Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Large-scale forest inventories of the United States and China reveal positive effects of biodiversity on productivity

Large-scale forest inventories of the United States and China reveal positive effects of... Background: With the loss of species worldwide due to anthropogenic factors, especially in forested ecosystems, it has become more urgent than ever to understand the biodiversity-ecosystem functioning relationship (BEFR). BEFR research in forested ecosystems is very limited and thus studies that incorporate greater geographic coverage and structural complexity are needed. Methods: We compiled ground-measured data from approx. one half million forest inventory sample plots across the contiguous United States, Alaska, and northeastern China to map tree species richness, forest stocking, and productivity at a continental scale. Based on these data, we investigated the relationship between forest productivity and tree species diversity, using a multiple regression analysis and a non-parametric approach to account for spatial autocorrelation. Results: In general, forests in the eastern United States consisted of more tree species than any other regions in the country. The highest forest stocking values over the entire study area were concentrated in the western United States and Central Appalachia. Overall, 96.4 % of sample plots (477,281) showed a significant positive effect of species richness on site productivity, and only 3.6 % (17,349) had an insignificant or negative effect. Conclusions: The large number of ground-measured plots, as well as the magnitude of geographic scale, rendered overwhelming evidence in support of a positive BEFR. This empirical evidence provides insights to forest management and biological conservation across different types of forested ecosystems. Forest timber productivity may be impaired by the loss of species in forests, and biological conservation, due to its potential benefits on maintaining species richness and productivity, can have profound impacts on the functioning and services of forested ecosystems. Keywords: Tree species diversity; Forest management; Biological conservation; Continental map of forest diversity; Spatial autocorrelation; Bootstrap Background become more urgent than ever to understand the BEFR Over the past two decades, there has been an extensive (Symstad et al. 2003). discussion (see Cardinale et al. 2012; Naeem et al. 2012 There is increasing evidence that supports a positive and references therein) over the biodiversity-ecosystem BEFR, which indicates the loss of biodiversity affects the functioning relationship (BEFR). The loss of biodiver- functioning of an ecosystem (Loreau et al. 2001; Hooper sity can greatly alter the characteristics and functioning et al. 2005; Cardinale et al. 2012; Naeem et al. 2012; Liang of an ecosystem, including its productivity (Liang et al. et al. 2015). There are several mechanisms which are 2015) and services (Hooper et al. 2005). With the loss thought to be the basis for a positive BEFR. Historically, it of species worldwide due to anthropogenic factors, cli- is thought that niche complementarity (Loreau et al. 2001; matic disturbance, altered disturbance regimes, and Cardinale et al. 2011; Reich et al. 2012), which constitutes biological invasions, etc. (Fleming et al. 2011), it has niche differentiation and species facilitation, as well as the sampling effects (Hooper and Vitousek 1997; Wardle * Correspondence: alpenbering@gmail.com 1 1999), i.e., the chance that a forest contains a more pro- Davis College of Agriculture, Natural Resources and Design, West Virginia ductive species increases with increasing species diversity, University, 26506 Morgantown, WV, USA Full list of author information is available at the end of the article © 2015 Watson et al. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. Watson et al. Forest Ecosystems (2015) 2:22 Page 2 of 16 are reasons for a positive BEFR. Recently, Liang et al. ground measurements, remote sensing data, and spatial (2015) developed a theoretical model to integrate comple- estimation. As variance is ignored within a raster, such mentarity and a new mechanism of diminishing marginal maps have limited value to BEFR studies. Maps showing productivity in quantifying the effects of biodiversity loss distribution of biodiversity, consisting of point data are on plant productivity. the most valuable, because spatially explicit biodiversity Most existing evidence of a positive BEFR comes from records can be matched with the forest productivity data grassland experiments at limited spatial scales (Loreau et al. at the same locations for the study of BEFR. Unfortu- 2001; Symstad et al. 2003; Cardinale et al. 2012). Since these nately, due to the cost of ground measurements, large- experiments usually do not include some key processes in- scale maps of forest biodiversity are extremely limited herent to natural ecosystems, such as the introduction of across the world (Köble and Seufert 2001) and to the best new species, decomposition of woody plant material, large of our knowledge, there is no map of forest biodiversity at standing crop, and the utilization of water by woody tree the national scale for the United States. species, it is difficult to extrapolate the results to greater The primary objective of this study was to investigate temporal and spatial scales (Symstad et al. 2003). the effects of biodiversity on forest timber productivity, BEFR studies in forested ecosystems are in general using extensive forest inventory data from the United limited to empirical analysis of observed data (Liang States of America and China. The secondary objective was et al. 2007, 2015; Paquette and Messier 2011; Zhang to synthesize these ground-measured data in mapping tree et al. 2012), primarily due to the complexity of forested species richness, forest stocking, and productivity over all ecosystems and the length of time it takes to derive esti- types of temperate forests across the United States. mates of growth and productivity. Although some argue otherwise (Chen et al. 2003; Vilà et al. 2003), the major- Methods ity of studies confirm a positive relationship between Data tree diversity and ecosystem productivity and services Data for this study came from two sources— Forest (Kelty 1989; Caspersen and Pacala 2001; Liang et al. Inventory and Analysis of the United States of America 2005, 2007, 2015; Lei et al. 2009b; Paquette and Messier (FIA, see Woudenberg et al. 2011) and the Forest Manage- 2011; Young et al. 2011; Zhang et al. 2012; Gamfeldt ment Planning Inventory (FMPI) from the Wangqing For- et al. 2013). BEFR also plays an important role in the estry Bureau in the Jilin Province of China (He et al. 2013). management of forest resources. A traditional view in silviculture is that clearcutting with artificial regeneration FIA (even-aged monoculture) optimizes forest productivity The FIA databases are a product of the United States (e.g., Assmann 1970), but it has been found that this Department of Agriculture (USDA), Forest Service FIA maxim does not generalize (Hasse and Ek 1981; Haight Program, which has established a network of permanent and Monserud 1990) and that the mixed-species stands sample plots across the United States to determine the could have higher long-term productivity (e.g., Haight and extent and status of the nation’s forests in regard to the Monserud 1990; Buongiorno et al. 1995; Liang et al. 2005, condition, volume, growth, and depletions (Woudenberg 2006). Most biodiversity studies in forested ecosystems et al. 2011). The FIA ground-based inventories have been have primarily used even-aged designs that lack structure conducted at different time periods by states, ranging complexity of natural uneven-aged forests (Leuschner from the 1960s to 2012. Following the passage of the 1998 et al. 2009). Thus, studies that incorporate greater geo- Farm Bill (Gillespie 1999), the FIA program in the new mil- graphic coverage and structural complexity are much lennium switched from a periodic inventory system to an needed for analyzing the BEFR in forested ecosystems. annual system in which a portion of the FIA plots in each Biodiversity studies of forested ecosystems, especially at state would be remeasured each year (Gillespie 1999). larger geographic scales, are not only of great value to The FIA program uses a 3-phase sampling scheme BEFR research, but also of strategic importance to the (Woudenberg et al. 2011). Phase 1 uses stratification to world’s energy security and economic development. Tree aggregate ground samples into groups to minimize vari- species provide an essential source of energy and financial ance using stratified estimation. Every FIA ground plot is income worldwide, especially for rural areas where liveli- assigned to a stratum of which a weight is based on its hoods depend heavily on forest resources (FAO 2012). proportion within the estimation unit. Phase 2 consists of The limited maps of biodiversity across the world’sfor- actual FIA ground plots, which follow a national standard ested ecosystems were developed using either raster data and are fixed-radius plots 0.40 ha in size. To protect the or point data. Biodiversity maps of raster data (Ricketts privacy of landowners, geographic coordinates of FIA 1999; Hernandez-Stefanoni and Ponce-Hernandez 2006; ground plots are “fuzzed” (Lister et al. 2002). The true plot Hernández-Stefanoni and Dupuy 2007; Kreft and Jetz locations are known to be within 1.61 km of the fuzzed 2007; Liang 2012) are mostly based on a combination of values under the periodic system, and 0.80 km under the Watson et al. Forest Ecosystems (2015) 2:22 Page 3 of 16 current plot design that was first used in the 1990s for the area is situated on the middle lower hill region of the last of the periodic inventories and all of the current an- Changbai Mountains in northeastern China. Elevation nual inventories (Woudenberg et al. 2011). The ramifica- ranges from 550 to 1,100 m A.S.L. with an annual rain- tions of using the fuzzed coordinates instead of real ones fall of 547 mm. The mean annual temperature is 3.9 °C. are provided in the discussion. Mean monthly maximum and minimum temperatures In the current FIA plot design, each permanent sample range between 22 °C and −37.5 °C, respectively. The study plot consists of four 0.02-ha subplots 7.32 m in radius. area was originally dominated by the mixed broad-leaved Subplots 2, 3, and 4 are situated around subplot 1 in a tri- Korean pine (Pinus koraiensis) forest type, with dark angular pattern, with a 36.58-m distance between the plot brown forest soil. Most primary forests, however, have centers of subplots 2, 3, and 4 and the center of subplot 1 been altered into other forest types such as spruce-fir dom- (Bechtold and Scott 2005). On a subplot, all trees greater inated mixed coniferous forests, birch-aspen mixed broad- than or equal to 12.70 cm in diameter at breast height leaved forests, or plantations consisting of larch, spruce, fir, (dbh) are measured. A 2.07-m radius microplot is located and pine after forestry practices and other disturbances. inside of each of the 4 subplots, in which all trees with a The purpose of FMPI is to assess forest resources and dbh smaller than 12.70 cm are tagged and measured. to supply information requirements for forest manage- FIA inventories are designed in a way such that state ment planning, spatial and functional patterns, and over- level sampling errors are met at the 67 % confidence limit all design of forestry at the FMU level (Lei et al. 2009a). (Woudenberg et al. 2011). Sampling error is kept ≤ 3 % for Sample plots, systematically designed with a 1 km × every 404,686 ha of timberland. It should be noted that 2 km grid, were measured with a 10-year interval. In despite the mandated sampling error to the area, the sam- total there are 1,389 plots distributed in the forest bur- pling errors applied to removals, volume, and total annual eau (Fig. 2). The 0.06-km plot is rectangular. In each growth are targeted by the FIA program, which is 5 % for plot, species and diameter of trees with a dbh above 7 3 every 3 × 10 m of growing stock on timberland for the 5 cm were recorded. Three to five average trees were se- 7 3 eastern United States and 10 % for every 3 × 10 m for lected to measure age and height for computing stand the western United States. A total of 475,892 FIA perman- age and height. Other factors measured are slope, aspect, ent sample plots distributed across the contiguous United elevation, soil type, soil depth, and management history. States and Alaska (Fig. 1) were used in this study. Forest mapping FMPI To study the forest productivity, stocking, and species The FMPI data were collected from permanent sample diversity across the United States, state and regional FIA plots at the forest management unit (FMU) level in the databases, obtained from different FIA regional offices, Wangqing Forestry Bureau (43°05’–43°40’ N, 123°56’– were compiled together into one nationwide master 131°04’ E) of China in 2007 (He et al. 2013). The study database (hereafter, master database). Using SQL queries Fig. 1 Forest types across the 48 contiguous U.S. states and Alaska, derived from both a forest type group map of the contiguous United States and a forest type group map of Alaska. With two sub-types within the Douglas-fir type (coastal Douglas-fir and interior Douglas-fir), we studied a total of 16 forest types across the United States. GCS_WGS_1984 projection for the main map and Alaska Albers Equal Area Conic projection for the inset Watson et al. Forest Ecosystems (2015) 2:22 Page 4 of 16 Fig. 2 Geographic distribution of the 1,389 FMPI plots in Northeastern China of the Microsoft Access program, we first extracted erroneous entries, the final master database consisted of values of 12 key attributes from individual state and re- 475,892 permanent sample plots distributed across the gional FIA databases − plot number, tree species, dbh, contiguous United States and Alaska. tree status, trees per unit area that a sample tree repre- Tree species diversity (N), in terms of species richness, sents, elevation, slope, forest type, site productivity class, was derived from the predesignated numeric codes latitude, longitude, and inventory year. The key attributes, named “SPCD” that identifies species of every single tree which were used to develop the seven final variables used in the FIA databases (Woudenberg et al. 2011). N repre- in this study (Table 1), were all directly measured in the sents the number of species among all the live trees on a field, with an exception of forest type, which was assigned permanent sample plot and across the United States, based on the plot location and the forest type map across and ranges from 1 to 24 (Table 2). the contiguous United States and Alaska at a 250-m spatial Basal area (B), the total cross sectional area of all the live resolution (Ruefenacht et al. 2008). We only selected these trees on a permanent sample plot, was derived from the 12 key attributes for the master database because 1) these FIA attributes dbh, tree status (ST), and trees per unit area attributes are essential for deriving forest productivity, (TPA) that a sample tree represents according to: basal area, and tree species diversity that were used in our study; 2) all individual state and regional databases have 3:14⋅dbh ⋅TPA ST¼1 B ¼ ð1Þ these attributes; and 3) all the redundant or unused attri- butes from individual databases were deleted to keep the master database within our computation and storage cap- where tree status is a field recorded code defining the acity. After removing inconsistent, missing, and apparently status of a tree: 0 indicates no status, 1 live, 2 dead, and Watson et al. Forest Ecosystems (2015) 2:22 Page 5 of 16 Table 1 Definition and units for variables used in this study Variable Units Short definition Long definition 3 −1 −1 C m ·ha ·yr Site productivity A measure of the potential timber growth that the site is capable of producing. It is based on the average annual increment of naturally occurring, fully stocked stands. 