Energy Crops in Regional Biogas Systems: An Integrative Spatial LCA to Assess the Influence of Crop Mix and Location on Cultivation GHG Emissions

Anaerobic digestion producing biogas is an important decentralized renewable energy technology used to mitigate climate change. It is dependent on local and regional feedstocks, which determine its sustainability. This has led to discussions on how to alter feedstock for biogas plants without compromising their GHG (Greenhouse gas) saving, one particular issue being the use of Maize silage (MS) as the dominant feedstock. To support this discussion, this paper presents an integrated life cycle assessment of energy crop cultivation for 425 biogas catchments in the region of Central Germany (CG). The simulations for the CG region showed that MS as an effective crop to mitigate GHG emissions per kilowatt hour (GHGculti) was context dependent. In some cases, GHGculti reductions were supported due to higher yields, and in other cases, this led to increased GHGculti. We show that the often-proposed strategy of substituting one crop for another needs to be adapted for strategies which take into account the crop mixtures fed into biogas plants and how they perform altogether, under the specific regional and locational conditions. Only in this way can the trade-offs for lower GHGculti be identified and managed.


Introduction
Anaerobic digestion producing biogas is one of the most important decentralized renewable energy technologies, which is heavily dependent on local and regional feedstocks. It is undergoing rapid expansion globally in order to combat climate change and to ensure a secure energy supply [1]. There are a variety of ways in which biogas can be produced, and this varies extensively between countries and even within countries. In Germany, for example, distinct regional differences can be observed for biogas production [2]. While financial incentives under the German Renewable Energy Sources Act have had a strong role in determining what feedstock or combinations of feedstocks are potentially used, German biogas plants are generally planned and built according to available feedstocks [3]. Approximately 92% of substrates used in biogas plants originate from agricultural sources [2], such as livestock manures (solid and liquid) and energy crops, which have a higher energy content [4][5][6].
The most prominent energy crop used is maize silage, contributing approx. 70% to the total mass of energy crops used in biogas plants in Germany. The next prevalent energy crops are grass and cereal (whole crop) silage [2]. Maize silage is the preferred crop due to its technological suitability. Additionally, its high dry matter and energy yields [6] promote it as the most effective crop to mitigate During the temporal scope of this study, there was an estimated 570 biogas plants in operation in the CG region. This includes plants using municipal organic wastes (e.g., domestic food waste), industrial wastes (e.g., distillery wastes), and agricultural plants (using energy crops and slurry) [2,30]. The main energy crops (EC) being used by the agricultural plants included; maize silage (MS), grass silage (GS), and small amounts of cereal grains (Cer)-Rye, Barley, and Triticale.

Overview of Life Cycle Approach-RELCA
The RELCA approach of O'Keeffe et al. [25] was applied (see Supplementary Material S1.1 for overview of modelling steps). The regional foreground was set to the eastern German region of Central Germany (CG), which consists of three federal states, or "Bundesländer" (Figure 1); Saxony, Saxony-Anhalt, and Thüringen. O'Keeffe et al. [10] have already used RELCA to assess the full regional life cycle for 425 agriculturally based biogas plants operating in CG. However, in this paper we focus only on the cultivation of energy crops being used within regional biogas catchments (no system expansion through credits). The scope for individual biogas catchments (n = 425) is from fieldto-field gate. However, the conversion step is indirectly taken into account through the assignment of crop catchment areas to biogas plants (see Section 2.6), as well as the use of the digestate from the associated biogas plants (i.e., part of fertilizing regime).

Functional Unit Combined with Spatial Indicators
The total GHG emission for cultivation (GHGculti) for each biogas plant is the product of all constituent grid cells within the biogas catchment (Bcat) calculated as kilogram equivalents of CO2 per (net) kilowatt hour electricity produced (kgCO2eq/kWhel) and for the analysis presented here, is the overall functional unit. However, we want to investigate the GHG performance of Bcats based on the shares of energy crops. To do this, we use two key spatial indicators in combination with GHGculti to compare across the Bcats and to help understand important underlying geographical factors.
The first is the emission intensity (EI) of the Bcats (kgCO2eq/ha). It differs from the CultiEmis (kgCO2eq/ha) estimated for the individual energy crops (Section 3.1), in that it is the medianed CultiEmis of the crop mixtures associated with a specific Bcat and is unique to the Bcat. If, for example, a Bcat has a high median EI, this means that based on the location, many of the constituent grid cells of the Bcat were associated with a high CultiEmis.

