated. The angle α0 was set at 72 degrees, which yielded a mean value of = 0.5 [37]. For each tree, the value of Wi was determined and the average for all trees was calculated. Wi and values close to 0 were associated with a regular neighbourhood of tree i, while values of Wi and close to 1 corresponded to irregularity of the spatial distribution in the neighbourhood of tree i.

Finally, calculation of also requires calculation of Ri. The exact value of for a Poisson distribution in each plot in a 4-tree sample was 1.799, as obtained by a simulation based on 106 trees. Values of Ri and plot mean close to 0 were associated with a regular tendency of the neighbourhood of tree i, while values of Ri and larger than 1.799 were associated with a tendency of the spatial distribution in the neighbourhood of tree i to be irregular (see more in [40]).

To exclude the edge effect, and therefore to enhance the accuracy of the estimates, the nearest-neighbour edge-correction concept (NN1) was applied, as proposed by [29], for calculating CE, and .

The hypothesis of complete spatial randomness (CSR) for the mean values of CE, and for each plot was tested by atwo-sided permutation test (here 10,000 permutations) If 1—P(ZCE), P(Z), and P(Z) are non-significantly small or non-significantly high (i.e. 0.01 < P < 0.99, at the 1% acceptance level), we can expect random effects and otherwise, directed forces. If the observed 1—P(ZCE), P(Z) or P(Z) values are smaller than 0.01, we can assume non-randomly acting diversifying forces (e.g. seed dispersal pattern, association with nutrient-rich patches) that will produce a clustered distribution. If the observed 1—P(ZCE), P(Z) or P(Z) values are larger than 0.99, we assume that non-randomly acting homogenizing forces (e.g. competition for light, water and nutrients) will yield a regular distribution [50], [51],[40]. After Bonferroni correction[52], the new (modified) critical P value (significance level* = 0.0002) was calculated by dividing the critical P value (here the significance level = 0.05) by the number of comparisons (hypotheses) (m = 216).

Spatial Structural Analysis by Ripley’s K(t)-function.

Ripley’s K(t)-function is used to characterize completely mapped spatial point process data. The mapped data are usually recorded in two or three dimensions and include the locations of all events in a predefined study area. Unlike other functions (e.g. mean nearest-neighbour distance or the cumulative distribution function of distance from random points to their nearest neighbours), this function preserves information about distances between all points in the pattern, thus enabling visualization of how point pattern distributions vary with scale. Ripley’s K(t)function is useful for summarizing point patterns, testing hypotheses about the patterns, estimating parameters and fitting models. Bivariate or multivariate K(t) functions can be used to describe relationships between two or more point patterns [53].

Ripley’s K-function [33], [35] was used to determine the scales at which the tree pattern in each plot tends to be regular, clumped or random. The function was used to describe the relationship between the spatial pattern of Picea chihuahuana and the spatial structure of the other tree species inside the 12 plots.

The univariate Ripley's K-function [53] can be estimated as (1) where A is the area of the study region, n is the number of observed points, wij(r) is an edge effect correction factor, δ(r) is an indicator function and dij is the distance between the i-th and j-th points.

Because of its hyperbolic behavior, the interpretation of the K-function is not straightforward and a modification, called the L-function, was proposed by Besag (1977) in order to normalize the function (Besag, 1977): (2)

Now, the expected value of the univariate L-function under CSR is 0 for all r, positive when the pattern tends to be clustered and negative when the pattern tends to be regular.

In order to test the deviation from randomness of the point pattern using the univariate L-function, a 99% simulation envelope of L(r) was computed, using the Monte Carlo Method [54], from 1,000 simulated CSR patterns with the same number of points contained inside a region with the same geometry.

The bivariate Ripley’s K-function [53] is estimated as (3) where ni and nj are the numbers of the points of type i and j respectively, wik,jl(r) is an edge effect correction factor and δ(r) is an indicator function and dik,jl is the distance between the k-th point of type i and the l-th point of type j.

Due to the edge effect, Kij and Kji are correlated, but not identical, and therefore the following means of estimating KB is recommended: (4)

Its associated bivariate L-function is defined as (5)

The expected value of the bivariate L-function under spatial independence is 0 for all r, positive when the two point processes tend to be aggregated and negative when the two point processes tend to be repulsive.

In order to generate the simulation envelope that corresponds to the hypothesis of spatial independence, the method holds the point pattern of type 1 and type 2 constant and then randomizes their relative position in each simulation. For more details, see Lotwick and Silverman [55].

The trees were grouped in two diameter classes to check i) whether clustering at small scales was caused by a high degree of aggregation of smaller trees (using the univariate L-function) and ii) whether young trees were clustered around the adults (using the bivariate L-function). To find the dcut (cut-off) that divides the population into smaller and larger individuals (determined by their DBH), we obtained the bivariate L-function for various values within the range 7 < dcut,DBH < 40 cm and we therefore chose the dcut where the aggregation patterns between the groups were visually more significant.

dcut,DBH,all corresponded to 23.2 cm, dcut,DBH,Pch to 29.3 cm.

