Effects of the German Renewable Energy Act and environmental, 1 social and economic factors: biogas plants adoption and agricultural 2 landscape change 3

18 Background: The German energy transition strategy calls for reform of the German energy sector. 19 Against this background, the Germany Renewable Energy Sources Act, or EEG, was issued in 2000 and is 20 widely regarded as a successful legislation for promoting bioenergy development, as more than 9000 21 biogas plants were built in Germany until 2017. However, the impact from different EEG periods on 22 regional biogas plants’ development and the long-term influence to regional landscape change are rarely 23 simultaneously studied. 24 Methods: This study aimed to quantitatively analyse the impact of the EEG on promotion the biogas 25 plant development in central Germany (CG) by using the event study econometric technique. A GIS-based 26 spatial analysis was further conducted to provide insight into the changes of the agricultural landscape, 27 which was resulted from the development of biogas plants during the EEG from 2000 to 2014. 28 Results: One of the main findings was that the EEGs had time-varying effects on motivating biogas 29 plants construction and selecting the plants size. The comparison between different EEG emendations 30 suggested that the EEG 2009 was the most successful one in market implementation. Besides, the 31 adoption of the biogas plant in CG was mainly driven by the farmer’s financial incentive and taken as an 32 investment to secure the farming business. At the landscape scale, the expansion of silage maize was 33 remarkable in CG from 2000 to 2014. The silage maize was intensively cultivated in the regions with high 34 biogas plant installed capacity. Since EEG 2009, the regional livestock number increased rapidly, which 35 was also associated with increasing pasture land area in CG. This phenomenon suggested a promising 36 regional animal farming and its potential of manure as biogas feedstock. 37 Conclusions: These findings imply that the policy makers should take the EEG 2009 emendation as 38 reference to promote the marketing of future new renewable energy technologies and cautions should be 39 paid on the potential land conflicts of agricultural-based bioenergy development. 40

applied method to analyze event impact in economic research, could shed light on evaluating 122 environmental relevant policy effectiveness [43][44][45]. As results, the impacts of each EEG version and 123 other determined factors on biogas plants in terms of their quantity and installed capacity could be 124 individually quantified. The quantified influence of each EEG made the comparison between different 125 EEG periods possible. The comparison results could be used to further assess the effectiveness of each 126 EEG emendation, helping policy makers in future policy design. Moreover, the pattern of agricultural 127 landscape utilization, such as cultivation area of regional energy crops and livestock farming, was 128 spatially analysed during each EEG period. In summary, by using the event study method, this study 129 where, in the below example regression equations system, the explanatory variables (X and Z) in the 166 process are for each n-th of the dependent variable (Y), and the time dummy variable (TD) indicates the 167 individual event date for each n-th of the dependent variable (Y) during the whole examined period from t 168 to t+m: 169 1, = 0,1 + 1,1 * 1 + 2,1 * 1 + * + 1, 2, +1 = 0,2 + 1,2 * 2 + 2,2 * 2 + +1 * +1 + 2, +1 ⁝ 170 , + = 0, + 1, * + 2, * + + * + + , + In this example, the variables X and Z on the right hand side are the variables that can explain the response 171 variable Y. Regarding the time dummy variable, TD is assigned to each certain n-th of the dependent 172 variable Y. For example, the first dependent variable for an event on date t is written as 1 Data preparation for event study technical potential, education level and GDP density, have significant impact on accepting biogas plants. 189 Therefore, these three variables were considered in this research to control their effects on adopting biogas 190 plants. Besides, the market prices of electricity and land were also believed to be correlated with the 191 decision of building biogas plant and choosing installed capacity. To compensate the high renewable 192 electricity production cost, there is guaranteed price provided by the EEG, which is higher than the market 193 electricity price [58]. As consequence, in Germany, we observed a rapid increase in the number of biogas 194 plants from 140 in 1992 to approximately 7,720 by the end of 2013 [21]. Meanwhile, due to competition 195 for land between biogas plants and traditional forms of agriculture, it was observed that land prices 196 increased substantially [59]. Based on these facts, it can be reasoned that if a farmer is experiencing low 197 market electricity and land prices, this farmer might have a stronger incentive to build biogas plant with a 198 higher installed capacity (using more land). Taking the advantages of the low market electricity and land 199 prices, this farmer can receive larger price difference between guaranteed tariffs for renewable electricity 200 and market price with lower transaction costs. This is a typical arbitrage behaviour driven by the financial 201 incentive, which can be understood as simultaneous purchase and sale of the same, or essentially similar, 202 goods for advantageously different prices [60]. Therefore, both land and market electricity prices should 203 have negative impact on adopting biogas plants. 204 Furthermore, the price received by the farmer for their agriculture production output might also play an 205 important role in biogas plant decision making. In a period of high prices for agriculture products, farmers 206 can gain more profits by increasing the number of livestock they breed or the area of land that they 207 cultivate. In this situation, they are reluctant to build a biogas plant, but focus more on their agriculture 208 business. In contrast, if the price they can obtain for their production is low, e.g., low milk prices, then 209 traditional agriculture is less profitable. In that example, the biomass grown on the former pasture land 210 could not be used for dairy feeding anymore. If the biomass is not put to an alternate use, e.g., as substrate 211 for biogas production, the land would fall out of agronomical production [59]. To summarize, an increase in the price of the agriculture production output received by the farmer should have a negative influence 213 on adopting biogas plants and on the selected installed capacity. 214 EEGs as well as biogas potential, education level, GDP density, electricity price, land price and price of 215 the agriculture production output were considered as explanatory variables for the decisions about 216 adopting a biogas plant and selecting its installed capacity. The annual county-level panel data were 217 collected on these variables for the period from 1995 to 2015.

