Potential of genetically modified oilseed rape for biofuels in Austria: Land use patterns and coexistence constraints could decrease domestic feedstock production.

Like other EU Member States, Austria will meet the substitution target of the EU European Renewable Energy Directive for transportation almost exclusively by first generation biofuels, primarily biodiesel from oilseed rape (OSR). Genetically modified (GM) plants have been promoted as a new option for biofuel production as they promise higher yield or higher quality feedstock. We tested implications of GM OSR application for biodiesel production in Austria by means of high resolution spatially explicit simulation of 140 different coexistence scenarios within six main OSR cropping regions in Austria (2400 km2). We identified structural land use characteristics such as field size, land use diversity, land holding patterns and the proportion of the target crop as the predominant factors which influence overall production of OSR in a coexistence scenario. Assuming isolation distances of 800 m and non-GM-OSR proportions of at least 10% resulted in a loss of area for cultivation of OSR in all study areas ranging from -4.5% to more than -25%, depending on the percentage of GM farmers and on the region. We could show that particularly the current primary OSR cropping regions are largely unsuitable for coexistence and would suffer from a net loss of OSR area even at isolation distances of 400 or 800 m. Coexistence constraints associated with application of GM OSR are likely to offset possible GM gains by substantially reducing farmland for OSR cultivation, thus contradicting the political aim to increase domestic OSR area to meet the combined demands of food, feed and biofuel production.

Introduction

Transportation is a major source of greenhouse gas emissions and its contribution is rising [1]. To reduce these emissions in Europe, a substitution target has been set in the Renewable Energy Directive [2] that defines that 10% of the fossil fuels used in the transportation sector have to be substituted with renewable sources by 2020. Like other EU Member States, Austria will meet the substitution target for transportation almost exclusively by first generation biofuels [3,4]. Since biodiesel, produced from oilseed rape (OSR), is by far the most important substitute in Austria (2009: 521,611 t biodiesel compared to only 99,628 t ethanol [5];), this study focuses on OSR production. First generation biofuels have come under scrutiny since their production requires huge area and is competing with food production [6]. Decreasing area for food production has been made responsible for rising food prices in recent years [7-9]. To compensate area loss for food production, biofuel production may trigger indirect land use changes by recruiting new farmland, partially by converting carbon-rich natural ecosystems, which can offset greenhouse gas (GHG) benefits of biofuels as compared to fossil fuels [4,10,11]. GHG efficiency could, however, be improved if yields could substantially be increased on site, i.e. if food, feed and fuels could be produced on essentially the same agricultural land as before. In this respect, genetically modified (GM) plants have been recommended as a new option for biofuel production [12,13] and this has been echoed in the political discussion concerning the potential risks and benefits of GM plants for biofuel production in Austria. Since these GM crops for biofuels are neither intended for feeding humans nor animals, but for feeding cars, public reluctance may be less pronounced and thus may open a back-door for GM crops even in countries that have advocated a restrictive GM policy to date, such as Austria.

Globally the commercial cultivation of GM OSR varieties is focused on a few countries (Canada, USA, Australia, Chile) and on cultivation of herbicide tolerant (HT) GM varieties [14]. In the EU, commercially important GM OSR varieties, like GT73 and Ms8xRf3, are authorised for import and processing, but not for cultivation so far. Even if current experience does not clearly indicate significant yield surplus by GM-HT OSR crops [15], management benefits and reduced production costs are strong arguments for farmers and will certainly raise the discussion for (non food) industrial use of GM OSR in the EU, e.g. for biofuel production.