All analyses were performed using the "Spatstat" package implemented in the free statistical application R[56][57].

The statistical tests for spatial tree pattern, null hypothesis, interpretation and related ecological questions are summarised in Table 2.

thumbnail
Table 2. Summary of statistical tests of spatial tree structure, null hypothesis, interpretation and related ecological questions; CE = neighbourhood-based Clark-Evans index, = Uniform Angle index, = Mean Directional index, LU(r) = univariate L-function, and LB(r) = bivariate L-function.

https://doi.org/10.1371/journal.pone.0140442.t002

Covariation analysis

The relationships between stand densities (N and G), relative frequency of 10 cm DBH class (fcd) and tree species diversities (the Hill numbers υsp,0, υsp,2, andυsp,∞ [46]), degree of latitude (lat) and spatial pattern indices (CE, and ) were measured by the covariation (C) described by Gregorius et al. [50]. This method can detect types of covariation that are monotonous but not necessarily linear. C ranges between -1 and 1, where C = 1 indicates an entirely positive covariation and C = -1 a strictly negative covariation. If the denominator is zero, C is undefined [50]. Formally, (6)

In order to test the possibility that the observed degrees of covariation C[N x CE], C[N x ], C[N x ], C[G x CE], C[G x ], C[G x ], C[fcd x CE], C[fcd x ], C[fcd x ], C[υsp x CE], C[υsp x ], C[υsp x ], C[lat x CE], C[lat x ] and C[lat x ] were only produced by random events rather than directed forces, a one-sided permutation test was performed (here 10,000 permutations) [58]. After Bonferroni correction, the new critical P value was 0.002.

In order to test whether the observed differences in the average spatial pattern indices (Diff) (CE, and ) between i) P. chihuahuana trees and all other tree species, ii) suppressed and dominant P. chihuahuana trees and iii) suppressed and dominant trees of all species in the plots were produced solely by random events rather than directed forces, a permutation test based on 10,000 randomly chosen reassignments was performed [58]. Loosely based on the BAL competition Index [59] [60], the dominant tree class used in the present study included all larger trees that together included 50% of the stand basal area. The suppressed trees included all smaller trees that together included the other 50% of the stand basal area. This permutation test constitutes a non-parametric approach, which among other uses enables comparison of two groups in terms of the mean values of some variable; however, unlike with the t test, the assumptions of normality and equality of variances do not need to be satisfied by the data [51]. After Bonferroni correction of the data, the new critical P value was 0.007.

Results

Most of the P. chihuahuana trees were randomly distributed (92% of cases), as confirmed by all measures used (and considering a significance level of 1%): the Clark-Evans index (CE), uniform angle index (), mean directional index () and univariate L-function. Based on CE, and , 8% of the plots display clusteringat the 1% significance level. The univariate L-function indicated CSR in all plots, mostly due to the low effective tree number (repetitions) for calculating the values (Table 3). The bivariate L-function showed that the number of Picea trees smaller than 29.3 cm DBH in the neighborhood of larger Picea trees (or equivalently the number of larger Picea trees in the neighbourhood of smaller Picea trees) was only larger than expected in TN and CV (Table 4). However, after Bonferroni correction of the data, all indices indicated random distribution of the P. chihuahuana trees in all plots.

Considering all tree species in each plot, the CE indicated CSR in 67% of the plots. The and indicate CSR in 92% of the plots at the 1% significance level (Table 5). The univariate L-function shows that the trees of all species in 42% of the plots are completely randomly distributed. For all trees and trees smaller than 23.2 cm DBH, a considerable proportion (42%) of the plots demonstrated clustering at smaller scales. For trees equal or larger than 23.2 cm DBH, all plots show CSR. For all trees, a smaller proportion (17%) of the plots indicate tree clustering at the intermediate and larger scales (Table 4, Fig 3 (above and centre)). The bivariate L-function demonstrated that the number of smaller trees in the neighbourhood of larger trees (or equivalently the number of larger trees in the neighbourhood of smaller trees) was larger than expected (Table 4). However, after Bonferroni correction, the data indicated that the trees in 92% of the plots were randomly distributed (P ≥ 0.0002).

thumbnail
Fig 3. Sample plots (50 x 50 m) illustrating the location of each tree, univariate and bivariate L-functions.