Functional Unit Combined with Spatial Indicators
The total GHG emission for cultivation (GHG culti ) for each biogas plant is the product of all constituent grid cells within the biogas catchment (Bcat) calculated as kilogram equivalents of CO 2 per (net) kilowatt hour electricity produced (kgCO 2eq /kWh el ) and for the analysis presented here, is the overall functional unit. However, we want to investigate the GHG performance of Bcats based on the shares of energy crops. To do this, we use two key spatial indicators in combination with GHG culti to compare across the Bcats and to help understand important underlying geographical factors.
The first is the emission intensity (EI) of the Bcats (kgCO 2eq /ha). It differs from the Culti Emis (kgCO 2eq /ha) estimated for the individual energy crops (Section 3.1), in that it is the medianed Culti Emis of the crop mixtures associated with a specific Bcat and is unique to the Bcat. If, for example, a Bcat has a high median EI, this means that based on the location, many of the constituent grid cells of the Bcat were associated with a high Culti Emis . The second indicator is the size of the biogas catchment area or the land area demand (LAD Bp ) per energetic output of the particular bioenergy plant (ha/MW el ). It is a form of direct land use footprint (Equation (1)). This spatial indicator is important, as the total GHG culti for each Bcat is the product of all constituent grid cells within the Bcat, i.e., total catchment area (LUTot Bp ) calculated as equivalents of CO 2 per (net) kilowatt hour electricity (NE BP ) produced (kgCO 2eq /kWhel). Therefore, the more land area being used (greater number of grid cells) then a greater cumulative sum of emissions is more likely, depending on the location.
Both EI and LAD of a Bcat are related to the yields of the associated catchment energy crops and installed capacity of biogas plant, but not always linearly, due to different locational factors such as climate, soils, and crop combinations (see S1.2 in Supplementary). The combination of these two spatial indicators, which are land and location based, will provide us with a better understanding of the mitigation potential of the biogas systems as a function of energetic output (kgCO 2eq /kWhel) because their product also equals the total GHG culti found for a particular biogas catchment (Equation (2)).

RELCA Step 1: Crop Allocation Modelling (CRAM)
The crop allocation modelling or CRAM approach of Wochele et al. [31] was implemented to determine the potential regional distribution of the different crop types; maize silage (MS), grass silage (GS), and cereal grains (Cer), which were used in regional biogas production for 2010/2011 (for more details, see S1.3 in Supplementary).

RELCA Step 2: Biogas Conversion Systems
In collaboration with the Deutsches Biomasseforschungszentrum (DBFZ), representative model agricultural biogas plants for the CG region were developed [10]. These model biogas plants were based on two characteristics relevant for the German Renewable Energy Sources Act (EEG): (1) Size classes (i.e., installed capacity kW), and (2) dominant feedstock (F i ), i.e., either animal manures or maize silage. Nine model biogas plants or clusters were identified for the CG region, and these were used to generate the LCIs (life cycle inventories) for 425 agricultural biogas plants (see S1.4 in Supplementary). For ease of discussion, the nine clusters described in O'Keeffe et al. [10] will be aggregated further into two classes, biogas catchments belonging to slurry dominant clusters (SL dom ), and energy crop dominant (EC dom ) clusters (Table 1). The SL dom clusters used relatively little energy crops (<20% EC of mass input), whereas the EC dom clusters used significantly higher energy crops (>60% EC) (see [10] and S1.4 in Supplementary for more detail).

RELCA Step 3: Biogas Catchment Modelling
Once the feedstock demand for each biogas plant was estimated, catchments were modelled using MATLAB generated scripts and the approach outlined in O'Keeffe et al. [25]. The demand (F i,p ) for each cultivated feedstock type (i.e., MS, Cer, GS, see S1.4 in Supplementary) was used to generate the catchments, which grew in size until the demands of all feedstock types were satisfied for each of the biogas plants (n = 425) modelled, in one simulation run (Equation (3)). However, if a grid cell from an allocated crop was closer to one biogas plant over another, that grid cell was allocated to the closest biogas plant, in order to avoid catchment area overlap [32]. The associated land area for each feedstock (LA Fi ) was then summed to calculate the total catchment area for each biogas plant (LUTot BP ), see S1.5 in Supplementary for overview of land demand modelled.

RELCA Step 4: Life Cycle Inventory for Biomass Management
Once the catchments had been delineated (Figure 1), the cultivation management and the associated cultivation emissions were calculated per hectare and per kilowatt electricity produced for each constituent grid cell of a Bcat using MATLAB scripts (see [25]). All flows relating to biomass cultivation were included for one harvest only, until just after harvesting for the biogas plant.
The operational year for biogas production was assumed to be from the point of harvest in autumn 2010, through to the autumn 2011, as this was the first-time frame for which the most complete data (land use and biogas technologies) was available.

Nitrogen (N) Management and Spreading of Digestate
The recommended nitrogen application and digestate rates were calculated assuming best farming practices and were based on yields and different geographical considerations (e.g., climate, soils, and agricultural suitable index assigned to a grid cell) for energy crops and grasslands (see S1.6 in Supplementary for details).

Other Farm Management-Auxiliary Inputs
The fertilizer application rates for phosphors (P) and potassium (K) were determined based on yields and off take. The inventory data for crop protection products and recommended dosages were also determined for the region (see S1.6 in Supplementary for details).

Field Machinery Operations
The fuel consumption relating to field operations were estimated using the online KTBL tool [33] and are outlined in [10].  national guidelines [34] and using the emission factors (EF) derived by Brocks, et al. [35] (see S1.7 in Supplementary for details).
2.8.3. Non-Regional-Indirect Cultivation Emissions (AuxillaryEmis) All non-regional, upstream emissions were attributed to each constituent grid cell of a Bcat and are outlined in [10].

