Case study sites were chosen across three European countries for investigating and demonstrating the practical applicability of bioenergy production on marginal lands. The selection was made to represent different types of marginal lands and different types of climate regimes (Fig. 1 and Table 1). At these sites different methods of bioenergy production were implemented as case studies. At each site soil profiles were analysed, soil samples were taken, and soil assessment according to the SQR framework was carried out.

Most of the Ukrainian sites and all investigated Greek sites show limitations for traditional agriculture due to natural constraints and in part due to anthropogenic degradation. The sites in the western part of Ukraine (Volyn and Lviv regions) and in the Poltava region represent the type of abandoned land which was formerly used for conventional agriculture and set aside due to different site-specific reasons. In Greece sites with naturally poor soil conditions (e.g. shallow soil depth) were selected in a mountainous region (Rhodope Mountains). The sites are currently in use for forestry or in some parts for low-intensity pasture systems.

In contrast, the Vinnitsa site in Ukraine as well as the two German sites has undergone the most severe anthropogenic disturbances. The Vinnitsa site was used as a municipal waste dump before; waste has been removed before preparing the site. In Lusatia (State of Brandenburg, eastern Germany) post-mining sites and former industrial or traffic areas represent types of marginal lands which can be frequently found particularly in central and eastern Europe.

With regard to climatic conditions (cf. Table 1) the selected field sites are within two gradients: (a) between subcontinental and continental and (b) between temperate and Mediterranean types of climate. The Ukrainian sites are characterized by continental climate conditions, whereas Lower Lusatia in eastern Germany has a subcontinentally influenced climate of the temperate zone. Northern Greece is part of the Mediterranean climate region with semiarid climatic conditions.

To assess soil quality (or conversely marginality) the Muencheberg Soil Quality Rating system (SQR; Mueller et al., 2007) was applied using the soil parameters derived from field and laboratory measurements. The SQR is designed to quantify the soil quality by a single value – theoretically ranging from 1 to 100 points – which is calculated on the basis of a set of basic and hazard indicators.

One set of eight basic indicators describes generic soil parameters (B 1: substrate, B 2: A horizon depth, B 3: topsoil structure, B 4: subsoil compaction, B 5: rooting depth, B 6: profile available water, B 7: wetness and ponding, B 8: slope and relief). Single scores are summarized and the resulting basic score has a range between 0 and 34 for arable land. Whereas 0 stands for absolutely infertile soils, 34 can be reached by best suited croplands. In a second step, 13 hazard indicators are examined including further factors influencing soil conditions and ecological functions (H 1: contamination, H 2: salinization, H 3: sodification, H 4: acidification, H 5: low total nutrient status, H 6: soil depth above hard rock, H 7: drought, H 8: flooding or extreme waterlogging, H 9: steep slope, H 10: rock at the surface, H 11: high percentage of coarse soil texture fragments, H 12: unsuitable soil thermal regime, H 13: miscellaneous hazards). For each hazard indicator the SQR guidelines provide multipliers on a ratio scale between 0.1 and 3. The lowest multiplier found (i.e. the most important hazard for the respective site) is used to calculate the final SQR score which is within a range between 0 and 100. Sites with a final score of 100 can be seen as sites with the best suitable soils for agriculture, whereas soils with SQR scores < 40 can be regarded as very poor or poor with regard to agricultural land use (Mueller et al., 2007). Such sites are classified here as “marginal”.

Soil profiles were investigated at each case study site in summer 2016. Profile descriptions were carried out according to the methodology of the German soil classification system (AG Boden, 2005). The soil horizon designations were transferred to international symbols according to Blume et al. (2016). Soil types were classified according to international standards provided by the World Reference Base for Soil Resources (WRB; IUSS Working Group WRB, 2007). Soil samples were taken from dominant soil horizons directly from the profile. From the selected horizons volumetric samples were taken by means of sampling rings for determining bulk density. In addition, mixed samples were taken from different depths at the study sites with different borers (depending on site conditions) and merged. Mixed soil samples were dried (3 days at 40 ∘C) and sieved (< 2 mm). Table 2 gives an overview of parameters assessed by means of field and laboratory methods as needed for soil assessment (SQR). According to the SQR method, hints for contaminations, particularly signs of artefact colour or odour, can be roughly tested by means of sensory analysis (Lichtfuss, 2004). Suspicious colours or odour that could indicate possible contamination with organic compounds were not detectable within any of the investigated soil profiles, so that further analysis in the laboratory was restricted to possible inorganic contamination with heavy metals. Measured values of relevant parameters are available from Table A1 (Appendix A).