Sample plots (50 x 50 m) illustrating the location of each tree of diameter at breast height (DBH) above 7 cm in the very rare and species-rich Picea chihuahuana tree community in Quebrada de los Durán (QD) and San José de las Causas (SJ), Durango (Mexico). S1 represents the P. chihuahuana trees (above), and S2–S7 represent the other tree species. For the point patterns in the QDand SJ plots, univariate L-functions are represented by black lines. The 99% simulation envelope (dashed red lines) for the CSR hypothesis was calculated via the Monte Carlo Method (Besag 1977), with 1,000 simulations (central). For the point pattern in the QD and SJ plots, the bivariate L-function is indicated represented by black lines. The 99% simulation envelope (dashed red lines) for the Independence hypothesis was calculated by the Random Shifting Method (Lotwick & Silverman, 1983), with 1,000 simulations (below).

https://doi.org/10.1371/journal.pone.0140442.g003

thumbnail
Table 3. Spatial structure of Picea chihuahuana M. in the 50 x 50 m plots in the 12 study locations, based on the neighbourhood-based Clark-Evans index (CE), Uniform Angle index (), and Mean Directional index () (P values estimated with 10,000 permutations) and univariate L-function

. The 99% simulation envelope (dashed red lines) for the CSR hypothesis was calculated by the Monte Carlo Method (Besag 1977), with 1,000 simulations (distance interval equals 0–12 m). N equals the number of Picea chihuahuana M. trees in the plot. 1) La Tinaja (TN), 2) El Ranchito (RC), 3) El Cuervo (CV), 4) Talayote (TY), 5) Las Trojas (TR), 6) El Venado (VN), 7) La Quebrada (LQ), 8) Paraje Piedra Rayada (PPR), 9) Quebrada de los Durán (Arroyo del Indio Ignacio) (QD), 10) Cebollitas (CB), 11) San José de las Causas (SJ), and 12) La Pista (Arroyo de La Pista) (LP).

https://doi.org/10.1371/journal.pone.0140442.t003

thumbnail
Table 4. Analysis of spatial tree structure in 50 x 50 m plots in the twelve locations including all tree species using the univariate (for all trees, smaller trees of all species [44](< 23.2 cm diameter at the breast height (DBH)), larger trees (≥ 23.2 m DBH), and bivariate versions of the L-function (spatial pattern of Picea chihuahuana (Pch) vs. other tree species, pattern of smaller vs. larger trees and of smaller Pch (< 29.3 cm DBH) vs. larger Pchtrees (≥ 29.3 cm DBH)).

The 99% simulation envelope (dashed red lines) for the CSR hypothesis (for the univariate L-function) and for spatial independence hypothesis (for the bivariate L-function) was calculated by the Monte Carlo Method (Besag 1977), with 1,000 simulations (distance interval equals 0–12 m).1) La Tinaja (TN), 2) El Ranchito (RC), 3) El Cuervo (CV), 4) Talayote (TY), 5) Las Trojas (TR), 6) El Venado (VN), 7) La Quebrada (LQ), 8) Paraje Piedra Rayada (PPR), 9) Quebrada de los Durán (Arroyo del Indio Ignacio) (QD), 10) Cebollitas (CB), 11) San José de las Causas (SJ), and 12) La Pista (Arroyo de La Pista) (LP).

https://doi.org/10.1371/journal.pone.0140442.t004

thumbnail
Table 5. Analysis of spatial tree structure in 50 x 50 m plots in the 12 locations including all tree species (species shown [44]) and based on the neighbourhood-based Clark-Evans index (CE), Uniform Angle index (), and Mean Directional index ().

P values estimated with 10,000 permutations. 1) La Tinaja (TN), 2) El Ranchito (RC), 3) El Cuervo (CV), 4) Talayote (TY), 5) Las Trojas (TR), 6) El Venado (VN), 7) La Quebrada (LQ), 8) Paraje Piedra Rayada (PPR), 9) Quebrada de los Durán (Arroyo del Indio Ignacio) (QD), 10) Cebollitas (CB), 11) San José de las Causas (SJ), and 12) La Pista (Arroyo de La Pista) (LP).

https://doi.org/10.1371/journal.pone.0140442.t005

The Picea and other tree species were not spatially segregated, i.e. Picea tended to be found in patches of other tree species excepting the location San Jose de las Causas (SJ) (Table 4, Fig 3 at the bottom left and right).

Analysis of the suppressed or immature trees versus the dominant or mature trees revealed that at the 1% significance level and according to CE, suppressed P. chihuahuana M. trees occurred in three plots (25%) and suppressed trees of all species occurred in five plots (41.7%), thus demonstrating clustering. However, dominant trees showed CSR in all 12 plots (Tables 6 and 7). The and values indicated CSR for all the plots where they were obtained. However, the mean and values for all plots of the suppressed trees (of all species) were statistically significantly higher than the dominant trees (of all species) (P = 0.0009 and P = 0.0006, respectively), also after Bonferroni correction (P < 0.007). Therefore, the tendency for clustering was significantly higher in the suppressed trees than in the dominant trees. and failed in some plots because of an insufficient number of trees(repetitions) for the calculations.

thumbnail
Table 6. Analysis of spatial structure of the suppressed and dominant trees in 50 x 50 m plots containing all Picea chihuahuana M. trees, in the 12 study locations, based on the neighbourhood-based Clark-Evans index (CE), Uniform Angle index (), and Mean Directional index ().

