Power Density

by Vaclav Smil

  Sugar cane is a superior feedstock: not only has it much higher yields than corn but because of its endophytic bacteria the most productive varieties do not require any nitrogen fertilizer. The crop's 2010 yield averaged nearly 72 t/ha worldwide and 79 t/ha in Brazil. Assuming that the global mean eventually rises to 80 t/ha and that the harvest would be converted to ethanol with Brazilian efficiency (82 L/t), the power density of cane-based ethanol production would be almost exactly 0.5 W/m2 (6,560 L x 24 MJ/L/31.536 Ms/104). Converting the entire global harvest (1.69 Gt in 2010) to ethanol would yield about 105 GW, merely 3% of the global liquid fuel demand in 2010. Production of the world's 2040 liquid fuel demand (4 TW) would require about 800 Mha of the crop, which would take slightly more than half of the global cropland and about 50% more than all the tropical and subtropical farmland suitable for cane cultivation. And supplying only half of all liquid fuel or replacing only all gasoline would still require 300-400 Mha of cane. The only way to get that much land would be to convert more tropical grasslands and forests to cane fields.

  But ethanol is not suitable for jetliners, and aviation has been one of the fastest-expanding sectors dependent on high-energy-density kerosene (jet fuel). In 2010 the worldwide demand for kerosene reached about 11.4 EJ, or roughly 360 GW (USEIA 2014a); by 2050 it is expected to reach 1 TW (ICAO 2010). Recent production power densities of biojet fuel range from just 0.06 W/m2 for a soybean-based substitute to 0.65 W/m2 for palm oil (Rossillo-Calle et al. 2012). Even the latter alternative would require about 57 Mha of palm oil plantations, 3.5 times their global 2010 area, inevitably leading to a further increase in the tropical deforestation that has accompanied the crop's recent expansion (Kongsager and Reenberg 2012).

  Basing the fuel on soybeans would need about 570 Mha of crop dedicated to biojet fuel, 5.5 times the 2010 global total, obviously a highly unlikely (if not impossible) extension. Turning to crops grown on marginal, nonarable land could provide only a partial solution: the much touted jatropha (Jatropha curcas, a hardy oilseed-bearing shrub or a small tree able to grow on and soils) would not produce more than 0.2 W/m2, and hence the world would need 180 Mha of it to satisfy the 2010 jet fuel demand, and 500 Mha in 2050, an area equal to slightly more than half of China's territory. Even if genetically improved cultivars were to double the yield, the likely jet fuel demand in 2050 would still call for covering roughly an Argentina - or Kazakhstan-sized area with jatropha.

  There is clearly room for increasing the yields of previously uncultivated plants (jatropha, switchgrass), but no yield or conversion gains can change the low power densities of the two leading biofuel crops. Global corn yield rose by 65% in the years between 1980 and 2010 and that of sugar cane increased by 30%; hence, even similar increases between 2010 and 2040 would still leave the power densities of corn-based ethanol lower than the recent densities of the US production (2010 yields averaged 5.2 t/ha worldwide and 9.6 t/ha in the United States). In Asia, where corn yields average only 50% of the US mean, the power density of corn-based ethanol would still be below 0.2 W/m2, while that of cane-based ethanol would remain below 0.7 W/m2. And while Novozymes' new enzymes promise to increase ethanol yield by up to 5% (Novozymes 2013), average American corn yields have recently declined by much more than that: they peaked at 10.3 t/ha in 2009, were at 9.2 t/ha in 2011, and widespread drought depressed them to just 7.7 t/ha in 2012 (FAO 2014).

  Future land claims imposed by high shares of liquid fuels derived from phytomass could be lowered but not massively reduced by producing cellulosic ethanol from crop wastes, from plants grown on marginal land, or from surplus phytomass. The annual production of crop residues rivals that of crops themselves: modern cereals have a residue:grain ratio of roughly 1:1, which means that the global output of straw and stover is about 2.5 Gt, and when residues from other crops are added that total rises to about 4 Gt (Smil 2013a). But crop residues are not wastes waiting to be converted to ethanol: in traditional agricultural societies they are still important sources of cooking fuel and animal feed, and their recycling is a key ingredient of proper agroecosystemic management as straws, stalks, and leaves return to soil nitrogen, phosphorus, and potassium, as well as many micronutrients; renew soil organic matter; help to retain moisture; and prevent wind and water erosion (Smil 1999, 2013a).