Since a substantial part of the EU public is opposing the use of GM crops for food production [16], coexistence rules are discussed to prevent or reduce contamination with GM material in conventional or organic crops to guarantee that different production systems (conventional, organic and GM) can co-exist side-by-side, and to ensure freedom of choice for producers and consumers (e.g. [17-19]). One important ex-ante measure is to separate GM and non-GM fields by isolation distances to reduced unintended processes like GM pollen flow (e.g. [20]). According to the European Directive 2001/18/EC the specifications for coexistence measures (e.g. isolation distances) for individual crop species are determined at the national level. Several European countries have already proposed such isolation distances for GM OSR: Slovakia: 400–600 m [21]; Luxembourg: 3000 m [21]; Latvia and Lithuania: 4000 m [22]; Ireland: 100 m (to conventional crops) to 500 m (to organic fields (pers. comm. Irish EPA [22]), Austria: 800 m to conventional and 1.000 m to organic fields (pers. comm. Austrian Agency for Health and Food Safety- AGES). Isolation distances do not only mitigate GM contamination but have significant impact on the availability of farmland where GM crops could be grown (e.g. [18,23-30]). The feasibility of planting GM crops could be negatively affected by increasing isolation distances, increasing proportion of non-GM crops of a specific species and decreasing field sizes.

However, little is known about the interacting effects of these variables and factors like land holding patterns and land use diversity have been largely neglected so far. While it is widely recognized that coexistence measures like isolation distances may restrict the individual freedom of choice for farmers, the potential implication for national or EU strategies that rely on increasing production, such as the Renewable Energy Directive, have not been discussed.

In order to assess the potential effect of GM crop application on the cultivable area and yield of OSR, we performed spatial explicit simulations of various coexistence scenarios (assuming different proportions of GM farmers, different regional proportion of OSR cultivation and different isolation distances) with high-resolution land use data. The study comprises six large regions (of about 400 km2) within the prior OSR cropping regions of Austria differing substantially in land use diversity, field size, farm size and land holding pattern. Based on the simulations, we address the question whether the use of GM OSR in a realistic coexistence scenario would be a viable option to increase the national production of OSR for biodiesel in Austria. Additionally, we discuss the effects and interactions of variable isolation distances, OSR cropping proportion, GM farmer proportion in the context of different field-and farm size, land holding pattern and land use diversity. Whereas the analysis focuses on OSR as the predominant crop for biofuel production in Austria, the approach can be easily applied to other crops and regions.

Methods

Data and study area

We used data derived from the IACS (integrated administration and control system) database (reference year 2008) of the Austrian Federal Ministry of Agriculture, Forestry, Environment and Water Management. The IACS is a framework of directives enacted by the European commission in order to control and standardize agro-policies among the Member States. The Austrian IACS system, managed by Agrarmarkt Austria, is the basic tool for the administration of agro-subsidies to farmers. The IACS GIS database is an annually updated high-resolution land use map. It covers individual field parcels and provides field-specific information on the cultivated crop and the management. It thus offered the opportunity for a fine-scale simulation of coexistence scenarios over large areas. Since IACS is primarily an administration tool for agro-subsidies, it contains only data on farms that actually receive payments and hence, does not cover the total land surface. In the context of our study, these limitations are negligible, as the database currently comprises about 82% of all farms but about 99% of the total arable land. We selected six study areas within the main OSR cropping regions in Austria, each with a size of about 400 km2 (Fig. 1. and Table 1.).

Fig. 1

Location of the study areas (1–6) within Austria. All study sites are situated in the Austrian lowlands, since cultivation of OSR in the Austrian Alps is not feasible. The shading indicates the actual proportion of OSR cropping within municipalities in 2008, ranging from low (light-grey) to high proportions (dark-grey) of OSR.

Table 1

Summary statistics of important characteristics of the six study areas; mean farm size: only acreage of annual crops; mean farm extent: mean area of the minimum bounding rectangle covering all fields of one farm.

Study area	Crop area [ha]	Whole area [ha]	% Crop area	Number of farms	Mean farm size [ha]	Mean field size [ha]	Mean farm extent [ha]	% OSR (2008)
1	18,445	40,131	45.96	1137	16.23	1.82	413.15	9.02
2	18,819	38,578	48.78	1452	12.98	1.53	330.72	6.85
3	23,233	40,986	56.69	841	27.63	1.93	1177.55	15.57
4	32,133	40,210	79.91	903	35.59	1.53	2479.37	9.37
5	29,873	40,290	74.15	696	42.94	3.26	2914.42	3.96
6	27,683	40,284	68.72	822	33.68	1.67	2875.83	13.91

The study areas differ in overall proportion of cropland, number of farms, mean farm size, field size, form and configuration, land holding pattern, and the spatial distribution of the field parcels (Table 1, Fig. 1.).

