Uncertainties of mapping aboveground forest carbon due to plot locations using national forest inventory plot and remotely sensed data

Abstract

Forest carbon sinks significantly contribute to mitigation of atmospheric concentrations of carbon dioxide. Thus, estimating forest carbon is becoming important to develop policies for mitigating climate change and trading carbon credits. However, a great challenge is how to quantify uncertainties in estimation of forest carbon. This study investigated uncertainties of mapping aboveground forest carbon due to location errors of sample plots for Lin-An County of China. National forest inventory plot data and Landsat TM images were combined using co-simulation algorithm. The findings show that randomly perturbing plot locations within 10 distance intervals statistically did not result in biased population mean predictions of aboveground forest carbon at a significant level of 0.05, but increased root mean square errors of the maps. The perturbations weakened spatial autocorrelation of aboveground forest carbon and its correlation with spectral variables. The perturbed distances of 800 m or less did not obviously change the spatial distribution of predicted values. However, when the perturbed distances were 1600 m or larger, the correlation coefficients of the predicted values from the perturbed locations with those from the true plot locations statistically did not significantly differ from zero at a level of 0.05 and the spatial distributions became random.     

EU-wide maps of growing stock and above-ground biomass in forests based on remote sensing and field measurements

The overall objective of this study was to combine national forest inventory data and remotely sensed data to produce pan-European maps on growing stock and above-ground woody biomass for the two species groups “broadleaves” and “conifers”. An automatic up-scaling approach making use of satellite remote sensing data and field measurement data was applied for EU-wide mapping of growing stock and above-ground biomass in forests. The approach is based on sampling and allows the direct combination of data with different measurement units such as forest inventory plot data and satellite remote sensing data. For the classification, data from the Moderate Resolution Imaging Spectroradiometer (MODIS) were used. Comprehensive field measurement data from national forest inventories for 98,979 locations from 16 countries were used for which tree species and growing stock estimates were available. The classification results were evaluated by comparison with regional estimates derived independently from the classification from national forest inventories. The validation at the regional level shows a high correlation between the classification results and the field based estimates with correlation coefficient r = 0.96 for coniferous, r = 0.94 for broadleaved and r = 0.97 for total growing stock per hectare. The mean absolute error of the estimations is 25 m³/ha for coniferous, 20 m³/ha for broadleaved and 25 m³/ha for total growing stock per hectare. Biomass conversion and expansion factors were applied to convert the growing stock classification results to carbon stock in above-ground biomass. As results of the classification, coniferous and broadleaved growing stock as well as carbon stock of the above-ground biomass is mapped on a wall-to-wall basis with a spatial resolution of 500 m × 500 m per grid cell. The mapped area is 5 million km², of which 2 million km² are forests, and covers the whole European Union, the EFTA countries, the Balkans, Belarus, the Ukraine, Moldova, Armenia, Azerbaijan, Georgia and Turkey.     

Estimating and mapping forest biomass in northeast China using joint forest resources inventory and remote sensing data

Being able to accurately estimate and map forest biomass at large scales is important for a better understanding of the terrestrial carbon cycle and for improving the effectiveness of forest management. In this study, forest plot sample data, forest resources inventory(FRI) data, and SPOT Vegetation(SPOT-VGT) normalized difference vegetation index(NDVI) data were used to estimate total forest biomass and spatial distribution of forest biomass in northeast China(with 1 km resolution). Total forest biomass at both county and provincial scales was estimated using FRI data of 11 different forest types obtained by sampling 1156 forest plots, and newly-created volume to biomass conversion models. The biomass density at the county scale and SPOT-VGT NDVI data were used to estimate the spatial distribution of forest biomass. The results suggest that the total forest biomass was 2.4 Pg(1 Pg = 10~(15) g), with an average of 77.2 Mg ha~(-1), during the study period. Forests having greater biomass density were located in the middle mountain ranges in the study area. Human activities affected forest biomass at different elevations, slopes and aspects. The results suggest that the volume to biomass conversion models that could be developed using more plot samples and more detailed forest type classifications would be better suited for the study area and would provide more accurate biomass estimates. Use of both FRI and remote sensing data allowed the down-scaling of regional forest biomass statistics to forest cover pixels to produce a relatively fineresolution biomass map.     

