Spatial variability of forest floor and topsoil thicknesses and their relation to topography and forest stand characteristics in managed forests of Norway spruce and European beech

Abstract

Soils play a significant role in climate regulation, especially due to soil organic carbon (SOC). The SOC pool is therefore modeled for various environments, and forest floor and topsoil thicknesses are important parameters for most of these models as they store most of the SOC. However, the forest floor and topsoil thicknesses show high spatial variability which is a result of multiple factors which are not agreed upon among scientists. Out of these factors, we choose topography parameters (elevation, slope, and topography wetness index) and forest stand characteristics (stand age, dominant tree species, and forest floor cover), and soil moisture, and we analyzed their relationship to the forest floor and topsoil thicknesses. The study was performed in a managed submontaneous forest in Central Europe dominated by Picea abies (L.) Karsten with small patches of Fagus sylvatica L. or other species. The thicknesses of the O horizons (Oi, Oe, Oa) and topsoil were measured at 221 sampling pits. Geographically weighted regression showed that the spatial variability of the overall forest floor plus topsoil thickness (OA) is responsible for 8% of its variability. The thickness of the OA is the most strongly controlled by forest floor cover explaining approximately 6% of its variability and soil moisture explaining 2–6% of the variability. The Oi + Oe horizon thickness is controlled only by forest floor cover explaining 10.7% of its variability, and the thickness of Oa + A horizon can be explained mainly by soil moisture in mineral horizon explaining 9% of the variability.

Annex

Annex 1: Location of the study catchment and the position of the sampling points

Annex 2: Soil type distribution in LIZ catchment

Map of soil types based on 30 randomly distributed sampling cores down to 70 cm and refined by 300 sampling pits down to 25 cm.

Annex 3: Soil density and particle size distribution

Soil type	Horizon complex	Number of samples	Bulk density (g/cm3)	Particle density (g/cm3)	Sand (%)	Silt (%)	Clay (%)	Textural class
	Oi + Oe	3	0.12	–	–	–	–
CMha	Oa + A	3	0.63	2.48	57	37	5	Sandy loam
	Mineral	3	1.22	2.51	53	40	8	Sandy loam
	Oi + Oe	3	0.14	–	–	–	–
CMdy	Oa + A	3	0.46	2.03	57	36	7	Sandy loam
	Mineral	3	1.17	2.45	56	37	7	Sandy loam
	Oi + Oe	3	0.09	–	–	–	–
PZet	Oa + A	3	0.62	2.17	56	38	9	Sandy loam
	Mineral	3	1.10	2.49	64	30	6	Sandy loam
	Oi + Oe	3	0.10	–	–	–	–
CMgl	Oa + A	3	0.43	2.3	56	40	3	Sandy loam
	Mineral	3	1.81	2.5	53	40	8	Sandy loam
	Oi + Oe	2	0.14	–	–	–	–
ST	Oa + A	2	0.11	1.52	49	33	18	Loam
	Mineral	2	1.02	2.5	59	33	8	Sandy loam
	Oi + Oe	2	0.04	–	–	–	–
GL	Oa + A	2	0.50	2.45	53	38	9	Sandy loam
	Mineral	2	2.13	2.61	70	22	8	Sandy loam

1. CMha—Haplic Cambisol, CMdy—Dystric Cambisol, CMgl—Gleyic Cambisol, PZet—Entic Podzol, GL—Gleysol, ST—Stagnosol

Annex 4: Soil edaphic categories in LIZ catchment

Retrieved from maps produced by ÚHUL at scale 1:10,000 and based on the forest site classification described by Viewegh (2003).

Annex 5: Stand age in LIZ catchment

Retrieved from maps produced by ÚHUL at scale 1:10,000.

