Environ Monit Assess (2010) 166:543561DOI 10.1007/s10661-009-1022-6

Implications of differing input data sources and approachesupon forest carbon stock estimation

Michael A. Wulder Joanne C. White Graham Stinson Thomas Hilker Werner A. Kurz Nicholas C. Coops Benit St-Onge J. A. (Tony) Trofymow

Received: 18 December 2008 / Accepted: 26 May 2009 / Published online: 11 June 2009 Her Majesty the Queen in Right of Canada 2009

Abstract Site index is an important forest inven-tory attribute that relates productivity and growthexpectation of forests over time. In forest in-ventory programs, site index is used in conjunc-tion with other forest inventory attributes (i.e.,height, age) for the estimation of stand volume.In turn, stand volumes are used to estimate bio-mass (and biomass components) and enable con-version to carbon. In this research, we explorethe implications and consequences of differentestimates of site index on carbon stock charac-terization for a 2,500-ha Douglas-fir-dominatedlandscape located on Eastern Vancouver Island,British Columbia, Canada. We compared site in-dex estimates from an existing forest inventoryto estimates generated from a combination of

M. A. Wulder (B) J. C. White G. Stinson W. A. Kurz J. A. (Tony) TrofymowCanadian Forest Service (Pacific Forestry Center),Natural Resources Canada, 506 West Burnside Rd.,Victoria, BC V8Z 1M5, Canadae-mail: mwulder@nrcan.gc.ca

T. Hilker N. C. CoopsDepartment of Forest Resource Management,University of British Columbia,Vancouver, British Columbia, Canada

B. St-OngeDepartment of Geography, Universit of Qubecin Montral, Montral, Quebec, Canada

forest inventory and light detection and rang-ing (LIDAR)-derived attributes and then exam-ined the resultant differences in biomass estimatesgenerated from a carbon budget model (CarbonBudget Model of the Canadian Forest Sector(CBM-CFS3)). Significant differences were foundbetween the original and LIDAR-derived site in-dices for all species types and for the resulting5-m site classes (p < 0.001). The LIDAR-derivedsite class was greater than the original site classfor 42% of stands; however, 77% of stands werewithin 1 site class of the original class. Differ-ences in biomass estimates between the modelscenarios were significant for both total standbiomass and biomass per hectare (p < 0.001); dif-ferences for Douglas-fir-dominated stands (rep-resenting 85% of all stands) were not significant(p = 0.288). Overall, the relationship between thetwo biomass estimates was strong (R2 = 0.92, p 200 non-ground LIDAR hits/ha)to estimate stand height. This subset of 1320 forestinventory polygons was used to compare the site-index values and subsequent CBM-CFS3 outputs.

Estimation of stand biomass and C stocks

Total stand biomass and C stocks and compo-nents (i.e., above- and below-ground biomass,dead wood, litter, and soil organic matter) wereestimated using the CBM-CFS3, an annual time-step model that uses forest inventory data alongwith merchantable volume yield tables to drivestand-level biomass C dynamics (Kurz et al. 1992;Kurz and Apps 1999; Kurz et al. 2002; Kull et al.2006). In this study, stand-level biomass carbondynamics were simulated using merchantable vol-ume yield tables derived for natural and managedstands in Coastal British Columbia, stratified byleading species and 5-m site index classes. Theyield tables used for managed Douglas-fir leadingstands (the most common stand type present inthe study area) are provided in Table 3. Otheryield tables were used for other stand typespresent in the study area, including hemlock-,red cedar-, amabilis-, and alder-dominated stands.These yield tables were developed by the BritishColumbia Ministry of Forests and Range usingthe Table Interpolation Program for Stand Yields(TIPSY) developed from the TASS individualtree growth model (Mitchell et al. 2000). In thisapplication of the CBM-CFS3, each forest coverpolygon in the forest inventory is treated as aseparate database record and is assigned a yieldtable on the basis of stand type (defined accordingto species mix) and 5-m site class (Kurz et al.2002). For CBM-CFS3, the continuous site-indexvalues for SIINV and SILIDAR were categorizedinto 5-m site classes (SC), with the site-index val-ues assigned to the closest 5-m interval (e.g., asite index of 12.4 is assigned a site class of 10; a

site index of 12.5 is assigned a site class of 15) togenerate SCINV and SCLIDAR.

Merchantable volume yields are converted inthe model to annual above-ground biomass car-bon (C) increments referenced to projected standage using stand-level biomass estimation models(Boudewyn et al. 2007) and assuming a C contentof 0.5 t C t1 biomass. Below-ground (root) bio-mass is estimated in the model as a function ofaboveground biomass and species group (Li et al.2003). CBM-CFS reports in units of tonnes C, andthis is multiplied by two to get biomass (tonnes).The entire study area has been logged at leastonce since 1920, so the entire stand history wassimulated for all stands present on the landscape.

