Secondary hardwood forests are managed with a number of different methods. Careful single tree selection cutting (that avoids “high-grading” or the selective removal of only the most valuable trees) is considered one of the most sustainable methods because it maintains species composition, forest biomass, and C sequestration over time [29]. No fertilizers or irrigation are typically applied to the forest and very little management is required between harvest events. Specialized equipment and skilled labor are required to extract biomass at multiple entry points. In forestlands of the eastern US, this management is most common in the northeastern region (primarily in Maine, New Hampshire, Vermont, Massachusetts, Connecticut, Rhode Island, New York, Pennsylvania, and West Virginia).

There are active and abandoned pine plantations, many dominated by Pinus taeda (loblolly pine), across much of the southern piedmont and coastal plain [30]. These forests could be either reclaimed if abandoned or diverted from pulp and paper products to biomass production if active. Irrigation and fertilizer are sometimes used in this intensive management system that typically includes complete harvests every 12–20 y.

Short-rotation woody crops are often comprised of poplar or willow species and can provide a consistent yield in a shorter turnaround time than pine plantations (3−13 y) [16]. Fertilizer and sometimes irrigation are applied with this management practice and all aboveground biomass is removed at harvest. In the eastern US, this management is currently most common in the Great Lakes region (primarily in Michigan, Minnesota, and Wisconsin).

Woody residues appropriate for bioenergy applications are created in a variety of operations. We defined forest residues to include logging slash (some bark, tree tops, limbs and leaves, etc.), thinning and tree removals associated with logging, clearing and fire treatments, or any other forest removals that are not directly associated with roundwood harvest [31]. Primary mill residues are the bark and wood remainders resulting from mill processing of stem wood into products such as boards or paper, and include sawdust, trimmings, cores, and pulp screenings [32]. Secondary mill residues are remainders following the use of wood in construction, manufacturing and retail (lumber yards). Urban residues include municipal solid waste (pallets, yard waste, chipped wood), utility trimmings, private tree company trimmings, and construction and demolition wood [33].

Low intensity biomass management practices in forests (partial harvest of site, e.g., single-tree selection, diameter limit cutting or patch cutting of a landscape) do not require inputs of fertilizer or irrigation that are characteristic of more intensive biomass production systems. This is one reason that best management practices (BMPs), which are regulated at the state level and usually voluntary, point to lower intensity harvesting as a means to preserve water quality and other ecosystem services (e.g., [27]. The mean biomass removal for low-intensity diameter-limit cutting in mixed hardwood forests with mature sawtimber is ~17% of the total aboveground biomass [14,26,29]. Although the harvest interval usually varies, 15-year intervals between low intensity harvests are common [15,29].

Previous studies that evaluate partial harvests relative to other managements generally conclude that increasing harvest intensity has a negative effect on forest C sequestration [14,15,18,51]. These studies assume site-level comparisons where a partially harvested site, for example, is compared to a site that was entirely clear-cut (e.g., [14,15]). These are informative for understanding harvest impacts, but must be evaluated in a slightly different context when considering landscape level processes. For example, at a landscape level, the impact of clear-cutting 17% of the forestland relative to denuding the entire landscape may be comparable to the difference estimated at the site-level between partial harvests and clear-cuts. Impacts of patch clear-cutting in larger woodsheds (the area that supplies a commercial wood mill) may be similar to partial harvests at the site-level, but patch clear-cutting requires long-term landscape planning to maintain aboveground biomass, C and community structure. Analysis of partial harvesting at the site level serves only as a proxy for landscape dynamics.

It is also evident that net C storage depends on the end-use of biomass that is harvested from the forest [15,18,52]. It has been estimated that biomass use for bioenergy results in greater CO₂ emissions than the use of biomass for wood products [52], mostly because wood products have a longer residence time than biomass that is immediately combusted for energy or heat. Thus, economic conditions and C residence times of harvested products can affect the sustainability of partial harvests from mature secondary forests [15].