Statistical and Spatial Analysis
Emission profiles were generated for each crop (MS, Cer, GS) and for each biogas cluster (see S1.4 in Supplementary for explanation of clusters) using MATLAB 2017b Statistics and Machine Learning Toolbox™ (MATLAB). The emission profiles for the individual energy crops ( Figure 2) refers to the distribution of Culti Emis (kgCO 2eq /ha) determined for each energy crop grid cell, estimated across all biogas catchments. The emission profiles of Figures 3 and 4 refer to the distribution of emissions determined for each of the Bcats analyzed within a cluster. Significant differences between different regional clusters were tested in R studio [36], using a Kruskal Wallis test (p < 0.05) and the post hoc test, Wilcoxon pairwise comparison, with a Holm correction, as the data was found to be nonparametric.

Contribution Analysis
For ease of discussion, we focus on nitrogen related cultivation emissions only, as O'Keeffe et al. [10] found these accounted for 46-59% of the total GHG balance (see S1.8 in Supplementary for additional contribution analysis).
Soil N2O contributes significantly to the GHG balances of different energy crops, and is dependent on the amount of nitrogen fertilizer applied and locational factors such as soil and climate [13,24,40]. For CG, soil N2O contributed between 58-88% of the CultiEmis (kgCO2eq/ha), depending on All crops were found to be significant from every other denoted by the different letters (at the 5% level). Extreme outliers were removed. Sustainability 2020, 12, 237 7 of 17 spatial indicators, EI (kgCO2eq/ha) and LAD (ha/MWel) to help understand important underlying geographical factors and to compare across the clusters. SLdom clusters had a large share of feedstock contributed from animal manures, whereas, for ECdom clusters a large share came from energy crops (Table 1). Therefore, for the purpose of clear and fair discussion, we will not make direct comparisons between the Bcats associated with ECdom and SLdom clusters. To support the visual representation, grid cells were aggregated using the ESRI ArcGIS polygon aggregate tool using a 2 km radius.

SLdom Biogas Catchments
For Bcats associated with SLdom clusters, the cultivation of energy crops contributed between 52-56% of the total emissions [10]. The GHGculti ranged from 0.086 to 0.137 kgCO2eq/kWhel. Based on MS shares in the crop mix, we would expect GHGCulti (high to low median) to be as follows: CL8 > CL4 > To support the visual representation, grid cells were aggregated using the ESRI ArcGIS polygon aggregate tool using a 2 km radius.
Sustainability 2020, 12, x FOR PEER REVIEW 9 of 17 CL7 > CL2 > CL1. However, we get the following order of: CL4 > CL8 > CL7 > CL1 > CL2, with Bcats associated with CL2 having the significantly lower GHGculti ( Figure 4A). Although the EIs of biogas catchments in CL2 are significantly higher than CL1 ( Figure 4C), the effect of significantly lower LADs ( Figure 4D) counteracts this, resulting in the overall significantly lower GHGculti for CL2 ( Figure 4A). The higher LADs of CL1 relates to the share of Cer in the Bcats feedstock mix. For CL1, Cer contributed approx. 60% of the total land area demand of the Bcats, unlike those of CL2, for which Cer contributed on average only 36%. What this means is that by requiring a greater area to supply the feedstock, the catchments of CL1 had a greater summed volume of GHG emissions per kilowatt hour ( Figure 4B), and leading to the increased GHGculti, emissions and it's poorer than expected performance. For CL4, although with significantly lower EI values, the slightly higher LAD values (not significant) than that of CL8 resulted in the overall median GHGculti of CL4 to be higher (see S1.10 in Supplementary for further discussion).

ECdom Biogas Catchments
For Bcats associated with ECdom clusters, the cultivation of energy crops was found to contribute between 64-67% of the total emissions [10]. Again, based on the share of MS in the feedstock mix of Using ESRI ArcGIS ® [37], a hotspot analysis (Getis Ord Gi) was carried out with a spatially weighted matrix (K nearest neighbor, n = 8), to identify the regional distribution of statistically higher ("hotspot") or lower ("coldspot") Culti Emis associated with the regional energy crop cultivation for biogas (kg CO 2eq /ha). A further hotspot analysis was done to identify the regional distribution of statistically higher or lower crop yields.
Furthermore, for SL dom and EC dom , those Bcats with the significantly lowest (less than the 10th percentile) and significantly highest GHG culti (kgCO 2eq /kWhel) (greater than 90th percentile) were determined. Summary statistics were generated using MATLAB 2017 [38] to profile their underlying geographical parameters. This was done to understand potential locational trade-offs between cultivation emissions, land use, and energy output. All boxplots, bar charts, were produced using MATLAB 2017 [38]. All maps were produced using ESRI ArcGIS ® [37].

Crop GHG Emission Profiles
The RELCA simulated median Culti Emis for MS was found to be significantly higher (median = 2708 kgCO 2eq /ha) than the median Culti Emis found for Cer and GS, of 1662 kgCO 2eq /ha and 1573 kg CO 2eq /ha, respectively ( Figure 2). These results are comparable to the literature, for which higher GHG emissions associated with MS cultivation for biogas has already been shown, in comparison with grass silage [4], sugar beet [8], and cereals [39]. Although the ranges and aspects included in such studies varies widely.