  This means that only a carefully assessed fraction of crop residues should be removed. Such a restriction would further lower the inherently low power densities of their production: these are (with a 50% removal rate) less than 0.08 W/m2 for Great Plains winter wheat, about 0.3 W/m2 for highyielding German wheat, and (with a nearly complete removal) less than 0.5 W/m2 for sugar cane bagasse. Roughly a third of stover (the residue of corn, America's largest crop) can be removed in conventional cropping, 70% with no-till cultivation; and a weighted mean of 40% translates to an annual harvest of some 80 Mt in terms of dry weight (Kadam and McMillan 2003).

  That would produce between 20 and 25 GL of ethanol, or just 3% of the US gasoline supply in 2010, and it would imply a final product power density of only 0.06 W/m2. As for any surplus woody phytomass, a review by Smeets and Faaij (2007) offers a corrective quantitative perspective. While they estimated that the world's theoretically available surplus wood (after satisfying the demand for traditional fuelwood and timber) is about 71 EJ (6.1 Gm3), the technical potential is 64 EJ, economic considerations reduce it to 15 EJ, and the inclusion of ecological criteria nearly halves it, to 8 EJ, or to only about 250 GW, an equivalent of less than 2% of the world's 2010 supply of fossil fuels.

  Most important, further adjustments of previously cited power densities are necessary to take into account the energy costs of crop cultivation and fuel production. Dijkman and Benders (2010) calculated the net power densities (they expressed them in GJ/ha/year) for actual bioethanol production from sugar beets and biodiesel production from rapeseed for three specific European cases. The lowest net power density, for Spanish bioethanol, was just 0.02 W/m2; the highest rate, for Dutch biodiesel, was 0.08 W/m2. These low rates mean that the replacement of crude derived liquid fuels in Europe would require (even if the mean value rose to 0.1 W/m2) feedstocks grown on an area of more than 600 Mha, six times larger than all arable land in EU-27 (FAO 2014).

  Similar order-of-magnitude calculations are easy to make for replacing fossil fuels by any kind (liquid or solid) of phytomass fuels. The world's 2012 fossil fuel consumption reached roughly 10.85 Gt of oil equivalent (BP 2014). That is (at 42 GJ/t) 455 EJ, or 14.45 TW, and if we assume that the 2050 demand will be just 40% higher (many forecasts indicate a much larger expansion of demand), then 20 TW are candidates for replacement. If only half of that rate were supplied by biofuels (with the total split between the combustion of woody phytomass to generate electricity and the conversion of phytomass to liquids or gases), then each of these biofuels would have to provide about 5 TW. Even if these conversions could be done with fairly high efficiencies-0.5 W/m2 for woody phytomass and 0.3 W/m2 for field crops-then crops for energy would have to be harvested from nearly 1.7 Gha of arable land, and wood for energy would require annual harvests of trees from 1 Gha.

  But that would mean that in 2050 the area of energy crops would have to be nearly 10% larger than the total area of arable land and permanent plantations in 2012, and the area of continuously harvested tree plantings would be equal to nearly 30% of all of today's closed forests (with canopies covering more than 40% of the ground). This would come on top of the higher future demand for food and wood and in a world of diminishing biodiversity. And yet some uncritical promoters of phytomass energies see no problems with this as they keep conjuring large areas of uncultivated land or assume that more forests and grasslands can be converted to croplands. Read (2008) envisioned having an additional 2.4 Gha of rain-fed arable land (compared to the 2013 total of about 1.55 Gha), mostly in the tropics. Marland and Obersteiner (2008, 335) concluded that "it is not now clear if this vision is a dream or a nightmare"-but I have no problem seeing it as truly nightmarish.

>   We are already harvesting a significant share (close to a fifth) of the biosphere's annual productivity (Smil 2013a), and any future massive increase in biofuel production would have to increase this intervention and would result not only in competition with food and timber production but also in the further weakening of environmental services, particularly those provided by mature forests. And perhaps the greatest irony is that those who would claim to displace fossil fuels by phytomass in order to reduce carbon emissions are apparently unaware that the expansion of land devoted to biofuels can bring the very opposite outcome.

  Fargione and co-workers (2008) showed that converting rain forests, peatlands, and grasslands in order to cultivate food crop-based biofuels in Brazil, Southeast Asia, and the United States could release 17-420 times (one to two orders of magnitude) more CO2 into the atmosphere than the annual greenhouse gas reductions resulting from the displacement of fossil fuels by these cultivated biofuels. Similarly, Searchinger and co-workers (2008) demonstrated that corn-based ethanol does not produce, as previously claimed, substantial CO2 savings; rather, its production nearly doubles CO2 emissions over thirty years and increases greenhouse gases for 167 years, while biofuels from often highly touted switchgrass increase emissions by 50%.