Spatial interpolation of in situ data by self-organizing map algorithms (neural networks) for the assessment of carbon stocks in European forests

Self-organizing maps (SOMs) are an advanced neural networks application. SOMs were applied for the spatially explicit estimation of forest carbon stocks for a test region in Thuringia (Germany). The approach utilizes in situ national forest inventory data and satellite remote sensing data (Landsat 7 ETM+) and provides maps showing a high-resolution spatial distribution of forest carbon stocks. The generated maps are compared to alternative estimates obtained by the k-nearest neighbour (kNN) method—a remote sensing based carbon assessment. Beside maps the SOM- and kNN-approaches were utilized to calculate statistical estimates of carbon stock and growing stock. The statistical estimates were validated by calculating bias and mean square errors with reference to in situ assessments.     

Estimating forest carbon fluxes for large regions based on process-based modelling, NFI data and Landsat satellite images

The aim of this study was to develop and evaluate a new approach for estimating forest carbon fluxes for large regions based on climate-sensitive process-based model, national forest inventory (NFI) data and satellite images. The approach was tested for Central Finland and Lapland with NFI field data and daily weather data from 2004 to 2008.The approach combines (1) a light use efficiency (LUE) model, (2) a process-based summary model for estimating gross primary production (GPP) and net primary production (NPP), and (3) the Yasso07 soil carbon model, which together allow the estimation of net ecosystem exchange (NEE). Landsat TM 5 satellite images were utilized to generalize the carbon fluxes obtained for field sample plots for all forested areas using the k-NN imputation method. The accuracy of the imputations was examined by leave-one-out cross validation and by comparing the imputed and simulated values with Eddy covariance (EC) measurements.RMSE of the k-NN imputations was slightly better in Central Finland than in Lapland, the bias staying at a similar level. Based on the EC comparisons, the approach seemed to work rather well with GPP estimates in both areas, but in the north the NEE estimates were remarkably biased. The main advantages of the approach include its applicability to basic NFI data and a high output resolution (30 m).The method proved to be a promising way to produce carbon flux estimates based on large-scale forest inventory data and could therefore be easily applied to the whole of Northern Europe. However, there are still drawbacks to the approach, such as lacking parameters for peat lands. One of the future goals is to integrate the approach with an interactive mapping framework, which could thereafter be utilized, for example, in climate change research.     

Effects of field plot configurations on the uncertainties of ALS-assisted forest resource estimates

The effects of field plot configurations on the uncertainties of plot-level forest resource estimates were analyzed using airborne laser scanner data, aerial photographs and field measurements. The aim was to select a field sample plot configuration that can be used for both large area and management inventories. Error estimates were evaluated at the plot level using six different training plot configurations. Additionally, separate plots with two different sizes were used for evaluation. Stem volume and five other forest resource characteristics were considered. The field measurement costs of the different plot configurations were also studied. RMSEs and mean deviations for airborne laser scanning ALS-assisted estimates were practically the same for the fixed radius plot, the two concentric plots and the angle count plot with a basal area factor of q = 1 for all three evaluation plot sizes. Angle count plots with basal area factors of q = 1.5 and 2 increased the RMSEs. For the former plot configurations, the RMSEs for the ALS-assisted estimates could be attributed to inaccuracy in the predicted relationships between the field data and ALS data, not to the training plot configuration. Tree measurements and costs can, therefore, be reduced from those of the Finnish management inventories without increasing RMSEs.     