Contribution Analysis
For ease of discussion, we focus on nitrogen related cultivation emissions only, as O'Keeffe et al. [10] found these accounted for 46-59% of the total GHG balance (see S1.8 in Supplementary for additional contribution analysis).
Soil N 2 O contributes significantly to the GHG balances of different energy crops, and is dependent on the amount of nitrogen fertilizer applied and locational factors such as soil and climate [13,24,40]. For CG, soil N 2 O contributed between 58-88% of the Culti Emis (kgCO 2eq /ha), depending on the crop. For the RELCA simulations, MS had the significantly highest range of soil N 2 O emissions, from 1.83-8.87 kg N 2 O_N/ha (median = 3.2 kg N 2 O_N/ha). Cer had a significantly lower N 2 O_N emission range, from 0.74-4.65 kg N 2 O_N ha −1 (median, 1.56 kg N 2 O_N/ha). GS was found to have the significantly lowest N 2 O_N emissions, ranging from 0.98-5.27 kg N 2 O_N/ha, with a median of 1.57 kg N 2 O_N/ha. The most up to date Tier 2 level approach was used, and these simulated values fall within the ranges found within the literature, and are therefore considered to be plausible (see S1.9 in Supplementary).
In comparison to Soil N 2 O, (chemical) fertilizer N contributed relatively little to the Culti Emis (4-12%) and was zero for GS and Cer, where the N demand had been met by the digestate applied. For MS, due to its higher N demand, only 26-52% of the N demand could be met with digestate. This meant that MS had the greater chemical N fertilizer demand. Therefore, with both the higher direct N emissions (Soil N 2 O) and higher indirect emissions (production of chemical N fertiliser) MS was found to have the highest Culti Emis (kgCO 2eq /ha).

Spatial Distribution of Culti Emis
In the CG region, there was a tendency for significantly higher Culti Emis in the central and western areas (hotspots) (Figure 3). This is where yields and hence N fertilizer application rates were relatively higher for all crops. In locations to the north, east, and south, a tendency for significantly lower Culti Emis was observed (cold spots). These were areas with significantly lower yields, and thus lower Sustainability 2020, 12, 237 9 of 17 N fertilizer application rates. The spatial trend of high and low Culti Emis corresponded with those found for the N 2 O emissions (see S1.7 in Supplementary). However, while in the north and in parts of the south had in general significantly lower N 2 O emissions, pockets of significantly higher N 2 O emissions were also found ( Figure 3). This is due to these areas having the geographical and climate conditions which resulted in higher N 2 O emission factors (EF) (see S1.7 in Supplementary). The spatial trends for all crops, MS, Cer, and GS had similar trends to that shown in Figure 3. 3.1.4. Hypothesis for GHG culti (kgCO 2eq /kWh el ) Based on Crops' Culti Emis (kgCO 2eq /ha) From the contribution and spatial analysis of the different crops' Culti Emis (kgCO 2eq /ha), we would now hypothesize that: (1) Those Bcats with a higher share of MS should have a tendency for higher GHG culti (kgCO 2eq /kWhel); and (2) Bcats found in the central west areas of the region should have higher GHG culti (kgCO 2eq /kWhel) than those operating in the north, east, and southern parts of the region. We test these two hypotheses in the next sections below.

Effect of Crop Mixtures on Biogas Catchments GHG culti
In the previous Section, we have investigated the GHG emissions associated with the cultivation of the different energy crops medianed across all Bcats for the CG region. In this Section, we want to investigate the GHG performance of Bcats as a function of net energetic output (i.e., kgCO 2eq /kWhel) and how crop shares, particularly MS, within the associated catchments can influence their GHG performance. To do this, we first aggregate the Bcats into Slurry dominant (SL dom ) and Energy Crop dominant (EC dom ) clusters (see S1.4 in Supplementary for descriptions). We then use the two key spatial indicators, EI (kgCO 2eq /ha) and LAD (ha/MW el ) to help understand important underlying geographical factors and to compare across the clusters. SL dom clusters had a large share of feedstock contributed from animal manures, whereas, for EC dom clusters a large share came from energy crops (Table 1). Therefore, for the purpose of clear and fair discussion, we will not make direct comparisons between the Bcats associated with EC dom and SL dom clusters.