  Many problems would accompany any large-scale phytomass harvesting for energy, among them the further destruction of natural ecosystems, demands for nutrients and water, the extension of monocultures, vulnerability to pests and diseases, and competition with land uses to grow food and feed. Expanded deforestation would be among the most likely consequences of any global-scale push for much increased biofuel production, and yet even without such pressures the world's forests have been in retreat: the best high-resolution global mapping showed that between 2000 and 2012 there was a net forest loss of 1.5 M km2, with rising losses in the tropics (Hansen et al. 2013).

  As I have demonstrated, the inherently low efficiency of phytomass production and subsequent energy losses arising from various fuel conversions limit the power densities of final forms of phytomass-based energy use to a fraction of 1 W/m2. This is the key reason why it is most unlikely that modern societies will soon come full circle and return from fossil fuels to phytomass fuels as the dominant source of their energy, and why any responsibly handled expansion of phytomass production and its least wasteful conversions will be able to make a non-negligible but inherently limited contribution to the world's primary energy supply.

  Metallurgical Charcoal

  As promised in the introductory chapter, I will now assess a surprisingly neglected aspect of energy transition from fossil to renewable sources, the replacement of coke in iron smelting by charcoal from tree plantations. This requires closer looks at modern blast furnaces and at prevailing plantation yields and charcoaling methods. The charcoal supply would have to energize the annual smelting of just a bit over 1 Gt of iron (the average rate for the years 2008-2012). I will use assumptions based on the best plausible case, Brazilian charcoal-based iron smelting, with recent data on eucalyptus plantation yields, charcoaling efficiencies, and blast furnace charges from Sampaio (2005), Swami and co-workers (2009), Pelaez-Samaniego and co-workers (2008), Piketty and co-workers (2009), Pereira and co-workers (2012), and Pfeifer, Sousa, and Silva (2012).

  In 2010 recycled metal accounted for about 25% of the global steel output, and only about 70 Mt came from the direct reduction of concentrated ores using natural gas. Iron smelted in charcoal-fueled blast furnaces has accounted for only about 0.5% of the total (Sampaio 2005), while iron from coke-fueled blast furnace iron reached 1.03 Gt in 2010 and surpassed 1.1 Gt in 2012 (World Steel Association 2013). The world's largest furnaces now have volumes in excess of 5,000 m3: ShougangJing Tang's furnace in Caofedian (blown in 2009) has a volume of 5,500 m3, ThyssenKrupp's Schwelgern 2 (since 1993) has a volume of 5,513 m3, and Japan's Japan's Oita 2 was enlarged to handle a volume of 5,775 m3 in 2004 (Hoffmann 2012; Smil 2008; ThyssenKrupp 2012).

  Each of these furnaces requires more than 2.5 Mt of coal equivalent to energize a daily output of more than 10,000 t of hot metal. I specify coal equivalent rather than coke, because in modern iron-making coke is partially substituted by coal dust, fuel oil, or natural gas blown directly into a furnace, and even by peletized plastic waste. The entire smelting operation has become much more energy-efficient, and specific energy requirements for coke (t coke/t of hot metal) have been steadily declining. The typical consumption of dry coke per tonne of hot metal declined from about 1 t in 1950 to just 0.6 t in 2000, and the best operations needed only about 450 kg of coke/kg of pig iron (de Beer, Worrell, and Blok 1998; Smil 2008).

  At the same time, the global blast-furnace output has been increasing, from less than 50 Mt in 1900 to 580 Mt in 2000, and then, mainly as a result of China's production surge, to 1.035 Gt in 2010, and the global demand for coke has reached record levels: in 2010, 900 Mt of coking coal were converted to about 650 Mt of coke. Hydrocarbons and coal dust directly injected into blast furnaces were equivalent to roughly another 100 Mt of coke, resulting in an annual energy input (30 GJ/t of coke) on the order of 22 EJ of fossil fuels. Metallurgical coke energizes the hightemperature melt (1,300-1,600°C) and acts as the reducing agent: its combustion generates CO2 (C + 02 -4 CO2), whose reduction yields CO (CO2 + C -4 2CO), and that gas reduces oxides into elemental iron (Fe2O3 + 3CO -4 2Fe + 3CO2). Coke also provides support for the heavy charge of iron ore and limestone (added to remove impurities), but it is sufficiently permeable to allow the ascent of reducing gases. In a world run solely by renewable energies, we would have to go back and replace all of these fuels by charcoal made from woody biomass.