2 −1 B m ·ha Stand basal area Total stand basal area of all the living trees from the most current measurement in FIA data. E m Plot elevation The vertical distance that a plot is located from sea level. Positive values indicate that the plot is located above sea level while negative values indicate that the plot was located below sea level. 70 % of the values were obtained from the DEM of the United States and 30 % from ground measurement. S degrees Slope The angle of slope in degrees; 5.6 % of the values were obtained from the DEM of the United States and 94.4 % from ground measurement. N Species richness The total number of different species of woody trees present on the plot. y degrees Latitude Latitude of the plot in NAD 83 x degrees Longitude Longitude of the plot in NAD 83 3 indicates a tree that has been removed (Woudenberg species composition to reduce the total number of forest et al. 2011). Over the contiguous United States and types from 30 to 15 (Fig. 1). 2 −1 Alaska, B ranges from 0 to 279.19 m ·ha (Table 2). Only one of the 15 forest types was divided into sub- The data for elevation (E) and slope (S), which were only types. The Douglas-fir forest type consisted of Douglas-fir used to develop a continuous measure of site productivity forests which grow near the Pacific coast and Douglas-fir based on site class (see Eq. 2 below), were mostly obtained forests which grow much further inland. Because of from the ground-measured FIA databases. However, 70 % substantial differences in the climate and other growing of FIA plots in the master database were missing elevation conditions (Hermann and Lavender 1990), the Douglas-fir records and 5.6 % were missing slope records. To retain forest type was separated into two subtypes, the coastal these plots for the estimation of site productivity, espe- Douglas-fir forests (west of the longitude 120° W) and the cially for those forest types with a relatively small sample inland Douglas-fir forests (east of the longitude 120° W). size (e.g., tropical and exotic hardwoods have only 408 The coastal Douglas-fir forests mainly consist of coast plots and western hardwoods 1,924 plots), we derived the Douglas-fir (Pseudotsuga menziesii var. menziesii), whereas missing data from the plot coordinates and the Digital Ele- the inland Douglas-fir forests are comprised mostly of an- vation Model (DEM) of the United States. For the eleva- other variety of Douglas-fir, namely Rocky Mountain or in- tion, 70 % of data were derived from the DEM of the terior Douglas-fir (P. menziesii var. glauca). United States, and 30 % from the FIA databases. For the Site productivity (C), a measure of the potential timber slope, 5.6 % of data came from the United States DEM growth that a plot is capable of sustaining, was derived and 94.4 % from the FIA databases. The DEM of the from a categorical attribute from the FIA databases contiguous United States (30-m resolution) and Alaska named “SITECLCD” which ranks site productivity in a (60-m resolution) was downloaded from the Geospatial hierarchical order from one to seven. Each code of Data Gateway site (http://datagateway.nrcs.usda.gov/, last SITECLCD denotes a range of productivity: 1 stands for accessed December 18, 2014). the most productive sites with a mean annual increment The forest type map across the contiguous United (MAI, see Hanson et al. 2003) greater than or equal to 3 −1 −1 3 −1 −1 States and Alaska was developed by Ruefenacht et al. 15.74 m ·ha ·yr ; 2 for 11.55–15.74 m ·ha ·yr ;3 3 −1 −1 3 −1 (2008) based on the FIA data. The map has a 250-m for 8.40–11.55 m ·ha ·yr ; 4 for 5.95–8.40 m ·ha · −1 3 −1 −1 3 spatial resolution, with an accuracy of ≈ 69 % for the 48 yr ; 5 for 3.50–5.95 m ·ha ·yr ; 6 for 1.40–3.50 m · −1 −1 contiguous U.S. states, and ≈ 78 % for Alaska. For simpli- ha ·yr , and 7 for the least productive sites with a 3 −1 −1 city, in this study, we grouped some of the original forest MAI less than 1.40 m ·ha ·yr (Woudenberg et al. types together based on similar geographic location and 2011). We converted the categorical attribute of SITECLCD Watson et al. Forest Ecosystems (2015) 2:22 Page 6 of 16 Table 2 Summary statistics by forest type. Std: Standard Deviation, n: total number of plots Productivity Number of Species Basal area Elevation (km) Slope Latitude Longitude 3 −1 −1 2 −1 (m ·ha ·yr ) (m ·ha ) (degrees) (degrees) (degrees) National (all forest types combined) Mean 4.84 6.00 19.13 0.61 6.29 38.20 −88.67 Std. 2.76 3.42 11.96 0.73 8.09 5.71 11.02 Max. 15.03 24.00 279.19 3.92 60.24 61.46 −67.00 Min. −0.50 1.00 0 −0.08 0 24.63 −153.86 n 475,892 475,892 475,892 475,892 475,892 475,892 475,892 Pinyon/juniper Mean 0 2.21 17.68 1.84 11.33 37.06 −110.00 Std. 1.27 1.41 13.27 0.51 9.72 2.84 5.56 Max. 12.41 19.00 130.34 3.35 57.17 48.14 −71.90 Min. −0.50 1.00 0 −0.07 0.00 29.27 −122.78 n 16,709 16,709 16,709 16,709 16,709 16,709 16,709 Douglas-fir (Coastal) Mean 8.43 4.24 23.53 0.18 5.63 45.23 −122.53 Std. 3.48 2.08 27.03 0.22 10.03 1.88 1.01 Max. 14.97 13.00 188.99 1.52 44.71 48.99 −120.00 Min. −0.48 1.00 0 0 0 42.00 −124.68 n 5,866 5,866 5,866 5,866 5,866 5,866 5,866 Douglas-fir (Interior) Mean 4.41 3.70 10.10 0.16 3.54 47.50 −118.26 Std. 2.35 2.00 9.05 0.23 7.74 1.57 0.83 Max. 12.39 12.00 58.71 1.77 43.38 48.99 −116.56 Min. −0.48 1.00 0 0.01 0 43.92 −119.99 n 1,211 1,211 1,211 1,211 1,211 1,211 1,211 Oak/pine Mean 5.61 6.65 17.60 0.52 4.50 33.49 −85.83 Std. 2.59 3.54 10.71 0.67 6.27 2.79 6.05 Max. 15.01 22.00 80.46 3.65 60.24 47.97 −68.82 Min. −0.48 1.00 0 −0.04 0 26.26 −103.00 n 19,023 19,023 19,023 19,023 19,023 19,023 19,023 Oak/gum/cypress Mean 5.91 6.23 20.59 0.53 1.53 32.29 −84.95 Std. 2.56 3.39 13.72 0.70 3.38 2.23 5.47 Max. 15.03 22.00 136.46 3.80 57.17 42.60 −71.40 Min. −0.47 1.00 0 0 0 24.63 −101.37 n 28,431 28,431 28,431 28,431 28,431 28,431 28,431 Watson et al. Forest Ecosystems (2015) 2:22 Page 7 of 16 Table 2 Summary statistics by forest type. Std: Standard Deviation, n: total number of plots (Continued) Productivity Number of Species Basal area Elevation (km) Slope Latitude Longitude 3 −1 −1 2 −1 (m ·ha ·yr ) (m ·ha ) (degrees) (degrees) (degrees) Elm/ash/cottonwood Mean 4.75 5.52 16.87 0.45 3.60 38.30 −91.30 Std. 3.15 3.27 11.04 0.57 6.17 5.37 6.55 Max. 15.03 22.00 130.62 3.61 43.53 61.44 −67.17 Min. −0.47 1.00 0 0 0 26.13 −153.46 n 11,742 11,742 11,742 11,742 11,742 11,742 11,742 Aspen/birch Mean 4.77 18.42 0.60 4.19 45.76 −90.76 −90.76 Std. 2.32 10.53 0.72 5.34 2.30 6.43 6.43 Max. 14.97 16.00 95.88 3.68 56.49 61.42 −67.23 Min. −0.49 1.00 0 0 0 33.62 −151.73 n 46,411 46,411 46,411 46,411 46,411 46,411 46,411 Southern pine Mean 5.97 5.87 17.61 0.53 2.89 32.65 −85.06 Std. 2.46 3.45 11.47 0.68 4.49 2.07 5.46 Max. 15.02 23.00 140.89 3.86 57.17 44.00 −70.00 Min. −0.47 1.00 0 −0.08 0 25.76 −101.23 n 102,844 102,844 102,844 102,844 102,844 102,844 102,844 Oak/hickory Mean 4.97 7.77 18.86 0.45 9.38 37.15 −85.40 Std. 2.47 3.45 9.82 0.58 9.07 3.18 5.80 Max. 15.02 24.00 87.62 3.92 57.17 48.99 −68.75 Min. −0.48 1.00 0 −0.03 0 25.61 −104.38 n 141,062 141,062 141,062 141,062 141,062 141,062 141,062 Maple/beech/birch Mean 4.03 6.22 21.85 0.44 5.88 44.16 −82.11 Std. 2.15 2.57 10.89 0.47 6.68 2.28 8.45 Max. 14.97 20.00 80.58 3.71 57.00 49.00 −67.11 Min. −0.47 1.00 0 0 0 34.81 −102.82 n 47,350 47,350 47,350 47,350 47,350 47,350 47,350 Tropical and exotic hardwoods Mean 3.82 3.14 14.72 0.54 0.95 28.44 −86.44 Std. 2.79 2.52 13.72 0.64 3.10 1.74 7.14 Max. 12.45 13.00 60.24 3.56 34.33 35.32 −80.11 Min. −0.46 1.00 0 0 0 25.79 −120.48 n 408 408 408 408 408 408 408 Watson et al. Forest Ecosystems (2015) 2:22 Page 8 of 16 Table 2 Summary statistics by forest type. Std: Standard Deviation, n: total number of plots (Continued) Productivity Number of Basal area Elevation (km) Slope Latitude Longitude 3 −1 −1 2 −1 (m ·ha ·yr ) Species (m ·ha ) (degrees) (degrees) (degrees) Spruce/fir and exotic softwoods Mean 3.59 4.79 20.33 0.48 2.73 46.54 −85.49 Std. 2.06 2.34 11.99 0.56 4.09 1.49 10.11 Max. 14.96 15.00 80.86 3.68 56.49 61.46 −67.00 Min. −0.46 1.00 0 0 0 38.28 −151.73 n 18,761 18,761 18,761 18,761 18,761 18,761 18,761 Northern pines Mean 4.38 4.69 19.75 0.44 3.87 44.82 −85.23 Std. 2.30 2.49 11.46 0.53 5.34 2.06 6.75 Max. 14.97 15.00 82.54 3.57 41.99 49.31 −67.58 Min. −0.45 1.00 0 0 0 26.33 −96.12 n 9,151 9,151 9,151 9,151 9,151 9,151 9,151 Western conifers Mean 3.80 2.75 24.55 1.86 15.54 43.00 −113.93 Std. 2.93 1.43 17.66 0.78 10.57 4.58 6.88 Max. 15.00 10.00 279.19 3.70 57.17 61.11 −95.06 Min. −0.50 1.00 0 0 0.00 31.80 −153.86 n 21,792 21,792 21,792 21,792 21,792 21,792 21,792 Western hardwoods Mean 2.61 2.08 15.70 0.94 10.56 36.87 −112.68 Std. 4.65 1.46 19.66 0.76 12.26 5.57 8.39 Max. 14.99 8.00 184.50 3.17 47.73 49.00 −99.21 Min. −0.48 1.00 0 −0.03 0 29.05 −124.60 n 1,924 1,924 1,924 1,924 1,924 1,924 1,924 Western oak Mean 1.23 2.35 14.38 1.34 16.87 36.79 −113.99 Std. 3.13 1.37 13.21 0.80 11.71 3.29 7.14 Max. 14.97 10.00 138.16 3.12 57.17 47.47 −98.46 Min. −0.49 1.00 0 0 0 29.57 −124.20 n 3,207 3,207 3,207 3,207 3,207 3,207 3,207 Chinese northeastern temperate forest Mean 3.06 6.86 23.07 0.76 10.67 43.38 130.45 Std. 1.39 2.53 9.90 0.23 7.70 0.12 0.25 Max. 15.30 13.00 75.28 7.70 36.00 43.65 131.05 Min. 0.08 1.00 0 0 0 43.11 129.97 n 1,385 1,385 1,385 1,385 1,385 1,385 1,385 to a continuous attribute of site productivity (C)byac- on previous findings that elevation and slope both have counting for the effects of elevation and slope with a lin- a profound impact on the productivity of forest sites ear model: (Stage and Salas 2007 and references therein). The model was calibrated with FIA data from the entire country by C ¼ α þ α ⋅SITECLCD þ α ⋅E þ α ⋅S ð2Þ 0 1 2 3 fitting observed plot-level values of SITECLCD, E,and S against the mid-point MAI values of the SITECLCD.The where α through α were coefficients to be estimated by model was examined for the level of significance of the co- 0 3 ordinary least squares (Table 3). This model was based efficients, the biological interpretation, and the normality Watson et al. Forest Ecosystems (2015) 2:22 Page 9 of 16 Table 3 Parameters of the site productivity model and a majority of forest dynamics studies (e.g., Shugart Independent variables Constant 1984) in terms of basal area effects. It should be noted, however, that we did not consider any abiotic factors in SITECLCD E S this study except for elevation and slope (Eq. 2) due to a Coef. −2.442 0.082 0.002 16.960 lack of soil and other environmental records. Subject to SE 0.000 0.001 0.000 0.002 the potential omitted-variable bias (Wooldridge 2000), P <0.0001 <0.0001 <0.0001 <0.0001 our results should be interpreted with caution. For in- R 0.98 stance, our results may address the effects of biodiversity dF 475,888 on productivity, but the abiotic causes of biodiversity variation across the study region and their potential con- of residual pattern. Then, we estimated the actual site prod- founding effect on BEFR were not directly detectable uctivity of 475,892 FIA plots using the values of SITECLCD, with this database. Nevertheless, as an inherent nature E,and S in Eq. 2, thereby making the site productivity of of abiotic factors, their potential effects especially on for- FIA plots a continuous variable across the country. est productivity are in general spatially autocorrelated The master database was created to produce nation- (Legendre 1993; Liang 2012). To this end, we also wide maps of forest stocking (basal area), tree species employed a geospatial model to account for spatial auto- richness, and site productivity. For the creation of each correlation and potential effects of abiotic factors. of these maps, two separate data layers were developed; Due to potential spatial autocorrelation, which can one for Alaska and one for the contiguous United States. bias tests of significance due to the violation of inde- Each plot represents a raster with size 0.05° for the con- pendence (Clifford et al. 1989) in the FIA data, we mea- tiguous United States and size 0.2° for Alaska. The GCS- sured the spatial autocorrelation function for C for each WGS-1984 projection was used for the contiguous forest type using a nonparametric approach (Bjørnstad United States and the Alaska Albers Equal Area Conic and Falck 2001). Due to computational constraints, it projection (NAD_1983 datum) was used for Alaska. was not possible to estimate the spatial autocorrelation function for most forest type’s entire set of data; for Diversity-productivity relationship example, the maximum number of sample locations For each of the 15 forest types across the contiguous for any one forest type was 141,062, while in 75 % of United States and Alaska, a multiple regression analysis the forest types the number of spatial locations exceeded was conducted to test the general effect of tree species 7,000. To overcome this constraint, we used a bootstrapping richness (N) on site productivity (C), with forest stocking sampling approach in which, for each forest type, up to ≈ (B) being accounted for: 2,000 spatial locations and their associated value of C were randomly selected, from which we estimated parameters of the spatial autocorrelation. This procedure was re- C ¼ β þ β ⋅N þ β ⋅N þ β ⋅N ⋅B ij ij ij ij 0;i 1;i 2;i ij 3;i peated independently 200 times, from which we estimated þ β ⋅B þ e ð3Þ ij ij 4;i the mean, and the 95 % confidence intervals using the 0.025 % and 0.975 % quantiles of the bootstrapped distri- 2 −1 where B represents forest stocking (m ·ha )and C bution of the parameters (Efron and Tibshirani 1993). 