Estimating forest data for analyses of forest production and utilization possibilities at local level by means of multi-source National Forest Inventory

The sample plot data of National Forest Inventories (NFI) are widely used in the analysis of forest production and utilization possibilities to support national and regional forest policy. However, there is an increasing interest for similar impact and scenario analyses for strategic planning at the local level. As the fairly sparse network of field plots only provides calculations for large areas, satellite image data have been applied to produce forest information for smaller areas. The aim of this study was to test the feasibility of generating forest data for a Finnish forest analysis tool, the MELA system, by means of the Landsat satellite imagery and the NFI sample plot data. The study was part of the preparation of a local forestry programme, where a strategic scenario analysis for the forest area of two villages (ca 8000 ha) was carried out. Management units that approximate forest stands were delineated by image segmentation. Stand volume and other parameters for each forest segment were estimated from weighted means of the NFI sample plots, where the individual sample plot weights were estimated by the k nearest neighbour (kNN) method. Two different spectral features were tested: single pixel values and average pixel values within a segment. The estimated forest data were compared with the forest data based on independent stand-level field assessments in two subareas, a national park and an area of forest managed for timber production.In the national park, the estimated mean volume of the growing stock from both spectral feature sets (about 160 m³ ha⁻¹) was clearly lower than that obtained from stand-level field assessment (186 m³ ha⁻¹). Using average pixel values within a segment resulted in a higher proportion of pine and a lower proportion of spruce volume than using single pixel values. It also resulted in an estimated felling potential nearly 10% higher over the first 10-year period in the scenario analysis of the area dedicated to timber production. However, the maximum long-term sustainable removal was at the same level (about 30,000 m³ year⁻¹) for both feature sets over the simulated 30-year period. The resulting annual felling area in the first 10-year period was 12% lower when the segment averages were applied, but the difference subsequently levelled off. The kNN approach in estimating initial forest data for scenario analyses at the local level was found promising.     

Empirical evaluation of confidence and prediction intervals for spatial models of forest structure in Jalisco,Mexico

In recent years there has been an increasing interest in developing spatial statistical models for data sets that are seemingly spatially independent.This lack of spatial structure makes it difficult,if not impossible to use optimal predictors such as ordinary kriging for modeling the spatial variability in the data.In many instances,the data still contain a wealth of information that could be used to gain flexibility and precision in estimation.In this paper we propose using a combination of regression analysis to describe the large-scale spatial variability in a set of survey data and a tree-based stratification design to enhance the estimation process of the small-scale spatial variability.With this approach,sample units(i.e.,pixel of a satellite image) are classified with respect to predictions of error attributes into homogeneous classes,and the classes are then used as strata in the stratified analysis.Independent variables used as a basis of stratification included terrain data and satellite imagery.A decision rule was used to identify a tree size that minimized the error in estimating the variance of the mean response and prediction uncertainties at new spatial locations.This approach was applied to a set of n=937 forested plots from a state-wide inventory conducted in 2006 in the Mexican State of Jalisco.The final models accounted for 62% to 82% of the variability observed in canopy closure(%),basal area(m2·ha-1),cubic volumes(m3·ha-1) and biomass(t·ha-1) on the sample plots.The spatial models provided unbiased estimates and when averaged over all sample units in the population,estimates of forest structure were very close to those obtained using classical estimates based on the sampling strategy used in the state-wide inventory.The spatial models also provided unbiased estimates of model variances leading to confidence and prediction coverage rates close to the 0.95 nominal rate.     

Effect of tree species on carbon stocks in forest floor and mineral soil and implications for soil carbon inventories

Forest soil organic carbon (SOC) and forest floor carbon (FFC) stocks are highly variable. The sampling effort required to assess SOC and FFC stocks is therefore large, resulting in limited sampling and poor estimates of the size, spatial distribution, and changes in SOC and FFC stocks in many countries. Forest SOC and FFC stocks are influenced by tree species. Therefore, quantification of the effect of tree species on carbon stocks combined with spatial information on tree species distribution could improve insight into the spatial distribution of forest carbon stocks.We present a study on the effect of tree species on FFC and SOC stock for a forest in the Netherlands and evaluate how this information could be used for inventory improvement. We assessed FFC and SOC stocks in stands of beech (Fagus sylvatica), Douglas fir (Pseudotsuga menziesii), Scots pine (Pinus sylvestris), oak (Quercus robur) and larch (Larix kaempferi).FFC and SOC stocks differed between a number of species. FFC stocks varied between 11.1 Mg C ha⁻¹ (beech) and 29.6 Mg C ha⁻¹ (larch). SOC stocks varied between 53.3 Mg C ha⁻¹ (beech) and 97.1 Mg C ha⁻¹ (larch). At managed locations, carbon stocks were lower than at unmanaged locations. The Dutch carbon inventory currently overestimates FFC stocks. Differences in carbon stocks between conifer and broadleaf forests were significant enough to consider them relevant for the Dutch system for carbon inventory.     