P values estimated with 10,000 permutations. N equals the tree number in the plot. 1) La Tinaja (TN), 2) El Ranchito (RC), 3) El Cuervo (CV), 4) Talayote (TY), 5) Las Trojas (TR), 6) El Venado (VN), 7) La Quebrada (LQ), 8) Paraje Piedra Rayada (PPR), 9) Quebrada de los Durán (Arroyo del Indio Ignacio) (QD), 10) Cebollitas (CB), 11) San José de las Causas (SJ), and 12) La Pista (Arroyo de La Pista) (LP).

https://doi.org/10.1371/journal.pone.0140442.t006

thumbnail
Table 7. Analysis of spatial structure of the suppressed and dominant trees in 50 x 50 m plots containing all tree species(species shown [44]),in the 12 study locations, based on the neighbourhood-based Clark-Evans index (CE), Uniform Angle index (), and Mean Directional index ().

P values estimated with 10,000 permutations. N equals the tree number in the plot. 1) La Tinaja (TN), 2) El Ranchito (RC), 3) El Cuervo (CV), 4) Talayote (TY), 5) Las Trojas (TR), 6) El Venado (VN), 7) La Quebrada (LQ), 8) Paraje Piedra Rayada (PPR), 9) Quebrada de los Durán (Arroyo del Indio Ignacio) (QD), 10) Cebollitas (CB), 11) San José de las Causas (SJ), and 12) La Pista (Arroyo de La Pista) (LP).

https://doi.org/10.1371/journal.pone.0140442.t007

The covariations C[CE x N], C[x N], C[ x N], C[CE x G], C[ x G], C[ x G], C[fcd x CE], C[fcd x ], C[fcd x ], C[CE x υsp], C[ x υsp], and C[ x υsp] were not statistically significant. The strongest covariation (C) between spatial pattern indices and tree density was C[ x N] with + 0.61 (P = 0.07) (i.e. clustering tended to be associated, but not significantly, with high stand density). The strongest covariation (C) between spatial pattern indices and tree species diversity was C[ x υsp,2] with + 0.51 (P = 0.12) (i.e. clustering tended to be associated, but not significantly, with high tree species diversity). None of the 12 uneven-aged forest plots showed a statistically significant regular spatial tree pattern.

No statistically significant differences (Diff) (P < 0.01) between the average spatial pattern indices (CE, and ) were observed for i) P. chihuahuana and all tree species or ii) suppressed and dominant P. chihuahuana trees.

When latitude was analyzed, we found that after Bonferroni correction, CE, but not and , decreased significantly from the southern to northern plots containing all tree species (C = -0.97, P = 0.0008).

Discussion and Conclusions

In this study, we analysed the fine-scale spatial tree patterns in a special forest tree community of P. chihuahuana M. in Mexico. We examined the spatial tree pattern and its relationships to tree stand density, vertical structure, tree species diversity, geographical latitude and tree dominance.

The findings demonstrate that the hypothesis of aggregated tree pattern is not applicable to the mean pattern measured by CE, and of the uneven-aged P. chihuahuana trees and P. chihuahuana tree community (Tables 3 and 5) and but to specific spatial scales measured by the univariate L-function, because 58% of the plots showed clustering at small (42%), intermediate and larger scales (17%) (Table 4). Frequent clustering at small scales was mainly caused by a high degree of aggregation of trees smaller than 23.2 cm DBH (Table 4) [42]. We also found that the mean and values for immature (suppressed) were significantly larger than the corresponding values for mature (dominant) trees (Table 7). We also observed that immature P. chihuahuana trees showed clustering in 25% of the plots, according to the CE index. As in many primeval forests, small (young) individuals are almost always located in groups (Table 4)—often as a result of tree fall gaps in the canopy as the dominant type of disturbance [61][62]. The common clustering at small scales in the P. chihuahuana tree community indicates that the forest patches were often created only by one canopy tree falling, as typically observed in species-rich tropical rain forests [63] and occasionally in old-growth (subalpine) spruce-fir forests. These forests tend to be horizontally structured, mainly because an initiating disturbance is followed by long periods when small-scale, low-intensity disturbances control tree regeneration [8]. Therefore, fire disturbance, which almost always homogenizes stands, is a rare event in the P. chihuahuana tree community and should not be necessary for or beneficial to the community dynamics. Moreover, fires may bring an end to this fragmented tree community in very small and isolated locations [1] [17]. In contrast, the clustering by small-scale disturbances may be mainly caused by insect attack, disease or windthrow, which may create patchiness and spatial heterogeneity within locations [8].

