Idaho forest growth response to post‐thinning energy biomass removal and complementary soil amendments

Utilization of woody biomass for biofuel can help meet the need for renewable energy production. However, there is a concern biomass removal will deplete soil nutrients, having short‐ and long‐term effects on tree growth. This study aimed to develop short‐term indicators to assess the impacts of the first three years after small‐diameter woody biomass removal on forest productivity to establish optimal biomass retention levels for mixed‐conifer forests in the Inland Northwest region, and to evaluate the ability of soil amendments to compensate for potential adverse effects from biomass removal. We examined impacts of four biomass retention‐level treatments at two study locations: full biomass removal (0x), full biomass retention (1x), double biomass retention (2x), and unthinned control. We combined biomass retention with four soil amendment treatments: biochar (B), fertilizer (F), fertilizer and biochar combined (FB), and an untreated control (C). We considered treatment effects on basal area and total stem volume growth for all trees per plot (plot trees) and for the six largest trees per plot (crop trees). Biomass removal had no effect on plot (P > 0.40) or crop tree growth (P > 0.65) compared to normal biomass retention. High biomass retention (2x) decreased plot tree growth as compared to normal biomass retention (1x) levels (P < 0.05) after three years. This growth difference was not explained by soil moisture, temperature, or nutrient uptake. While there were strong tree growth differences between study locations, patterns of biomass and amendment treatment responses did not differ. Fertilizer increased basal area growth and total volume growth (P < 0.10) as expected, because nitrogen is limiting in the region. Biochar had no effect on tree growth (P > 0.47). Initial findings after three years suggest removing small‐diameter biomass for biofuel feedstocks is feasible in the Inland Northwest without negative impacts on tree growth.