For testing the suitability of SQR scores for bioenergy crops mean biomass data were collected for all case study regions. As plantations with bioenergy crops were established at all investigated case study sites later on, direct correlations between field-specific SQR scores and biomass yields could not be investigated yet. Thus, regional project partners provided data on average biomass yields from adjacent field sites with soil conditions comparable to the respective case study sites. These data for the same bioenergy crops as cultivated on the case study sites were used as an estimate of local biomass yields. The available biomass yield data were roughly separated into two groups: biomass yields from woody bioenergy crops (willow, poplar, black locust, and pine) and from grass-like species (miscanthus, switchgrass). The non-parametric Spearman rank coefficient between SQR final scores and mean biomass yields from woody bioenergy crops (n=16) was calculated and tested for significance using the R statistical package (R Commander). The small dataset for grass-like bioenergy crops (n=4) did not allow for calculating correlation coefficients.

SQR score calculations and correlation statistics were performed using MS Excel statistical functions. For deriving typical sets of soil constraints for different types of marginal lands, the investigated sites were grouped according to the relevance of their assigned SQR hazard indicators. For this purpose, the assessed SQR hazard multipliers with values between 0.1 (highest influence of this hazard indicator) and 3 (no influence of this hazard indicator) were transformed to a scale of influence (values ranging between 1.0 – hazard of highest relevance for a site – and 0 – no hazard influence detectable). These influence values for the individual hazard indicators were calculated following Eq. (1): Influence=1-x3 with x is equal to value of hazard indicator multipliers.

The resulting transformed values (as summarized in Table A2, Appendix A) were subject to a cluster analysis (Ward method, squared Euclidian distance) using R statistical package (R Commander). For each cluster, mean values of influence were calculated for all hazard indicators.

An extensive search was carried out in order to obtain geographical data for the algorithm components for European countries. The required geospatial datasets include the basic and hazard indicators for the calculation of the SQR score (Table 4), the screening criteria for the selection of bioenergy crops (Table 3), and additional information regarding protected areas and current land cover (Table 4).

Pan-European datasets of the European Soil Data Center (ESDAC) have been primarily used, whereas data from the HWSD were used for areas or parameters not covered by the ESDAC datasets, especially for Ukraine. The resolution of the original input datasets varied from 250 m to 5 km. A uniform cell size of 500 m was applied to all datasets before the analysis. The resolution was selected following the resolution of the geospatial data available for soil texture classes from ESDAC. The selection was based on the fact that soil texture is itself one of the basic indicators for the calculation of SQR (B 1) and also a parameter for the calculation of two additional basic indicators (B 5 & B 6). Thus, the application of its resolution was selected to reflect substrate variations across Europe. Resampling for discrete data (e.g. land use) was performed using the nearest resampling algorithm, whereas bilinear interpolation was applied for continuous data.

The coordinate reference system is ETRS89-LAEA Europe, EPSG:3035. Latitude of origin: 52∘ N, longitude of origin (central meridian): 10∘ E

Each raster dataset was reclassified based on the SQR field manual, the SQR assessment scheme according to BGR (2010), and adaptations made by Brandenburg University of Technology Cottbus-Senftenberg within the SEEMLA project.

The SEEMLA algorithm was the basis for the development of a GIS toolset to quantify marginal land in Europe and evaluate its potential for biomass production for bioenergy through raster analysis. The necessary geospatial datasets were accordingly processed to allow for implementation of the appropriate SQR calculations and subsequent compilations.

Raster analysis was applied, after the datasets were reclassified in compliance with the SQR classes for basic and hazard indicators. The next step was to apply elimination criteria to derive the suitability of marginal lands for biomass production. The last function of the GIS toolset cross-references the site conditions with crop species demands to determine appropriate marginal lands for cultivating selected crop species.

The outputs of the GIS toolset can be used to produce various thematic maps such as SQR or marginal-land map for Europe, most important hazard indicator, marginal-land suitability per bioenergy crop, and any desired combination of the resulted layers.

The application was developed as a toolbox for ESRI ArcGIS desktop (v.10.2.2 or newer). The GIS outputs include both raster (cell size 500×500 m) and vector datasets and corresponding maps. The standardized outputs include the following information:

mapping and quantification of marginal lands in Europe using the Muencheberg SQR system,

The SQR assessment of soil quality clearly shows that most of the selected marginal-land sites can be in fact considered as “poor” or even “very poor”. With only one exception (abandoned arable land in Poltava region, Ukraine: SQR value: 55.0) all final SQR scores are below the threshold of 40 (Fig. 4). The mean basic score is 17.1 indicating medium soil conditions in general and the mean final SQR score of 22.5 clearly reveals the low quality of soils of marginal lands. For comparison, a mean SQR score of 64 for arable land in Germany was determined by Hennings et al. (2016). In general, the SQR methodology seems be a generally applicable technique for identifying marginal lands based on soil parameters.

The post-industrial sites (former railway sites, Germany) reached the lowest basic score values as a result of a weak natural potential of the prevailing artificially created substrates with very young and undeveloped soils. Most of the anthropogenically degraded soils and all mountainous and naturally poor soils (Greece) can be classified as “very poor” with final SQR values below 20, whereas former arable sites usually exhibit considerably better soil-quality conditions with SQR scores only slightly below 40. Sites with similar SQR values are frequently found in German low mountain ranges and in the eastern German lowlands (BGR, 2013) at sites which are in use for agriculture. Thus, abandoned arable lands offer at least some minor potential for traditional agricultural land use options.