The rare clustering at larger scales was mainly affected by the low tendency of aggregation of canopy trees (Table 4), as also reported by Malik et al. [64], Christensen [65] and Whipple [66] for uneven-aged populations. Therefore, the overall random patterns were a result of shift from initial aggregation to a random distribution [67]). Hence, Lepš and Kindlmann [42] postulated that i) an observed random pattern does not represent evidence of independence of individuals and ii) the intensity of spatial pattern should not be considered a measure of community organization.

Self-thinning due to intra and inter-specific competition-induced mortality was probably the main cause of the decrease in aggregation intensity [42], [67] during the course of population development in the P. chihuahuana tree community. However, environmental heterogeneity, uneven-age distributions, insufficient competition, limited seed dispersal and random germination may have prevented the presence of a significantly regular pattern of the mature trees in the tree community under study [68]. The aforementioned factors, particularly insufficient competition in the plot may also have favoured clustering in the northern locations (plots).

The number of P. chihuahuana trees in the neighbourhood of other tree species (or the number of trees from other species in the neighbourhood of P. chihuahuana) was not expected. The spatial distribution of P. chihuahuana trees was independent of the patches of other tree species, except in the San Jose de las Causas (SJ) location (Table 4, Fig 3). In SJ, spruces had a repulsive pattern to other species, similarly to a study in an old growth spruce-fir forest in Changbaishan Natural Reserve, China [41]. We therefore assume that there was a similar but weaker inter- and intraspecific competition between the trees at the species level [68] and that P. chihuahuana can tolerate partial shade conditions. The bivariate L-function showed that smaller trees (of all species) often grew in the neighbourhood of larger trees (of all species) (Table 4), typically in uneven-aged forests (Fig 2) and probably as a result of some shade-tolerant frequent tree species under mature canopy (such as Abies durangensis, Cupressus lindleyi and Juniperus deppeana) and the slender shaped crowns of the mature canopy trees in this community [69]. The smaller P. chihuahuana trees, did not, however, generally grow more frequently in the surroundings of larger Picea trees (Table 4), perhaps only for statistical reasons (such as few tree/repetitions) or due to no special preferences of the young (small) Picea trees for mature canopy Picea trees. In this study, P. chihuahuana regeneration was occasionally found on horizontal dead trees.

The plots in which a clustered structure was observed tended to be associated (not significantly) with a large number of trees because of the presence of a greater number of understory trees. Understory trees often displayed a tendency to grouping (Table 4).

Although the aggregation indices were not associated with the diameter distribution, the results showed that none of the 12 uneven-aged forest plots under study displayed a statistically significant regular spatial tree pattern; however, 58% showed clustering in specific spatial scales at the 0.1% significant level (Table 4). No covariation (C) between aggregation indices and diameter distributions was observed, because the 12 diameter distributions scarcely varied in their reverse J-shaped form (Table 2)

The cluster structure was weakly positively related to higher tree species diversity, probably due to a combination of the accumulation effect [13] [70] and increasing competition in denser plots [71]. While the accumulation effect resulted in higher diversity, the self-thinning processes led to saturation in tree species diversity [72].The high tree species diversity in the P. chihuahuana community [13] may also provoke clustering at smaller scales (in small gaps) because the lifespan and dimension of each tree species are often different. The probability that two or more trees of different species would fall at the same time and create a gap is lower than the probability of the same happening with trees of the same species.

We conclude that satisfactory understanding of spatial forest structure is essential for the sustainable conservation of this unique mixed uneven-aged Picea forest [20]. Our measures of spatial tree structure, particularly and failed in several plots because of an insufficient number of trees (repetitions) for the calculations. Therefore, we recommend use of larger sample plot sizes (> 0.25 ha) in uneven-aged and species-rich forest ecosystems to detect less obvious, but important, relationships between spatial tree pattern and functioning and diversity in these forests.

Supporting Information

Acknowledgments

We thank Sergio Leonel Simental-Rodríguez and Carmen Zulema Quiñones-Pérez for their helpful assistance.

Author Contributions

Conceived and designed the experiments: CW. Performed the experiments: CW. Analyzed the data: CW JMBP. Contributed reagents/materials/analysis tools: CW. Wrote the paper: CW JMBP FCC ACP JHMG.