232
Denoted as , (mWel./km 2 ), this variable first used the numbers of cattle ( , ) and pigs ( , ) 233 and the cultivated areas of maize ( , ) and grassland ( , ) in county x in year t together 234 with the electricity generation rates of those feedstocks ( , , and ) to calculate the 235 theoretical biogas plant electricity generation ability of county x for year t. Then, the yearly county-level 236 electricity generation ability was adjusted according to the proportions of substrates and divided by the 237 area of the county to obtain the biogas potential per squared kilometre. 238 As examined in the studies conducted in other countries, the education level of the people played a vital 239 role in adopting biogas plants in terms of the ability to foresee the benefits and to operate the biogas plant 240 (2) 244 , (%) was the ratio of high school graduates with university entrance permission to the total number 245 of high school graduates from county x in year t. In Germany, students can complete their high school 246 education in three types of schools, i.e., Realschule, Berufsschule and Gymnasium. Among these, only the 247 graduates from Gymnasium can take the university entrance test (German: Abitur) and receive university 248 entrance permission (German: allgemeine Hochschulreife). The graduates from the other two types of 249 high school are not allowed to go directly to the university. Therefore, the higher the , , the higher 250 the education level in county x in year t. 251 The third variable derived from the collected data is the GDP density, written as , : 252 The descriptive statistics of all selected variables were summarized in Table 2. 257 period of a biogas plant varied strongly from two months to two years according to the size of the plant 265 and that construction technology has developed rapidly in the last two decades, the average building 266 period of a biogas plant was assumed to be approximately one year based on the data from BiogasWorld 267  In contrast with Model I, the response variable in this model was on the basis of individual biogas plants. 296 . was the installed capacity of the i-th biogas plant in county x in the year y. The set of independent 297 variables was the same as before to explore the effects of the environmental, social and economic factors 298 and the EEGs on the installed capacity of the operating biogas plant.

Data and methods for land use change analysis 300
The regional land use change pattern was derived from Corine land cover (CLC) change-maps which are other vegetated land (OV), and wetlands and waterbodies (WW) (see Table 3). To understand the 308 developments of land use in CG during the study period, four change area matrices were prepared using 309  Table 3 Land use types in current study 312