Annex 6 Geographically weighted regression computation details

In GWR 4.09, the fixed distance of Gaussian Kernel was left to find out by Golden selection search with the criterion of minimum AICc (AIC with a correction for small sample sizes). The categories of the categorical explanatory variables (forest floor cover, dominant tree species and forest stand age) were recoded as dummy variables. From the dummy variables representing the same categorical explanatory variable one was always omitted from the analysis due to multicollinearity (see Fotheringham et al. 2002). However, all of the dummy variables representing a particular categorical explanatory variable were analyzed and interpreted as one phenomenon.

The data were supposed spatially variable. On the other hand, no justification was found for the spatial variability of explanatory variables. For instance, slope or dominant tree species should influence the forest floor and topsoil thicknesses in the same way over the whole area. Furthermore, this assumption was verified by a geographical variability test embodied in GWR 4.09. No significant geographical variability was found for most explanatory variables with an exception of the forest floor cover. However, the data of explanatory variables was assumed to be geographically invariable as a whole. The explanatory variables were then classified to those with local influence (Intercept) and those with global influence (all the others) (see Fotheringham et al. 2002).

Local variables are computed locally from neighbor values within the bandwidth and weighted by their distance. As a result, there is not an exact estimation of their value in the model, the estimation could be slightly different at every location (Fotheringham et al. 2002). Locally computed intercept works similarly as an interpolation in this case, and substitutes a role of an autoregressive parameter which deals with the spatial autocorrelation. The spatial independency of the model residuum was verified by Moran’s I. The advantage of using the local intercept instead of the model without intercept but employing an autoregressive parameter is at first that it can be computed by Ordinary least squares method compared to the latter model which should be performed by Maximum likelihood computation and the second advantage is the interpretation which is more intuitive for the models with the intercept. The use of the local intercept and the autoregressive parameter at the same model is not possible due to their multicollinearity and a model combining a global intercept with an autoregressive parameter was less flexible.

The two above mentioned models which were run separately for Oi + Oe, Oa + A, and OA horizons can be written as follows:

$$\begin{aligned} {\text{FLTH}}_{i} & = \alpha_{0} \left( {u_{i} ,v_{i} } \right) + \alpha_{1} {\text{WETIN}} + \alpha_{2} {\text{COV}} + \alpha_{3} {\text{AGE}} + \alpha_{4} {\text{TREE}} + \alpha_{5} {\text{ELEV}} + \alpha_{6} {\text{MOIST}} + \, \varepsilon_{i} \\ {\text{FLTH}}_{i} & = \, \alpha_{0} \left( {u_{i} ,v_{i} } \right) + \alpha_{1} {\text{SLOPE}} + \alpha_{2} {\text{CURV}} + \alpha_{3} {\text{TREE}} + \, \alpha_{5} {\text{MOIST}} + \alpha_{6} {\text{AGE}} + \alpha_{7} {\text{ELEV}} + \varepsilon_{i} \\ \end{aligned}$$

where FLTH_i = the thickness of a modeled horizon complex at location i, COV = forest floor cover, TREE = dominant tree species; AGE = forest stand age, ELEV = elevation, WETIN = wetness index, MOIST = moisture in the mineral soil measured in the field, SLOPE = slope, CURV = relief curvature, α₁–α₇ are model estimates, α₀ is intercept estimate computed locally.

Annex 7: OA thickness variability

Annex 8: Variability of the thickness of Oi + Oe horizons' complex

Annex 9: Variability of the thickness of Oa + A horizon's complex

Annex 10: OA thickness variability at stands of oligotrophic edaphic category

Annex 11: The variability of the thickness of Oi + Oe horizon's complex at stands of oligotrophic edaphic category

Annex 12: The variability of the thickness of Oa + A horizon's complex at stands of oligotrophic edaphic category

Annex 13: OA thickness variability at stands of hydric edaphic category

Annex 14: The variability of the thickness of Oa + A horizon's complex at stands of hydric edaphic category

Annex 7–14: Oi—organic litter horizon, Oe—fragmented organic horizon, Oa—humified organic horizon, A—topsoil, OA—forest floor plus topsoil in total