The effect of soil on spatial variation of the herbaceous layer modulated by overstorey in an Eastern European poplar-willow forest

The tree species composition can influence the dynamics of herbaceous species and enhance the
spatial heterogeneity of the soil. But there is very little evidence on how both overstorey structure
and soil properties affect the spatial variation of the herb layer. The aim of this study is to evaluate the factors of the soil and overstorey structure by which it is possible to explain the fine-scale
variation of herbaceous layer communities in an Eastern European poplar-willow forest. The research was conducted in the “Dnipro-Orils’kiy” Nature Reserve (Ukraine). The research polygon
(48º30’51”N, 34º49”02”E) was laid in an Eastern European poplar-willow forest in the floodplain
of the River Protich, which is a left inflow of the River Dnipro. The site consists of 7 transects.
Each transect was made up of 15 test points. The distance between rows in the site was 3 m. At
the site, we established a plot of 45×21 m, with 105 subplots of 3×3 m organized in a regular grid.
The adjacent subplots were in close proximity. Vascular plant species lists were recorded at each
3×3 m subplot along with visual estimates of species cover using the nine-degree Braun-Blanquet
scale. Within the plot, all woody stems ≥ 1 cm in diameter at breast height were measured and
mapped. Dixon’s segregation index was calculated for tree species to quantify their relative spatial
mixing. Based on geobotanical descriptions, a phytoindicative assessment of environmental factors according to the Didukh scale was made. The redundancy analysis was used for the analysis
of variance in the herbaceous layer species composition. The geographic coordinates of sampling
locations were used to generate a set of orthogonal eigenvector-based spatial variables. Two measurements of the overstorey spatial structure were applied: the distances from the nearest tree of
each species and the distance based on the evaluation of spatial density of point objects, which
are separate trees. In both cases, the distance matrix of sampling locations was calculated, which
provided the opportunity to generate eigenvector-based spatial variables. A kernel smoothed intensity function was used to compute the density of the trees’ spatial distribution from the point
patterns’ data. Gaussian kernel functions with various bandwidths were used. The coordinates of sampling locations in the space obtained after the conversion of the trees’ spatial distribution
densities were used to generate a set of orthogonal eigenvector-based spatial variables, each of
them representing a pattern of particular scale within the extent of the bandwidth area structured
according to distance and reciprocal placement of the trees. An overall test of random labelling reveals the total nonrandom distribution of the tree stems within the site. The unexplained variation
consists of 43.8%. The variation explained solely by soil variables is equal to 15.5%, while the variation explained both by spatial and soil variables is 18.0%. The measure of the overstorey spatial
structure, which is based on the evaluation of its density enables us to obtain different estimations
depending on the bandwidth. The bandwidth affects the explanatory capacity of the tree stand. A
considerable part of the plant community variation explained by soil factors was spatially structured. The orthogonal eigenvector-based spatial variables (dbMEMs) approach can be extended
to quantifying the effect of forest structures on the herbaceous layer community. The measure of
the overstorey spatial structure, which is based on the evaluation of its density, was very useful in
explaining herbaceous layer community variation.