3 −1 −1 represents site productivity (m ·ha ·yr ), i forest Since C for most forest types revealed initially some level type (i=1,2,3,…, 15), and j permanent sample plot of spatial autocorrelation, as determined by 95 % confi- number within the ith forest type. e represents the ran- dence intervals that did not include 0 for the estimate of dom error term. Coefficients (β’s) were estimated with the local spatial autocorrelation (Bjørnstad and Falck ordinary least squares, based on two assumptions: the 2001), we sought to detrend the data using a second-order polynomial spatial model fit to values of C for each forest sample is random, and the error term is of zero condi- tional mean and homoskedasticity. The full quadratic type according to: terms of N and B were incorporated in Eq. 3 to study 2 2 C ¼ x þ y þ xy þ x þ y ð4Þ ;y how site productivity (C) changes in response to changes in species richness (N) while keeping basal area in which x and y represent the longitude and latitude, (B)constant atits sample mean forthatforesttype. respectively, of the spatial location corresponding to the We employed Eq. 3 to study BEFR across different for- value of C. In this case, we used all values of C for each est types, with an underlying assumption that biotic fac- forest type as opposed to a randomly chosen subset. We tors, namely species diversity and basal area, have direct used mixed stepwise regression to determine the appro- causal effects on forest productivity. This assumption priate model. Ultimately, 13 of the 15 forest types exhib- was supported by recent BEFR studies (e.g., Zhang et al. ited significant spatial autocorrelation in values of C as 2012; Liang et al. 2015) in terms of biodiversity effects, ascertained by significance of at least one parameter Watson et al. Forest Ecosystems (2015) 2:22 Page 10 of 16 Table 4 Estimates of spatial autocorrelation and the appropriate spatial detrending model for all 15 forest types Forest type Spatial autocorrelation estimates Detrending model 1 2 3 Local spatial autocorrelation Range of spatial autocorrelation (km) Parameters n Pinyon/juniper 0.33 1,302.7 All 16,709 Douglas-fir 0.45 260.2 All 7,077 Oak/pine 0.21 490.0 All 19,023 Oak/gum/cypress 0.16 597.4 All 28,431 Elm/ash/cottonwood 0.51 617.5 All 11,742 Aspen/birch 0.13 536.6 All 46,411 Southern pine 0.19 617.0 All 102,844 2 2 Oak/hickory 0.14 524.0 x, y, x , y 141,062 Maple/beech/birch NS NS NS 47,350 Tropical and exotic hardwoods 0.35 404.6 All 408 2 2 Northern pines 0.08 219.7 x, y, x , y 9,151 Spruce/fir and exotic softwoods NS NS NS 18,761 Western conifers 0.39 973.2 x, xy, y 21,792 2 2 Western hardwoods 0.65 1,167.8 x, x , y 1,924 Western oak 0.37 990.7 x, y, xy, y 3,207 Empirical mean (from 200 bootstrapped simulations) of the estimate of local autocorrelation as the distance between sampling locations approaches 0 Mean distance (from 200 bootstrapped simulations) of the lag distance at which the estimate of local autocorrelation = 0 3 2 2 Full model parameters =x, y, xy, x , and y NS not significant value in Eq. 4 (Table 4). For these 13 forest types, we autocorrelation in the residuals based upon the 95 % con- then obtained the detrended residuals from Eq. 4.To fidence intervals. Thus, the spatially-detrended residuals ensure that the second-order polynomial spatial model for these 13 forest types were used in all follow-up ana- adequately removed the spatial autocorrelation, we then lyses. For the remaining two forest types that did not ex- estimated the spatial autocorrelation of the residuals ob- hibit significant spatial autocorrelation (maple/beech/ tained from Eq. 4 for these 13 forest types using the boot- birch, and spruce/fir and exotic softwoods, Table 4), we strap sampling approach described above. In all forest used values of C in all follow-up analyses. All analyses type cases, there was no significant local spatial were conducted in R 2.14.0. Fig. 3 Richness of woody plant species across the 48 contiguous U.S. states and Alaska, derived from FIA ground measurements completed between 1968 and 2011 Watson et al. Forest Ecosystems (2015) 2:22 Page 11 of 16 2 −1 Fig. 4 Total forest stand basal area (m ·ha ) across the 48 contiguous U.S. states and Alaska, derived from FIA ground measurements completed between 1968 and 2011 Results values were concentrated in the western United States Geographic distribution of tree species richness, forest and Central Appalachia (Fig. 4). productivity and stocking Forests with high site productivity were generally dis- The southeastern region (with an exception of the Gulf tributed in the western slopes of the mountain ranges in and Atlantic Coastal plains) and the Appalachian Moun- northern and central California, western Oregon and tains showed the highest tree species richness, whereas Washington, northern Idaho, the southeastern United the majority of the mountainous regions of the western States (except Florida), southern Michigan, and the United States, the black hills region of South Dakota, states of Illinois and Indiana. The areas with the most western Texas, northwestern Minnesota, and the state of notable overall low values of site productivity were the Florida had the lowest tree species richness (Fig. 3). In southern Rocky Mountains and western Texas (Fig. 5). general, forests in the eastern United States consisted of Oak/hickory and southern pine forests had the overall more tree species than those in the central and western highest levels of species richness, and the western hard- parts of the country including the state of Alaska. Over woods, western conifers and western oak forests had the contiguous United States, the highest forest stocking the lowest (Table 2). The mean productivity over all of 3 −1 −1 Fig. 5 Forest productivity (m ·ha ·yr ) across the 48 contiguous U.S. states and Alaska, derived from FIA ground measurements completed between 1968 and 2011 Watson et al. Forest Ecosystems (2015) 2:22 Page 12 of 16 3 −1 −1 3 −1 −1 the forest types studied here was 4.84 m ·ha ·yr . western oak forest type, 0.9 m ·ha ·yr for the south- 3 −1 −1 The most productive forest type was oak/gum/cypress ern pine forest type, 1.6 m ·ha ·yr for the spruce/fir 3 −1 −1 whereas the least productive was pinyon/juniper. Aver- and exotic softwoods forest type, 0.6 m ·ha yr for 3 −1 −1 age species richness for all forest types combined is 6.00 the oak/hickory forest type, and 1.5 m ·ha ·yr for with a standard deviation of 3.42 (Table 2). the western conifers forest type (Fig. 6). The forests that showed a flat or negative diversity-productivity relation- Diversity-productivity relationship ship were the coastal Douglas-fir subtype with a decline 3 −1 −1 Throughout all the 15 forest types in the United States, 12 of −1.0 m ·ha ·yr , the northern pines forest type plus the interior Douglas-fir subtype showed a positive re- (flat), the tropical and exotic hardwoods forest type lationship between species richness and site productivity, (flat), and the western hardwoods forest type with a de- 3 −1 −1 and only the coastal Douglas-fir subtype, northern pines, cline of −1.5 m ·ha ·yr (Fig. 6). tropical and exotic hardwoods, and western hardwoods The FMPI data from Northeastern China conformed forest types had a negative or insignificant relationship to the positive diversity-productivity relationship (Table 5). Species richness was highly significant for all (Table 5). As species richness increased from 1 to 12, but the tropical and exotic hardwoods forest type. Over productivity of the northeastern temperate forests in 3 −1 −1 all the study areas, 96.4 % of sample plots (477,281) con- China improved from 2.5 to 4.0 m ·ha ·yr , a 60 % formed to a positive effect of species richness on site increase from the base value (Fig. 7). productivity, and only 3.6 % (17,349) showed an insig- nificant or negative effect. Discussion Based on the estimated coefficients of Eq. 3, when B The findings largely support, from the perspectives of was kept constant at its sample mean, as species richness forested ecosystems over a large geographic scale, a posi- increased from 1 to the 75th percentile values, site prod- tive biodiversity-ecosystem functioning relationship 3 −1 −1 uctivity was expected to increase by 1.2 m ·ha ·yr (BEFR). This is consistent with other experiments based 3 −1 −1 for the pinyon/juniper forest type, 4.0 m ·ha ·yr for generally on controlled field experiments with herb- 3 −1 −1 the interior Douglas-fir subtype, 1.0 m ·ha ·yr for aceous species (see Cardinale et al. 2012 and references 3 −1 −1 the oak/pine forest type, 1.6 m ·ha ·yr for the oak/ therein). Evidence from this study was unique because it 3 −1 −1 gum/cypress forest type, 0.3 m ·ha ·yr for the elm/ was based upon almost half a million ground-measured 3 −1 −1 ash/cottonwood forest type, 1.0 m ·ha ·yr for the forest inventory plots from a large geographic scale. Rec- 3 −1 −1 aspen/birch forest type, 1.8 m ·ha ·yr for the ognizing that different forest types vary in regards to Table 5 Parameters of site productivity models by forest type with predictor variables of species richness, basal area, elevation, and slope Forest type n AIC BIC Coefficients Const. N N N·B B Pinyon/juniper 16,709 47112 47158 −0.398 *** 0.209 *** −0.012 *** 0.005 *** −0.010 *** Douglas-fir (Coastal) 5,866 29214 29241 −0.024 −0.011 *** 0.002 *** Douglas-fir (Interior) 1,211 5259 5284 −2.735 *** 0.809 *** −0.035 ** 0.031 *** Oak/pine 19,023 86805 86853 −1.599 *** 0.205 *** −0.008 *** −0.001 * 0.047 *** Oak/gum/cypress 28,431 128225 128266 −1.058 *** 0.142 *** −0.002 *** 0.019 *** Elm/ash/cottonwood 11,742 55755 55800 −1.394 *** 0.244 *** −0.010 *** −0.003 *** 0.048 *** Aspen/birch 46,411 200909 200953 −0.865 *** 0.105 *** −0.004 *** 0.040 *** Southern pine 102,844 457216 457273 −1.070 *** 0.093 *** −0.001 ** −0.001 *** 0.040 *** Oak/hickory 141,062 636358 636397 −0.798 *** 0.040 *** 0.026 *** Maple/beech/birch 47,350 206442 206495 3.703 *** −0.033 * 0.013 *** −0.004 *** 0.024 *** Tropical and exotic hardwoods 408 1700 1712 −0.409 ** 0.028 *** Northern pines 9,151 40244 40280 −1.469 *** 0.161 *** −0.007 *** 0.075 *** Spruce/fir and exotic softwoods 18,761 79280 79327 2.011 *** 0.479 *** −0.025 *** −0.002 ** 0.008 * Western conifers 21,792 101207 101247 −1.409 *** 0.184 *** 0.003 *** 0.029 *** Western hardwoods 1,924 8041 8075 −0.386 ** 0.446 ** −0.125 *** 0.010 *** −0.017 * Western oak 3,207 15060 15090 −1.311 *** 0.409 *** −0.037 * 0.015 *** Chinese northeastern temperate forest 1,385 4768 4789 1.288 *** 0.397 *** −0.017 *** −0.0007 0.004 Level of significance: < 0.001: ***; 0.001: **; 0.01: * Watson et al. Forest Ecosystems (2015) 2:22 Page 13 of 16 3 −1 −1 Fig. 6 Sensitivity of stand productivity (m ·ha ·yr ) to species richness for 15 forest types (each panel represents one type of forest) across the 48 contiguous U.S. states and Alaska. Solid lines represent predicted means of different forest types and broken lines the 95 % confidence interval of the predicted means, with stand basal area being kept constant at its sample mean how species diversity affects productivity, we categorized growth, this study quantified site productivity as potential all the ground-measured plots into 15 forest types, and timber growth that a site could sustain. Our results indi- analyzed the diversity-productivity relationship specific cate that intensively managed coastal Douglas-fir forests to each forest type. The number of ground-measured feature a negative effect of diversity on potential timber plots, as well as the magnitude of geographic scale, ren- growth presumably because these stands are artificially dered overwhelming evidence in support of a positive maintained in an early stage of stand development (stem tree species diversity-timber productivity relationship. exclusion) where current annual increment is nearly opti- The negative diversity productivity relationship for the mized at a low diversity. The inland Douglas-fir forests coastal Douglas-fir subtype that was found (Fig. 6b) con- conformed to the positive biodiversity-forest productivity tradicts with the positive net basal area change in associ- relationship as they are less intensively managed. It should ation with tree species diversity reported by Liang et al. be noted, however, that this implication was only applicable (2007) for the same subtype. The main reason for this to Douglas-fir forests, which are in general low in tree spe- discrepancy may be in the measure of site productivity. cies diversity. Southern pine forests, in spite of high man- While Liang et al. (2007) measured site productivity by net agement intensity, still show a positive tree species annual basal area change, which represents actual forest diversity-timber productivity relationship (Fig. 6j). A Watson et al. Forest Ecosystems (2015) 2:22 Page 14 of 16 FIA plots can be treated as a normally distributed ran- dom process. Furthermore, as most forests across the United States are distributed on relatively flat surfaces, bias in elevation that is caused by fuzzed coordinates was limited. Slope data is the most sensitive to plot loca- tions, but the accuracy of the site productivity model was maintained as 94 % of the data were measured in the field and only 6 % were estimated using the fuzzed coordinates. Conclusion Over all the study areas, 96.4 percent of sample plots 3 −1 −1 (477,281) showed a positive effect