References

  1. 1. Ledig FT, Mapula L M, Bermejo V B, Reyes H V, Flores-López C, Capó-Arteaga MA. Locations of endangered spruce populations in Mexico and the demography of Picea chihuahuana. Madroño. 2000; 47: 71–88.
  2. 2. Norma Oficial Mexicana. NOM-059-ECOL-2010. Protección ambiental -Especies nativas de México de flora y fauna silvestres- Categorías de riesgo y especificaciones para su inclusión, exclusión o cambio. Lista de especies en riesgo. Diario Oficial de la Federación (Segunda sección),2010. p 1–77.
  3. 3. Wehenkel C, Sáenz-Romero C. Estimating genetic erosion using the example of Picea chihuahuana Martínez. Tree Genet.Genomes. 2012; 8(5): 1085–1094.
  4. 4. Gordon AG. Ecology of Picea chihuahuana Martínez.Ecology. 1968; 49: 880–896.
  5. 5. Sánchez CJ. Picea chihuahuana, conífera en Peligro de Extinción. Cienc Forest. 1984; 9: 51–63.
  6. 6. Narváez FR. Contribution to the Knowledge of the Ecology of Picea chihuahuana.Bachelor thesis (Biology), Faculty of Biological Sciences, Universidad Autónoma de Nuevo León, Monterrey, Nuevo León, México, 1984.
  7. 7. Quiñones-Pérez CZ, Sáenz-Romero C, Wehenkel C. Genetic diversity and conservation of Picea chihuahuana Martínez: A review. Afr J Biotechnol. 2014; 13(28): 2786–2795.
  8. 8. Parish R., Antos J. A., and Fortin M-J.. Stand development in an old-growth subalpine forest in southern interior British Columbia. Can. J. Forest Res. 1999; 29(9): 1347–1356.
  9. 9. Instituto Nacional de Estadística, Geografía e Informática (INEGI). Estadísticas históricas de México 2014. 2015; Available: http://www.inegi.org.mx/prod_serv/contenidos/espanol/bvinegi/productos/nueva_estruc/HyM2014/22.%20MedioAmbiente.pdf accessed 25 August 2015.
  10. 10. Ledig FT, Jacob-Cervantes V, Hodgskiss PD, Eguiluz-Piedra T. Recent evolution and divergence among populations of a rare Mexican endemic, Chihuahua spruce, following Holocene climatic warming. Evolution. 1997; 51: 1815–1827.
  11. 11. Jaramillo-Correa JP, Beaulieu J, Ledig FT, Bousquet J. Decoupled mitochondrial and chloroplast DNA population structure reveals holocene collapse and population isolation in a threatened Mexican-endemic conifer. Mol Ecol. 2006; 15: 2787–2800. pmid:16911200
  12. 12. Quiñones-Pérez CZ, Sáenz-Romero C, Wehenkel C. Influence of neighbouring tree species on AFLP variants of endangered Picea chihuahuana populations on the Sierra Madre Occidental, Northeastern México. Pol J Ecol. 2014; 62(1): 69–79.
  13. 13. Simental-Rodríguez SL, Quiñones-Pérez CZ, Moya D, Hernández-Tecles E, López-Sánchez CA, Wehenkel C The Relationship between Species Diversity and Genetic Structure in the Rare Picea chihuahuana Tree Species Community, Mexico. Plos One. 2014; 9(11): e111623. pmid:25375134
  14. 14. Wehenkel C, Sáenz-Romero C, Jaramillo-Correa JP. (2015): Estimating genetic erosion in threatened conifers: the example of Picea chihuahuana Martínez. In: Ahuja M. R. and Jain S.M. (eds.): Genetic Erosion and Biodiversity, Springer SBM, The Netherlands, 2015. in press.
  15. 15. Aguilar-Soto V, Melgoza-Castillo A, Villarreal-Guerrero F, Wehenkel C, Pinedo-Alvarez C. Modeling the potential distribution of Picea chihuahuana Martinez, an endangered species on the Sierra Madre Occidental, Mexico, Forests. 2015; 6: 692–707.
  16. 16. Ledig FT, Rehfeldt GE, Sáenz-Romero C, Flores-López C. Projections of suitable habitat for rare species under global warming scenarios. Am J Bot. 2010; 97: 970–987. pmid:21622467
  17. 17. Ledig FT. Climate Change and Conservation. Acta Silv. Lign. Hung. 2012; 8: 57–74.
  18. 18. Mendoza-Maya E, Espino-Espino J, Quiñones-Pérez CZ, Flores-López C, Wehenkel C, Vargas-Hernández JJ, et al. Proposal for conservation of three endangered species of Mexican spruce. Fitotécnica. 2015, 38(3): 235–247.
  19. 19. Spies TA. Forest Structure: A Key to the Ecosystem. Northwest Science. 1998; 72(2): 34–39.
  20. 20. Pommerening A. Evaluating structural indices by reversing forest structural analysis. Forest Ecol. Manage. 2006; 224(3): 266–277.