SL dom Biogas Catchments
For Bcats associated with SL dom clusters, the cultivation of energy crops contributed between 52-56% of the total emissions [10]. The GHG culti ranged from 0.086 to 0.137 kgCO 2eq /kWhel. Based on MS shares in the crop mix, we would expect GHG Culti (high to low median) to be as follows: CL8 > CL4 > CL7 > CL2 > CL1. However, we get the following order of: CL4 > CL8 > CL7 > CL1 > CL2, with Bcats associated with CL2 having the significantly lower GHG culti ( Figure 4A).
Although the EIs of biogas catchments in CL2 are significantly higher than CL1 ( Figure 4C), the effect of significantly lower LADs ( Figure 4D) counteracts this, resulting in the overall significantly lower GHG culti for CL2 ( Figure 4A). The higher LADs of CL1 relates to the share of Cer in the Bcats feedstock mix. For CL1, Cer contributed approx. 60% of the total land area demand of the Bcats, unlike those of CL2, for which Cer contributed on average only 36%. What this means is that by requiring a greater area to supply the feedstock, the catchments of CL1 had a greater summed volume of GHG emissions per kilowatt hour ( Figure 4B), and leading to the increased GHG culti, emissions and it's poorer than expected performance. For CL4, although with significantly lower EI values, the slightly higher LAD values (not significant) than that of CL8 resulted in the overall median GHG culti of CL4 to be higher (see S1.10 in Supplementary for further discussion).

EC dom Biogas Catchments
For Bcats associated with EC dom clusters, the cultivation of energy crops was found to contribute between 64-67% of the total emissions [10]. Again, based on the share of MS in the feedstock mix of these Bcats, we would expect the GHG culti to rank, from high to low, in the following order: CL6 = CL9 > CL3 > CL5. However, they have the opposite rank of: CL5 > CL3 > CL6 > CL9. Therefore, to pick one clear comparison, why is CL5 performing so poorly and CL9 so well?
The median EI of the Bcats of CL5 ( Figure 5C) was the highest (non-significant), but these catchments also had the lowest median LADs ( Figure 5D), but were not low enough to counteract the higher EIs ( Figure 5C), thus resulting in the poorer than expected performance of CL5. In the case of CL9, the associated biogas catchments had relatively lower EIs ( Figure 5C) and although the LADs were slightly higher than for CL5 ( Figure 5D), the lower EIs counteracted this and led to the lower GHG culti of CL9 (see S1.10 in Supplementary). Sustainability 2020, 12, x FOR PEER REVIEW 10 of 17 > CL3 > CL5. However, they have the opposite rank of: CL5 > CL3 > CL6 > CL9. Therefore, to pick one clear comparison, why is CL5 performing so poorly and CL9 so well?
The median EI of the Bcats of CL5 ( Figure 5C) was the highest (non-significant), but these catchments also had the lowest median LADs ( Figure 5D), but were not low enough to counteract the higher EIs ( Figure 5C), thus resulting in the poorer than expected performance of CL5. In the case of CL9, the associated biogas catchments had relatively lower EIs ( Figure 5C) and although the LADs were slightly higher than for CL5 ( Figure 5D), the lower EIs counteracted this and led to the lower GHGculti of CL9 (see S1.10 in Supplementary).  Figure 5B, MS emission profiles are to be read from the axis on the right-hand side (red), with Cer and GS to be read from the axis to the left (blue). Clusters were not found to be significantly different (at the 5% level).

The Effect of Crop Mixtures and the Role of MS on GHGculti Performance
From the analysis of biogas catchments (Bcats) associated with SLdom and ECdom clusters, we see that the overall GHGculti of these Bcats is not entirely dependent on the emission intensity of the dominant energy crop, MS, but is related to the share of the crops in the feedstock mix. This is particularly apparent with the use of crops like cereals, as their relatively low yields result in greater catchment sizes (i.e., bigger LAD values). Additionally, further underlying locational factors play a role in the GHGculti performance, and we analyze this next.

Location Effects on Biogas Catchments
The locational hypothesis based on CultiEmis (kgCO2eq/ha) of crops in Section 3.1.3, is in general the opposite of what we see for GHGculti. From the CultiEmis analysis, we did gain some insights into potential EIs for the regional Bcats, with catchments found in areas denoted blue having low EIs and catchments found in the red areas having higher EIs ( Figure 6). However, as seen in Section 3.2, the overall GHGculti of a Bcat not only depends on its EI, but also on the volume of cumulative emissions, i.e., the land areal demand (LAD) of the catchment. Location plays an important role in the balance between these two indicators and hence on the GHGculti.

North, East, and Southern Border Locations (Poorest Performing Bcats)
In the north, east, and southern border locations of the region, RELCA simulations showed Bcats (SLdom and ECdom) to be operating with significantly higher GHGculti (greater than 90th percentile). The  Figure 5B, MS emission profiles are to be read from the axis on the right-hand side (red), with Cer and GS to be read from the axis to the left (blue). Clusters were not found to be significantly different (at the 5% level).

The Effect of Crop Mixtures and the Role of MS on GHG culti Performance
From the analysis of biogas catchments (Bcats) associated with SL dom and EC dom clusters, we see that the overall GHG culti of these Bcats is not entirely dependent on the emission intensity of the dominant energy crop, MS, but is related to the share of the crops in the feedstock mix. This is particularly apparent with the use of crops like cereals, as their relatively low yields result in greater catchment sizes (i.e., bigger LAD values). Additionally, further underlying locational factors play a role in the GHG culti performance, and we analyze this next.