We simulated the partial harvest of a mature mixed hardwood forest at 15 y intervals using a version of the PnET model [35,36,38,39], PnET-CN_sat [37] which has been used to simulate different harvesting practices in the Fernow Experimental Forest, WV [14]. The simulated forest is assumed to be unmanaged (with no harvest events) for 100 y prior to harvest simulations, with mixed age classes and high species diversity that is typical of temperate hardwood forests. We estimated annual wood production and net C sequestration assuming partial harvests with removal rates of 10%, 17%, 34%, or 51% of aboveground biomass. Only the removal rates varied, not the intervals between harvests. This allowed us to isolate the effect of increasing harvest intensity (although an interaction between harvest interval and intensity does exist). Net C sequestration included net biomass and soil accumulation as well as C emitted as a result of the final end use as bioenergy.

Modeled wood production in response to partial harvesting was maintained over a 60-y simulation period, with periodic fluctuations in response to harvest and climate variation (Figure 1a). Net C storage in forest biomass and soil increased over the sixty year period with 10%, 17%, or 34% biomass removal events, but the managed ecosystem only remained a net sink of C throughout the entire 60-y simulation period in the case of 10% and 17% removals (Figure 1b). Removals of 34% biomass resulted in a gradual gain of C over the long-term, but was still a net source of C to the atmosphere after 60 y. In contrast, removals of 51% caused the forest to be a net source of C to the atmosphere throughout most of the simulation period (Figure 1b).

Model simulations were parameterized for a forest in West Virginia, near the southern boundary of the northeast region, but the forest biomass estimated is comparable to that of forests in other parts of the northeast. The average biomass simulated before harvest was ~100 Mg ha⁻¹. Excluding Maine, which has a much lower mean merchantable biomass of 53 Mg ha⁻¹ as a result of extensive management, the mean biomass reported for the other northeastern states is ~101 Mg ha⁻¹, ranging from 91 Mg ha⁻¹ in New Hampshire to 115 Mg ha⁻¹ in Connecticut, which has limited forest management [26]. The level of active management of forests throughout the northeast varies widely, but responses to harvest that were estimated for West Virginia should apply to mixed hardwood forests across the region.

Fertilizers are often used in more intensively managed forestry production, i.e., plantations and short-rotation woody cropping systems. Irrigation is used infrequently (usually only at the time of planting) but sometimes supplements precipitation during periods of drought. We analyzed results from 24 forested sites measured in 13 independent studies of loblolly pine plantations in the southeastern US [53,54,55,56,57,58,59,60,61,62,63,64] to determine the effect of water (including both precipitation and irrigation), nitrogen (N), and phosphorus (P) inputs on biomass production. Using a 3-way ANOVA, we tested the effects of water, N, P, and the interactive effects of these variables on annual wood growth of loblolly pine plantations.

Based on literature, we found that annual wood growth was significantly affected by the interaction of phosphorus and water inputs (p = 0.0320) as well as the interaction of phosphorus and nitrogen inputs (p = 0.0253). The interactive effect of all three variables was not significant (p = 0.1976). If each variable was considered separately, the main effects of water and P, while very small, were both significant (p_water = 0.0396, p_P = 0.0345; Figure 2) but the effect of N alone was not (p = 0.1137; Figure 2). Irrigation in managed loblolly pine plantations accounted for only 14% of total water inputs on average. Thus, P appears to be the most important input to loblolly pine plantations for increased growth, although the effect of P depends on N additions (p_pxn = 0.0253). The range of P and N that was applied in the studies analyzed here was 8–112 kgP ha⁻¹ y⁻¹ and 36–195 kgN ha⁻¹ y⁻¹, respectively. This range describes experimental fertilizer applications and does not necessarily characterize the most common practices in pine plantations.

We simulated annual wood production and net C stored in pine plantations that were completely harvested at either 12 or 20 y intervals using the PnET-II version of the PnET model. PnET-II has been used to simulate pine production in the southeastern US [30,43]. The simulation assumed no manipulations of water, N, or P. Wood production was maintained at a mean of ~7 Mg ha⁻¹ y⁻¹ during three harvest cycles that were simulated over 60 y (data not shown), but wood production declined to a mean of ~4 Mg ha⁻¹ y⁻¹ in response to 12-y harvest cycles over the 60 y period (Figure 3a). This loss could be offset by nutrient additions, which, if optimized, could stimulate long-term production by up to 32% (Figure 2c and Figure 3a). Without this nutrient supplementation, net C storage (including net biomass and soil accumulation as well as C emitted as a result of the final end use as bioenergy) declined over the 60 y with a 12 y harvest rotation (Figure 3b) as well as with a 20-y harvest rotation. In the near-term, the pine plantation was a sink of C, but the long-term perspective indicates that the system will become a net source of C to the atmosphere if harvests continue at regular intervals.