  21. 21. Gadow Kv, Zhang CY, Wehenkel C, Pommerening A, Corral-Rivas JJ, Korol M, et al. Forest Structure and Diversity, Chapter 2, In: Pukkala T, Gadow Kv, editors. Continuous Cover Forestry. Book Series Managing Forest Ecosystems 23, Springer Science+Business Media B.V; 2012.pp. 29–83.
  22. 22. Moeur M. Characterizing spatial patterns of trees using stem-mapped data. For Sci. 1993; 39: 756–775.
  23. 23. Miller TE, Weiner J. Local density variation may mimic effects of asymmetric competition on plant size variability. Ecology. 1989; 70: 1188–1191.
  24. 24. Kenkel NC, Hoskins JA, Hoskins WD. Local competition in a naturally established jack pine stand. Can J Bot. 1989; 67: 2630–2635.
  25. 25. Newton PF, Jolliffe PA. Assessing processes of intraspecific competition within spatially heterogeneous black spruce stands. Can J Forest Res. 1998; 28: 259–275.
  26. 26. Condit R, Ashton PS, Baker P, Bunyavejchewin S, Gunatilleke S, Gunatilleke N, et al. Spatial patterns in the distribution of tropical tree species. Science 2000; 288(5470): 1414–1418. pmid:10827950
  27. 27. Kuuluvainen T., Pukkala T. Effect of crown shape and tree distribution on the spatial distribution of shade. Agr Forest Meteorol. 1987; 40: 215–231.
  28. 28. Pretzsch H. Zum Einfluß des Baumverteilungsmusters auf den Bestandszuwachs. Allg Forst Jagdztg. 1995; 166: 190–201.
  29. 29. Guntis B, Didzis E, Ligita L, Iluta L, Guntis T, Didzis T. Age and spatial structure of natural Pinus sylvestris stands in Latvia. Scand J Forest Res. 2005; 20: 471–480.
  30. 30. Li L, Huang ZL, Ye WH, Cao HL, Wei SG, Wang ZG, et al. Spatial distributions of tree species in a subtropical forest of China. Oikos. 2009; 118(4): 495–502.
  31. 31. Clark PJ, Evans FC.Distance to nearest neighbour as a measure of spatial relationships in populations.Ecology 1954; 35: 445–453.
  32. 32. Diggle PJ. Statistical Analysis of Spatial Point Patterns.Academic Press, London; 1983.
  33. 33. Ripley BD. Modeling spatial patterns (with discussion).Journal of the Royal Statistical Society. Series B 1977; 39: 172–212.
  34. 34. Stoyan D, Stoyan H. Fractals, random shapes and point fields: methods of geometrical statistics. Wiley, Chichester; 1994.
  35. 35. Upton G, Fingleton B. Spatial data analysis by example.Vol. 1: Point Pattern and Quantitative Data. John Wiley & Sons; 1985.
  36. 36. Gadow Kv, Hui GY, Albert M. Das Winkelmass—ein Strukturparameter zur Beschreibung der Individualverteilung in Waldbeständen. Cent bl gesamte Forstwes. 1998; 115(1): 1–9.
  37. 37. Hui GY, Gadow Kv. Das Winkelmass—Herteilung des Optimalen Standardwinkels. Allg Forst Jagdz. 2002; 10: 173–177.
  38. 38. Aguirre O, Hui GY, Gadow Kv, Jimenez J. Comparative Analysis of Natural Forest Sites in Durango, Mexico. For Ecol Manage. 2003; 183: 137–145.
  39. 39. Illian J, Penttinen A, Stoyan H, Stoyan D. Statistical Analysis and Modelling of Spatial Point Patterns. John Wiley & Sons Ltd, Chichester; 2008.
  40. 40. Corral-Rivas JJ, Wehenkel C, Castellanos BH, Vargas-Larreta B, Diéguez-Aranda U. A permutation test of spatial randomness: application to nearest neighbour indices in forest stands. J Forest Res. 2010; 15: 218–225.
  41. 41. Chen J, Bradshaw GA. Forest structure in space: a case study of an old growth spruce-fir forest in Changbaishan Natural Reserve, PR China. Forest Ecol. Manage., 1999; 120(1): 219–233.
  42. 42. Lepš J, Kindlmann P. Models of the development of spatial pattern of an even-aged plant population over time. Ecol Model. 1987; 39(1): 45–57.
  43. 43. Sáenz-Romero C, Rehfeldt GE, Crookston NL, Duval P, St-Amant R, Beaulieu J. Spline models of contemporary, 2030, 2060 and 2090 climates for Mexico and their use in understanding climate-change impacts on the vegetation. Clim Chang. 2010; 102: 595–623.
  44. 44. Quiñones-Pérez CZ, Simental-Rodríguez SL, Sáenz-Romero C, Jaramillo-Correa JP, Wehenkel C. Spatial genetic structure in the very rare and species-rich Picea chihuahuana tree community (Mexico). Silvae Genet. 2014; 63(4): 149–159.
  45. 45. Wehenkel C, Corral-Rivas JJ, Hernández-Díaz JC, Gadow Kv. Estimating Balanced Structure Areas in multi-species forests on the Sierra Madre Occidental, Mexico. Annals of Forest Science, 2011; 68: 385–394.