Results

Stand height and site-index estimation

Mean stand heights were determined for eachforest stand. Although no field measures wereavailable for comparison at this study site, Coopset al. (2007), working in a similar area with similarspecies, data, and methods for establishing meanplot height from LIDAR, found that LIDAR esti-mates of plot height were consistently lower thanfield-measured plot heights. As both the site indexand modeled biomass estimates were generatedfrom different data sources, they were expectedto vary. The distribution of SIINV and SILIDAR forall 1320 stands is shown in Fig. 2. Regressions ofSIINV on SILIDAR produced a low overall R2 valueof 0.03 (p< 0.001; Fig. 3) and the stand-level dif-ferences between SIINV and SILIDAR were foundto be statistically significant (dependent samples,two-tailed t-test, = 0.05, t = 9.224, p < 0.001,df = 1, 319). Figure 4 characterizes the differencesbetween site index estimates by leading species;statistically significant differences were found forall leading species except western red cedar lead-ing stands, which also had the largest standarderror (Table 4). SILIDAR are greater than the SIINVfor all stands sampled and for all species groups.

The site-index values were reclassified into5-m site classes for input into CBM-CFS3. The re-lationship between SCINV and SCLIDAR was weak(R2 = 0.02, p < 0.001), and there was a significant

Environ Monit Assess (2010) 166:543561 553

Fig. 2 Comparison of thedistribution of site indexestimates in the originalforest inventory and thesite index estimates madeusing LIDAR-derivedmean stand heights

difference between the two site class estimates(dependent samples two-tailed t-test, = 0.05,t = 8.972, p < 0.001, df = 1, 319) (Table 4).Figure 5 shows the distribution of site class differ-ences (SCINVSCLIDAR); the LIDAR-derived siteclasses were greater than the corresponding inven-tory site classes for 42% of stands. The majorityof stands (77%) were within 1 class of the siteclass estimated from the original site-index values.Figure 6 illustrates the differences in site classby leading species (sample sizes in Table 1) and

indicates that for all species, SCLIDAR was greaterthan SCINV. Amabilis fir dominated stands hadthe greatest difference in site class, followed bywestern hemlock; Douglas-fir dominated standshad the smallest difference in site class.

Estimation of stand biomass and C stocks

Two separate model runs were conductedusing the CBM-CFS3 model; all parameters wereidentical with the exception of site class. Figure 7

Fig. 3 Scatterplotshowing SIINV versusSILIDAR

554 Environ Monit Assess (2010) 166:543561

Fig. 4 Comparison of thedistribution of site indexestimates in the originalinventory and those madeusing LIDAR-derivedmean stand heights, byleading species. Samplesizes are provided inTable 1. Mean, mean SE (box), and mean 1SD (whisker)

Table 4 Comparison ofsite index, site class, andbiomass/ha, by leadingspecies

Leading species Mean SD t Df p

Site indexAmabilis fir SIINV 22.22 2.11 23.791 8 0.000

SILIDAR 45.67 2.75Western red cedar SIINV 22.88 7.44 2.286 8 0.052

SILIDAR 26.22 5.93Red alder SIINV 24.42 2.12 12.633 228 0.000

SILIDAR 30.19 6.75Douglas-fir SIINV 32.62 4.21 2.060 1045 0.039

SILIDAR 33.02 5.21Western hemlock SIINV 26.22 5.12 25.512 26 0.000

SILIDAR 40.34 4.96

Site classAmabilis fir SIINV 22.78 2.64 18.81 8 0.000

SILIDAR 45.56 3.00Western red cedar SIINV 21.67 8.29 2.683 8 0.027

SILIDAR 26.67 5.59Red alder SIINV 25.09 1.55 11.922 228 0.000

SILIDAR 30.37 6.85Douglas-fir SIINV 32.60 3.04 2.167 1045 0.030

SILIDAR 33.01 5.53Western hemlock SIINV 27.78 6.41 16.901 26 0.000

SILIDAR 40.74 4.94

biomass/ha (tonnes)Amabilis fir SIINV 109.55 0.45 3.811 8 0.005

SILIDAR 228.48 94.03Western red cedar SIINV 180.09 37.08 2.395 8 0.043

SILIDAR 209.40 20.13Red alder SIINV 134.59 49.75 8.914 228 0.000

SILIDAR 150.72 63.64Douglas-fir SIINV 268.82 98.64 1.061 1045 0.288

SILIDAR 271.41 110.29Western hemlock SIINV 200.94 36.03 16.098 26 0.000

SILIDAR 349.95 32.97

Environ Monit Assess (2010) 166:543561 555

Fig. 5 Distribution of thesite class differencevalues (original siteclassLIDAR-derivedsite class)