Data and method for agricultural landscape pattern analysis 313
For the agricultural landscape pattern analysis, the whole studied period was from 1995 to 2014. Apart 314 from the above defined four sub-periods of EEGs, the Non-EEG period was introduced covering the years 315 from 1995 to 2000. Considering CLC changed-maps only report arable land change as one land use type, 316 which limits the analysis for silage maize area changes during various EEGs, the detailed agricultural 317 crops area was derived from regional crop statistic record to facilitate evaluation. To relate dynamic 318 changes in the agricultural landscape to biogas plant installed capacity density, we adopted the approach 319 from Csikos et al. [29]. Three impact zones (A, B and C), representing different density categories (low, 320 medium and high) of the installed capacity (kWel./km 2 ) of biogas plants, were delineated using the Kernel 321 Density tool within ArcGIS 10.7 software. The separation of the single installed capacity density class into 322 three intervals followed the Jenks natural breaks classification method. After that, the impact zones layer 323 was overlaid with the county-level map of CG to assign each county to a zone. Once this classification 324 was completed, based on the collected data on regional crop distribution, the changes in agricultural 325 landscape pattern from the Non-EEG to EEG IV time periods were analysed for each impact zone. There 326 was a two-level evaluation: 1) at the utilised agricultural area (UTA) level, the ratios of arable land and 327 pasture land to UTA were calculated; 2) at the arable land level, the areas of CG cultivated with dominant 328 agricultural crops, such as wheat, rye, triticale, maize, sugar beet and rapeseed, were divided by total 329 arable land. This provided a general impression of the change in the cultivation pattern of energy crops 330 between Non-EEG and EEG IV. In addition, considering that silage maize is the main feedstock for 331 biogas plants, a detailed analysis of silage maize area related to total arable land within each impact zone 332 was further explored. 333 Results 334

Impact of EEGs and environmental, social and economic factors on biogas plants 335
For each of the Model I and II, we first regressed the dependent variable only on the independent 336 environmental social and economic variables and then we ran the whole model. The results of the 337 multivariate regression with time dummy variables for both runs of each model were summarized in Table  338 4. In addition, the results of analysis of variance based on the first and second runs for both Model I and II 339 are summarized in Appendix Table 6. The multicollinearity of the independent variables was checked 340 using variance inflation factor and the result could refer to Appendix Table 7. The regression residuals of 341 both models were also examined using regression diagnostic plots and reported in Appendix Fig. 9 and 10. 0.41 0.97 Sample size 700 1016 Note: "***","**","*" and " · " denote 99.9%, 99%, 95% and 90% confidence levels, respectively. In the parentheses are the standard error. The interpretation of the coefficients of Environmental-social-economic variables in Model I:

Maize expansion influenced by biogas plant development 391
In total, three impact zones were identified in the studied area with value ranges of 0-3.80 kWel./km 2 for 392 impact zone A, 3.80-10.00 kWel./km 2 for impact zone B, and 10.00-18.70 kWel./km 2 for impact zone C 393 (see Fig. 4). If a certain impact zone occupied more than 50% of the area of a county, this county was 394 assigned that impact zone. If no impact zone represented more than 50% of a county's area, the county 395 was considered to be an impact zone B county. There were 12, 32 and 6 counties categorized into impact 396 zones A, B and C in CG (see Fig. 4 a). From Non-EEG to EEG IV, the overall proportion of arable land in 397 the UTA decreased from 72.06% to 53.59%, within which impact zone A showed a relatively higher 398 proportion of arable land than impact zones B and C. After EEG III, there was a steep decline in the 399 proportion of arable land in each impact zone (see Fig. 4 b). In contrast, the proportion of grassland area in 400 the UTA showed a rapid increase in EEG III. In the study area, the grassland proportion increased from 401 21.59% to 45.91% across the whole examined period. The highest proportion of grassland was observed 402 in impact zone C, with a mean value of 32.98% during all EEG periods. In comparison, impact zones A 403 and B showed mean values of 22.05% and 29.40%, respectively during the EEG periods (see Fig. 4 c). 404 The results of the change analysis for the areas of major agricultural crops showed that the areas of the 410 energy crops silage maize and rapeseed increased by 57.85% and 76.49%, respectively from Non-EEG to 411 EEG IV. Wheat, occupied around 45.79% of the area of major agriculture crops, was the largest 412 proportion among all of the crops during the whole studied period. Compared to the Non-EEG period, the 413 cultivated area of wheat in EEG IV increased by 37.53%, while rye, triticale and sugar beet shared a 414 declined trend over the same period. Among them, the area of triticale decreased significantly by 25.74% 415 (see Fig. 5 a). 416 In CG, the proportions of total arable land planted with silage maize in EEGs I to IV were higher than in 417 the Non-EEG period. Compared to the proportion of silage maize in the Non-EEG period, the proportions 418 in EEGs I to IV were 5.14%, 13.63%, 59.48% and 132.09% higher. However, when it comes to specific 419 impact zones, the patterns differ. In Non-EEG period, the highest proportion of silage maize was detected 420 in impact zone C, with a mean value of 7.28%. In comparison, impact zones A and B showed mean values 421 of 5.81% and 5.89%, respectively. In impact zone C, a slight decline of 3.21% in the proportion of silage 422 EEG Impact on regional landscape pattern a) c) b)

