In our model, feedstock prices are exogenous variables and therefore independ­ent from production scale. This assumption is based on the fact that transportation costs are the main driver for raw material prices and the costs per unit increase with the scale of a plant as transport routes become longer. The rationale behind this assumption is that each company aims to operate at the optimal production scale in the light of (a) the tension between scale benefits and (b) increasing cost of capital associated with transportation costs. Cost advantages driven by learning

and scale are significant endogenous parameters and therefore main determinants in our calculation model.

For each type of biofuel, we assume learning-based cost-reduction potentials which diminish over time applied to all process steps as defined in Fig. 3. In regard to second-generation biofuels, we estimate learning curve effects of 40, 30 and 20 % for the corresponding time frames 2005-2010, 2010-2015 and 2015-2020. This in return leads to progress coefficients of 60, 70 and 80 %, respectively. For first-generation bioethanol, progress coefficients of 70, 80 and 85 % were estimated. In autoregressive time series models, these progress coefficients are sequentially multiplied with previous values to derive operational and total production costs for specific points in time. Based on our scale size estimates for the different types of biofuel (see Fig. 2), scale effects were incorporated into biofuel conversion costs depending on the output of biofuel product (10, 50, 100, 250 and 250 kilotonnes per year). Table 4 is one example for the application of our assumptions in order to calculate conversion costs. It represents the results for first-generation bioethanol.

1.1.2 Total Production Costs

Depending on the type of raw material, different numbers of litres of biofuel can be produced from one tonne of feedstock. A conversion factor was implemented in order to translate prices for one tonne of raw material into the prices per litre of produced biofuel. Production costs were calculated as the sum of raw material costs and conversion costs. For a better comparison, energy density factors (in Millijoule per litre, MJ/L) were taken into account and normalised to the average energy den­sity of fossil fuel. The results were adjusted production costs, based on the specific density of biofuels. Reference scenarios were calculated for 2015 and 2020. This model enables the calculation for different production scales in place and planned or hypothetical scales (e. g. simulation of not yet realised production scales).

As previously mentioned, the price of fossil fuel is the decisive factor for biofuel market success. Therefore, it is essential that biofuel production costs can compete

Table 4 Modelling of conversion costs for first-generation ethanol Scale Investment Depreciation Operational costs Total costs = operational expenses plus depreciation 2005 2010 2015 2020 2005 2010 2015 2020 (kt> (ml) (m Euro) (m (Cent/1) (m Euro) (Cent/1) (m Euro) (Cent/1) (m (Cent/1) (m Euro) (Cent/1) (m Euro) (Cent/1) (m Euro) (Cent/1) (m Euro) (Cent/1) (m (Cent/1) Euro/ Euro) Euro) year) Process step 1: enzymatic hydrolysis Learning curve 0.70 effect 2005-2010 Learning curve 0.80 effect 2010-2015 Learning curve 0.85 effect 2015-2010 10 13 10 0.50 3.95 6.00 47.40 4.20 33.18 3.36 26.54 2.86 22.56 6.50 51.35 4.70 37.13 3.86 30.49 3.36 26.51 50 63 35 1.75 2.77 12.00 18.96 8.40 13.27 6.72 10.62 5.71 9.02 13.75 21.73 10.15 16.04 8.47 13.38 7.46 11.79 100 127 50 2.50 1.98 18.00 14.22 12.60 9.95 10.08 7.96 8.57 6.77 20.50 16.20 15.10 11.93 12.58 9.94 11.07 8.74 250 316 75 3.75 1.19 24.00 7.58 16.80 5.31 13.44 4.25 11.42 3.61 27.75 8.77 20.55 6.49 17.19 5.43 15.17 4.79 500 633 100 5.00 0.79 30.00 4.74 21.00 3.32 16.80 2.65 14.28 2.26 35.00 5.53 26.00 4.11 21.80 3.44 19.28 3.05 Process step 2: fermentation Learning curve 0.70 effect 2005-2010 Learning curve 0.80 effect 2010-2015 Learning curve 0.85 effect 2015-2010 10 13 15 0.75 5.93 9.00 71.10 6.30 49.77 5.04 39.82 4.28 33.84 9.75 77.03 7.05 55.70 5.79 45.74 5.03 39.77 50 63 53 2.63 4.15 18.00 28.44 12.60 19.91 10.08 15.93 8.57 13.54 20.63 32.59 15.23 24.06 12.71 20.07 11.19 17.68 100 127 75 3.75 2.96 27.00 21.33 18.90 14.93 15.12 11.94 12.85 10.15 30.75 24.29 22.65 17.89 18.87 14.91 16.60 13.12 250 316 113 5.63 1.78 36.00 11.38 25.20 7.96 20.16 6.37 17.14 5.41 41.63 13.15 30.83 9.74 25.79 8.15 22.76 7.19 500 633 150 7.50 1.19 45.00 7.11 31.50 4.98 25.20 3.98 21.42 3.38 52.50 8.30 39.00 6.16 32.70 5.17 28.92 4.57

effect

2005-2010

Learning curve 0.80

effect 2010-2015

Learning curve 0.85

effect 2015-2010

with those of fossil fuels. This was the main focus for our comparative analysis. To compare production costs, historical prices for raw materials were extrapolated in the course of reference scenarios of the fossil fuel price. The identification of eco­nomically promising biofuel technologies was then enabled through modelling of projections for technological advancements in respect to production scale and learn­ing effects. In other words, our approach enables the comparison of different biofu­els’ production costs while considering the specific development state, economies of scale in context of realistic scenarios for the market prices for biomass. Plausibility checks based on current data as well as consistency of the results across production technologies enhanced the accuracy of the results. At the same time, we assessed the comparability of data and performed corresponding adjustments if necessary.