Three-Dimensional Mapping of Forest Soil Carbon Stocks Using SCORPAN Modelling and Relative Depth Gradients in the North-Eastern Lowlands of Germany

Abstract

To cope with the challenges in forest management that are contemporarily caused by climate change, data on current chemical and physical soil properties are more and more necessary. For this purpose, we present a further amalgam of depth functions and SCORPAN modelling to provide data at arbitrary depth layers. In this concept, regionalisation is split up into the modelling of plot totals and the estimation of vertical distributions. The intended benefits by splitting up are: consistency between estimates on plot level and depth layer level, avoidance of artificial depth gradients, straightforward interpretation of covariates in the sense of pedogenetic processes, and circumnavigation of the propagation of uncertainties associated with separation between horizons during field sampling. The methodology was tailored to the circumstances within the north-eastern lowlands and the utilisation of current inventory data of the National Forest Soil Inventory (NFSI) in Brandenburg (Germany). Using the regionalisation of soil organic carbon (SOC) as an example, the application is demonstrated and discussed in detail. The depth to groundwater table and terrain parameters related to the catchment area were the main factors in SOC storage. The use of kriging did not improve the model performance. The relative depth gradients of SOC were especially distinguished by tree species composition and stand age. We suppose that interesting fields of application may be found in scenario-based modelling of SOC and when SOC serves as a basis for hydrological modelling.

Spatial 3D mapping of forest soil carbon stocks in Hesse, Germany

“…Validation for depth increments often showed worse performance than the full model. The pattern of low performance at the top, best performance at intermediate depths, and again lower performance at lower‐profile depths was also reported by others (Nauman and Duniway, 2019; Russ et al., 2021). Ma et al.…”

“…50%; Jobbágy and Jackson, 2001), fits the proportion of NFSI observations (67%), and slightly lower than reported for terrestrial soils of Brandenburg (approx. 74%; Russ et al., 2021). Increasing SOC stocks in the subsoil is currently discussed as a mean to reduce the increase of atmospheric CO 2 concentrations (Whitmore et al., 2015), because SOC in subsoils generally has high mean residence times.…”

“…In this case, AVE and R 2 measures are well comparable (Li, 2017). Achieved R 2 values in spatial modeling of OC contents or stocks were reported to range from 0.1 to 0.7, but mostly between 0.3 and 0.5 (de Brogniez et al., 2015; Gomes et al., 2019; Ma et al., 2021; Mulder et al., 2016; Nussbaum et al., 2014; Russ et al., 2021). Large‐scale studies often had lower R 2 values around 0.2 to 0.3, whereas small‐scale studies often reach R 2 values around 0.5 (Chakan et al., 2017; Nussbaum et al., 2014; Russ et al., 2021).…”

“…However, especially sampling of forest soils is challenging. Division of the forest floor and the top of the mineral soil may introduce sampling errors (Jansen et al., 2005; Russ et al., 2021). Moreover, effects of former, often unknown, management are strongest at the topmost parts of soil profiles (Poeplau et al., 2020; Springob et al., 2001).…”

“…Russ et al. (2021) reported a proportion of 30% of FFC for Brandenburg. The federal‐state Brandenburg is characterized by high proportions of sandy, acidic soils and coniferous forest stands (Scots pine), which explains this finding.…”