The SQR system was primarily invented for traditional agricultural sites and its scores are well correlated with agricultural crop yields (Hennings et al., 2016; Mueller et al., 2010, 2016). By definition marginal lands are not suitable for agriculture and are described as “surplus land” (Dauber et al., 2012). The basic aim of the SQR system is to evaluate soil productivity functions related to traditional agricultural land use (Mueller et al., 2007, 2010) so that the assessment of land marginality is not within the original focus of the method. For validating the reliability of the SQR method for bioenergy crops, correlations between (estimated) biomass yields and SQR final scores were tested. Table 5 gives an overview of bioenergy crops cultivated at the studied sites and ranges of the estimated biomass yields as reported by regional project partners for sites with similar degrees of marginality. The estimated local biomass yields of marginal sites used here are in the majority of cases clearly below biomass yields as reported in literature. For example, Clifton-Brown et al. (2001) found mean biomass yields for miscanthus in Europe of 18 t DM ha-1 a-1. For woody biomass from short rotation plantations in Europe, Djomo et al. (2015) published an average of 9.3 t DM  ha-1 a-1.

Figure 5 illustrates the relationship between SQR final scores and estimated biomass yields for the investigated test sites, and separately for grass-like and woody bioenergy crops. A trend of growing biomass yield with decreasing marginality (increasing SQR scores) is visible from the figure for grass-like bioenergy crops, which are cultivated at two of the Ukrainian test sites. The correlation, however, of woody bioenergy crops with final SQR scores as an indicator of site marginality turns out to be strong and statistically significant (Fig. 5). It can be preliminarily concluded, therefore, that SQR scores are suitable to represent the productivity, or conversely the marginality, of soils with regard to yield potentials of bioenergy crops. Further, it becomes visible from this analysis that poor and also very poor sites with regard to soil conditions provide at least a certain potential for biomass production. The authors are aware of the necessity of an increasing number of investigations to be able to derive transferable results and trends. The presented results are valid for perennial bioenergy crops, mainly for fast-growing tree species. Effects of soil quality on the performance of annual bioenergy crops have not been considered. In addition, further research might be needed to analyse relations between soil quality and characteristics of biomass with regard to its later use in power plants or bio-refineries.

SQR hazard indicators are site properties which are able to critically affect the total soil quality (Mueller et al., 2007). Analysing the importance of the different hazard indicators of marginal lands allows for identifying generic factors of marginality which superimpose other properties. Previous studies (BGR, 2013; Hennings et al., 2016) showed that large areas of agricultural land are generally not limited by any hazard indicator. For German agricultural lands, Hennings et al. (2016) calculated an amount of 61.5 % without any ecological constraint according to the SQR assessment. The remaining 38.5 % of arable lands in Germany show different ecological limitations and about 6 % of the arable lands can be seen as clearly marginal with final SQR scores below 40. Against this background, the relevance of individual hazard indicators for marginality of the selected case study sites presented here was further elaborated. By clustering the investigated case study sites considering their hazard indicators, four groups of marginal lands could be revealed (Fig. 6). The average influence values of the respective hazard indicators for each cluster are shown in Table 6.

Cluster 1 mainly represents the post-mining sites studied in Lower Lusatia (Germany). Soils at these sites are predominantly affected by higher acidification potentials and a low nutritional status. The same is true for one of the western Ukrainian sites, which is therefore part of this cluster.

The large cluster 2 integrates the further Ukrainian abandoned former arable lands, also including the former municipal waste dump sites in the Vinnitsa region. The main hazards related to these soils are the danger of seasonal over-wetting due to flooding or extreme waterlogging. Several of these sites are situated in depressions of the flat landscapes and/or exhibit a dense layer in the subsurface. The second relevant hazard found here was the insufficient nutritional status of the soils. Both factors can be seen as the reason for abandoning these former agricultural sites during the last two decades.

Extreme soil conditions were found at the German post-industrial site (former railway area), which was grouped into cluster 3. Main hazards here are the very high amount of coarse particles above and in the soil substrate. These particles consist of remains of railroad ballast as well as rubble from former buildings at this site. This composition can be seen as typical for brownfields. The investigated site showed no signs of contamination.

Cluster 4 contains the Greek sites, which were already characterized as marginal sites with naturally poor soil conditions. With its Mediterranean climate regime, Greece is generally prone to droughts during the summer season. Furthermore, soil substrates are in part extremely stony both at the surface and within the profile. It is assumed that the geology of the Rhodope Mountains in northern Greece with frequent ore deposits provides slightly elevated background contents of heavy metals (Cr, Ni) and arsenic (As) so that the contamination hazard indicator gained higher influence. In addition, as a result of previous erosion, the remaining soil profiles are mainly shallow with low depths above hard rock. This is consistent with conclusions by Hennings et al. (2016) who found low soil depth as an important SQR hazard indicator in mountain regions of Germany.