Grassland
Arable land maize was observed in EEG II relative to EEG I. After EEG II, the proportion of silage maize increased 423 strongly in each impact zone, especially in the EEG III period. Temporally, the proportion of arable land 424 represented by silage maize increased significantly after the EEG was introduced. Spatially, strong 425 increases were identified in impact zones A and B, while impact zone C showed a mild increment (see Fig.  426  to the individual farm feedstock availability, which was resulted from farming business volume (farm-452 scale plants) [70]. In high biogas technical potential area, the feedstock availability of each farm was 453 enough to support a small to medium size farm-scale biogas plant. Whereas in low biogas potential 454 counties, a larger biogas plant might be built and fed by several farms, since it is still not economic to 455 build biogas plant for each farm even after considering the cost on feedstock transportation [70]. is positive. In fact, in Germany where the GDP density is higher than most of countries in this world, the 471 regional GDP density factor says less about the economy, but rather indicates whether a county is rural 472 (Landkreis) or urban (Kreisfreie Stadt). The mean yearly GDP density of 12 cities in CG from 1995 to 473 2013 was 29.49 M€/km 2 , whereas that of all rural regions was only 2.48 M€/km 2 . The negative influence 474 of GDP density could be thus interpreted to mean that the biogas plants, especially those with a high 475 installed capacity, tend to be built in rural areas. 476 Land price had its expected effect on the decision to build a biogas plant. In contrast, the market electricity 477 price had an unexpected strong positive effect on the number of constructed biogas plants CG. The result 478 did not directly support that the decision to build a biogas plant was driven by arbitrage incentives. 479 However, it should be pointed out that the biogas plant operators did not arbitrage by purchasing and 480 selling the electricity. Compared to the amount of green electricity that the biogas plants owners sold to 481 national grids, the amount of electricity they bought back for their daily farming use was considerable 482 small. The plants owners actually arbitraged the difference between biogas production cost and EEG FIT 483 for electricity generated from biogas. The positive impact of market electricity price further indicated that 484 in the period of high market electricity price, the farmers had even stronger financial incentive to arbitrage 485 the price difference to subsidize their own electricity consumption. This conclusion was supported by 486 Scheftelowitz et al. [23], who stated that the financial incentives, provided by EEG, have the potential to Austria, where farmers tended to intensify the energy production, when the electricity price went up [77]. 493 Therefore, it could be concluded that the adoptions of biogas plants in CG were strongly driven by the 494 farmers' financial incentive. 495 Increases in the price of agriculture products could reduce willingness to build plants but did not influence 496 their installed capacity. This implicated that the biogas plants in CG were mainly built to be an investment 497 to secure the farm business. If agricultural products are profitable, farmers aim to increase their 498 agricultural production capacity by investing more money and increasing the production area. Under those conditions, they are reluctant and also lack the money to build and operate biogas plants. In periods with 500 low prices for agricultural products, they are willing to build biogas plants to reduce their daily business 501 costs on electricity to survive. This finding went in line with Thiering [28] and Fuchs et al. [78]. However, 502 the biogas plant size did not vary with the price index, supporting the hypothesis that the installed capacity 503 depends on the farmer's own farming business volume. 504