Introduction
There has been a global demand to reduce reliance on fossil fuels and a growing interest in using woody biomass for energy production. The Energy Independence and Security Act (2007) requires the United States to replace 36 billion gallons a year of fossil fuels with biofuels including 16 billion gallons coming from cellulosic feedstocks (Schnepf & Yacobucci, 2015). Small-diameter woody biomass thinning operations produce abundant forest residues which can be used for bioenergy. There are many benefits to using forest biomass for biofuel, as it can reduce the risk of wildfire and disease (Schroeder, 2008;Evans & Finkral, 2009) and provides a source of income as a secondary commercial product. However, little is known of the impacts removal operations will have on soil properties and short-and long-term forest productivity.
Forest harvest impact research has focused on comparing whole-tree to bole-only harvest removal. Studies have found whole-tree harvests can reduce soil nutrient availability (Johnson & Todd, 1998;Saarsalmi et al., 2010), but others found no impact on nitrogen (N) or site productivity (Voldseth et al., 2011;Slesak et al., 2017). Powers et al. (2005) found that biomass removals during clear-cut harvesting result in declines in soil carbon, but not in forest growth through 10 years. The impacts of nutrient removal from whole-tree harvesting are site specific and depend on such factors as nutrient capital, soil type, parent material, climate and moisture regime, forest type, and harvest technique (Slesak et al., 2017). The impacts of small-diameter biomass removal from forest sites during precommercial thinning operations are not as well understood as harvest removal impacts, because few studies examine the removal of biomass from thinning operations. Helmisaari et al. (2011) found lower 10-and 20-year volume growth increments from whole-tree harvesting after thinning compared to conventional harvesting after thinning, finding a negative correlation between amount of residue harvested and volume growth. However, these stands were thinned at 44 years. Precommercial thinning typically fells biomass when stands are relatively young, about 10-15 years for Douglas-fir stands (Reukema, 1975). Precommercial thinning is used to reduce tree competition for resources such as light, water, and soil nutrients and usually results in higher nutrient and water availability (Brockley, 2005;Chase et al., 2016). While it is difficult to make generalizations that can be applied in every forest type, biomass harvesting from nutrient limited sites will have a proportionally greater impact than harvesting from nutrient abundant sites (Thiffault et al., 2011). Small-diameter biomass removal will likely impact site conditions less than whole-tree harvesting (Kabrick et al., 2013;Klockow et al., 2014) as smaller quantities of biomass will be removed, but soil impacts will depend on the harvest method (e.g., skidder, forwarder), how much slash remains after harvesting, soil conditions, and climatic regime (Page-Dumroese et al., 2010).
Thinning activities in the western United States are estimated to generate 155 Mt of woody biomass that can be used for bioenergy (Rummer et al., 2005). Often during harvesting, whole trees are moved to a central logging deck where they are delimbed and detopped (Polagye et al., 2007). However, depending on the age of the stand, trees can be left where they are felled or piled and burned for wildfire mitigation (Kalabokidis & Omi, 1998). An alternative to burning woody biomass after thinning is to use the residues for bioenergy production (U.S. Department of Energy, 2011). Removing biomass from forests and using it as a biofuel source provides a secondary source of income that also reduces the risk of wildfire and disease (Schroeder, 2008;Evans & Finkral, 2009).
Thinning is a common silvicultural treatment in many areas of the world and can be accompanied by the addition of fertilizer (Crane, 1982;Donald, 1987;Valinger, 1992). Nitrogen is considered to be the main limiting nutrient in western forest soils (Edmonds et al., 1989;Coleman et al., 2014) and biomass removal can decrease nitrogen pools (Elliot et al., 1998). Fertilizing thinned stands has the potential to increase tree production (Footen et al., 2009), and may compensate for loss of nutrients from biomass removal (Helmisaari et al., 2011). In the northwestern United States, Douglas-fir (Pseudotsuga menziesii (Mirb.) Franco) forests typically respond to nitrogen fertilization for up to ten years (Coleman et al., 2014) and one study found nitrogen fertilization of Douglas-fir plantations increased site productivity for 15-22 years in low-quality sites (Footen et al., 2009). But growth benefits of fertilization may not be initially detectable.
In addition to fertilizer increasing post-thinned stand productivity, the application of biochar may also mitigate the effects of biomass removal. Depending on the feedstock and pyrolysis method, biochar can be fine-grained and highly porous with high water-holding capacity and nutrient retention (Laird, 2008;Page-Dumroese et al., 2017). On forest sites, biochar can increase ground cover after disturbance, increase soil moisture, and decrease soil compaction associated with harvesting (Page-Dumroese et al., 2017). Biochar can also decrease greenhouse gas emissions (Laird, 2008). The carbon in biochar is resistant to decomposition when added as a soil amendment (Amonette et al., 2009), so the benefits are likely long term.
In order to understand the impacts of small-diameter biomass removal on residual forest growth, there is a need for regional, site-specific studies (Smith et al., 2000). Our goal was to evaluate the impacts of various levels of biomass removal on forests in the Inland Northwest while maintaining or enhancing site productivity. The objectives were to develop short-term indicators of the impact on forest productivity of smalldiameter woody biomass removal, establish ideal biomass retention levels for the region, and evaluate the ability of soil amendments to compensate for potential adverse effects from biomass removal, such as decreased growth. We hypothesized tree growth would be higher on plots with retained biomass compared to complete biomass removal. Any differences in growth in sites with biomass removal would be mitigated by fertilizer and biochar amendments, as fertilizer would replace nitrogen lost, and biochar would replace carbon lost and increase water availability during dry summers.

Site characteristics
Two northern Idaho study locations, Pitwood (46.983104, À116.483943) and the University of Idaho Experimental Forest (UIEF) (46.849512, À116.845068), were chosen in different forest types (Fig. 1). The 1981-2010 mean annual temperature at Pitwood was 6.6°C, while at UIEF it was 7.8°C. Mean annual precipitation at Pitwood was 106 cm, while at UIEF it was 74 cm (PRISM Climate Group). Both sites were regenerated mixed-conifer forests and had predominately silt loam soils that contain volcanic ash. disjunctum (Rupr.) Ching), and ladyfern (Athyrium filix-femina (L.) Roth). Its elevation ranges from approximately 990 to 1060 meters above sea level. The site area has a slope of 20-60% with a northern aspect. The soil is deep, well drained, and formed from material weathered from fine-grained quartzite with a thick mantle of volcanic ash. It is classified as ashy over loamy, amorphic over isotic, frigid Typic Udivetrand (Flewsie soil series; National Cooperative Soil Survey 2013). Pitwood has been managed as a commercial forest with a rotation period of approximately 40 years, with the last harvest occurring in 1992. It was broadcast burned in 1993 and planted in 1994 with Douglas-fir. There is considerable natural regeneration of associated species (Table 1).
These soils are deep loess with a small amount of volcanic ash in the surface. The parent material is granite bedrock. The UIEF was used for grazing cattle, recreation, and research, and managed for commercial forestry. This stand was harvested in the late 1980s as a seed-tree harvest, leaving ponderosa pine for forest regeneration, with a harvest rotation period of 45-50 years. Seed trees were harvested in 2012. Pretreatment site characteristics including stand composition are noted in Table 1.