of species richness on Fig. 7 Sensitivity of stand productivity (m ·ha ·yr ) to species richness for the northeastern temperate forests in China. Solid lines site productivity, and only 3.6 percent (17,349) had an represent predicted means of different forest types and broken lines insignificant or negative effect. the 95 % confidence interval of the predicted means, with stand The results of this study suggest that maintaining spe- basal area being kept constant at its sample mean cies diversity is an important means to maintain forest productivity, which is supported by an array of forestry studies (Kelty 1989; Caspersen and Pacala 2001; Liang et al. 2005, 2007; Lei et al. 2009b; Young et al. 2011; possible explanation is southern pine forests by nature Zhang et al. 2012; Gamfeldt et al. 2013). These results consist of more tree species than Douglas-fir forests. Also, should assist landowners in making management deci- the niche complementarity effect (see Loreau and Hector sions that are relevant to the specific forest types that 2001 and references therein) that contributes to the posi- they respectively manage. These findings also imply that tive BEFR could therefore be more prominent in the south- productivity of forests across the United States may be ern pine forests. impaired by the loss of both woody and non-woody Due to the use of observational data, the observed plant species in forested ecosystems (Fleming et al. trends are subject to the usual caveats of multicollinear- 2011; Liang et al. 2015), and that biological conserva- ity. However, multicollinearity does not lead to a biased tion, due to its potential benefits in maintaining forest diversity-productivity relationship (Goldberger 1991), productivity, can have profound impacts on the prod- even though it may lead to difficulties in quantifying the uctivity of selected services that can be obtained from variance of predicted means. The partial effects of diver- forests across the United States. sity may be more uncertain due to multicollinearity. Nevertheless, multicollinearity was not excessive in the Competing interests present case as the Variance Inflation Factor of tree spe- The authors declare that they have no competing interests. cies richness (N) was estimated to be 1.08 using the “car” package of R (Fox and Weisberg 2011), and all the Authors’ contributions JVW, JL, and XL compiled the data. JVW, JL, and CEA conducted mapping. forest types show very low standard errors in predicted JVW, JL, and PCT performed regression analysis. All authors contributed to site productivity (Fig. 6). Another statistical caveat of this the writing, and read and approved the final manuscript. study, due to the use of observational data, is in the causal relationship between diversity and productivity. Acknowledgements Without a controlled experiment, it is difficult to deter- We are obliged to Karen Waddell and Charles J. Barnett for assistance with FIA data. We thank Mo Zhou, Jacquelyn Strager, George Merovich, and Eric mine the cause and effect in the diversity-productivity King for statistical and mapping assistance. This research is supported in relationship. Compared to controlled experiments, this parts by the United States Department of Agriculture McIntire-Stennis Act empirical evidence provides insights to forest manage- Fund WVA00104, and by the Division of Forestry and Natural Resources, West Virginia University. ment and biological conservation that are of a much broader applicability, both in terms of forest type and Author details geographic scale. Davis College of Agriculture, Natural Resources and Design, West Virginia University, 26506 Morgantown, WV, USA. Forest Service, U.S. Department of The fuzzed FIA plot coordinates could affect the ac- Agriculture, Northern Research Station, 26505 Morgantown, WV, USA. School curacy of the estimated site productivity (C), as 70 % of of Environmental and Forest Sciences, University of Washington, 98195 plot elevation records and 6 % of plot slope records were Seattle, WA, USA. Research Institute of Forest Resource Information Techniques, Chinese Academy of Forestry, 100091 Beijing, P. R. China. obtained from a DEM using the fuzzed coordinates. Nevertheless, the impact should be small, as the differ- Received: 26 December 2014 Accepted: 23 June 2015 ences between fuzzed coordinates and true locations of Watson et al. Forest Ecosystems (2015) 2:22 Page 15 of 16 References Kelty MJ (1989) Productivity of New England hemlock/hardwood stands as affected Assmann E (1970) The principles of forest yield study. Pergamon Press, Oxford, UK by species composition and canopy structure. For Ecol Manage 28:237–257 Bechtold WA, Scott CT (2005) The forest inventory and analysis plot design. In: Köble R, Seufert G (2001) Novel maps for forest tree species in Europe. In: the enhanced forest inventory and analysis program: national sampling design Hjorth J, Raes F, Angeletti G (eds) Proceedings of the 8th European and estimation procedures. USDA Forest Service, Asheville, NC, pp 27–42 symposium on the physico-chemical behaviour of air pollutants: a changing Bjørnstad ON, Falck W (2001) Nonparametric spatial covariance functions: atmosphere. Commission of the European Communities, Directorate General estimation and testing. Environ Ecol Stat 8:53–70 Telecommunications, Information Industries and Innovation. Torino, Italy, pp 17–20 Kreft H, Jetz W (2007) Global patterns and determinants of vascular plant Buongiorno J, Peyron JL, Houllier F, Bruciamacchie M (1995) Growth and management of mixed-species, uneven-aged forests in the French Jura: diversity. Proc Natl Acad Sci U S A 104:5925–5930 implications for economic returns and tree diversity. For Sci 41:397–429 Legendre P (1993) Spatial autocorrelation: trouble or new paradigm? Ecology 74:1659–1673 Cardinale BJ, Duffy JE, Gonzalez A, Hooper DU, Perrings C, Venail P, Narwani A, Mace GM, Tilman D, Wardle DA, Kinzig AP, Daily GC, Loreau M, Grace JB, Lei X, Tang M, Lu Y, Hong L, Tian D (2009a) Forest inventory in China: status and challenges. Int For Rev 11:52–63 Larigauderie A, Srivastava DS, Naeem S (2012) Biodiversity loss and its impact on humanity. Nature 486:59–67 Lei X, Wang W, Peng C (2009b) Relationships between stand growth and Cardinale BJ, Matulich KL, Hooper DU, Byrnes JE, Duffy E, Gamfeldt L, Balvanera P, structural diversity in spruce-dominated forests in New Brunswick, Canada. O’Connor MI, Gonzalez A (2011) The functional role of producer diversity in Can J For Res 39:1835–1847 ecosystems. Am J Bot 98:572–592 Leuschner C, Jungkunst HF, Fleck S (2009) Functional role of forest diversity: pros Caspersen JP, Pacala SW (2001) Successional diversity and forest ecosystem and cons of synthetic stands and across-site comparisons in established function. Ecol Res 16:895–903 forests. Basic Appl Ecol 10:1–9 Chen HYH, Klinka K, Mathey AH, Wang X, Varga P, Chourmouzis C (2003) Are Liang J (2012) Mapping large-scale forest dynamics: a geospatial approach. mixed-species stands more productive than single-species stands: an empirical Landscape Ecol 27:1091–1108 test of three forest types in British Columbia and Alberta. Can J For Res Liang J, Buongiorno J, Monserud RA (2005) Growth and yield of all-aged 33:1227–1237 Douglas-fir/western hemlock stands: a Matrix model with stand diversity effects. Can J For Res 35:2369–2382 Clifford P, Richardson S, Hémon D (1989) Assessing the significance of the correlation between two spatial processes. Biometrics 45:123–134 Liang J, Buongiorno J, Monserud RA (2006) Bootstrap simulation and response Efron B, Tibshirani RJ (1993) An introduction to the bootstrap. Chapman & Hall, surface optimization of management regimes for Douglas-fir/western New York hemlock stands. For Sci 52:579–594 FAO (2012) State of the world’s forests 2012. Food and Agriculture Organization Liang J, Buongiorno J, Monserud RA, Kruger EL, Zhou M (2007) Effects of diversity of the United Nations, Rome, Italy, p 46 of tree species and size on forest basal area growth, recruitment, and Fleming R, Brown N, Jenik J, Kahumbu P, Plesnik J (2011) Emerging perspectives mortality. For Ecol Manage 243:116–127 on forest biodiversity. In: United Nations Environment Program (ed) UNEP Liang J, Zhou M, Tobin PC, McGuire AD, Reich PB (2015) Biodiversity influences year book 2011. UNEP, Nairobi, Kenya, pp 46–59 plant productivity through niche-efficiency. Proc Natl Acad Sci U S A 112:5738–5743 Fox J, Weisberg S (2011) An R companion to applied regression. Sage, Thousand Lister A, Scott C, King S, Hoppus M, Butler B, Griffith D (2002) Strategies for Oaks, CA preserving owner privacy in the national information management system Gamfeldt L, Snall T, Bagchi R, Jonsson M, Gustafsson L, Kjellander P, Ruiz-Jaen MC, of the USDA Forest Service’s forest inventory and analysis unit. In: McRoberts Froberg M, Stendahl J, Philipson CD, Mikusinski G, Andersson E, Westerlund RE (ed) The fourth annual forest inventory and analysis symposium. USDA B, Andren H, Moberg F, Moen J, Bengtsson J (2013) Higher levels of multiple Forest Service, New Orleans, LA, pp 163–166 ecosystem services are found in forests with more tree species. Nat Commun 4:1340 Loreau M, Hector A (2001) Partitioning selection and complementarity in Gillespie AJ (1999) Rationale for a national annual forest inventory program. J For biodiversity experiments. Nature 412:72–76 97:16–20 Loreau M, Naeem S, Inchausti P, Bengtsson J, Grime JP, Hector A, Hooper DU, Goldberger AS (1991) A course in econometrics. Harvard University Press, Huston MA, Raffaelli D, Schmid B, Tilman D, Wardle DA (2001) Biodiversity Cambridge, MA and ecosystem functioning: current knowledge and future challenges. Science 294:804–808 Haight RG, Monserud RA (1990) Optimizing any-aged management of mixed-species Naeem S, Duffy JE, Zavaleta E (2012) The functions of biological diversity in an stands: II. Effects of decision criteria. For Sci 36:125–144 age of extinction. Science 336:1401–1406 Hanson EJ, Azuma DL, Hiserote BA (2003) Site index equations and mean annual Paquette A, Messier C (2011) The effect of biodiversity on tree productivity: from increment equations for Pacific Northwest Research Station forest inventory temperate to boreal forests. Global Ecol Biogeogr 20:170–180 and analysis inventories, 1985–2001. U.S. Dept. of Agriculture, Forest Service, Pacific Northwest Research Station, Portland, OR Reich PB, Tilman D, Isbell F, Mueller K, Hobbie SE, Flynn DFB, Eisenhauer N (2012) Hasse WD, Ek AR (1981) A simulated comparison of yields for even- versus Impacts of biodiversity loss escalate through time as redundancy fades. uneven-aged management of northern hardwood stands. J Environ Manage Science 336:589–592 12:235–246 Ricketts TH (1999) Terrestrial ecoregions of North America: a conservation He P, Zhang H, Lei X, Li X (2013) Estimation of spatial distribution of tree species assessment. Island Press, Washington, DC diversity based on Universal Krige Model. J Central-South Univ For Technol Ruefenacht B, Finco MV, Nelson MD, Czaplewski R, Helmer EH, Blackard JA, 33:67–71 Holden GR, Lister AJ, Salajanu D, Weyermann D, Winterberger K (2008) Conterminous US and Alaska forest type mapping using forest inventory and Hermann RK, Lavender DP (1990) Douglas-fir. In: Burns RM, Honkala BH (eds) analysis data Photogramm. Eng Remote Sensing 74:1379–1388 Silvics of North America. U.S. Department of Agriculture, Forest Service, Shugart HH (1984) A theory of forest dynamics: the ecological implications of Washington DC, pp 527–540 forest succession models. Springer Verlag, New York Hernández-Stefanoni JL, Dupuy JM (2007) Mapping species density of trees, shrubs and vines in a tropical forest, using field measurements, satellite Stage AR, Salas C (2007) Interaction of elevation, aspect, and slope in models of multiespectral imagery and spatial interpolation. Biodivers Conserv forest species composition and productivity. For Sci 53:486–492 16:3817–3833 Symstad AJ, Chapin FS, Wall DH, Gross KL, Huenneke LF, Mittelbach GG, Hernandez-Stefanoni JL, Ponce-Hernandez R (2006) Mapping the spatial variability of Peters DP, Tilman D (2003) Long-term and large-scale perspectives on plant diversity in a tropical forest: comparison of spatial interpolation methods. the relationship between biodiversity and ecosystem functioning. Environ Monit Assess 117:307–334 Bioscience 53:89–98 Hooper DU, Chapin FS, Ewel JJ, Hector A, Inchausti P, Lavorel S, Lawton JH, Vilà M, Vayreda J, Gracia C, Ibáñez JJ (2003) Does tree diversity increase wood Lodge DM, Loreau M, Naeem S, Schmid B, Setala H, Symstad AJ, Vandermeer production in pine forests? Oecologia 135:299–303 J, Wardle DA (2005) Effects of biodiversity on ecosystem functioning: a Wardle DA (1999) Is “sampling effect” a problem for experiments investigating consensus of current knowledge. Ecol Monogr 75:3–35 biodiversity-ecosystem function relationships? Oikos 87:403–407 Hooper DU, Vitousek PM (1997) The effects of plant composition and diversity on Wooldridge JM (2000) Introductory econometrics: a modern approach. South-Western ecosystem processes. Science 277:1302–1305 College Publishing, Cincinnati, OH Watson et al. Forest Ecosystems (2015) 2:22 Page 16 of 16 Woudenberg S, Conkling B, O’Connell B, LaPoint E, Turner J, Waddell K, Boyer D, Christensen G, Ridley T (2011) The forest inventory and analysis database: description and users manual version 51 for phase 2. US Department of Agriculture, Forest Service, Fort Collins, CO, p 336 Young B, Liang J, Stuart Chapin F III (2011) Effects of species and tree size diversity on recruitment in the Alaskan boreal forest: a geospatial approach. For Ecol Manage 262:1608–1617 Zhang Y, Chen HYH, Reich PB (2012) Forest productivity increases with evenness, species richness and trait variation: a global meta-analysis. J Ecol 100:742–749 Submit your manuscript to a journal and benefi t from: 7 Convenient online submission 7 Rigorous peer review 7 Immediate publication on acceptance 7 Open access: articles freely available online 7 High visibility within the fi eld 7 Retaining the copyright to your article Submit your next manuscript at 7 springeropen.com http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png "Forest Ecosystems" Springer Journals

Large-scale forest inventories of the United States and China reveal positive effects of biodiversity on productivity

Loading next page...