  46. 46. Hill MO. Diversity and evenness: a unifying notation and its consequences. Ecology. 1973; 54: 427–432.
  47. 47. Corral-Rivas JJ, Pommerening A, Gadow Kv, Stoyan D. An analysis of two directional indices for characterizing the spatial distribution of forest. In Corral-Rivas JJ, editor. Models of tree growth and spatial structure for multi-species, uneven-aged forests in Durango (Mexico). PhD. Thesis. University of Göttingen, 2006; pp. 106–121.
  48. 48. Albert M. Analyse der eingriffsbedingten Strukturveränderung und Durchforstungsmodellierung in Mischbeständen. Doctoral dissertation, University of Göttingen, Germany, 1999.
  49. 49. Hui GY, Hu YB. Measuring species spatial segregation in mixed forest. For Res. 2001; 14: 23–27.
  50. 50. Gregorius HR, Degen B, König A. Problems in the analysis of genetic differentiation among populations a case study in Quercus robur. Silvae Genet. 2007; 56: 190–199.
  51. 51. Wehenkel C, Corral-Rivas JJ, Castellanos-Bocaz HA. Is there selection by species diversity in Picea abies L.? Plant Ecol. 2010; 208: 47–54.
  52. 52. Hochberg Y. A sharper Bonferroni procedure for multiple tests of significance. Biometrika 1988; 75 (4): 800–802.
  53. 53. Dixon PM. Ripley's K function. Encyclopedia of Environmetrics; 2002.
  54. 54. Besag J, Diggle PJ. Simple Monte Carlo tests for spatial pattern. Applied statistics. 1977; 327–333.
  55. 55. Lotwick HW, Silverman BW. Methods for analyzing spatial processes of several types of points. Journal of the Royal Statistical Society, Series B. 1982; 39: 172–212.
  56. 56. Baddeley A, Turner R. spatstat: An R Package for Analyzing Spatial Point Patterns. Journal of Statistical Software, 2005; 12(6): 1–42. URL http://www.jstatsoft.org/v12/i06/.
  57. 57. Development Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria, 2014. URL http://www.R-project.org/.
  58. 58. Manly BFJ. Randomization, Bootstrap and Monte Carlo methods in biology. Chapman and Hall, London; 1997.
  59. 59. Wykoff WR. A Basal Area Increment Model for Individual Conifers in the Northern Rocky Mountains. For Sci. 1990; 36(4): 1077–1104.
  60. 60. Schröder J, Gadow Kv. Testing a new competition index for maritime pine in northwestern Spain. Can J For Res. 1999; 29(2): 280–283.
  61. 61. Shugart HH. A Theory of Forest Dynamics.Springer, New York—Berlin—Heidelberg–Tokyo; 1984.
  62. 62. Whitmore TC. Canopy gaps and the two major groups of forest trees. Ecology. 1989; 70: 536–538.
  63. 63. Hubbell SH, Foster RB, O´Brien ST, Harms KE, Condit R, Wechsler B, et al. Light-gap disturbances, recruitment limitation, and tree diversity in a Neotropical forest. Science. 1999; 283: 554–557. pmid:9915706
  64. 64. Malik AR, Anderson DJ, Myescough PJ. Studies on structure in plant communities. VII. Field and experimental analyses of Atriplex oesicaria populations from the Riverine Plain of New South Wales. Aust J Bot. 1976; 24: 265–280.
  65. 65. Christensen NL. Changes in structure, pattern and diversity associated with climax forest maturation in Piedmont, North Carolina. Am. Midl. Nat., 1977; 97: 176–188.
  66. 66. Whipple SA. Population dispersion patterns of trees in a southern Louisiana hardwood forest. Bull. Torrey Bot. Club. 1980; 107: 71–76.
  67. 67. Getzin S, Dean C, He FA, Trofymow J, Wiegand K, Wiegand T. Spatial patterns and competition of tree species in a Douglas—fir chronosequence on Vancouver Island. Ecography. 2006; 29(5): 671–682.
  68. 68. Kenkel NC. Patterns of self-thinning in jack pine: testing the random mortality hypothesis. Ecology. 1988; 69: 1017–1024.
  69. 69. Schütz JP. Silvicultural tools to develop irregular and diverse forest structures. Forestry. 2002; 75(4): 329–337.
  70. 70. Gotelli NJ, Colwell RK. Quantifying biodiversity: procedures and pitfalls in the measurement and comparison of species richness. Ecol Lett. 2001; 4: 379–391.
  71. 71. Begon M, Harper JL, Townsend CR. Ecology: Individuals, populations and communities. Blackwell Science. Third edition. USA; 1996.
  72. 72. Silva-Flores R, Perez-Verdin G, Wehenkel C. Relationship between diversity of tree species and climatic factors in the Sierra Madre Occidental, Mexico. Plos One. 2014; 9(8): e105034. pmid:25127455