Biomass treatments
Two replicate stands, each approximately 6-10 ha in size, were selected at each location. To install the study plots, each stand was divided into four biomass treatment whole plots, including three thinned treatments with varying biomass retention levels and one unthinned control. The three biomass retention plots at each stand were thinned in spring 2013. Plots were thinned to approximately 5 m spacing and~400 trees per hectare (Table 1). Stands were thinned from below, removing the smaller and less desirable trees. Douglas-fir was retained at Pitwood, with primarily western redcedar being removed. Ponderosa pine was retained at the UIEF upper stand after the thinning, while a mix of species was retained in the lower stand (Table 1). The thinned treatment whole plots were divided into three biomass retention levels: full biomass removal (0x), full biomass retention (1x), and double biomass retention (2x). In the 2x treatment, all thinning residues were retained and all the residues from 0x plots were added. The four whole-plot treatments are represented in Fig. 2.

Soil amendments
Soil amendment treatments were randomly assigned within each whole plot to one of four 40 9 40 m split plots (Fig. 2). As a result, there were a total of 64 study plots (2 locations 9 2 replicate stands 9 4 thinning treatments 9 4 amendments). A 20 9 20 m measurement plot was assigned in the middle of each plot, leaving a 10-m treatment buffer surrounding each plot. The amendment treatments were fertilizer (F), biochar (B), fertilizer and biochar (FB), and a control with no amendments (C). Biochar was obtained from Evergreen Forest Products (Tamarack, ID, USA) and was produced from mixed-conifer mill residue at 890°C. Biochar was applied to the top of the O horizons and added in the amount of 2.5 Mg ha À1 to each B and FB plot. It was hydraulically sprayed on UIEF plots in the upper stand in October 2013 and the lower stand in April 2014.
It was applied manually at Pitwood in May 2014. Nitrogen fertilizer (urea) was applied at a rate of approximately 630 kg ha À1 to the F and FB plots in November 2013 at UIEF and May 2014 at Pitwood.

Biomass retention
To quantify the amount of woody debris left in the biomass retention treatments, total aboveground standing biomass was calculated from tree diameters using an allometric equation specific to each species group (Jenkins et al., 2003) for the pretreatment and post-treatment stands ( Table 2). The difference between post-treatment aboveground total biomass and pretreatment standing biomass was an estimate of the amount cut in tons per hectare in each plot, and the amount deposited Table 1 Pre-and post-treatment site characteristics of stands at two experimental locations. Pretreatment characteristics were measured in 2012, and post-treatment characteristics were based on plots that were thinned and measured in 2013 TPA, trees per acre; QMD, quadratic mean diameter; BA, basal area; SDI, standard density index (Reineke, 1933); RD, Curtis's relative density (Curtis, 1982); DF, Douglas-fir; GF, grand fir; WH, western hemlock; RC, western redcedar; LP, lodgepole pine; PP, ponderosa pine; WL, western larch. for each of the 1x plots. To estimate the debris left in the 2x plots, we added the amount cut to the average amount cut in the 0x plots within the same replicate stand. The 0x plots were assumed to have 0 tons per hectare of biomass added, as the cut debris was manually moved to the 2x plots.
We estimated the amount of nitrogen in the DWD. The approximate proportions of foliage, branches, bole wood, and bark of the cut trees were estimated from whole-tree ratios (Ares et al., 2007). The biomass of each component was multiplied by their respective nitrogen concentration (Ares et al., 2007) and summed to calculate the DWD nitrogen content in each plot. Nitrogen content in each plot was averaged for each biomass treatment at each location (Table 2).