 
/lp/springer-journals/large-scale-forest-inventories-of-the-united-states-and-china-reveal-q02libTqdY

References (54)

Publisher
Springer Journals
Copyright
2015 Watson et al.
eISSN
2197-5620
DOI
10.1186/s40663-015-0045-4
Publisher site
See Article on Publisher Site

Abstract

Background: With the loss of species worldwide due to anthropogenic factors, especially in forested ecosystems, it has become more urgent than ever to understand the biodiversity-ecosystem functioning relationship (BEFR). BEFR research in forested ecosystems is very limited and thus studies that incorporate greater geographic coverage and structural complexity are needed. Methods: We compiled ground-measured data from approx. one half million forest inventory sample plots across the contiguous United States, Alaska, and northeastern China to map tree species richness, forest stocking, and productivity at a continental scale. Based on these data, we investigated the relationship between forest productivity and tree species diversity, using a multiple regression analysis and a non-parametric approach to account for spatial autocorrelation. Results: In general, forests in the eastern United States consisted of more tree species than any other regions in the country. The highest forest stocking values over the entire study area were concentrated in the western United States and Central Appalachia. Overall, 96.4 % of sample plots (477,281) showed a significant positive effect of species richness on site productivity, and only 3.6 % (17,349) had an insignificant or negative effect. Conclusions: The large number of ground-measured plots, as well as the magnitude of geographic scale, rendered overwhelming evidence in support of a positive BEFR. This empirical evidence provides insights to forest management and biological conservation across different types of forested ecosystems. Forest timber productivity may be impaired by the loss of species in forests, and biological conservation, due to its potential benefits on maintaining species richness and productivity, can have profound impacts on the functioning and services of forested ecosystems. Keywords: Tree species diversity; Forest management; Biological conservation; Continental map of forest diversity; Spatial autocorrelation; Bootstrap Background become more urgent than ever to understand the BEFR Over the past two decades, there has been an extensive (Symstad et al. 2003). discussion (see Cardinale et al. 2012; Naeem et al. 2012 There is increasing evidence that supports a positive and references therein) over the biodiversity-ecosystem BEFR, which indicates the loss of biodiversity affects the functioning relationship (BEFR). The loss of biodiver- functioning of an ecosystem (Loreau et al. 2001; Hooper sity can greatly alter the characteristics and functioning et al. 2005; Cardinale et al. 2012; Naeem et al. 2012; Liang of an ecosystem, including its productivity (Liang et al. et al. 2015). There are several mechanisms which are 2015) and services (Hooper et al. 2005). With the loss thought to be the basis for a positive BEFR. Historically, it of species worldwide due to anthropogenic factors, cli- is thought that niche complementarity (Loreau et al. 2001; matic disturbance, altered disturbance regimes, and Cardinale et al. 2011; Reich et al. 2012), which constitutes biological invasions, etc. (Fleming et al. 2011), it has niche differentiation and species facilitation, as well as the sampling effects (Hooper and Vitousek 1997; Wardle * Correspondence: alpenbering@gmail.com 1 1999), i.e., the chance that a forest contains a more pro- Davis College of Agriculture, Natural Resources and Design, West Virginia ductive species increases with increasing species diversity, University, 26506 Morgantown, WV, USA Full list of author information is available at the end of the article © 2015 Watson et al. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. Watson et al. Forest Ecosystems (2015) 2:22 Page 2 of 16 are reasons for a positive BEFR. Recently, Liang et al. ground measurements, remote sensing data, and spatial (2015) developed a theoretical model to integrate comple- estimation. As variance is ignored within a raster, such mentarity and a new mechanism of diminishing marginal maps have limited value to BEFR studies. Maps showing productivity in quantifying the effects of biodiversity loss distribution of biodiversity, consisting of point data are on plant productivity. the most valuable, because spatially explicit biodiversity Most existing evidence of a positive BEFR comes from records can be matched with the forest productivity data grassland experiments at limited spatial scales (Loreau et al. at the same locations for the study of BEFR. Unfortu- 2001; Symstad et al. 2003; Cardinale et al. 2012). Since these nately, due to the cost of ground measurements, large- experiments usually do not include some key processes in- scale maps of forest biodiversity are extremely limited herent to natural ecosystems, such as the introduction of across the world (Köble and Seufert 2001) and to the best new species, decomposition of woody plant material, large of our knowledge, there is no map of forest biodiversity at standing crop, and the utilization of water by woody tree the national scale for the United States. species, it is difficult to extrapolate the results to greater The primary objective of this study was to investigate temporal and spatial scales (Symstad et al. 2003). the effects of biodiversity on forest timber productivity, BEFR studies in forested ecosystems are in general using extensive forest inventory data from the United limited to empirical analysis of observed data (Liang States of America and China. The secondary objective was et al. 2007, 2015; Paquette and Messier 2011; Zhang to synthesize these ground-measured data in mapping tree et al. 2012), primarily due to the complexity of forested species richness, forest stocking, and productivity over all ecosystems and the length of time it takes to derive esti- types of temperate forests across the United States. mates of growth and productivity. Although some argue otherwise (Chen et al. 2003; Vilà et al. 2003), the major- Methods ity of studies confirm a positive relationship between Data tree diversity and ecosystem productivity and services Data for this study came from two sources— Forest (Kelty 1989; Caspersen and Pacala 2001; Liang et al. Inventory and Analysis of the United States of America 2005, 2007, 2015; Lei et al. 2009b; Paquette and Messier (FIA, see Woudenberg et al. 2011) and the Forest Manage- 2011; Young et al. 2011; Zhang et al. 2012; Gamfeldt ment Planning Inventory (FMPI) from the Wangqing For- et al. 2013). BEFR also plays an important role in the estry Bureau in the Jilin Province of China (He et al. 2013). management of forest resources. A traditional view in silviculture is that clearcutting with artificial regeneration FIA (even-aged monoculture) optimizes forest productivity The FIA databases are a product of the United States (e.g., Assmann 1970), but it has been found that this Department of Agriculture (USDA), Forest Service FIA maxim does not generalize (Hasse and Ek 1981; Haight Program, which has established a network of permanent and Monserud 1990) and that the mixed-species stands sample plots across the United States to determine the could have higher long-term productivity (e.g., Haight and extent and status of the nation’s forests in regard to the Monserud 1990; Buongiorno et al. 1995; Liang et al. 2005, condition, volume, growth, and depletions (Woudenberg 2006). Most biodiversity studies in forested ecosystems et al. 2011). The FIA ground-based inventories have been have primarily used even-aged designs that lack structure conducted at different time periods by states, ranging complexity of natural uneven-aged forests (Leuschner from the 1960s to 2012. Following the passage of the 1998 et al. 2009). Thus, studies that incorporate greater geo- Farm Bill (Gillespie 1999), the FIA program in the new mil- graphic coverage and structural complexity are much lennium switched from a periodic inventory system to an needed for analyzing the BEFR in forested ecosystems. annual system in which a portion of the FIA plots in each Biodiversity studies of forested ecosystems, especially at state would be remeasured each year (Gillespie 1999). larger geographic scales, are not only of great value to The FIA program uses a 3-phase sampling scheme BEFR research, but also of strategic importance to the (Woudenberg et al. 2011). Phase 1 uses stratification to world’s energy security and economic development. Tree aggregate ground samples into groups to minimize vari- species provide an essential source of energy and financial ance using stratified estimation. Every FIA ground plot is income worldwide, especially for rural areas where liveli- assigned to a stratum of which a weight is based on its hoods depend heavily on forest resources (FAO 2012). proportion within the estimation unit. Phase 2 consists of The limited maps of biodiversity across the world’sfor- actual FIA ground plots, which follow a national standard ested ecosystems were developed using either raster data and are fixed-radius plots 0.40 ha in size. To protect the or point data. Biodiversity maps of raster data (Ricketts privacy of landowners, geographic coordinates of FIA 1999; Hernandez-Stefanoni and Ponce-Hernandez 2006; ground plots are “fuzzed” (Lister et al. 2002). The true plot Hernández-Stefanoni and Dupuy 2007; Kreft and Jetz locations are known to be within 1.61 km of the fuzzed 2007; Liang 2012) are mostly based on a combination of values under the periodic system, and 0.80 km under the Watson et al. Forest Ecosystems (2015) 2:22 Page 3 of 16 current plot design that was first used in the 1990s for the area is situated on the middle lower hill region of the last of the periodic inventories and all of the current an- Changbai Mountains in northeastern China. Elevation nual inventories (Woudenberg et al. 2011). The ramifica- ranges from 550 to 1,100 m A.S.L. with an annual rain- tions of using the fuzzed coordinates instead of real ones fall of 547 mm. The mean annual temperature is 3.9 °C. are provided in the discussion. Mean monthly maximum and minimum temperatures In the current FIA plot design, each permanent sample range between 22 °C and −37.5 °C, respectively. The study plot consists of four 0.02-ha subplots 7.32 m in radius. area was originally dominated by the mixed broad-leaved Subplots 2, 3, and 4 are situated around subplot 1 in a tri- Korean pine (Pinus koraiensis) forest type, with dark angular pattern, with a 36.58-m distance between the plot brown forest soil. Most primary forests, however, have centers of subplots 2, 3, and 4 and the center of subplot 1 been altered into other forest types such as spruce-fir dom- (Bechtold and Scott 2005). On a subplot, all trees greater inated mixed coniferous forests, birch-aspen mixed broad- than or equal to 12.70 cm in diameter at breast height leaved forests, or plantations consisting of larch, spruce, fir, (dbh) are measured. A 2.07-m radius microplot is located and pine after forestry practices and other disturbances. inside of each of the 4 subplots, in which all trees with a The purpose of FMPI is to assess forest resources and dbh smaller than 12.70 cm are tagged and measured. to supply information requirements for forest manage- FIA inventories are designed in a way such that state ment planning, spatial and functional patterns, and over- level sampling errors are met at the 67 % confidence limit all design of forestry at the FMU level (Lei et al. 2009a). (Woudenberg et al. 2011). Sampling error is kept ≤ 3 % for Sample plots, systematically designed with a 1 km × every 404,686 ha of timberland. It should be noted that 2 km grid, were measured with a 10-year interval. In despite the mandated sampling error to the area, the sam- total there are 1,389 plots distributed in the forest bur- pling errors applied to removals, volume, and total annual eau (Fig. 2). The 0.06-km plot is rectangular. In each growth are targeted by the FIA program, which is 5 % for plot, species and diameter of trees with a dbh above 7 3 every 3 × 10 m of growing stock on timberland for the 5 cm were recorded. Three to five average trees were se- 7 3 eastern United States and 10 % for every 3 × 10 m for lected to measure age and height for computing stand the western United States. A total of 475,892 FIA perman- age and height. Other factors measured are slope, aspect, ent sample plots distributed across the contiguous United elevation, soil type, soil depth, and management history. States and Alaska (Fig. 1) were used in this study. Forest mapping FMPI To study the forest productivity, stocking, and species The FMPI data were collected from permanent sample diversity across the United States, state and regional FIA plots at the forest management unit (FMU) level in the databases, obtained from different FIA regional offices, Wangqing Forestry Bureau (43°05’–43°40’ N, 123°56’– were compiled together into one nationwide master 131°04’ E) of China in 2007 (He et al. 2013). The study database (hereafter, master database). Using SQL queries Fig. 1 Forest types across the 48 contiguous U.S. states and Alaska, derived from both a forest type group map of the contiguous United States and a forest type group map of Alaska. With two sub-types within the Douglas-fir type (coastal Douglas-fir and interior Douglas-fir), we studied a total of 16 forest types across the United States. GCS_WGS_1984 projection for the main map and Alaska Albers Equal Area Conic projection for the inset Watson et al. Forest Ecosystems (2015) 2:22 Page 4 of 16 Fig. 2 Geographic distribution of the 1,389 FMPI plots in Northeastern China of the Microsoft Access program, we first extracted erroneous entries, the final master database consisted of values of 12 key attributes from individual state and re- 475,892 permanent sample plots distributed across the gional FIA databases − plot number, tree species, dbh, contiguous United States and Alaska. tree status, trees per unit area that a sample tree repre- Tree species diversity (N), in terms of species richness, sents, elevation, slope, forest type, site productivity class, was derived from the predesignated numeric codes latitude, longitude, and inventory year. The key attributes, named “SPCD” that identifies species of every single tree which were used to develop the seven final variables used in the FIA databases (Woudenberg et al. 2011). N repre- in this study (Table 1), were all directly measured in the sents the number of species among all the live trees on a field, with an exception of forest type, which was assigned permanent sample plot and across the United States, based on the plot location and the forest type map across and ranges from 1 to 24 (Table 2). the contiguous United States and Alaska at a 250-m spatial Basal area (B), the total cross sectional area of all the live resolution (Ruefenacht et al. 2008). We only selected these trees on a permanent sample plot, was derived from the 12 key attributes for the master database because 1) these FIA attributes dbh, tree status (ST), and trees per unit area attributes are essential for deriving forest productivity, (TPA) that a sample tree represents according to: basal area, and tree species diversity that were used in our study; 2) all individual state and regional databases have 3:14⋅dbh ⋅TPA ST¼1 B ¼ ð1Þ these attributes; and 3) all the redundant or unused attri- butes from individual databases were deleted to keep the master database within our computation and storage cap- where tree status is a field recorded code defining the acity. After removing inconsistent, missing, and apparently status of a tree: 0 indicates no status, 1 live, 2 dead, and Watson et al. Forest Ecosystems (2015) 2:22 Page 5 of 16 Table 1 Definition and units for variables used in this study Variable Units Short definition Long definition 3 −1 −1 C m ·ha ·yr Site productivity A measure of the potential timber growth that the site is capable of producing. It is based on the average annual increment of naturally occurring, fully stocked stands. 