shows differences in the total biomass per hectarefor the entire study area as modeled from 1920to 2006. Total biomass does not differ greatlyuntil the 1940smost of the stands sampledwere established in the 1940s, and when growthis modeled for these stands, the higher LIDARsite-index values (representing more favorablesite conditions) result in greater estimates ofbiomass. Biomass estimates prior to 1940 wereunaffected by differences in site index becausesite index was not used to assign yield tables to

the old-growth stands that were present priorto logging. Instead, old-growth biomass in 1920was estimated using stand-level merchantablevolume estimates derived from a 1919 timbercruise as described by Trofymow et al. (2008).Old-growth biomass increments were simulatedby assigning natural stand yield tables in a mannerthat provided appropriate starting volumesand subsequent annual increments that wereconsistent with our assumptions about old-growth biomass increments in these ecosystems.

Fig. 6 Distribution of siteclass difference values, byleading species (originalsite indexLIDAR siteindex). Mean, mean SE(box), and mean 1 SD(whisker)

556 Environ Monit Assess (2010) 166:543561

Fig. 7 Relationshipbetween total study areabiomass (in tonnes/ha) asestimated from the twoseparate CBM-CFS3model runs. A model runestimated total annualbiomass for each yearfrom 1920 to 2006 andmodel inputs wereidentical with theexception of siteindex class

Once the stands were harvested, the old-growthstand attributes and natural stand yield tableassociations were replaced with stand attributesprovided in the forest inventory data and theresulting managed stand yield table associations.Our biomass estimates for stands that were on siteprior to the stands present in 2006 had no effecton our estimates of stand biomass in 2006. Asexpected, the different site class estimates used inthe two model scenarios contribute to increasinglylarger differences in biomass estimates over time.

Differences in biomass estimates between thetwo model scenarios in 2006 were statistically sig-

nificant for both total stand biomass (dependentsamples two-tailed t-test, = 0.05, t = 6.26, p /JPEG2000ColorACSImageDict > /JPEG2000ColorImageDict > /AntiAliasGrayImages false /CropGrayImages true /GrayImageMinResolution 150 /GrayImageMinResolutionPolicy /Warning /DownsampleGrayImages true /GrayImageDownsampleType /Bicubic /GrayImageResolution 150 /GrayImageDepth -1 /GrayImageMinDownsampleDepth 2 /GrayImageDownsampleThreshold 1.50000 /EncodeGrayImages true /GrayImageFilter /DCTEncode /AutoFilterGrayImages true /GrayImageAutoFilterStrategy /JPEG /GrayACSImageDict > /GrayImageDict > /JPEG2000GrayACSImageDict > /JPEG2000GrayImageDict > /AntiAliasMonoImages false /CropMonoImages true /MonoImageMinResolution 600 /MonoImageMinResolutionPolicy /Warning /DownsampleMonoImages true /MonoImageDownsampleType /Bicubic /MonoImageResolution 600 /MonoImageDepth -1 /MonoImageDownsampleThreshold 1.50000 /EncodeMonoImages true /MonoImageFilter /CCITTFaxEncode /MonoImageDict > /AllowPSXObjects false /CheckCompliance [ /None ] /PDFX1aCheck false /PDFX3Check false /PDFXCompliantPDFOnly false /PDFXNoTrimBoxError true /PDFXTrimBoxToMediaBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXSetBleedBoxToMediaBox true /PDFXBleedBoxToTrimBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXOutputIntentProfile (None) /PDFXOutputConditionIdentifier () /PDFXOutputCondition () /PDFXRegistryName () /PDFXTrapped /False

/Description > /Namespace [ (Adobe) (Common) (1.0) ] /OtherNamespaces [ > /FormElements false /GenerateStructure false /IncludeBookmarks false /IncludeHyperlinks false /IncludeInteractive false /IncludeLayers false /IncludeProfiles true /MultimediaHandling /UseObjectSettings /Namespace [ (Adobe) (CreativeSuite) (2.0) ] /PDFXOutputIntentProfileSelector /NA /PreserveEditing false /UntaggedCMYKHandling /UseDocumentProfile /UntaggedRGBHandling /UseDocumentProfile /UseDocumentBleed false >> ]>> setdistillerparams> setpagedevice