Tree growth
We measured trees in three to six randomly selected 20 9 20 m plots within each replicate stand at the end of 2012 growing season. Residual trees within all 20 9 20 m plots were measured in the fall of 2013 through fall 2016. Diameter at breast height (DBH) for all trees greater than 2.5 cm DBH was measured with a D-tape, and species were identified and tagged with a unique tree ID at all UIEF plots and at all thinned plots at Pitwood. In the unthinned plots at Pitwood, which contained high stem densities, all trees with a DBH of 13 cm or larger were measured in the 20 9 20 m plot and all trees with a diameter of 2.5 cm or larger were measured within a 5 9 5 m subplot. Tree heights were measured annually at UIEF with TruPulse laser rangefinder (Laser Technology Inc, Centennial, CO, USA). Tree heights at Pitwood were measured at least once in each plot. In any one year, tree heights were measured in an average of two plots in each of the four biomass treatments. Missing tree heights for volume calculations were estimated using allometric equations specific to tree species (Wykoff et al., 1982) and validated against measured tree heights.
The six largest live trees in each plot were designated as crop trees in 2016. Tree growth was calculated as basal area growth (m 2 ha À1 ) and total volume growth (m 3 ha À1 ) for each of the 64 plots each year from 2013 to 2016 at the plot level and for crop trees. Plot and crop tree total volume was calculated from tree diameters and heights using species-specific allometric equations (Browne, 1962).

Foliar nutrient analysis
Three branches were collected from one to three trees of the dominant species in each plot. For each branch, 20 ponderosa pine needles and 100 Douglas-fir needles were removed for analysis. Needles were weighed and analyzed for nitrogen concentration using Kjeldahl digestion analysis at Harris Lab (Lincoln, NE, USA). Relative concentrations and content were calculated using the unthinned and unamended control plots as the baseline for each species.

Soil moisture and temperature
Soil moisture and temperature were measured seasonally at the 15 cm soil depth with a soil moisture probe (HydroSense II, Campbell Scientific, Logan, UT, USA) and a soil temperature probe (6000-09TC Soil Probe Thermocouple, Li-Cor, Lincoln, NE, USA) in three random locations in each split plot every nonwinter season from spring 2014 through fall 2016, typically in April, July, and October. Additionally, one or more Em5b data loggers (Decagon Devices, Pullman, WA, USA) were installed in biomass treatment blocks. Soil moisture and temperature readings were recorded continuously at 4-h intervals throughout each day at 15 cm using MAS-14-20 moisture and RT-1 temperature sensors (Decagon Devices). Soil temperature and moisture averages were calculated from spot measurements, and were checked for accuracy using the data logger measurements.

Statistical analysis
An analysis of covariance (ANCOVA) was performed for both the plot level and crop tree basal area growth and total stem volume growth as the response variables. Each growth year increment and the three-year periodic annual increment (PAI) for basal area growth were analyzed separately as response variables. Initial basal area as measured in 2013 was used as covariate using the 'aov' function in the 'stats' package in R (R Development Core Team 2015). Each model included three factors with varying levels including location (n = 2), biomass retention/thinning treatment (n = 4), and soil amendment (n = 4), as well as interactions between factors. If interactions were not significant, they were eliminated from the model. The same model was used for the needle nitrogen concentration, without a covariate.
A mixed-effects linear model with an initial BA covariate and a random factor of split plot within whole plot for each stand was completed using the 'lme' function in the 'nlme' package (Pinheiro et al., 2014) to evaluate differences within factors and had comparable results of significance to the ANCO-VA. The random factor did not improve the model, so we used the ANCOVA for the final analysis. Growth means were calculated using the least squared estimates derived in the 'lsmeans' package (Lenth, 2016).
A regression analysis was used to calculate growth response associated with the amount of residual downed woody debris. We used a quadratic equation y = x + x 2 + error, where y is basal area growth and x is downed woody debris in tons per hectare. This was completed using the 'lm' function in the 'stats' package (R Development Core Team 2015).
retention/thinning treatments and soil amendment treatments using the 'TukeyHSD' function in the 'stats' package (R Development Core Team 2015). Differences within factors were assumed to be significant at a = 0.1. Seasonal soil moisture and soil temperature differences between whole-plot biomass treatments and split-plot soil amendment treatments were analyzed using an analysis of variance via the 'aov' function in the 'stats' package (R Development Core Team 2015) with moisture or temperature as the response variable and location (n = 2), biomass treatment (n = 4), soil amendment treatment (n = 4), and season (n = 3) as factors, as well as interactions.