  With very similar charcoal and coke energy densities, straightforward replacement of the energy used in primary iron smelting would have called for approximately 750 Mt of charcoal in 2010. But charcoal is a much softer material than coke: depending on the wood species, its compressive strength varies between 10 and 50 kg/cm2, compared to 130-160 kg/cm2 for coke, and hence charcoal could not support the massive iron ore and limestone charges without getting crushed. The desirable bulk density of metallurgical charcoal is at least 0.4 g/cm3, charcoal from eucalyptus has 0.53-0.59 g/cm3 (Pereira et al. 2012), but many wood species yield charcoal densities of only 0.28-0.4 g/cm3.

  That is why Brazilian iron-makers cannot replace all coke by charcoal: economies of scale, competition with foreign producers, and basic technical considerations (charging large furnaces with friable charcoal would cause serious equipment damage) make that impossible (NCIB 2012). As a result, the internal volume of modern charcoal-fueled Brazilian blast furnaces is an order of magnitude smaller than that of the coke-fired units deployed around the world: the biggest Brazilian furnace has just 568 m3 (Pfeifer, Sousa, and Silva 2012). The construction of larger numbers of smaller furnaces (and the ensuing acceptance of higher metal prices) would not change the necessity of harvesting large amounts of woody phytomass. We can calculate the resulting wood demand based on Brazilian commercial practices.

  Most of Brazil's steel comes from modern enterprises that integrate cokebased smelting of pig iron with the production of continuously cast steel, but about a third comes from operations that use charcoal in small blast furnaces concentrated in the states of Para, Minas Gerais, and Mato Grosso do Sul (Uhlig 2011). An increasing share of charcoal is made from eucalyptus, whose plantations covered nearly 5 Mha in 2011 (Pereira et al. 2012)but as much as third of the total charge may be much cheaper wood that comes from the illegal cutting of natural forests (Monteiro 2006). Uhlig (2011) estimated that as much as 15% of the Amazon's deforestation could be ascribed to charcoaling. These shares are uncertain because of significant disparities in nationwide estimates of the country's charcoal production (Ghilardi and Steierer 2011).

  About 80% of the fuel is made in small brick-and-mud semicircular kilns commonly called rabo quente, hot-tail (fig. 8.1). These 2.5-m-high beehive structures, usually grouped in rows by dozens, are stacked with air-dried wood, set alight, an
d let to smolder for five to seven days; three days after they are extinguished, men enter the kilns and unload the charcoal. The working conditions in many of these charcoaling operations have been described as akin to slave labor that is also highly hazardous (Greenpeace 2013). Workers removing charcoal from ovens are exposed to high temperatures, dust, and smoke; there is long-term exposure to uncontrolled emissions of nitrogen and sulfur oxides, benzene, methanol, phenols, naphthalene, and polycyclic aromatic hydrocarbons (Kato et al. 2005). In mass terms these small kilns convert only between 22% and 27% of wood into charcoal, and the average nationwide rate is no more than 25% (Bailis et al. 2013; Swami et al. 2009).

  Figure 8.1

  Brazilian eucalyptus plantation. International Forestry Resources and Institutions (IFRI 2012).

  An average blast furnace charge of 450 kg C/kg of pig iron could be supplied by about 500 kg of coke or 630 kg of charcoal (Sampaio 2005). The minimum theoretical requirement would thus be about 650 Mt of charcoal, which means that the global rate of charcoal production would have to increase at least 14-fold compared to the year 2010, when the FAO put its output at 47 Mt (FAO 2014). With an average charcoaling efficiency of 25%, the global iron smelting at the 2010 rate of 1.035 Gt would annually consume 2.6 Gt of wood. Another (official) Brazilian source gives average rates of 2.2 m3 of charcoal and 4.4 m3 of wood per tonne of pig iron (Secretaria de Estado de Meio Ambiente do Para 2008). With a mean eucalyptus wood density of 0.55 t/m3, that translates into 2.42 t of wood per tonne of pig iron and 2.5 Gt of roundwood that would be needed to smelt the world's 2010 iron production.

  The real total would have to be higher because of the transportation and handling losses that would inevitably arise during the export of tropical charcoal to temperate climates, particularly the fuel made from less dense woods than eucalyptus clones. An alternative would be to transport entire logs and set up massive charcoaling facilities in high-income countries, but that would require an unprecedented trade in wood. In 2010, all wood traded globally amounted to about 170 Mt (FAO 2014); exporting most of the wood needed to make charcoal would mean an order of magnitude increase in shipments.