Study areas 1 and 2, located in Upper Austria, are dominated by mixed agriculture (annual crops, meadows, pastures, and forests). Cropland does not exceed 50%. Farms are small and spatially compact (mean extent ca. 2 × 2 km, Fig. 2(1), (2)), which leads to a clustered land holding pattern. In 2008, OSR cultivated area proportion ranged between 6.9 and 9.0% of arable land according to the IACS database.

Fig. 2

Examples of the land holding pattern of typical farms within the six study areas. Numbers correspond with the study areas numbers. Coordinates of farm centroids: (1): 13.415 E 48.201 N; (2): 13.902 E 48.332 N; (3): 15.676 E 48.771 N; (4): 16.743 E 48.544 N; (5): 16.842 E 48.222 N; (6): 17.005 E 47.983 N. Grey shading: crop fields; black shaded: crop fields belonging to one particular farm; un-shaded: other land use.

Study area 3 is located in the northern part of Lower Austria and is characterized by approx. 57% cropland interspersed with forests and meadows. Fields are predominantly small, narrowly elongated and distributed over a comparably large area (mean ca. 3.4 × 3.4 km, Fig. 2(3)) around their farmstead. Study area 3 shows the highest proportion of OSR cultivation (15.6% in 2008) of all study areas.

Study area 4 and 6 are intensively used agricultural areas in Lower Austria with a high proportion of annual crops (80 and 69%). The mean farm size is about twice the farm size in study area 1 and 2. Fields are small and scattered over large areas which results in large farm extents (mean ca. 5.2 × 5.2 km, Fig. 2(4), (6)) and a scattered land holding pattern. OSR production covered intermediate (9.4%, study area 4) and high (13.9%, study area 6) proportions in 2008.

Study area 5, located in Lower Austria, is one of the most intensively used agricultural regions in Austria and is dominated by annual crops (74%). It is the region with the largest farms, fields and farm extents (mean: ca. 5.4 × 5.4 km, Fig. 2(5)) of all study regions but the lowest OSR proportion (<4%).

Simulation

Scenario definitions

Our spatially explicit simulations are based on a set of core specifications that apply to all scenarios considered. First, all farms were either conventional farms or GM farms. We excluded intra-farm coexistence and we did not allow GM farmers to cultivate non-GM OSR within the isolation zone. Then, we applied one isolation distance for each scenario, i.e. we did not discriminate between conventional and organic farming in terms of different isolation distances. This simplified the simulation scripts and facilitated the interpretation of the model outcomes. Since the discussion about appropriate isolation distances for OSR is still controversial and proposed distances range from 50 m (lower limit according to [31]) to 4000 m (as required e.g. by Latvia), we varied isolation distances between 400 and 1600 m across our scenarios. The tested isolation distances cover the draft technical specification for coexistence of OSR crops in Austria, but are also relevant for the prediction of realistic coexistence scenarios beyond Austria.

We further assumed that a certain proportion of non-GM OSR cultivation being randomly distributed over all farms and crop fields in the whole study areas. We assumed that OSR will only be cultivated on crop fields and excluded all other land use types such as meadows, forests or vineyard from our simulations. Reflecting the current situation in the EU, we assume property right is with the non-GM farmers, i.e. the non-GM farmer has the right to produce non-GM crops and GM farmers have to take measures to protect that right [30,32] Hence, GM cropping is restricted to the area beyond the isolation distance around non-GM fields.

Monte Carlo simulations

Based on the above specifications, we used a three-step approach to develop the scenario families. First, we randomly selected varying proportions (5, 10, 15, 20, 25, 30 and 35%) of farms as GM farms. Secondly, we randomly selected varying proportions of fields (5, 10, 15, 20 and 25%) as OSR fields. Finally, we applied different isolation distances (400, 800, 1200 and 1600 m) around non-GM OSR fields and selected all fields of GM farmers that are outside the resulting isolation area as remaining area for GM OSR production.

Consequently, we performed Monte Carlo simulations [33] of 7 × 5 × 4 = 140 different scenarios with 100 iterations each for the six test regions. This resulted in 84,000 simulation runs over an area of approximately 6 × 400 = 2400 km2.