Downed wood estimates and stand growth
The amount of biomass felled during thinning operations differed between locations and in turn affected residual tree growth. Pitwood had 181% more downed woody debris in the 1x plots and approximately 119% more downed debris in the 2x plots (Table 2) compared to UIEF. We used the downed woody debris estimates to calculate the quadratic growth relationship between downed woody debris and basal area growth at the plot level and crop tree level in the thinned treatments (Figs S1 and S2). The regression results were not significant (P > 0.1), but stand growth was affected by biomass treatment as a categorical variable in the ANCOVA (Table 3). Basal area and volume growth were also sensitive to study location, soil amendment, and initial basal area at both the plot level and crop tree level (Table 3). However, stand growth response to biomass treatments did not vary between locations and were consistent at all levels of amendment treatments (i.e., no two-or three-way interactions).

Stand growth response to location
Pitwood basal area growth was greater than UIEF at the plot level and at the crop tree level for all growth increments (Table 4). The plot-level basal area periodic annual increment (PAI) at Pitwood was 107% greater than at UIEF. Crop tree PAI basal area growth at Pitwood was 44% greater than UIEF (Table 4). Three-year PAI plot-level growth response to initial basal area depended on location (I 9 L interaction) and was associated with a steeper response to initial basal area at Pitwood than at UIEF because larger Pitwood trees had greater basal area growth compared to smaller trees at Pitwood, while the growth rate of UIEF of smaller trees and larger trees was more consistent (Fig. 3).
The plot-level PAI volume growth at Pitwood was 119% greater than UIEF and Pitwood crop tree PAI volume growth was 68% greater than UIEF (Table 5). As with basal area, three-year volume PAI response to initial basal area depended on location (I 9 L interaction) at the plot level (Table 3) due to a steeper response to initial basal area at Pitwood than at UIEF.

Basal area response to biomass treatment
Response of basal area to biomass treatments depends on mortality as well as tree growth. Significant mortality did not occur in the thinned plots, but did occur in the unthinned plots, especially during the 2016 growing season, where the unthinned plot basal area showed a decrease in growth rate (Fig. 4a). The basal area of the six largest living trees in each plot represents the crop tree basal area in each year (Fig. 4b).
The slope of the lines between each year shows relative growth among treatments was higher in thinned plots (Fig. 4).
Plot tree growth was affected by biomass treatment starting at the second year of growth (Table 3). The basal area PAI in 1x was 30% greater than 2x, and 44% greater than unthinned plots (Fig. 5d). Growth in the 1x was not significantly greater than 0x in any year increment (Fig. 5a-c), or for the PAI growth (Fig. 5d). A regression analysis for tree growth by coarse woody residue loads as a continuous variable did not find any significant correlations (Figs S1 and S2).
Biomass treatments contributed to significant changes in crop tree growth each year after thinning (Table 3; Fig. 6a-c). Most of the differences were due to diminished growth in unthinned plots. Basal area PAI in 1x was 50% greater than in unthinned (Fig. 6d), while 0x growth was 44% greater than unthinned (Fig. 6d).

Volume growth response to biomass treatment
Plot-level tree volume was affected by biomass treatment starting the first year after thinning ( Table 3). The PAI volume growth in the 1x treatments was 45% greater than the 2x treatment growth, and 52% greater than the unthinned plots (Fig. 7d). Volume growth in the 1x plots was not significantly greater than the 0x treatment plots in any year (Fig. 7a-c).
Crop tree volume growth was affected by biomass treatment beginning the first year after thinning (Table 3; Fig. 8a-c). In the third year, crop tree volume growth in the 1x treatments was 32% greater than the 2x treatment growth, and 63% greater than the unthinned plots (Fig. 8c).

Basal area response to soil amendments
Soil amendments did not affect plot basal area growth during the first two years post-thin (Table 3; Fig. 9a, b), Table 3 Analysis of covariance results for plot basal area and volume growth. Results include F-statistic (F) and P-values (P) for first three years after thinning and the periodic annual increment (PAI) at the plot level and for crop trees. Initial basal area (basal area in 2013) was used as a covariate. Three-year PAI is the difference in basal area for the three years from 2013 to 2016. Only significant interactions were included in the final models for each growth variable but showed a fertilizer effect after the third year (Fig. 9c). Fertilizer-only (F) plots had 39% higher basal area PAI compared to unamended control (C) plots ( Fig. 9d). Biochar did not affect basal area growth within the plots in any year. Soil amendments did not affect crop tree basal area in the first year (Table 3). After three years, crop trees in F plots had 23% greater PAI basal area growth compared to the unamended plots, and trees in FB increased basal area growth 27% as compared to the control (Fig. 10d). Biochar did not diminish or improve residual tree growth in any year increment compared to the unamended control plots. Standard errors are shown in parentheses. Same letters within each time period and measurement indicate no differences between locations at a = 0.1. Fig. 3 Average plot-level basal area growth over the three years after thinning by initial basal area and location. Location and initial basal area interaction is significant (P = 0.03).  Volume growth response to soil amendments Volume growth responded to fertilizer at the plot level and crop tree level. None of the soil amendments affected tree growth during the first growing season after thinning (Table 3). At the plot level, fertilized plots had 19% (F) and 25% (FB) greater PAI volume growth than the control plots after three years (Fig. 11d). Crop trees in FB plots had 28% greater PAI growth compared to unamended control plots (Fig. 12d). Biochar-only plots did not diminish or improve plot or crop tree basal area growth compared to the unamended control plots.