2 −1 B m ·ha Stand basal area Total stand basal area of all the living trees from the most current measurement in FIA data. E m Plot elevation The vertical distance that a plot is located from sea level. Positive values indicate that the plot is located above sea level while negative values indicate that the plot was located below sea level. 70 % of the values were obtained from the DEM of the United States and 30 % from ground measurement. S degrees Slope The angle of slope in degrees; 5.6 % of the values were obtained from the DEM of the United States and 94.4 % from ground measurement. N Species richness The total number of different species of woody trees present on the plot. y degrees Latitude Latitude of the plot in NAD 83 x degrees Longitude Longitude of the plot in NAD 83 3 indicates a tree that has been removed (Woudenberg species composition to reduce the total number of forest et al. 2011). Over the contiguous United States and types from 30 to 15 (Fig. 1). 2 −1 Alaska, B ranges from 0 to 279.19 m ·ha (Table 2). Only one of the 15 forest types was divided into sub- The data for elevation (E) and slope (S), which were only types. The Douglas-fir forest type consisted of Douglas-fir used to develop a continuous measure of site productivity forests which grow near the Pacific coast and Douglas-fir based on site class (see Eq. 2 below), were mostly obtained forests which grow much further inland. Because of from the ground-measured FIA databases. However, 70 % substantial differences in the climate and other growing of FIA plots in the master database were missing elevation conditions (Hermann and Lavender 1990), the Douglas-fir records and 5.6 % were missing slope records. To retain forest type was separated into two subtypes, the coastal these plots for the estimation of site productivity, espe- Douglas-fir forests (west of the longitude 120° W) and the cially for those forest types with a relatively small sample inland Douglas-fir forests (east of the longitude 120° W). size (e.g., tropical and exotic hardwoods have only 408 The coastal Douglas-fir forests mainly consist of coast plots and western hardwoods 1,924 plots), we derived the Douglas-fir (Pseudotsuga menziesii var. menziesii), whereas missing data from the plot coordinates and the Digital Ele- the inland Douglas-fir forests are comprised mostly of an- vation Model (DEM) of the United States. For the eleva- other variety of Douglas-fir, namely Rocky Mountain or in- tion, 70 % of data were derived from the DEM of the terior Douglas-fir (P. menziesii var. glauca). United States, and 30 % from the FIA databases. For the Site productivity (C), a measure of the potential timber slope, 5.6 % of data came from the United States DEM growth that a plot is capable of sustaining, was derived and 94.4 % from the FIA databases. The DEM of the from a categorical attribute from the FIA databases contiguous United States (30-m resolution) and Alaska named “SITECLCD” which ranks site productivity in a (60-m resolution) was downloaded from the Geospatial hierarchical order from one to seven. Each code of Data Gateway site (http://datagateway.nrcs.usda.gov/, last SITECLCD denotes a range of productivity: 1 stands for accessed December 18, 2014). the most productive sites with a mean annual increment The forest type map across the contiguous United (MAI, see Hanson et al. 2003) greater than or equal to 3 −1 −1 3 −1 −1 States and Alaska was developed by Ruefenacht et al. 15.74 m ·ha ·yr ; 2 for 11.55–15.74 m ·ha ·yr ;3 3 −1 −1 3 −1 (2008) based on the FIA data. The map has a 250-m for 8.40–11.55 m ·ha ·yr ; 4 for 5.95–8.40 m ·ha · −1 3 −1 −1 3 spatial resolution, with an accuracy of ≈ 69 % for the 48 yr ; 5 for 3.50–5.95 m ·ha ·yr ; 6 for 1.40–3.50 m · −1 −1 contiguous U.S. states, and ≈ 78 % for Alaska. For simpli- ha ·yr , and 7 for the least productive sites with a 3 −1 −1 city, in this study, we grouped some of the original forest MAI less than 1.40 m ·ha ·yr (Woudenberg et al. types together based on similar geographic location and 2011). We converted the categorical attribute of SITECLCD Watson et al. Forest Ecosystems (2015) 2:22 Page 6 of 16 Table 2 Summary statistics by forest type. Std: Standard Deviation, n: total number of plots Productivity Number of Species Basal area Elevation (km) Slope Latitude Longitude 3 −1 −1 2 −1 (m ·ha ·yr ) (m ·ha ) (degrees) (degrees) (degrees) National (all forest types combined) Mean 4.84 6.00 19.13 0.61 6.29 38.20 −88.67 Std. 2.76 3.42 11.96 0.73 8.09 5.71 11.02 Max. 15.03 24.00 279.19 3.92 60.24 61.46 −67.00 Min. −0.50 1.00 0 −0.08 0 24.63 −153.86 n 475,892 475,892 475,892 475,892 475,892 475,892 475,892 Pinyon/juniper Mean 0 2.21 17.68 1.84 11.33 37.06 −110.00 Std. 1.27 1.41 13.27 0.51 9.72 2.84 5.56 Max. 12.41 19.00 130.34 3.35 57.17 48.14 −71.90 Min. −0.50 1.00 0 −0.07 0.00 29.27 −122.78 n 16,709 16,709 16,709 16,709 16,709 16,709 16,709 Douglas-fir (Coastal) Mean 8.43 4.24 23.53 0.18 5.63 45.23 −122.53 Std. 3.48 2.08 27.03 0.22 10.03 1.88 1.01 Max. 14.97 13.00 188.99 1.52 44.71 48.99 −120.00 Min. −0.48 1.00 0 0 0 42.00 −124.68 n 5,866 5,866 5,866 5,866 5,866 5,866 5,866 Douglas-fir (Interior) Mean 4.41 3.70 10.10 0.16 3.54 47.50 −118.26 Std. 2.35 2.00 9.05 0.23 7.74 1.57 0.83 Max. 12.39 12.00 58.71 1.77 43.38 48.99 −116.56 Min. −0.48 1.00 0 0.01 0 43.92 −119.99 n 1,211 1,211 1,211 1,211 1,211 1,211 1,211 Oak/pine Mean 5.61 6.65 17.60 0.52 4.50 33.49 −85.83 Std. 2.59 3.54 10.71 0.67 6.27 2.79 6.05 Max. 15.01 22.00 80.46 3.65 60.24 47.97 −68.82 Min. −0.48 1.00 0 −0.04 0 26.26 −103.00 n 19,023 19,023 19,023 19,023 19,023 19,023 19,023 Oak/gum/cypress Mean 5.91 6.23 20.59 0.53 1.53 32.29 −84.95 Std. 2.56 3.39 13.72 0.70 3.38 2.23 5.47 Max. 15.03 22.00 136.46 3.80 57.17 42.60 −71.40 Min. −0.47 1.00 0 0 0 24.63 −101.37 n 28,431 28,431 28,431 28,431 28,431 28,431 28,431 Watson et al. Forest Ecosystems (2015) 2:22 Page 7 of 16 Table 2 Summary statistics by forest type. Std: Standard Deviation, n: total number of plots (Continued) Productivity Number of Species Basal area Elevation (km) Slope Latitude Longitude 3 −1 −1 2 −1 (m ·ha ·yr ) (m ·ha ) (degrees) (degrees) (degrees) Elm/ash/cottonwood Mean 4.75 5.52 16.87 0.45 3.60 38.30 −91.30 Std. 3.15 3.27 11.04 0.57 6.17 5.37 6.55 Max. 15.03 22.00 130.62 3.61 43.53 61.44 −67.17 Min. −0.47 1.00 0 0 0 26.13 −153.46 n 11,742 11,742 11,742 11,742 11,742 11,742 11,742 Aspen/birch Mean 4.77 18.42 0.60 4.19 45.76 −90.76 −90.76 Std. 2.32 10.53 0.72 5.34 2.30 6.43 6.43 Max. 14.97 16.00 95.88 3.68 56.49 61.42 −67.23 Min. −0.49 1.00 0 0 0 33.62 −151.73 n 46,411 46,411 46,411 46,411 46,411 46,411 46,411 Southern pine Mean 5.97 5.87 17.61 0.53 2.89 32.65 −85.06 Std. 2.46 3.45 11.47 0.68 4.49 2.07 5.46 Max. 15.02 23.00 140.89 3.86 57.17 44.00 −70.00 Min. −0.47 1.00 0 −0.08 0 25.76 −101.23 n 102,844 102,844 102,844 102,844 102,844 102,844 102,844 Oak/hickory Mean 4.97 7.77 18.86 0.45 9.38 37.15 −85.40 Std. 2.47 3.45 9.82 0.58 9.07 3.18 5.80 Max. 15.02 24.00 87.62 3.92 57.17 48.99 −68.75 Min. −0.48 1.00 0 −0.03 0 25.61 −104.38 n 141,062 141,062 141,062 141,062 141,062 141,062 141,062 Maple/beech/birch Mean 4.03 6.22 21.85 0.44 5.88 44.16 −82.11 Std. 2.15 2.57 10.89 0.47 6.68 2.28 8.45 Max. 14.97 20.00 80.58 3.71 57.00 49.00 −67.11 Min. −0.47 1.00 0 0 0 34.81 −102.82 n 47,350 47,350 47,350 47,350 47,350 47,350 47,350 Tropical and exotic hardwoods Mean 3.82 3.14 14.72 0.54 0.95 28.44 −86.44 Std. 2.79 2.52 13.72 0.64 3.10 1.74 7.14 Max. 12.45 13.00 60.24 3.56 34.33 35.32 −80.11 Min. −0.46 1.00 0 0 0 25.79 −120.48 n 408 408 408 408 408 408 408 Watson et al. Forest Ecosystems (2015) 2:22 Page 8 of 16 Table 2 Summary statistics by forest type. Std: Standard Deviation, n: total number of plots (Continued) Productivity Number of Basal area Elevation (km) Slope Latitude Longitude 3 −1 −1 2 −1 (m ·ha ·yr ) Species (m ·ha ) (degrees) (degrees) (degrees) Spruce/fir and exotic softwoods Mean 3.59 4.79 20.33 0.48 2.73 46.54 −85.49 Std. 2.06 2.34 11.99 0.56 4.09 1.49 10.11 Max. 14.96 15.00 80.86 3.68 56.49 61.46 −67.00 Min. −0.46 1.00 0 0 0 38.28 −151.73 n 18,761 18,761 18,761 18,761 18,761 18,761 18,761 Northern pines Mean 4.38 4.69 19.75 0.44 3.87 44.82 −85.23 Std. 2.30 2.49 11.46 0.53 5.34 2.06 6.75 Max. 14.97 15.00 82.54 3.57 41.99 49.31 −67.58 Min. −0.45 1.00 0 0 0 26.33 −96.12 n 9,151 9,151 9,151 9,151 9,151 9,151 9,151 Western conifers Mean 3.80 2.75 24.55 1.86 15.54 43.00 −113.93 Std. 2.93 1.43 17.66 0.78 10.57 4.58 6.88 Max. 15.00 10.00 279.19 3.70 57.17 61.11 −95.06 Min. −0.50 1.00 0 0 0.00 31.80 −153.86 n 21,792 21,792 21,792 21,792 21,792 21,792 21,792 Western hardwoods Mean 2.61 2.08 15.70 0.94 10.56 36.87 −112.68 Std. 4.65 1.46 19.66 0.76 12.26 5.57 8.39 Max. 14.99 8.00 184.50 3.17 47.73 49.00 −99.21 Min. −0.48 1.00 0 −0.03 0 29.05 −124.60 n 1,924 1,924 1,924 1,924 1,924 1,924 1,924 Western oak Mean 1.23 2.35 14.38 1.34 16.87 36.79 −113.99 Std. 3.13 1.37 13.21 0.80 11.71 3.29 7.14 Max. 14.97 10.00 138.16 3.12 57.17 47.47 −98.46 Min. −0.49 1.00 0 0 0 29.57 −124.20 n 3,207 3,207 3,207 3,207 3,207 3,207 3,207 Chinese northeastern temperate forest Mean 3.06 6.86 23.07 0.76 10.67 43.38 130.45 Std. 1.39 2.53 9.90 0.23 7.70 0.12 0.25 Max. 15.30 13.00 75.28 7.70 36.00 43.65 131.05 Min. 0.08 1.00 0 0 0 43.11 129.97 n 1,385 1,385 1,385 1,385 1,385 1,385 1,385 to a continuous attribute of site productivity (C)byac- on previous findings that elevation and slope both have counting for the effects of elevation and slope with a lin- a profound impact on the productivity of forest sites ear model: (Stage and Salas 2007 and references therein). The model was calibrated with FIA data from the entire country by C ¼ α þ α ⋅SITECLCD þ α ⋅E þ α ⋅S ð2Þ 0 1 2 3 fitting observed plot-level values of SITECLCD, E,and S against the mid-point MAI values of the SITECLCD.The where α through α were coefficients to be estimated by model was examined for the level of significance of the co- 0 3 ordinary least squares (Table 3). This model was based efficients, the biological interpretation, and the normality Watson et al. Forest Ecosystems (2015) 2:22 Page 9 of 16 Table 3 Parameters of the site productivity model and a majority of forest dynamics studies (e.g., Shugart Independent variables Constant 1984) in terms of basal area effects. It should be noted, however, that we did not consider any abiotic factors in SITECLCD E S this study except for elevation and slope (Eq. 2) due to a Coef. −2.442 0.082 0.002 16.960 lack of soil and other environmental records. Subject to SE 0.000 0.001 0.000 0.002 the potential omitted-variable bias (Wooldridge 2000), P <0.0001 <0.0001 <0.0001 <0.0001 our results should be interpreted with caution. For in- R 0.98 stance, our results may address the effects of biodiversity dF 475,888 on productivity, but the abiotic causes of biodiversity variation across the study region and their potential con- of residual pattern. Then, we estimated the actual site prod- founding effect on BEFR were not directly detectable uctivity of 475,892 FIA plots using the values of SITECLCD, with this database. Nevertheless, as an inherent nature E,and S in Eq. 2, thereby making the site productivity of of abiotic factors, their potential effects especially on for- FIA plots a continuous variable across the country. est productivity are in general spatially autocorrelated The master database was created to produce nation- (Legendre 1993; Liang 2012). To this end, we also wide maps of forest stocking (basal area), tree species employed a geospatial model to account for spatial auto- richness, and site productivity. For the creation of each correlation and potential effects of abiotic factors. of these maps, two separate data layers were developed; Due to potential spatial autocorrelation, which can one for Alaska and one for the contiguous United States. bias tests of significance due to the violation of inde- Each plot represents a raster with size 0.05° for the con- pendence (Clifford et al. 1989) in the FIA data, we mea- tiguous United States and size 0.2° for Alaska. The GCS- sured the spatial autocorrelation function for C for each WGS-1984 projection was used for the contiguous forest type using a nonparametric approach (Bjørnstad United States and the Alaska Albers Equal Area Conic and Falck 2001). Due to computational constraints, it projection (NAD_1983 datum) was used for Alaska. was not possible to estimate the spatial autocorrelation function for most forest type’s entire set of data; for Diversity-productivity relationship example, the maximum number of sample locations For each of the 15 forest types across the contiguous for any one forest type was 141,062, while in 75 % of United States and Alaska, a multiple regression analysis the forest types the number of spatial locations exceeded was conducted to test the general effect of tree species 7,000. To overcome this constraint, we used a bootstrapping richness (N) on site productivity (C), with forest stocking sampling approach in which, for each forest type, up to ≈ (B) being accounted for: 2,000 spatial locations and their associated value of C were randomly selected, from which we estimated parameters of the spatial autocorrelation. This procedure was re- C ¼ β þ β ⋅N þ β ⋅N þ β ⋅N ⋅B ij ij ij ij 0;i 1;i 2;i ij 3;i peated independently 200 times, from which we estimated þ β ⋅B þ e ð3Þ ij ij 4;i the mean, and the 95 % confidence intervals using the 0.025 % and 0.975 % quantiles of the bootstrapped distri- 2 −1 where B represents forest stocking (m ·ha )and C bution of the parameters (Efron and Tibshirani 1993). 