Location Effects on Biogas Catchments
The locational hypothesis based on Culti Emis (kgCO 2eq /ha) of crops in Section 3.1.3, is in general the opposite of what we see for GHG culti . From the Culti Emis analysis, we did gain some insights into potential EIs for the regional Bcats, with catchments found in areas denoted blue having low EIs and catchments found in the red areas having higher EIs ( Figure 6). However, as seen in Section 3.2, the overall GHG culti of a Bcat not only depends on its EI, but also on the volume of cumulative emissions, i.e., the land areal demand (LAD) of the catchment. Location plays an important role in the balance between these two indicators and hence on the GHG culti .

North, East, and Southern Border Locations (Poorest Performing Bcats)
In the north, east, and southern border locations of the region, RELCA simulations showed Bcats (SL dom and EC dom ) to be operating with significantly higher GHG culti (greater than 90th percentile). The Bcats with the significantly highest GHG culti are found in both blue and red demarcated areas with differing locational factors for their significantly higher GHG culti ( Figure 6).
All biogas catchments in the blue locations (N, S, E) were found to have overall lower yields (MS 28% lower than the median) (see S1.11 in Supplementary for locational profiles) and lower N fertilizer demand. This combined with lower N 2 O EFs resulted in Bcats in these locations to have lower EIs (kgCO 2eq /ha). However, the lower yields, led to relatively larger LADs (see S1.11 in Supplementary). This resulted in these Bcats to have significantly high GHG Culti , the opposite GHG performance expected when basing it on the feedstock alone, as in Section 3.1.3.
Bcats found in the red locations (north and south) had differing locational factors leading to their significantly higher GHG Culti (Figure 6). In the north (mostly EC dom catchments), the MS yields were found to be lower than the median yields (<17%) and with higher N 2 O EFs; these Bcats had higher associated EIs. Therefore, the combination of lower yields resulted in larger LAD values (i.e., greater catchment areas, 6% less than the median) and higher EIs, leading to significantly higher GHG culti emissions than the median. Bcats in the southern border locations had the highest N 2 O EFs, but with relatively low LADs, due to relatively good MS and GS yields (i.e., close to the median). However, for these Bcats, the EIs were so high they could not be buffered by low LADs, and this resulted in the significantly higher GHG culti (see S1.11 in Supplementary). Bcats with the significantly highest GHGculti are found in both blue and red demarcated areas with differing locational factors for their significantly higher GHGculti ( Figure 6). All biogas catchments in the blue locations (N, S, E) were found to have overall lower yields (MS 28% lower than the median) (see S1.11 in Supplementary for locational profiles) and lower N fertilizer demand. This combined with lower N2O EFs resulted in Bcats in these locations to have lower EIs (kgCO2eq/ha). However, the lower yields, led to relatively larger LADs (see S1.11 in Supplementary). This resulted in these Bcats to have significantly high GHGCulti, the opposite GHG performance expected when basing it on the feedstock alone, as in Section 3.1.3.
Bcats found in the red locations (north and south) had differing locational factors leading to their significantly higher GHGCulti (Figure 6). In the north (mostly ECdom catchments), the MS yields were found to be lower than the median yields (<17%) and with higher N2O EFs; these Bcats had higher associated EIs. Therefore, the combination of lower yields resulted in larger LAD values (i.e., greater catchment areas, 6% less than the median) and higher EIs, leading to significantly higher GHGculti emissions than the median. Bcats in the southern border locations had the highest N2O EFs, but with relatively low LADs, due to relatively good MS and GS yields (i.e., close to the median). However, for these Bcats, the EIs were so high they could not be buffered by low LADs, and this resulted in the significantly higher GHGculti (see S1.11 in Supplementary).

Central and South-Central Locations (Best Performing Bcats)
RELCA simulations showed Bcats to be operating with significantly lower GHGculti (less than 10th percentile), mostly in the central and south-central areas of the region. The opposite of what was expected. The central area was predominantly demarcated red for higher CultiEmis, the southcentral area identified as blue, significantly lower CultiEmis. For all Bcats with significantly low GHGculti emission, regardless of the location type (i.e., red or blue), they had relatively higher yields for MS and Cer (see S1.11 in Supplementary).

Central and South-Central Locations (Best Performing Bcats)
RELCA simulations showed Bcats to be operating with significantly lower GHG culti (less than 10th percentile), mostly in the central and south-central areas of the region. The opposite of what was expected. The central area was predominantly demarcated red for higher Culti Emis , the southcentral area identified as blue, significantly lower Culti Emis . For all Bcats with significantly low GHG culti emission, regardless of the location type (i.e., red or blue), they had relatively higher yields for MS and Cer (see S1.11 in Supplementary).
For Bcats in the south-central part of the region (blue), although associated with a relatively higher N fertiliser demand, lower N 2 O EFsresulted in lower emission intensities (EI) (S1.11 in Supplementary). This, combined with lower LADs, resulted in the most optimum conditions and hence the significantly lower GHG culti (KgCO 2eq /kWh el ). Bcats in the central areas (red) were found to have much higher yields and hence lower LADs, and could counteract the higher catchment EIs. This resulted in these biogas catchments to have significantly lower GHG culti (see S1.11 in Supplementary).
It must also be noted, that although highly dependent on EC feedstock, EC dom catchments in these central locations had GHG culti emissions similar to some of the worst performing SL dom plants, which had a much lower demand for energy crops. We mention this not to compare between them, but rather to point out the effect location can have on the GHG Culti performance. What this shows, is that from a GHG culti point of view, if the conditions are optimal, a higher volume of energy crops does not necessarily result in an overall higher GHG culti (see S1.11 in Supplementary). However, it must also be noted that a higher volume of energy crops could also result in a higher environmental burden (e.g., reduced soil quality, eutrophication), and this should also be a consideration for examining alternative feedstocks (this analysis is outside the scope of this paper).