Noteworthy is that this scenario of a clear-cut plantation assumed that the same site was repeatedly and completely harvested at each interval. This should be distinguished from landscape-level patch cutting that uses clear-cuts at this frequency on only a small portion of the land while the remainder and majority is left un-harvested. In this landscape design, an individual land parcel can be un-harvested for ≥80 y between clearing events.

Fast growing broadleaf trees such as Populus spp. (poplar) and Salix spp. (willow) and inter-specific hybrids, are the most common species considered as short-rotation woody crops [65,66,67]. In this system, known as short-rotation coppice, monocultures of these species or hybrids are planted, cut to the ground after 3–20 y, and then allowed to re-grow. Yields following coppice are typically much greater than in the first cycle, since re-growth is fueled from reserves in the root system. When considering areas that intersect with eastern US forestland, short-rotation woody crops are most common in the Great Lakes region. Because most data for this system have been collected outside of the eastern US, this section of our review first summarizes general observations of large scale field trials conducted in temperate zones of Europe and the Pacific Northwest, summarizing the main factors impacting yield. We then focus on the potential production and sustainability of short rotation woody crops specifically in the Great Lakes Region.

A review of historic data (available in the Biofuel Ecophysiological Traits and Yield Database (BETY-db [68]), collected from 11 countries for poplar, and 5 countries for willow, indicated a wide range in yields of poplar and willow. Averaged globally, poplar and willow have a mean annual growth of 7.1 (n = 663, sd = 4.7) and 7.3 (n = 349, sd = 4.5) Mg ha⁻¹ yr⁻¹, respectively. In the US, similar yields of 7.7 (n = 277, sd = 4.5) and 9.4 (n = 68, sd = 3.8) Mg ha⁻¹ yr⁻¹ have been achieved for poplar and willow, respectively. The annual growth of poplar varied greatly across different climates, soils and also depended on the choice of clones and initial planting densities. In the northwestern US, annual growth ranged from 11.4 to 24.3 Mg ha⁻¹ yr⁻¹ in 4-year-old poplar stands under different planting conditions [46,69,70]. The yield of 4 y old hybrid poplars in Puyallup, WA was 35 Mg ha⁻¹ [70]. In the lower Midwest USA, poplar clones yielded up to 70 Mg ha⁻¹ over 5 years (14 Mg ha⁻¹ y⁻¹ ) without irrigation and fertilization [71]. There has been considerable recent effort to breed for sustainable higher yielding varieties of both poplars and willows, and a continued improvement in yield for this system should be expected [72].

Although there are fewer examples of poplar and willow crops in the eastern US relative to other parts of the world, they are considered a potentially important biomass crop for the region [73]. In the Great Lakes Region, annual growth of poplar ranged from 3.5 to 12.8 Mg ha⁻¹ y⁻¹ [45,47]. Many poplar stands in the Great Lakes region are harvested at an older age (9 y) relative to those discussed above. Field trials of willow plantations have mostly been conducted in the UK to date, where the average annual growth is 8.1 Mg ha⁻¹ y⁻¹ [74]. There was however a willow trial in Tully, NY, USA where the annual growth was 16.3 Mg ha⁻¹ y⁻¹ for a willow clone (SV1) in its fifth growing season [75]. Over a three-year period in the Great Lakes region, another willow clone (Sx-61) has an annual growth of 12.7 Mg ha⁻¹ y⁻¹ [76]. New hybrids of willow are currently being developed in the US for bioenergy and are expected to have even greater yields (e.g., [77]).

Based on literature, we found that the annual growth of poplar and willow was positively correlated with annual precipitation and was not affected by N fertilization (Figure 4). Thus, water availability appeared to be the key factor impacting production of poplar and willow crops. The range of N that was applied in the studies analyzed here was 50–336 kgN ha⁻¹ y⁻¹ for poplar and 10–336 kgN ha⁻¹ y⁻¹ for willow. Although there is no evidence that N has a significant effect on biomass production, N does reduce the time required to reach maximum production. Further, it seems unlikely that 50 Mg removals of biomass every 5 y could be sustained without N additions. Kopp et al. [75] found that, with N additions of 336 kgN ha⁻¹ y⁻¹, the cycle from cutting to harvest was reduced by one year in both poplar and willow. However, it is not clear that such extreme fertilizer additions are necessary to achieve this gain.