3 −1 −1 represents site productivity (m ·ha ·yr ), i forest Since C for most forest types revealed initially some level type (i=1,2,3,…, 15), and j permanent sample plot of spatial autocorrelation, as determined by 95 % confi- number within the ith forest type. e represents the ran- dence intervals that did not include 0 for the estimate of dom error term. Coefficients (β’s) were estimated with the local spatial autocorrelation (Bjørnstad and Falck ordinary least squares, based on two assumptions: the 2001), we sought to detrend the data using a second-order polynomial spatial model fit to values of C for each forest sample is random, and the error term is of zero condi- tional mean and homoskedasticity. The full quadratic type according to: terms of N and B were incorporated in Eq. 3 to study 2 2 C ¼ x þ y þ xy þ x þ y ð4Þ ;y how site productivity (C) changes in response to changes in species richness (N) while keeping basal area in which x and y represent the longitude and latitude, (B)constant atits sample mean forthatforesttype. respectively, of the spatial location corresponding to the We employed Eq. 3 to study BEFR across different for- value of C. In this case, we used all values of C for each est types, with an underlying assumption that biotic fac- forest type as opposed to a randomly chosen subset. We tors, namely species diversity and basal area, have direct used mixed stepwise regression to determine the appro- causal effects on forest productivity. This assumption priate model. Ultimately, 13 of the 15 forest types exhib- was supported by recent BEFR studies (e.g., Zhang et al. ited significant spatial autocorrelation in values of C as 2012; Liang et al. 2015) in terms of biodiversity effects, ascertained by significance of at least one parameter Watson et al. Forest Ecosystems (2015) 2:22 Page 10 of 16 Table 4 Estimates of spatial autocorrelation and the appropriate spatial detrending model for all 15 forest types Forest type Spatial autocorrelation estimates Detrending model 1 2 3 Local spatial autocorrelation Range of spatial autocorrelation (km) Parameters n Pinyon/juniper 0.33 1,302.7 All 16,709 Douglas-fir 0.45 260.2 All 7,077 Oak/pine 0.21 490.0 All 19,023 Oak/gum/cypress 0.16 597.4 All 28,431 Elm/ash/cottonwood 0.51 617.5 All 11,742 Aspen/birch 0.13 536.6 All 46,411 Southern pine 0.19 617.0 All 102,844 2 2 Oak/hickory 0.14 524.0 x, y, x , y 141,062 Maple/beech/birch NS NS NS 47,350 Tropical and exotic hardwoods 0.35 404.6 All 408 2 2 Northern pines 0.08 219.7 x, y, x , y 9,151 Spruce/fir and exotic softwoods NS NS NS 18,761 Western conifers 0.39 973.2 x, xy, y 21,792 2 2 Western hardwoods 0.65 1,167.8 x, x , y 1,924 Western oak 0.37 990.7 x, y, xy, y 3,207 Empirical mean (from 200 bootstrapped simulations) of the estimate of local autocorrelation as the distance between sampling locations approaches 0 Mean distance (from 200 bootstrapped simulations) of the lag distance at which the estimate of local autocorrelation = 0 3 2 2 Full model parameters =x, y, xy, x , and y NS not significant value in Eq. 4 (Table 4). For these 13 forest types, we autocorrelation in the residuals based upon the 95 % con- then obtained the detrended residuals from Eq. 4.To fidence intervals. Thus, the spatially-detrended residuals ensure that the second-order polynomial spatial model for these 13 forest types were used in all follow-up ana- adequately removed the spatial autocorrelation, we then lyses. For the remaining two forest types that did not ex- estimated the spatial autocorrelation of the residuals ob- hibit significant spatial autocorrelation (maple/beech/ tained from Eq. 4 for these 13 forest types using the boot- birch, and spruce/fir and exotic softwoods, Table 4), we strap sampling approach described above. In all forest used values of C in all follow-up analyses. All analyses type cases, there was no significant local spatial were conducted in R 2.14.0. Fig. 3 Richness of woody plant species across the 48 contiguous U.S. states and Alaska, derived from FIA ground measurements completed between 1968 and 2011 Watson et al. Forest Ecosystems (2015) 2:22 Page 11 of 16 2 −1 Fig. 4 Total forest stand basal area (m ·ha ) across the 48 contiguous U.S. states and Alaska, derived from FIA ground measurements completed between 1968 and 2011 Results values were concentrated in the western United States Geographic distribution of tree species richness, forest and Central Appalachia (Fig. 4). productivity and stocking Forests with high site productivity were generally dis- The southeastern region (with an exception of the Gulf tributed in the western slopes of the mountain ranges in and Atlantic Coastal plains) and the Appalachian Moun- northern and central California, western Oregon and tains showed the highest tree species richness, whereas Washington, northern Idaho, the southeastern United the majority of the mountainous regions of the western States (except Florida), southern Michigan, and the United States, the black hills region of South Dakota, states of Illinois and Indiana. The areas with the most western Texas, northwestern Minnesota, and the state of notable overall low values of site productivity were the Florida had the lowest tree species richness (Fig. 3). In southern Rocky Mountains and western Texas (Fig. 5). general, forests in the eastern United States consisted of Oak/hickory and southern pine forests had the overall more tree species than those in the central and western highest levels of species richness, and the western hard- parts of the country including the state of Alaska. Over woods, western conifers and western oak forests had the contiguous United States, the highest forest stocking the lowest (Table 2). The mean productivity over all of 3 −1 −1 Fig. 5 Forest productivity (m ·ha ·yr ) across the 48 contiguous U.S. states and Alaska, derived from FIA ground measurements completed between 1968 and 2011 Watson et al. Forest Ecosystems (2015) 2:22 Page 12 of 16 3 −1 −1 3 −1 −1 the forest types studied here was 4.84 m ·ha ·yr . western oak forest type, 0.9 m ·ha ·yr for the south- 3 −1 −1 The most productive forest type was oak/gum/cypress ern pine forest type, 1.6 m ·ha ·yr for the spruce/fir 3 −1 −1 whereas the least productive was pinyon/juniper. Aver- and exotic softwoods forest type, 0.6 m ·ha yr for 3 −1 −1 age species richness for all forest types combined is 6.00 the oak/hickory forest type, and 1.5 m ·ha ·yr for with a standard deviation of 3.42 (Table 2). the western conifers forest type (Fig. 6). The forests that showed a flat or negative diversity-productivity relation- Diversity-productivity relationship ship were the coastal Douglas-fir subtype with a decline 3 −1 −1 Throughout all the 15 forest types in the United States, 12 of −1.0 m ·ha ·yr , the northern pines forest type plus the interior Douglas-fir subtype showed a positive re- (flat), the tropical and exotic hardwoods forest type lationship between species richness and site productivity, (flat), and the western hardwoods forest type with a de- 3 −1 −1 and only the coastal Douglas-fir subtype, northern pines, cline of −1.5 m ·ha ·yr (Fig. 6). tropical and exotic hardwoods, and western hardwoods The FMPI data from Northeastern China conformed forest types had a negative or insignificant relationship to the positive diversity-productivity relationship (Table 5). Species richness was highly significant for all (Table 5). As species richness increased from 1 to 12, but the tropical and exotic hardwoods forest type. Over productivity of the northeastern temperate forests in 3 −1 −1 all the study areas, 96.4 % of sample plots (477,281) con- China improved from 2.5 to 4.0 m ·ha ·yr , a 60 % formed to a positive effect of species richness on site increase from the base value (Fig. 7). productivity, and only 3.6 % (17,349) showed an insig- nificant or negative effect. Discussion Based on the estimated coefficients of Eq. 3, when B The findings largely support, from the perspectives of was kept constant at its sample mean, as species richness forested ecosystems over a large geographic scale, a posi- increased from 1 to the 75th percentile values, site prod- tive biodiversity-ecosystem functioning relationship 3 −1 −1 uctivity was expected to increase by 1.2 m ·ha ·yr (BEFR). This is consistent with other experiments based 3 −1 −1 for the pinyon/juniper forest type, 4.0 m ·ha ·yr for generally on controlled field experiments with herb- 3 −1 −1 the interior Douglas-fir subtype, 1.0 m ·ha ·yr for aceous species (see Cardinale et al. 2012 and references 3 −1 −1 the oak/pine forest type, 1.6 m ·ha ·yr for the oak/ therein). Evidence from this study was unique because it 3 −1 −1 gum/cypress forest type, 0.3 m ·ha ·yr for the elm/ was based upon almost half a million ground-measured 3 −1 −1 ash/cottonwood forest type, 1.0 m ·ha ·yr for the forest inventory plots from a large geographic scale. Rec- 3 −1 −1 aspen/birch forest type, 1.8 m ·ha ·yr for the ognizing that different forest types vary in regards to Table 5 Parameters of site productivity models by forest type with predictor variables of species richness, basal area, elevation, and slope Forest type n AIC BIC Coefficients Const. N N N·B B Pinyon/juniper 16,709 47112 47158 −0.398 *** 0.209 *** −0.012 *** 0.005 *** −0.010 *** Douglas-fir (Coastal) 5,866 29214 29241 −0.024 −0.011 *** 0.002 *** Douglas-fir (Interior) 1,211 5259 5284 −2.735 *** 0.809 *** −0.035 ** 0.031 *** Oak/pine 19,023 86805 86853 −1.599 *** 0.205 *** −0.008 *** −0.001 * 0.047 *** Oak/gum/cypress 28,431 128225 128266 −1.058 *** 0.142 *** −0.002 *** 0.019 *** Elm/ash/cottonwood 11,742 55755 55800 −1.394 *** 0.244 *** −0.010 *** −0.003 *** 0.048 *** Aspen/birch 46,411 200909 200953 −0.865 *** 0.105 *** −0.004 *** 0.040 *** Southern pine 102,844 457216 457273 −1.070 *** 0.093 *** −0.001 ** −0.001 *** 0.040 *** Oak/hickory 141,062 636358 636397 −0.798 *** 0.040 *** 0.026 *** Maple/beech/birch 47,350 206442 206495 3.703 *** −0.033 * 0.013 *** −0.004 *** 0.024 *** Tropical and exotic hardwoods 408 1700 1712 −0.409 ** 0.028 *** Northern pines 9,151 40244 40280 −1.469 *** 0.161 *** −0.007 *** 0.075 *** Spruce/fir and exotic softwoods 18,761 79280 79327 2.011 *** 0.479 *** −0.025 *** −0.002 ** 0.008 * Western conifers 21,792 101207 101247 −1.409 *** 0.184 *** 0.003 *** 0.029 *** Western hardwoods 1,924 8041 8075 −0.386 ** 0.446 ** −0.125 *** 0.010 *** −0.017 * Western oak 3,207 15060 15090 −1.311 *** 0.409 *** −0.037 * 0.015 *** Chinese northeastern temperate forest 1,385 4768 4789 1.288 *** 0.397 *** −0.017 *** −0.0007 0.004 Level of significance: < 0.001: ***; 0.001: **; 0.01: * Watson et al. Forest Ecosystems (2015) 2:22 Page 13 of 16 3 −1 −1 Fig. 6 Sensitivity of stand productivity (m ·ha ·yr ) to species richness for 15 forest types (each panel represents one type of forest) across the 48 contiguous U.S. states and Alaska. Solid lines represent predicted means of different forest types and broken lines the 95 % confidence interval of the predicted means, with stand basal area being kept constant at its sample mean how species diversity affects productivity, we categorized growth, this study quantified site productivity as potential all the ground-measured plots into 15 forest types, and timber growth that a site could sustain. Our results indi- analyzed the diversity-productivity relationship specific cate that intensively managed coastal Douglas-fir forests to each forest type. The number of ground-measured feature a negative effect of diversity on potential timber plots, as well as the magnitude of geographic scale, ren- growth presumably because these stands are artificially dered overwhelming evidence in support of a positive maintained in an early stage of stand development (stem tree species diversity-timber productivity relationship. exclusion) where current annual increment is nearly opti- The negative diversity productivity relationship for the mized at a low diversity. The inland Douglas-fir forests coastal Douglas-fir subtype that was found (Fig. 6b) con- conformed to the positive biodiversity-forest productivity tradicts with the positive net basal area change in associ- relationship as they are less intensively managed. It should ation with tree species diversity reported by Liang et al. be noted, however, that this implication was only applicable (2007) for the same subtype. The main reason for this to Douglas-fir forests, which are in general low in tree spe- discrepancy may be in the measure of site productivity. cies diversity. Southern pine forests, in spite of high man- While Liang et al. (2007) measured site productivity by net agement intensity, still show a positive tree species annual basal area change, which represents actual forest diversity-timber productivity relationship (Fig. 6j). A Watson et al. Forest Ecosystems (2015) 2:22 Page 14 of 16 FIA plots can be treated as a normally distributed ran- dom process. Furthermore, as most forests across the United States are distributed on relatively flat surfaces, bias in elevation that is caused by fuzzed coordinates was limited. Slope data is the most sensitive to plot loca- tions, but the accuracy of the site productivity model was maintained as 94 % of the data were measured in the field and only 6 % were estimated using the fuzzed coordinates. Conclusion Over all the study areas, 96.4 