Spatial prediction of forest stand variables

This study aims at the development of a model to predict forest stand variables in management units (stands) from sample plot inventory data. For this purpose we apply a non-parametric most similar neighbour (MSN) approach. The study area is the municipal forest of Waldkirch, 13 km north-east of Freiburg, Germany, which comprises 328 forest stands and 834 sample plots. Low-resolution laser scanning data, classification variables as well rough estimations from the forest management planning serve as auxiliary variables. In order to avoid common problems of k-NN-approaches caused by asymmetry at the boundaries of the regression spaces and distorted distributions, forest stands are tessellated into subunits with an area approximately equivalent to an inventory sample plot. For each subunit only the one nearest neighbour is consulted. Predictions for target variables in stands are obtained by averaging the predictions for all subunits. After formulating a random parameter model with variance components, we calibrate the prior predictions by means of sample plot data within the forest stands via BLUPs (best linear unbiased predictors). Based on bootstrap simulations, prediction errors for most management units finally prove to be smaller than the design-based sampling error of the mean. The calibration approach shows superiority compared with pure non-parametric MSN predictions.     

The inventory of carbon stock in New Zealand’s post-1989 planted forest for reporting under the Kyoto protocol

The United Nations Framework Convention on Climate Change (UNFCCC) requires reporting net carbon stock changes and anthropogenic greenhouse gas emissions, including those related to forests. This paper describes the design and implementation of a nation-wide forest inventory of New Zealand’s planted post-1989 forests that arose from Land Use, Land-Use Change and Forestry activities (LULUCF) under Article 3.3 of the Kyoto Protocol. The majority of these forests are planted with Pinus radiata, with the remainder made up of other species exotic to New Zealand. At the start of the project there was no on-going national forest inventory that could be used as a basis for calculating carbon stocks and meet Good Practice Guidelines.A network of ground-based permanent sample plots was installed with airborne LiDAR (Light Detection and Ranging) for double sampling using regression estimators to predict carbon in each of the four carbon pools of above- and below-ground live biomass, dead wood and litter. Measurement, data acquisition and quality assurance/control protocols were developed specifically for the inventory, carried out in 2007 and 2008. Plots were located at the intersection of a forest with a 4 km square grid, coincident with an equivalent 8 km square grid established over the indigenous forest and “grassland with woody biomass” (Other Wooded Land). Planted tree carbon within a ground plot was calculated by an integrated system of growth, wood density and compartment allocation models utilising the data from measurements of trees and shrubs on the plots. This system, called the Forest Carbon Predictor, predicts past and future carbon in a stand and is conditioned so that the calculated basal area and mean top height equals that obtained by conventional mensuration methods at the time of the plot measurement. Mean per hectare carbon stocks were then multiplied by an estimate of the total area of post 1989 forests obtained from wall to wall mapping using a combination of satellite imagery and ortho-photography.The network of permanent samples plots and LiDAR double sampling methodology was designed to be simple and robust to change over time. In the future, using LiDAR should achieve sampling efficiencies over using ground plots alone and reduces any problems regarding restricted access on the ground. The network is to be remeasured at the end of commitment period 1, 2012, and the carbon stocks re-estimated in order to calculate change.     