Foliar nitrogen
Relative foliar nitrogen concentrations increased after nitrogen fertilization, but the response was not consistent between locations (Fig. 13). The F and FB plots had equal foliar nitrogen concentrations at Pitwood, but FB plots had higher concentrations than F plots at UIEF.
There were no differences in foliar nitrogen between biomass treatments, or any interactions (P > 0.05).

Soil moisture and temperature
Neither the level of biomass nor soil amendment affected soil moisture or temperature at the 15 cm   depth. There were highly significant seasonal and locational effects on soil moisture and soil temperature. UIEF was warmer in the spring and summer and cooler in the fall compared to Pitwood. Generally, in spring soil temperature was mild (10.8 AE 0.3°C) with high moisture (22.9 AE 0.4), summer was hot (15.3 AE 0.4°C) and dry (4.4 AE 0.2 %VMC), while fall was cool (9.1 AE 0.11°C) with moderate moisture (14.7 AE 0.3). No seasonal differences in soil moisture or temperature were found between biomass treatments or soil amendment treatments (Figs S3-S6).

Response to biomass retention level and thinning
The biomass treatments provided an opportunity to compare different woody debris retention levels with the goal of determining best management practices for leaving enough woody biomass to sustain site productivity. During the three years following thinning, removing all of the thinned biomass from the forest did not significantly affect basal area or total volume growth compared to the slash retention plots at both the plot level and the crop tree level. This indicates that over the relatively short study period, these two   sites may be resilient to biomass removal during precommercial thinning, and the results support the use of small-diameter biomass for biofuel feedstocks. In contrast, some long-term studies have shown negative effects of biomass removal that were not apparent during the first five years (Voldseth et al., 2011), or not measured during the first 5 years (Helmisaari et al., 2011) so it is possible that long-term impacts may be different than our three-year results indicate. Other long-term studies also show few or no impacts of whole-tree removal (Slesak et al., 2017), which removes greater amounts of carbon and nutrients than precommercial thinning, so it is likely that smaller diameter biomass removal will not impact forest growth on these sites. The impacts of biomass removal during thinning in combination with an intensive whole-tree harvest remain an open question. Previous harvests occurred at both locations in the late 1980s and early 1990s, but quantities of residual biomass retained are unknown. Site preparation practices during this time often involved broadcast burning. Remnants of decaying, charred logs were visible at both locations. Inputs of dead wood from these harvest operations undoubtedly contribute nutrients to soil (Chen & Xu, 2005;Klockow et al., 2013). The lack of impact of biomass removal on growth might not be replicated on a site that had been whole-tree harvested prior to planting or naturally regenerated (Johnson & Todd, 1998;Saarsalmi et al., 2010).
We expected to see an increase of tree growth with increased slash loading. The reduced basal area and volume growth in the 2x plots compared to other thinned plots with less biomass retention was unexpected. Decreased tree growth associated with greater slash loadings has not been widely shown, but possible explanations include an increased risk of pests and disease, slash influencing soil moisture and temperature, and nutrient immobilization. Greater slash retention increases the risk of pests and disease (Fettig et al., 2007;Schroeder, 2008). While it is possible insect damage or disease occurred, significant mortality was not observed in thinned plots. Greater biomass loads have been shown to increase soil moisture and decrease summer soil temperature following a clear-cut harvest (Devine & Harrington, 2007;Harrington et al., 2013), which could affect residual tree growth, and has actually been shown to increase Douglas-fir growth in young stands (Harrington et al., 2013), yet our study did not show significant moisture and temperature differences between the biomass treatments at either site (Figs S3-S6). Although we may have seen a different moisture and temperature response had we measured vertical spatial variation within treatments, we chose to focus the limited number of ports per data loggers on horizontal variation at a consistent depth within measured plots. Nutrient immobilization is another possible explanation for the decrease in growth in 2x plots. Wood decay fungi may have decreased nutrients available to tree roots by translocating and accumulating soil nitrogen in the woody debris. The presence of woody residue has been shown to decrease nitrogen and phosphorus leaching in some soils (Carlyle et al., 1998). However, other studies showed increased logging debris retention increased nutrient leaching (Strahm et al., 2005;Wall, 2008) and also increased total soil N, but not foliar N in Douglasfir stands (Slesak et al., 2010). Our study did not find significant biomass impacts on foliar nitrogen.