EEG performance in promoting biogas plants 505
All EEGs together have the ultimate goal of increasing the contribution of renewable energy to total 506 electricity consumption in Germany [8][9][10][11][12]. Therefore, these acts have created many advantageous 507 conditions for the access of biogas plants to electricity markets and grids as well as provided a secure 508 investment and financing of biogas plants through remuneration [15]. The remuneration policies of EEGs 509 I to IV are summarized in the Appendix Table 11. 510 The EEG I had a one-track biogas plant remuneration system and the subsidy categories were set with an installed capacity of up to 150 kWel. was specified in the new remuneration tariff. Stronger 517 subsidy policy and of EEG II led to a better performance than EEG I. In the EEG III, the basic 518 remuneration plus premium subsidy model was introduced. Aside from the premium for renewable 519 resources inherited from EEG II, subsidies for manure, landscaping material, among others as substrates 520 were introduced. This was obviously to attract farmers who still hesitated to adopt a biogas plant because 521 their farming type did not apply to previous EEGs subsidy schema. Besides, for each installed capacity 522 category and substrate type, there was a unique allowance amount combined by basic remuneration plus 523 premium subsidy. This could help the potential biogas plants operators to place themselves into the most 524 attractive remuneration categories according to their own farming situation. Despite the general decrease 525 in the magnitude of the per-unit subsidy, the increased reward scopes for feedstock types, more tailored 526 remuneration categories and less remuneration degression rate led to a strong increase of marginal 527 acceptance of biogas plants in CG. Therefore, the EEG III was the most successful emendation in 528 prompting the adoption of biogas plants in CG. In the EEG IV, to encourage the use of manure which has 529 relatively low electricity generation ratio in biogas production, a new subsidy program is added for 530 extremely small livestock waste based biogas plants (up to 75 kWel.). Different from the EEG III 531 remuneration programs, the EEG IV focused more on supporting the biogas plants with straw, landscaping 532 material, and manure. The result of the EEG IV, was generally satisfactory. 533 In terms of installed capacity, compared to EEG I, EEG III and IV had significantly impacts on reducing was the evidence of the success of these two emendations. 543

Maize expansion and changes of agricultural landscape 544
The landscape analysis indicated an increase in demand on substrates (e.g. silage maize or grass, etc.) 545 resulted from growing number of biogas plants in CG. To decrease the pressure on permanent grassland, 546 the EU's Common Agricultural Policy reform in 2013 regulated "greening" obligations to finance farmers 547 to conduct environmentally sound farming practices, such as crop diversification and maintaining 548 ecologically rich landscape features [81]. Besides a higher premium for biogas production using manure, 549 the EEG IV also introduced the "maize cap", which emphasized that the share of feedstock represented by 550 maize and cereal grain kernels must not exceed 60% of the total mass [11]. All of these regulatory forces resulted in a visible increase in pasture area during 2012-2018 in the CLC change-maps as well as an 552 expansion of permanent grassland. This indicated for the future development of the agricultural-based 553 biogas industry that animal manure should be more favourable than energy crops as a substrate. Another 554 possible reason was the complementarity of biogas and livestock production could lead to an additional 555 intensification of land use and more investments in livestock production [82]. Germany is known as the 556 largest milk and pork producer in EU and after France the second-largest producer of beef and veal [83]. 557 As shown in Appendix Fig. 12, the prices of dairy products and meat showed upwards trends in the last 558 two decades and might attract more farmer to start animal farming. The evidence could be found in the 559 increasing numbers of livestock, especially cattle and pigs, in CG (see Fig. 7). Moreover, the German 560 export value of animal products increased significantly from 2000 and exceeded the import value in 2005 561 (see Fig. 8a). Apart from this, a widely discussed concern was the indirect land-use effects (iLU, leakage) that resulted 569 from meeting a given demand of feedstock for bioenergy production. The report showed that the digestion 570 of manure for biogas production had no iLUC at all, while the production of ethanol/biodiesel from energy 571 crops did have iLUC [85,86]. Public concern has been raised regarding the potential for iLUC due to 572 energy crop cultivation and transportation for biogas production. For example, the evaluation of biogas 573 production sustainability in Italy would be different if iLUC was considered, since maize, the major 574 substrate for biogas production, was partially imported from other countries [87]. In addition, the findings 575 from Britz and Delzeit [88] showed the land demand in Germany for biomass used for biogas production 576 could reduce Germany's exports, or the another way around, increase its imports of agricultural goods. 577 Based on their research, the comparison between EEG III and IV suggested the increasing demand on land 578 for biogas production, which consequently attributed to agricultural products (e.g. cereals, oilseeds and 579 animal products) price increase in EU market. This also accompanied with agricultural land use expansion 580 in the EU outside of Germany. In addition, we found a higher import value of vegetable products in 581 Germany after EEG I, which also supports this statement (see Fig. 8b). This indicated the potential iLUC 582 of arable land utilisation in other countries. We did not consider iLUC in the current study, not only 583 because it was beyond our research focus, but also with the consideration of the large uncertainty for 584 iLUC calculation, which resulted from missing of standard evaluation approach. However, this would be a 585 very crucial aspect if direct land use changes and iLUC were linked to evaluate the environmental impacts 586 and sustainability of biogas plants, considering aspects such as GHG emissions and global warming 587 potential. 588 In Germany, there has been an upward tendency in the area of organically farmed land, which increased 589 from 2,721.39 to 15,213.14 km 2 from 1994 to 2018. Meanwhile, the number of organic farms increased 590 from 5,866 to 38,713. In 2018, the total area of organic farming in CG made up around 13.40% of the total 591 area of organic farms in Germany, but the UTA of CG occupied 17.10% of the national total UTA [89]. 592 Therefore, it is high feasibility for future organic farming development in CG. Additionally, as concluded in previous study, there was a large potential for the use of biogas slurry as a fertilizer for organic farming 594 systems [90]. In this way, the booming of biogas plants during the EEGs could facilitate the development 595 of organic farming in CG. Moreover, as pointed out in Siegmeier et al. [91], biogas-production-integrated 596 organic farms may further contribute to renewable resource supplies without an additional need for land. 597 More importantly, this could simultaneously increase food production and reduce GHG emissions from 598 livestock manure. These concepts together could be tailored into an effective approach for sustainable 599 biogas plant management in an agricultural landscape by considering potential land use change and 600 environmental impacts. 601