A national level forest resource assessment for Burkina Faso - A field based forest inventory in a semiarid environment combining small sample size with large observation plots

Even though considerable parts of the global tropical forests are located in Africa, reliable data on African forest resources is limited. While this is widely recognized for tropical moist forests, it also holds for tropical dry forests. To partially fill the gap a forest inventory was carried out in Burkina Faso, West Africa. In this paper we present a methodological approach and sample based estimates of the tree and forest resources including estimates of (1) land cover classes, (2) species composition, and (3) above ground tree carbon stocks. Following the land classification of the Food and Agriculture Organization of the United Nations (FAO), the forest cover of Burkina Faso was estimated as 42.6% (116,847 km²). For the classes “other wooded land”, “other land” and “other land with tree cover” the estimates were 1.6%, 53.6%, and 9.1%, respectively. We found notable differences to the estimates published by FAO, in particular when considering the classes “forest” and “other wooded land” separately, but lesser so when the two classes are combined. That points to a major issue in applying these class definitions in semiarid environments. Given the relatively small sample size (n = 46 field observed plots), relative standard errors (SE%) of area estimates are high (around 9% for the larger area classes). Aboveground tree carbon stocks were estimated to be 6.640, 5.580 and 7.222 Mg ha⁻¹ for “forest”, “other wooded land” and “other land with tree cover”, respectively (SE% around 18% for all three estimates). Availability of biomass models is very limited for all classes, in particular when it comes to shrubs. Furthermore, it was estimated that the most abundant tree species in Burkina Faso is Vittelaria paradoxa, the “shea butter tree” which is a multi-use tree species of high relevance for rural livelihoods.To our knowledge this study is the first field-based forest inventory on national level in Burkina Faso where the estimation of errors was possible on statistical grounds, and done. The results of this study revealed major issues that should be taken into account when doing similar studies, including carbon monitoring and accounting: increasing the sample size will lead to smaller standard errors (at a higher costs, of course), but will not solve the crucial points (1) of non-availability of suitable biomass models, in particular for shrub lands and (2) of implementation issues regarding the definition of land cover types.     

广西森林资源连续清查角规样地体系评价

广西森林资源连续清查(以下简称"广西连清)"角规样地体系,是我国唯一的以点抽样理论为基础,以固定角规样地为监测载体的省(区)级森林资源连续清查体系。广西连清第7次复查,除了增设的方形样地调查以外,还对原有的角规样地进行了复查,因此本文得以用同时进行调查的方形样地调查结果作为参照对象,对角规样地体系的优点和存在的问题进行定性和定量相结合的综合分析评价。分析评价结果表明,与方形样地比较,角规样地除了具有显著的隐蔽性外,外业工作量只相当于方形样地的53%,两套体系的活立木蓄积量差异仅为3.16%,角规样地和方形样地体系总蓄积量抽样精度分别为94.47%与94.57%,均达到国家森林资源连续清查技术规定要求(≥90%)。角规样地复位率大于规定的98%,样木复位率远大于规定的95%,达97%以上,能满足林木蓄积生长量和消耗量监测的要求。角规样地体系的不足主要是由于漏测木和进测木的存在,导致森林资源的现状估计值偏低,且动态估计精度明显低于方形样地体系。     

Analysis of spatio-temporal changes in forest biomass in China

Forests play a central role in the global carbon cycle.China's forests have a high carbon sequestration potential owing to their wide distribution,young age and relatively low carbon density.Forest biomass is an essential variable for assessing carbon sequestration capacity,thus determining the spatio-temporal changes of forest biomass is critical to the national carbon budget and to contribute to sustainable forest management.Based on Chinese for-est inventory data (1999-2013),this study explored spatial patterns of forest biomass at a grid resolution of 1 km by applying a downscaling method and further analyzed spatio-temporal changes of biomass at different spatial scales.The main findings are:(1) the regression relationship between forest biomass and the associated influencing factors at a provincial scale can be applied to estimate biomass at a pixel scale by employing a downscaling method;(2) for-est biomass had a distinct spatial pattern with the greatest biomass occurring in the major mountain ranges;(3) forest biomass changes had a notable spatial distribution pattern;increase (i.e.,carbon sinks) occurred in east and southeast China,decreases (i.e.,carbon sources) were observed in the northeast to southwest,with the largest biomass losses in the Hengduan Mountains,Southern Hainan and Northern Da Hinggan Mountains;and,(4) forest vegetation functioned as a carbon sink during 1999-2013 with a net increase in biomass of 3.71 Pg.     