percent of sample plots 3 −1 −1 (477,281) showed a positive effect of species richness on Fig. 7 Sensitivity of stand productivity (m ·ha ·yr ) to species richness for the northeastern temperate forests in China. Solid lines site productivity, and only 3.6 percent (17,349) had an represent predicted means of different forest types and broken lines insignificant or negative effect. the 95 % confidence interval of the predicted means, with stand The results of this study suggest that maintaining spe- basal area being kept constant at its sample mean cies diversity is an important means to maintain forest productivity, which is supported by an array of forestry studies (Kelty 1989; Caspersen and Pacala 2001; Liang et al. 2005, 2007; Lei et al. 2009b; Young et al. 2011; possible explanation is southern pine forests by nature Zhang et al. 2012; Gamfeldt et al. 2013). These results consist of more tree species than Douglas-fir forests. Also, should assist landowners in making management deci- the niche complementarity effect (see Loreau and Hector sions that are relevant to the specific forest types that 2001 and references therein) that contributes to the posi- they respectively manage. These findings also imply that tive BEFR could therefore be more prominent in the south- productivity of forests across the United States may be ern pine forests. impaired by the loss of both woody and non-woody Due to the use of observational data, the observed plant species in forested ecosystems (Fleming et al. trends are subject to the usual caveats of multicollinear- 2011; Liang et al. 2015), and that biological conserva- ity. However, multicollinearity does not lead to a biased tion, due to its potential benefits in maintaining forest diversity-productivity relationship (Goldberger 1991), productivity, can have profound impacts on the prod- even though it may lead to difficulties in quantifying the uctivity of selected services that can be obtained from variance of predicted means. The partial effects of diver- forests across the United States. sity may be more uncertain due to multicollinearity. Nevertheless, multicollinearity was not excessive in the Competing interests present case as the Variance Inflation Factor of tree spe- The authors declare that they have no competing interests. cies richness (N) was estimated to be 1.08 using the “car” package of R (Fox and Weisberg 2011), and all the Authors’ contributions JVW, JL, and XL compiled the data. JVW, JL, and CEA conducted mapping. forest types show very low standard errors in predicted JVW, JL, and PCT performed regression analysis. All authors contributed to site productivity (Fig. 6). Another statistical caveat of this the writing, and read and approved the final manuscript. study, due to the use of observational data, is in the causal relationship between diversity and productivity. Acknowledgements Without a controlled experiment, it is difficult to deter- We are obliged to Karen Waddell and Charles J. Barnett for assistance with FIA data. We thank Mo Zhou, Jacquelyn Strager, George Merovich, and Eric mine the cause and effect in the diversity-productivity King for statistical and mapping assistance. This research is supported in relationship. Compared to controlled experiments, this parts by the United States Department of Agriculture McIntire-Stennis Act empirical evidence provides insights to forest manage- Fund WVA00104, and by the Division of Forestry and Natural Resources, West Virginia University. ment and biological conservation that are of a much broader applicability, both in terms of forest type and Author details geographic scale. Davis College of Agriculture, Natural Resources and Design, West Virginia University, 26506 Morgantown, WV, USA. Forest Service, U.S. Department of The fuzzed FIA plot coordinates could affect the ac- Agriculture, Northern Research Station, 26505 Morgantown, WV, USA. School curacy of the estimated site productivity (C), as 70 % of of Environmental and Forest Sciences, University of Washington, 98195 plot elevation records and 6 % of plot slope records were Seattle, WA, USA. Research Institute of Forest Resource Information Techniques, Chinese Academy of Forestry, 100091 Beijing, P. R. China. obtained from a DEM using the fuzzed coordinates. Nevertheless, the impact should be small, as the differ- Received: 26 December 2014 Accepted: 23 June 2015 ences between fuzzed coordinates and true locations of Watson et al. Forest Ecosystems (2015) 2:22 Page 15 of 16 References Kelty MJ (1989) Productivity of New England hemlock/hardwood stands as affected Assmann E (1970) The principles of forest yield study. Pergamon Press, Oxford, UK by species composition and canopy structure. For Ecol Manage 28:237–257 Bechtold WA, Scott CT (2005) The forest inventory and analysis plot design. In: Köble R, Seufert G (2001) Novel maps for forest tree species in Europe. In: the enhanced forest inventory and analysis program: national sampling design Hjorth J, Raes F, Angeletti G (eds) Proceedings of the 8th European and estimation procedures. USDA Forest Service, Asheville, NC, pp 27–42 symposium on the physico-chemical behaviour of air pollutants: a changing Bjørnstad ON, Falck W (2001) Nonparametric spatial covariance functions: atmosphere. Commission of the European Communities, Directorate General estimation and testing. Environ Ecol Stat 8:53–70 Telecommunications, Information Industries and Innovation. Torino, Italy, pp 17–20 Kreft H, Jetz W (2007) Global patterns and determinants of vascular plant Buongiorno J, Peyron JL, Houllier F, Bruciamacchie M (1995) Growth and management of mixed-species, uneven-aged forests in the French Jura: diversity. Proc Natl Acad Sci U S A 104:5925–5930 implications for economic returns and tree diversity. For Sci 41:397–429 Legendre P (1993) Spatial autocorrelation: trouble or new paradigm? Ecology 74:1659–1673 Cardinale BJ, Duffy JE, Gonzalez A, Hooper DU, Perrings C, Venail P, Narwani A, Mace GM, Tilman D, Wardle DA, Kinzig AP, Daily GC, Loreau M, Grace JB, Lei X, Tang M, Lu Y, Hong L, Tian D (2009a) Forest inventory in China: status and challenges. Int For Rev 11:52–63 Larigauderie A, Srivastava DS, Naeem S (2012) Biodiversity loss and its impact on humanity. Nature 486:59–67 Lei X, Wang W, Peng C (2009b) Relationships between stand growth and Cardinale BJ, Matulich KL, Hooper DU, Byrnes JE, Duffy E, Gamfeldt L, Balvanera P, structural diversity in spruce-dominated forests in New Brunswick, Canada. O’Connor MI, Gonzalez A (2011) The functional role of producer diversity in Can J For Res 39:1835–1847 ecosystems. Am J Bot 98:572–592 Leuschner C, Jungkunst HF, Fleck S (2009) Functional role of forest diversity: pros Caspersen JP, Pacala SW (2001) Successional diversity and forest ecosystem and cons of synthetic stands and across-site comparisons in established function. Ecol Res 16:895–903 forests. Basic Appl Ecol 10:1–9 Chen HYH, Klinka K, Mathey AH, Wang X, Varga P, Chourmouzis C (2003) Are Liang J (2012) Mapping large-scale forest dynamics: a geospatial approach. mixed-species stands more productive than single-species stands: an empirical Landscape Ecol 27:1091–1108 test of three forest types in British Columbia and Alberta. Can J For Res Liang J, Buongiorno J, Monserud RA (2005) Growth and yield of all-aged 33:1227–1237 Douglas-fir/western hemlock stands: a Matrix model with stand diversity effects. Can J For Res 35:2369–2382 Clifford P, Richardson S, Hémon D (1989) Assessing the significance of the correlation between two spatial processes. Biometrics 45:123–134 Liang J, Buongiorno J, Monserud RA (2006) Bootstrap simulation and response Efron B, Tibshirani RJ (1993) An introduction to the bootstrap. Chapman & Hall, surface optimization of management regimes for Douglas-fir/western New York hemlock stands. For Sci 52:579–594 FAO (2012) State of the world’s forests 2012. Food and Agriculture Organization Liang J, Buongiorno J, Monserud RA, Kruger EL, Zhou M (2007) Effects of diversity of the United Nations, Rome, Italy, p 46 of tree species and size on forest basal area growth, recruitment, and Fleming R, Brown N, Jenik J, Kahumbu P, Plesnik J (2011) Emerging perspectives mortality. For Ecol Manage 243:116–127 on forest biodiversity. In: United Nations Environment Program (ed) UNEP Liang J, Zhou M, Tobin PC, McGuire AD, Reich PB (2015) Biodiversity influences year book 2011. UNEP, Nairobi, Kenya, pp 46–59 plant productivity through niche-efficiency. Proc Natl Acad Sci U S A 112:5738–5743 Fox J, Weisberg S (2011) An R companion to applied regression. Sage, Thousand Lister A, Scott C, King S, Hoppus M, Butler B, Griffith D (2002) Strategies for Oaks, CA preserving owner privacy in the national information management system Gamfeldt L, Snall T, Bagchi R, Jonsson M, Gustafsson L, Kjellander P, Ruiz-Jaen MC, of the USDA Forest Service’s forest inventory and analysis unit. In: McRoberts Froberg M, Stendahl J, Philipson CD, Mikusinski G, Andersson E, Westerlund RE (ed) The fourth annual forest inventory and analysis symposium. USDA B, Andren H, Moberg F, Moen J, Bengtsson J (2013) Higher levels of multiple Forest Service, New Orleans, LA, pp 163–166 ecosystem services are found in forests with more tree species. Nat Commun 4:1340 Loreau M, Hector A (2001) Partitioning selection and complementarity in Gillespie AJ (1999) Rationale for a national annual forest inventory program. J For biodiversity experiments. Nature 412:72–76 97:16–20 Loreau M, Naeem S, Inchausti P, Bengtsson J, Grime JP, Hector A, Hooper DU, Goldberger AS (1991) A course in econometrics. Harvard University Press, Huston MA, Raffaelli D, Schmid B, Tilman D, Wardle DA (2001) Biodiversity Cambridge, MA and ecosystem functioning: current knowledge and future challenges. Science 294:804–808 Haight RG, Monserud RA (1990) Optimizing any-aged management of mixed-species Naeem S, Duffy JE, Zavaleta E (2012) The functions of biological diversity in an stands: II. Effects of decision criteria. For Sci 36:125–144 age of extinction. Science 336:1401–1406 Hanson EJ, Azuma DL, Hiserote BA (2003) Site index equations and mean annual Paquette A, Messier C (2011) The effect of biodiversity on tree productivity: from increment equations for Pacific Northwest Research Station forest inventory temperate to boreal forests. Global Ecol Biogeogr 20:170–180 and analysis inventories, 1985–2001. U.S. Dept. of Agriculture, Forest Service, Pacific Northwest Research Station, Portland, OR Reich PB, Tilman D, Isbell F, Mueller K, Hobbie SE, Flynn DFB, Eisenhauer N (2012) Hasse WD, Ek AR (1981) A simulated comparison of yields for even- versus Impacts of biodiversity loss escalate through time as redundancy fades. uneven-aged management of northern hardwood stands. J Environ Manage Science 336:589–592 12:235–246 Ricketts TH (1999) Terrestrial ecoregions of North America: a conservation He P, Zhang H, Lei X, Li X (2013) Estimation of spatial distribution of tree species assessment. Island Press, Washington, DC diversity based on Universal Krige Model. J Central-South Univ For Technol Ruefenacht B, Finco MV, Nelson MD, Czaplewski R, Helmer EH, Blackard JA, 33:67–71 Holden GR, Lister AJ, Salajanu D, Weyermann D, Winterberger K (2008) Conterminous US and Alaska forest type mapping using forest inventory and Hermann RK, Lavender DP (1990) Douglas-fir. In: Burns RM, Honkala BH (eds) analysis data Photogramm. Eng Remote Sensing 74:1379–1388 Silvics of North America. U.S. Department of Agriculture, Forest Service, Shugart HH (1984) A theory of forest dynamics: the ecological implications of Washington DC, pp 527–540 forest succession models. Springer Verlag, New York Hernández-Stefanoni JL, Dupuy JM (2007) Mapping species density of trees, shrubs and vines in a tropical forest, using field measurements, satellite Stage AR, Salas C (2007) Interaction of elevation, aspect, and slope in models of multiespectral imagery and spatial interpolation. Biodivers Conserv forest species composition and productivity. For Sci 53:486–492 16:3817–3833 Symstad AJ, Chapin FS, Wall DH, Gross KL, Huenneke LF, Mittelbach GG, Hernandez-Stefanoni JL, Ponce-Hernandez R (2006) Mapping the spatial variability of Peters DP, Tilman D (2003) Long-term and large-scale perspectives on plant diversity in a tropical forest: comparison of spatial interpolation methods. the relationship between biodiversity and ecosystem functioning. Environ Monit Assess 117:307–334 Bioscience 53:89–98 Hooper DU, Chapin FS, Ewel JJ, Hector A, Inchausti P, Lavorel S, Lawton JH, Vilà M, Vayreda J, Gracia C, Ibáñez JJ (2003) Does tree diversity increase wood Lodge DM, Loreau M, Naeem S, Schmid B, Setala H, Symstad AJ, Vandermeer production in pine forests? Oecologia 135:299–303 J, Wardle DA (2005) Effects of biodiversity on ecosystem functioning: a Wardle DA (1999) Is “sampling effect” a problem for experiments investigating consensus of current knowledge. Ecol Monogr 75:3–35 biodiversity-ecosystem function relationships? Oikos 87:403–407 Hooper DU, Vitousek PM (1997) The effects of plant composition and diversity on Wooldridge JM (2000) Introductory econometrics: a modern approach. South-Western ecosystem processes. Science 277:1302–1305 College Publishing, Cincinnati, OH Watson et al. Forest Ecosystems (2015) 2:22 Page 16 of 16 Woudenberg S, Conkling B, O’Connell B, LaPoint E, Turner J, Waddell K, Boyer D, Christensen G, Ridley T (2011) The forest inventory and analysis database: description and users manual version 51 for phase 2. US Department of Agriculture, Forest Service, Fort Collins, CO, p 336 Young B, Liang J, Stuart Chapin F III (2011) Effects of species and tree size diversity on recruitment in the Alaskan boreal forest: a geospatial approach. For Ecol Manage 262:1608–1617 Zhang Y, Chen HYH, Reich PB (2012) Forest productivity increases with evenness, species richness and trait variation: a global meta-analysis. J Ecol 100:742–749 Submit your manuscript to a journal and benefi t from: 7 Convenient online submission 7 Rigorous peer review 7 Immediate publication on acceptance 7 Open access: articles freely available online 7 High visibility within the fi eld 7 Retaining the copyright to your article Submit your next manuscript at 7 springeropen.com

Journal

"Forest Ecosystems"Springer Journals

Published: Dec 1, 2015

Keywords: Ecology; Ecosystems; Forestry

There are no references for this article.