Location Types and Opportunities for Altering Feedstock Mix in CG
From the spatial analysis of the biogas catchments we identified two extreme location types, as well as locational trade-offs for biogas plants operating within the CG region. These trade-offs are between lowering the land areal demand of a biogas catchment (i.e., lower LAD values due to higher yields) versus lowering its emission intensity (i.e., reducing the nitrogen fertilizer demands, often resulting in lower yields), which all depend on either favorable or unfavorable geographical and climatic conditions (e.g., leading to high or low N 2 O emissions). Determining location types and their associated trade-offs can help to identify opportunities for altering feedstock mix for biogas production, and thus help to reduce the GHG culti (kgCO 2eq /kWhel) for the region. We provide two brief examples of this.
Location type 1-Low EI and high yields: These optimum locations have the best opportunity for diversification of feedstock mix, as these areas are not entirely restricted by balancing the trade-offs of increasing catchment size (i.e., increase in LAD) and the potential for generating higher EIs (e.g., higher nitrogen fertilizer for higher yields). Therefore, crops such as grass silage (GS) with lower yields than MS could be an option for locations like these.
Location type 2-High EI potential and low yields: In areas like these, the mitigation potential of the biogas (field-to-CHP) production can drop by 11-20%, depending on the feedstock mix, e.g., SL dom or EC dom . Bcats associated with EC dom clusters are far more sensitive to these least optimum locations, due their higher dependency on energy crops. Therefore, it could be recommended that for such locations, crops which have a relatively similar or a higher yield than MS (both dry matter and methane), but with a relatively lower N demand should be considered. In this way, aiming to optimize between potentially reducing the EI, while trying to maintain or limit the increase in the LADs of the biogas catchments. The use of cereal grains should be done sparingly in these location types, if used at all, due to the relatively low yields (i.e., increases LAD). Additionally, where possible, another option could be to reduce the dependency on energy crops altogether and substituting instead with locally available residual/by-product feedstocks (e.g., domestic organic waste, molasses).
For the Central German (CG) region simulations, it was identified that the majority of biogas plants and their catchments were found in location types which had, relative for the region, average to high EIs and average to high yields (resulting in a large LAD range). The challenge for altering feedstock mixtures in these areas would be to find the trade-off between lowering the nitrogen fertilizer to reduce the EIs, while not generating too much additional land areal demand (i.e., higher LAD values due to lower yields). In these locations, Maize silage could support a lowering of the GHG culti with its higher yields; however, the dominance of MS would still need to be reduced in catchment locations tending towards higher EIs. The reason being that the high N fertilizer demand of Maize silage could exacerbate the EIs of the biogas catchments so much that the lower LAD value has no effect in reducing the GHG culti . This was observed for some biogas catchments.
For the CG region, several crops have been proposed by the regional agricultural authorities for diversifying and enhancing the feedstock mix of biogas plants and their catchments. These include crops such as whole crop silage (cereal), sugar beet, and sorghum [29,41,42]. These crops were selected in general due their compatibility to the crop rotations of farmers and their comparable methane yields to MS; albeit in some cases it could be between 15%-30% lower than MS [29]. In the case of sugar beet, this was identified as having potentially much higher methane yields and naturally much higher harvest yields (> +20t/ha), and is seen to potentially have a bigger role in the future [11]. However, sugar beet is not without its own agronomic challenges, such as suitability within crop rotations and soil loss [43].
Non-annual crops proposed included grassland and Silphie. The latter could be, once planted, harvested each year for up to 15 years. However, if farmers would dedicate large areas to Silphie, while removing the associated land out of rotation for other cropping products (e.g., livestock systems, market crops), is difficult to answer. With regards to grassland, although the methane yields are slightly lower, in some cases in CG, the yields were comparable to MS, and indeed the emission contributions were 2-3% lower than its energetic contribution. That being said, some studies have suggested that due to the relative higher methane yields and the lower costs of producing MS, GS is not an economically competitive option [4].
However, to test the viability of changing feedstock mixtures being feed into biogas plants would require further, more in-depth investigations (i.e., field studies, more detailed scenario analysis) to determine the feasibility of using these different crops in different proportions within a biogas catchment, while also taking into account the important locational factors which would influence EIs and LADs of the catchments.
We can show for the CG region possible points where trade-offs and optimizations need further research, however, for other regions these conditions will more than likely differ. This is due to the regional and local dependency of feedstocks [24]. Therefore, a similar study would also need to be carried out to identify the ranges, location types, and potential trade-offs (i.e., balancing between LADs and EIs) for other regions. Additionally, future studies should also consider further integral factors, apart from GHG savings, such as developing cropping strategies to support, for example, a greater agro-diversity and maintaining good soil quality. Naturally too, the economic viability of such feedstock configurations would also be an important consideration.