Simulation outputs are presented by box-plots (Figs. 4 and 5). The boxes represent 50% of the data showing the lower quartile, the median and the upper quartile. The whiskers indicate the smallest and largest observations. Each single box-bar represents the statistic of 100 simulation runs for one scenario of one study area. A sequence of six bars (separated by dashed vertical lines) represents the outcomes for the six study regions for one single scenario. In Fig. 3, the y-axis of the box-plots shows net balance of OSR area for a study region, i.e. the combined area of conventional and GM OSR (assuming total allocation of available land for GM cropping) as percent of total arable land per study region. Fig. 4 shows the available area for GM OSR as percentage of GM farm area. To condense the results and to facilitate interpretations, we excluded scenarios based on large isolation distances (>800 m) from the figures, as they only continue trends but do not provide additional information. However, all results are provided in the Supplementary Material (SOM Table S1).

Fig. 3

Scheme of Monte Carlo simulations: The map shows a detail of our study area to illustrate the spatial operations during the simulation; (a) dark grey: all fields of randomly selected GM farms; bold black bordered: randomly selected non-GM OSR fields; white: other crop fields. (b) Situation after applying the isolation distance around non-GM OSR fields; light grey: blocked areas of GM farms; dark grey: remaining area for GM OSR cropping; white: other crop fields.

Fig. 4

Relation between total resulting OSR area (non-GM OSR + GM OSR, assuming complete allocation of available area for GM OSR) and the proportion of non-GM OSR for four scenarios with different isolation distances (400 and 800 m) and different proportions of GM farmers (10 and 20%); the horizontal line indicates the proportion of non-GM OSR as the threshold below which a net loss of OSR area occurs; the boxes represent 50% of the data showing the lower quartile, the median and the upper quartile. The whiskers indicate the smallest and largest observations. Each single box-bar represents the statistic of 100 simulation runs for one scenario of one study area. A sequence of six bars (separated by dashed vertical lines) represents the outcomes for the six study areas (1, 2, 3, 4, 5, and 6) for one scenario.

We used R [34] for random selections, statistics and plots. Spatial operations were implemented by ESRI ARC-GIS 9.3 Python scripts.

Results

Effect of the proportion of GM farmers, isolation distances and non-GM OSR proportion

Our simulations (Figs. 4 and 5) showed that only at low isolation distances (400 m) and a low proportion of non-GM OSR cultivation (5%), a positive net balance of area could be reached in all study regions, i.e. the combined area of conventional and GM OSR is equal or higher than the assumed overall OSR proportion. As the proportion of non-GM OSR increased (10%, Fig. 4b), study areas 3, 4 and 6 suffered from a net loss of total cultivable land for OSR as GM farmers were not able to allocate at least 10% of their land for GM OSR cultivation. A further increase to 15% non-GM OSR proportion resulted in a negative total OSR production area balance for all study areas. At isolation distances of 800 m, only 3 study areas achieved a positive balance, and this only at lowest proportions of non-GM OSR cropping (5%). A further increase to 10% non-GM OSR proportion induced a negative OSR production area balance in all regions. Generally, isolation distances beyond 800 m always resulted in an almost total exclusion of GM production and caused high values of area loss, (SOM) provided that OSR was being cultivated on at least 5% of the farmland area.

Scenarios with isolation distances of 800 m and 10% non-GM OSR proportion caused an area loss of −8.6 to −9.9% at low levels of GM farmers proportion (10%) but could increase to −17.8 to −24.6% at higher proportions of GM farmers (25%), depending on the region (SOM).

Differences owing to land use patterns between regions

To assess the effect of land use patterns, we compared the performance of different regions within the same scenarios (Figs. 4 and 5). Despite isolation distance, proportion of the target crop and proportion of GM farmers were assumed equal, model outcomes differed distinctly between the regions: In study area 1, farmers would be able to allocate about four times more land for GM OSR cropping than in study area 4, which is unsuitable for OSR coexistence in most scenarios. At low isolation distances (400 m), study areas 1, 2 and 5 reached higher total OSR area than study areas 3, 4 and 6.