We estimated from calculations of nitrogen concentrations in tree components (Ares et al., 2007) that an average of 44 kg ha À1 of nitrogen was present in the logging residue of the 1x treatments, and 188 kg ha À1 of N was present in the 2x treatments at the time of harvest and residue deposit. Nitrogen demand is expected to increase with tree growth throughout the stand development. However, our study only looked at the first three years postharvest, during that time the proportion of nitrogen released from the woody debris may be relatively small compared to long-term decomposition effects. In Douglas-fir forests, woody debris has been shown to both immobilize and release nitrogen in the first two years postharvest depending on factors such as soil temperature and residue mass (Piatek, 2007). The increased mass of debris in the 2x plots may have contributed to a decrease in nitrogen mineralization during our study period. If we had reduced the size of the woody debris in the slashed treatments, we may have seen a faster response. Mulched residue can increase nitrogen availability by 32% after three to five years, and provide added benefit of fuel reduction (Rhoades et al., 2012).
We could not find studies where slash was doubled to evaluate aboveground responses, or studies with slash loading greater than 70 tons per hectare to compare with our 2x treatment. Slash levels at Pitwood averaged 158 tons per hectare. Decreased growth may be due to the excessive slash level, which was due to the overstocked pretreatment conditions and the unique 2x slash treatment. Average 2x slash levels at UIEF were around 72 tons per hectare, which is high compared with other reports. Decreased growth response in 2x plots was consistent between locations, which makes the underperforming 2x growth of greater interest, as Pitwood 2x plots had 119% more slash retention in tons per hectare than UIEF 2x plots.
The results of this study suggest large amounts of slash can decrease tree growth at these two sites. In addition, removing all thinning biomass residue on the soil surface did not change tree growth as compared to retaining it. Assessing an ideal biomass retention level for this region was an objective of this study. There was a peak slash level of~30 tons per hectare of slash at UIEF and~60 tons per hectare at Pitwood for growth response using a second-order polynomial. Unfortunately, the regression results were not significant, possibly due to wide variation in slash levels, the accuracy of our method of quantifying downed debris, and limited replicates. The trend of our regression model suggests that an ideal biomass retention level is between zero and more than the full biomass retention level. This indicates best management practices in Inland Northwest forests may include full biomass removal or retention as convenient during thinning operations and, where necessary, at a level that limits a fuel source for wildfire. It should be noted our study sites had very productive soils, and therefore, our results may not transfer to low-productivity sites.
There are advantages to using thinned forest stands to study biomass removal impacts. Empirical field trials with small-diameter thinned stands give results sooner than clear-cut studies that monitor subsequent seedling growth. Significant growth deficiencies in seedlings established after timber harvesting may take decades to detect in the western United States, while thinned stands can show responses within one to five growing seasons. Retained trees in thinned stands quickly occupy available growing space, and respond sensitively to changes in site resource availability (Binkley & Reid, 1984;Grady & Hart, 2006;Roberts & Harrington, 2008). Our study used this time advantage to quickly obtain information on the growth responses to biomass retention, which is most relevant to forest bioenergy production systems.