Outlook and implications for future study 602
Beyond the current study, we advise considering the costs of agricultural waste disposal in future studies, 603 since it also plays a vital role in farmers' decision making on adopting biogas plants. In Germany, among 604 agricultural wastes, the disposal of livestock excrement is most strictly regulated. In each region, there is 605 an upper limit for livestock waste production set to maintain the soil nutrient cycle in the regional 606 landscape. Once the upper limit is reached, the excess slurry must be transported out of the region to avoid 607 potential groundwater pollution. Therefore, so-called liquid manure exchanges (German: Güllebörse) 608 operate in Germany to facilitate the transaction of agricultural waste. For instance, after the exchange is 609 negotiated, surplus livestock excrement from livestock-intensive regions is transported to regions where 610 the farmers mainly practice agriculture and need the liquid manure as free fertilizer. This process 611 generates transaction and waste transport costs. If the cost of agricultural waste disposal is high, the meat 612 and dairy farm owners are prone to building their own biogas plants to dispose of the waste and gain extra 613 profits. However, data on the cost of agricultural waste disposal could not be collected properly in our 614 study, since the transportation cost varies from deal to deal depending on the type of waste, transport 615 distance, etc., and the transaction cost is not easy to quantify. Additionally, in the current study, the 616 importance of spatial analysis was emphasized. Considering the regional heterogeneity caused by spatial 617 characteristics, e.g. topographic, soil, climatic, and other social-economic variations, the EEGs impacts on 618 different regions in Germany still contain large discrepancy. Therefore, to understand policy effectiveness on national level, future studies should take both environmental, social and economic factors, as well as 620 spatial-temporal regional agricultural landscape change into account. In addition, if time series data on 621 crop distribution could be obtained or generated using models, it would be possible to conduct more 622 detailed landscape analysis such as resource optimization and trade-off analysis between the 623 environmental costs and economic gain of adopting biogas plants. 624

Conclusion 625
Two research goals were pursued in our study. The first was to deploy the event study technique to  Note: "***","**","*" and " · " denote 99.9%, 99%, 95% and 90% confidence levels, respectively. Source: Helmholtz-Centre for Environmental Research (UFZ) EE-Monitor.  Note: The total balance is difference between contribution from others and contribution to others. Positive value indicates that this land type increase its area in the examined period, while negative value means loss.

Environmental-social-economic Variables
To