We simulated annual wood production and net C sequestration of poplar plantations in Mondovi, WI, USA, that were completely harvested at either 8 or 4 y intervals using the ED model [40,41,42]. ED has been validated against poplar production in multiple field sites in the Great Lake Region and also used to predict the annual growth of poplar for the 48 contiguous states of the US [44]. The simulation assumed no manipulations of water, N, or P. Annual wood production in the first harvesting cycle was comparable to the observed mean yield of two poplar clones collected in a field trial established in Mondovi, WI, USA [47]. If grown in 8 y cycles, the model simulations of net C storage (including net biomass and soil accumulation) were negative for the first 20 y and recovery occurred after about 3 harvesting cycles (Figure 5a). With a 4 y harvest cycle, the net C storage continued to decline over the 60 y period (Figure 5b). In the longer harvesting cycle, the poplar plantation was a net sink of C after about 25 y, but in the shorter harvesting cycle, the poplar plantation was a net source of C to the atmosphere for the entire period of time tested. Carbon benefits will likely be increased with more recently bred high-productivity lines [72,77].

In bioenergy management, there are four kinds of sustainability that must be considered: climate change mitigation, environmental sustainability, economic sustainability, and energy security. Inevitably there are tradeoffs that do not allow all four aspects of sustainability to be maximized simultaneously, but we will briefly discuss how our results relate to each one. Results from this study directly inform objectives related to climate change mitigation and environmental sustainability (in C storage, 1 unit C = 3.67 units CO₂). Analysis of harvest costs and legislation are required to thoroughly address economic and energy security sustainability. While such analysis is beyond the scope of this study, we will provide brief examples of how economic and policy instruments can affect the sustainability of biomass production in the forested region of the Eastern US.

Although terrestrial C sequestration can be a large sink, climate change mitigation requires more holistic strategies that are not limited in scope to simple manipulations of C. Nitrogen and P management, for example, can have important implications for not only production and C sequestration but also for GHG emissions associated with manufacturing fertilizer. Application of fertilizers could boost C sequestration in the plantation system for example (Figure 3), but can also lead to large GHG fluxes. Based on the rates of N applied to plantations and cropping systems reviewed here (36–195 kgN ha⁻¹ y⁻¹), the additional GHG emission associated with N fertilizer manufacturing would be between 31 and 291 kg CO_2eq ha⁻¹ [89,90]. Based on the range of P applications analyzed here (8–112 kgP ha⁻¹ y⁻¹ ), the additional GHG flux associated with P fertilizer manufacturing would be between 5 and 66 kg CO_2eq ha⁻¹ [90]. Over a 60-year period, the combined application of N and P could result in an emission that is up to 14 Mg CO_2eq ha⁻¹ greater than a management system with no input requirements (i.e., partial harvest). This GHG emission would be in addition to the terrestrial C fluxes we have described (e.g., Figure 3 and Figure 5) and reported in Table 2. This estimate does not include emissions of N₂O that would occur at the time of fertilizer application, another significant source of GHG in managed landscapes.

In addition to C sequestration and GHG mitigation, environmental sustainability also depends on maintaining or enhancing biodiversity, ecosystem function, and water quality. Low intensity partial harvests, at a site-level or landscape-level, by definition allow the retention of aboveground biomass (e.g., 83% biomass remains standing after a 17% harvest). Therefore, properly designed harvest regimes may maintain the inherent diversity of forest plant communities that mediate biogeochemical cycling and provide habitat for vertebrate and invertebrate wildlife of multiple trophic levels. Nationwide, BMPs often invoke low intensity partial harvests rather than intensive plantations or even-age management at the landscape level to accomplish environmental benefits, primarily maintenance of water quality [27]. However, landscape-level patch cutting on a small percentage of land while maintaining a mature forest across the majority of a landscape may be sustainable because this practice limits the number of entry points for machinery that can lead to increased soil disturbance and incur additional costs. Such management requires long-term planning if the goal is to maintain a continuous biomass supply.