Application of an Integrated Regional Assessment
While approaches to tackle the complexity of biobased energy systems are challenging [44,45], we have started in this paper to do so at a regional scale and in a spatially integrative manner. It is clear from our analysis that conducting LCA's to estimate the Culti Emis for individual crops alone (kgCO 2eq /ha) does not provide a sufficient indication of the potential GHG culti (kgCO 2eq /kWh el ) generated by the catchments of the regional biogas plants. Instead what we see is that the cultivation emissions of regional biogas systems are dependent on their catchment configurations (e.g., share of MS and other energy crops) and location, e.g., yields and soil N 2 O EFs.
We also show that MS as an effective crop to mitigate GHG emissions per kilowatt hour produced was context dependent, based on location. In some cases, MS could support GHG reductions with its higher yields. In other cases, the nitrogen demand for such high yields was combined with unfavorable geographic factors, and this led to increased GHG culti (kgCO 2eq /kWh el ).
Therefore, we conclude from our integrative approach that the often proposed strategy of simply substituting one crop for another (i.e., based on Culti Emis alone) [4,8] needs to be adapted, and instead strategies should be promoted which take into account the configurations of crops being fed into a biogas plant and how they perform altogether, under the specific regional and locational conditions. Only in this way can the trade-offs for lower GHG culti emissions be identified and managed properly.
While the approach presented here has limitations, i.e., only one time point and not taking into account the potential effects of crop rotations (see also S1.13 in Supplementary). It is the first time that so many biogas plants and their catchments (n = 425) have been modelled for an entire region. This in itself is valuable, as it provides us with a useful regional overview to support a more integrative understanding of these systems. Furthermore, the use of spatial indicators in combination with the functional unit of electricity production also helps to provide a solution to the LCA community's debate about best functional unit for bioenergy. Additionally, it allows us to see a more complete picture of these biogas systems and to debate more clearly about the trade-offs for such energy production. Thus, the ideas and concepts presented in this paper could be a useful step in carrying out more robust and integrated assessments of biobased systems in a future bioeconomy.

Influence of Energy Crops and Location on Overall GHG Balances
From the spatial and more detailed analysis in this paper, we can add further insight into the previous set of results presented in [10], the overall GHG balances (i.e., they included cultivation, transport, and conversion). It was found that biogas catchments associated with SL dom clusters in the least optimum location types (type 2) showed a drop in their potential GHG mitigation potential. This might not be a major issue for the overall balances, as with the application of GHG credits, O'Keeffe et al. [10] estimated a mitigation potential for these biogas plants of over a 100%. However, according to the RED II (EU Renewable Energy Directive), default values for plants that use approx. 80% slurry and 20% MS should have a GHG mitigation potential of 120% [46]. In these non-optimum locations, it can drop below this (c. −5%), which in the future, if more stringent regulations are proposed, could become an issue for these locations to consider what and how they should produce feedstock for their biogas plants.
Finally, O'Keeffe et al. [10] proposed that biogas plants and catchments associated with larger EC dom clusters, while lucrative in supplying energy to the grid, should be targeted first to improve the GHG mitigation potential of the overall future electricity supplied by the regional biogas systems. This is mainly as these biogas plants have much higher cultivation emissions. From the more detailed analysis in this paper, it is clear that such biogas catchments are highly sensitive to locations with high EIs and low yields, due to their high dependency on energy crops. Therefore, adapting Bcats in these location types first, in a manner which would not affect their electricity supply to the grid, could be a good way to start improving the mitigation potential of the overall regional biogas system.
Additionally, it must be also noted, that while we focus here on the GHG environmental dimension of sustainability, future studies should consider further integral factors, apart from GHG savings, such as cropping strategies to support, a greater agro-diversity, and maintaining good soil quality.

Conclusions
While approaches to tackle the complexity of biobased energy systems are challenging, we have started in this paper to do so at a regional scale and in a spatially integrative manner. It is the first time that so many biogas plants and their catchments (n = 425) have been modelled for an entire region. This in itself is valuable, as it provides us with a useful regional overview to support a more integrative understanding of these systems. We show that for altering the feedstock supply of biogas plants, the often-proposed strategy of simply substituting one crop for another (i.e., based on Culti Emis ) needs to be adapted. Instead, strategies should be promoted which take into account the configurations of crops being feed into a biogas plant and how they perform altogether, under the specific regional and locational conditions. Only in this way can the trade-offs for lower GHG culti emissions be identified and managed properly. This is particularly important for the future use of maize silage. Furthermore, the use of integrated approaches, as shown in this paper, allows us to see a more complete regional picture of these biogas systems and to debate more clearly about the trade-offs for such renewable energy production. Thus, the ideas and concepts presented in this paper could be a useful step in carrying out more robust and integrated assessments of biobased systems in a future bioeconomy.