Response to soil amendments
This study supports previous studies that found fertilizing forest soil increases forest productivity, especially in N-limited environments. Because N is limited in most forest soils in the Inland Northwest, fertilization was expected to increase tree growth compared to the control (Coleman et al., 2014). Fertilized plots showed increased plot and crop tree growth compared to the unfertilized control during the study. While biomass removal did not result in decreased growth in the first three years of the study, it is reasonable to assume fertilizer would be beneficial to Inland Northwest forest productivity if longer-term research finds biomass removal impacts forest growth later in the rotation cycle. Fertilizer has been shown to mitigated impacts of biomass removal in other regions (Powers et al., 2005).
Biochar had no negative impacts on tree growth, indicating it can be used for carbon sequestration without negatively affecting aboveground processes. Biochar is highly variable and agricultural studies often find different growth responses (Spokas et al., 2011) ranging from increased growth (Graber et al., 2010;Hossain et al., 2010;Carter et al., 2013), diminished growth (Asai et al., 2009;Hammond et al., 2013), or no response (Schulz et al., 2014). The lack of biochar affect may be due to the existing volcanic ash content in the soils. Andisols and andic property soils found on both study sites produce highly productive forests when compared to other mineral soil orders (Meurisse et al., 1991). The porous quality of volcanic ash-influenced soil increases water-holding capacity which can increase tree and understory productivity (Meurisse et al., 1991). The increased water-holding capacity in ash-influenced soils is an important factor for forest productivity in the Inland Northwest as the region experiences warm, dry summers with a prolonged summer drought (McDaniel et al., 2005). Biochar amended soil may not increase tree growth because without biochar, these soils already support a high level of biomass production associated with adequate soil moisture retention during dry seasons. We expected biochar to alter soil water-holding capacity and result in higher seasonal soil moisture (Page-Dumroese et al., 2017), but we could not detect any differences in soil moisture (Figs S5 and S6). This may be due to the low quantity of biochar added, the nature of the soil, or other factors not yet identified. Studies have examined various levels of biochar applications in forestry, ranging from 2 Mg ha À1 to 47 Mg ha À1 (Thomas & Gale, 2015;Page-Dumroese et al., 2017). This study applied 2.5 Mg ha À1 of biochar to plots, while other biochar field studies with significant findings applied 7.8 Mg ha À1 (Ma et al., 2016), 10 Mg ha À1 (Brantley et al., 2015), and 47-105 Mg ha À1 (Ippolito et al., 2015). It is possible that increasing the rate of biochar would have resulted in increased impacts, yet application at higher rates, which were done in agricultural contexts, may be impractical in forestry practices using current technology.

Management implications
Forest management objectives aim to provide commercial products along with other ecosystem services including biodiversity, fire reduction, pest control, wildlife habitat, clean water, and climate change mitigation (Thompson et al., 2011). The finding that removing all of the thinning residues did not impact forest growth in the first three years after thinning indicates that the studied soils and forest types may be resilient to this type of harvesting and could be a source of bioenergy feedstocks without impacting forest productivity. Full biomass removal for biofuel production from precommercial thinning would provide a secondary monetary crop and will decrease fossil fuel consumption, potentially aiding in climate change mitigation. Another benefit of biomass removal for this fire-prone region is decreased wildfire risk. Alternatively, while leaving thinned residue did not provide growth benefits within the first three years, there may be longer-term impacts because woody debris decomposition can provide a slow release of nutrients or add to the organic matter pool (Fahey et al., 1991). Leaving some woody debris decreases erosion and protects seedlings from exposure, thereby increasing survival rates (Jurgensen et al., 1997). It also increases biodiversity by providing structure and forage for wildlife and fungi (Franklin et al., 2002;Berch et al., 2011). Forest managers will need to evaluate the benefits and detriments of biomass removal when managing for cellulose biofuel production. We found that adding nitrogen fertilizer to these nitrogen-limiting forests increased forest growth. If long-term impacts of biomass removal are shown later, fertilization may be an important practice when managing forests for biomass fuel. A follow-up study is recommended for these measurement plots to assess the long-term impacts of small-diameter biomass removal, as well as similar studies in other soil and forest types to more fully understand impacts of biomass removal and the global capacity for cellulose biofuel systems.

Supporting Information
Additional Supporting Information may be found online in the supporting information tab for this article: Figure S1. PAI basal area growth response to downed woody debris by location with fitted quadratic curve. Figure S2. PAI crop tree basal area growth response to downed woody debris by location with fitted quadratic curve. Figure S3. Seasonal temperature differences by biomass treatment. Figure S4. Seasonal temperature differences by soil amendment. Figure S5. Seasonal moisture differences by biomass treatment. Figure S6. Seasonal moisture differences by soil amendment.