Since long-term economic sustainability depends on the biological potential of a managed forest or plantation system, it is important to weigh short-term economic gains against both the biological and economic potential of a forest in the long-term. To do this, economic analyses should distinguish total standing biomass from incremental biomass changes in a forest to reconcile short-term goals with long-term sustainability. Even though a lower amount of biomass per unit area is harvested in low-intensity partial harvests than from clear-cut plantations, the sustainability of production and the value per unit product are greater than high-intensity timbering (Figure 7). The harvest of many small diameter stems is more expensive per unit mass than the harvest of larger stems that yield the same biomass in fewer stems [91]. If one assumes a landscape level approach to management, small areas of the landscape might be harvested intensively but large-diameter trees can be cultivated by staggering the location and increasing the interval of harvest at a single location to 80+ years.

Increasing interest in bioenergy is expanding markets for forest materials that have historically had little commercial value, such as small-diameter trees and logging residues [17,21,81]. Some of the forest-derived material that could be used for bioenergy already is being extracted from the forest (e.g., living or dead trees removed for fuel reduction or forest health); bioenergetic utilization of this feedstock is limited primarily by transportation costs and processing infrastructure [81]. To maintain ecological integrity and ensure sustainability in eastern forests, some proportion of the available biomass in any given stand should remain on site, and this residual material should consist of multiple biomass categories (e.g., large dead wood, small-diameter trees, logging residues). However, precise quantities are mostly unknown and are likely to be highly site-specific [20,21]. Thus we recommend careful consideration of the tradeoffs between economic benefits and local ecology when managing residues.

The estimated 176 Tg y⁻¹ of biomass that could be produced relatively sustainably from woody feedstocks in the eastern US is enough to generate ~16 billion gallons of ethanol annually. This is equivalent to 4% of liquid fuel consumption in the US in 2009 [92]. The degree to which energy security objectives can be met by harvesting wood resources in the eastern US depends on the amount of land that is available to provide biomass. Abandoned lands in the eastern US total anywhere from 10.8 to 20.4 Mha depending on the degree to which abandonment of crops was offset by transitions to pasture or vice versa. For each scenario that was discussed here, we can only roughly estimate the potential biomass production from abandoned lands.

A payback time is defined as the time it takes to recuperate a lost pool of C plus the loss in the continued stimulation of C sequestration that may have occurred by displacing the baseline condition [93,94]. If a new management practice is to be introduced for biomass production, then the initial conditions of forestlands must be considered. The initial conditions define the trajectory of C storage and allow for a more accurate estimate of payback times. All of the scenarios modeled in the results were analyzed without accounting for differences in land use practices prior to agricultural abandonment. The exception to this is the partial harvesting scenario, which assumes a mature closed-canopy forest prior to time zero.

Payback times for an initial disturbance or land use change are handled differently in different disciplines [95,96]. One reason for disagreement about the method for calculating payback times associated with bioenergy agriculture is that the conversion efficiency of the bioenergy products (energy per unit C) is not equal to the conversion efficiency of the fossil fuel technology that is being replaced. Displacement of fossil fuels can reduce or eliminate payback times [97]. With a displacement of 4% of transportation fuels in the US, the payback would be reduced by 25 Tg C annually (emissions that would no longer be released from fossil fuels). Among the harvesting practices reviewed in this study, pine plantations are the most likely to result in C emissions. With pine plantations on abandoned lands in the southeast included as a source of biomass in the eastern US, we estimated that C would be emitted to the atmosphere at a rate of 1.7 Tg of C y⁻¹. Although we conclude this is unsustainable relative to alternative practices, it should be noted that emissions from intensive forest management for biomass are an order of magnitude less than the emissions currently produced from fossil fuel use.

Our analysis describes only the terrestrial fluxes that would be inputs to a life-cycle analysis of bioenergy production from wood resources. There are many other components that ultimately must be considered when evaluating the production chain associated with each scenario [95]. Upstream of resource production, one would consider land use change, machinery, land preparation, fertilizer and equipment manufacturing. Downstream, one must consider co-products, fuel conversion processes, transportation, distribution, and infrastructure costs. The focus of this study was limited to terrestrial ecosystem fluxes of GHG.