Discussion

Coexistence constraints and land use patterns

Our study advances existing knowledge as it empirically and quantitatively tests the implications of GM OSR production under a wide range of scenarios. Overall, our simulations confirm previously established relationships between available land for GM cropping, isolation distance and proportion of the target crop or field size. The mechanism behind the effect of the isolation distance is straightforward: The larger the isolation distance, the smaller the available area for GM cropping [23-30]. At low levels of OSR percentage (i.e. resulting in low overlaps of the isolation zones), the production area difference between two isolation distances is determined by geometrical constraints and can be calculated as (r1/r2)2 (with r = isolation distance). The higher the percentage of non-GM OSR, the higher the overlap of the isolation areas and the smaller the relative effect of increasing the isolation distances. Similarly, increasing area of the non-GM target crop implies that the isolation zone increases rapidly in a non-linear manner and prohibits GM production over large areas. With decreasing field size, the non-GM crop is spread out over larger areas and thus precludes GM production accordingly [24]. On the other hand, increasing proportions of GM farmers result in a moderate increase of the available area for GM cropping.

Our simulations show that land use patterns substantially affect available cropping areas: For example, GM farmers in study area 1 would be able to allocate about four times more land for GM OSR cropping than farmers in study area 4, which has about the same effect as halving the isolation distance from 800 m to 400 m and even higher than the effect of halving the non-GM OSR cropping proportion. At low isolation distances (400 m), study areas 1, 2 and 5 show consistently higher values of available area for GM OSR cropping than study areas 3, 4 and 6. Considering the reasons for this clustering, one might suppose larger farms to have larger land resources unaffected by neighbours available for GM cropping. By contrast, feasibility of coexistence is highest in the study areas with the smallest farms (1 and 2). Low proportions of arable land (i.e. higher land use diversity) increases the mean distance between individual fields and hence decreases the probability that a field will be within the isolation distance of a neighbouring field. Similarly, farm extent affects the probability of a field being within the range of the isolation distance of a neighbouring field of the same crop. If farms are spatially compact, fields of a farm are predominantly adjoined by fields from the same farm, thus core areas emerge outside the range of fields from neighbouring farms. Such zones frequently emerged in study areas 1 and 2, although they were not very large. By contrast, in other study areas, the majority of fields were surrounded by fields belonging to other farms.

Increasing field size or clustering of GM fields has been shown [27,35] to significantly mitigate the effect of isolation distances. If fields are small, the same proportion of the target crop is distributed over a larger number of fields and the isolation area increases. If this is accompanied by a high level of spatial inter-dispersion of fields form different farms (as in study areas 3, 4 and 6), the suppression effect is amplified. On the other hand, study area 5 shows that larger field sizes can compensate for the suppressive effect of high field inter-dispersion as the same area proportion of the target crop is distributed over a smaller number of fields, which results in a smaller total isolation area. However, the average field size in Austria is small compared to some other farming areas of e.g. Czech Republic, Eastern Parts of Slovakia (e.g. Bratislavský kraj or Západné Slovensko), Eastern Germany (e.g. Mecklenburg-Vorpommern) or Hungary (e.g. Közép-Dunántúl) where fields are much larger [36] and the effect of isolation distance could be much smaller than shown in our simulation.

Potential gains of GM cultivation for biofuel production

In the context of biofuel production, strong pro-arguments for GM OSR application are based on the assumption of increased yields. Higher yields may improve the GHG balance of biodiesel, increase national production and would relax the competition with food production. In terms of this consideration, effects on yield by the GM traits currently available or potentially commercialised within the next years need to be taken into account. However, the information on yield gains by available GM traits as reviewed by Graef et al. [15] and Brookes and Barfoot [37] is somewhat contradictory: herbicide tolerant GM OSR varieties show moderately increased, similar or even decreased yields in different years and areas compared to conventional crops. At best, production gains obtained from GM varieties are calculated to be less than 10% [38]. Hence, the reasons for the use of GM crops are the simplification of weed management, more timing flexibility and reduced tillage management rather than increased yields [15,39]. More marked benefits could be expected from GM OSR varieties with enhanced nitrogen use efficiency [40], or lower methyl-bromide emissions [12]. However, these traits are still in early development and will not be available for commercialisation in the EU within the next years [41]. Since rapeseed cake, representing a valuable additional income stream from OSR production, cannot be sold to feed markets when made from GM varieties, advantages of a production segregation into GM crops for biofuels and conventional crops for food appear very limited.

Arable land and yield loss

The simulations indicated that those regions with highest natural or economic potential for OSR production (study area 3, 4 and 6) featured a structural configuration that is poorly suited for coexistence of GM and conventional crops even at short isolation distances (400 or 800 m). On the other hand, in study area 1 and 2, which would allow higher proportions of GM cropping at low levels of non-GM OSR proportion and small isolation distances, the proportion of arable land is small, which results in a smaller total area for OSR and a low contribution to the total national production. Study area 5 performed relatively well in terms of coexistence and had a high total cover of arable land but the current proportion of OSR (<4%) indicates its relative unimportance for OSR production. Without a large increase in OSR prices, it is not likely that OSR would gain significant importance in study area 5 in the future. On the other hand, study areas 3, 4 and 6, which are today's preferential growing areas for OSR, are also the regions with the lowest potential for GM OSR cropping. There, coexistence without net loss of OSR area is not possible at all, not even with the lowest isolation distance of 400 m. Introduction of GM OSR while maintaining the current OSR proportion of about and above 10% will lead to a net loss of OSR production area in Austria. Assuming moderate isolation distances of 800 m, as currently proposed for Austria, and a non-GM OSR proportion of 10% would result in an area loss ranging from −8.6% to more than −24.6%, depending on the percentage of GM farmers and on the region.

Minimum isolation distances of 400 and 800 m are most probably at or below the minimum requirement for an effective prevention of GM contamination of neighbouring conventional or organic OSR fields and might thus not be sufficient to ensure coexistence. An analysis of cross-fertilisation indicates that relevant outcrossing frequencies can still be detected at distances of at least 1000 m under conditions favouring pollen-flow [42]. However, the specification of coexistence measures is a political issue [18] and does not necessarily fully reflect scientific results.

So far, our considerations are based on the assumption that GM farmers would not grow non-GM OSR within the isolation zone. We excluded intra-farm coexistence from our assumptions since it seems impracticable in Austria. The need for separate storage and separate machinery or high cleaning effort as well as the effort for volunteer control to meet the tolerance threshold of 0.9% renders intra-farm coexistence unfeasible for small farms. Related, but not identically, we assumed GM farmers not to plant non-GM OSR within the isolation zones. This assumption is based on the assumption of a favourable market for GM-free crops where GM-free crops achieve price premiums compared to GM OSR for industrial biofuel production [28,29]. Even when being produced from non-GM seeds, conventional OSR grown within the isolation zone could not be traded with a GM-free premium anymore and would only receive the lower prices of GM OSR without benefitting from any GM yield gains. With significant price penalties for GM-contaminated OSR, GM adopters would switch to other crops within the isolation zone, which leads to a substantial area loss of OSR. If this loss cannot be compensated by GM gains, large scale GM OSR application could substantially reduce domestic production and thus counteract national or European strategies like the EU substitution target.

Conclusions

Our study assumes a coexistence scenario of GM free food production and GM application for industrial use as biofuel feedstock. We analysed different land holding patterns, farm sizes and percentages of crop area, and explored a wide range of isolation distances between GM and non-GM crops and proportions of OSR cropping scenarios in their effect on total OSR production potential. Whereas our analysis focused on OSR as the predominant crop for biofuel production in Austria, the approach can be easily applied to other crops and regions.

We showed quantitatively that different farming, ownership and land use patterns substantially affects the potential total area of conventional-and GM OSR, and that even within a small country the production area potentials may vary considerably between regions. Possible yield gains to be expected from future GM crops would typically be offset by substantial losses of available farmland area for non-GM OSR. We thus conclude that under the current regulatory requirement and crop production conditions, GM OSR application for biofuel feedstock production is not a viable option, neither for Austria nor for countries with similar land holding and land use patterns.