Satellite-based radar mapping of British forest age and Net Ecosystem Exchange using ERS tandem coherence

A method of mapping forest age structure using satellite remote sensing data in combination with ground data is used to form an age structure map of mainland Britain’s forests. Age structure is then used to demonstrate a method of calculating Net Ecosystem Exchange (NEE) for a region of forest. Synthetic Aperture Radar (SAR) coherence data provides an indication of forest biomass, which is related to forest age using forest management data at five independent locations. Coherence data is sensitive to time varying environmental effects and hence requires extensive calibration of the function relating coherence to forest age. The calibration approach appropriate in Britain makes use of the extensive ground GIS data available. Age structure information for nearly 3 million hectares of forest is generated, of which 70% is privately owned and age information is otherwise unavailable. The resulting map has a spatial resolution that can differentiate individual forest stands and provides detailed regional age and biomass estimates. Comparison of stand age estimates with ground data is has potential to provide growth rate and felling information. The age structure map is used in combination with an exemplary function relating forest age to NEE to estimate atmospheric carbon exchange for England, Scotland and Wales. This method predicts threefold higher (10.87 M t a⁻¹) forest carbon uptake than national inventory figures. The remote sensing data also indicates age estimates that conflict with the ground data in Wales, which is explainable by the introduction of partial felling practices in this region during the time period of the coherence acquisitions.     

INFOCARB: A regional scale forest carbon inventory (Provincia Autonoma di Trento,Southern Italian Alps)

The aim of this inventory (acronym: INFOCARB) was to measure the organic carbon stored in the forest ecosystems of the Trento region (Provincia Autonoma di Trento, Northern Italy) in both above- and belowground pools, according to the Kyoto protocol and IPCC requirements. A total of 150 forest sampling points were selected on the entire regional area (6206 km²) with a statistical sampling approach, based on the timber volume as a proxy variable for a stratified sampling. Each sampling point was located with a GPS receiver and a 600 m² circular plot was delimited around each point. Inside the plots, the biomass of trees, shrubs and herbaceous vegetation was measured, while litter was collected in systematically placed subplots. Topsoil (down to 30 cm depth) was sampled with the excavation method on three systematically located pits, to determine the organic carbon content, the bulk density and the volume occupied by stones and roots.     

Improving the estimate of forest biomass carbon storage by combining two forest inventory systems

Quantifying forest carbon storage and its spatial distribution at regional scales is critical for the creation of greenhouse gases inventories, the evaluation of forest services and carbon-oriented forest management. The plot-based forest inventory (PBFI) and stand-based forest inventory (SBFI) collect extensive information on trees and stands respectively, and together, provide an opportunity to improve the regional estimates of forest carbon. In this study, we applied the SBFI to overcome the spatial extent limits of the PBFI in neighboring plots and improve the regional carbon estimation. We found that the forests in Sichuan Province reserved a total of 624.2?Tg?C in biomass and featured a large spatial heterogeneity, with high values in natural forests and low values in plantations. We found that the solo use of PBFI derived a slightly higher (46.63?Mg?C/ha) estimation on average compared with the integrated method (43.6?Mg?C/ha). However, when considering the spatial distribution, the PBFI generated an overestimation of young forests located between 3000and 4000?m in elevation, and an underestimation in mature forests. The spatially explicit biomass carbon estimation could be helpful in guiding regional forest management and biodiversity conservation.     

Estimating aboveground biomass of Pinus densata-dominated forests using Landsat time series and permanent sample plot data

Southwest China is one of three major forest regions in China and plays an important role in carbon sequestration.Accurate estimations of changes in aboveground biomass are critical for understanding forest carbon cycling and promoting climate change mitigation.Southwest China is characterized by complex topographic features and forest canopy structures,complicating methods for mapping aboveground biomass and its dynamics.The integration of continuous Landsat images and national forest inventory data provides an alternative approach to develop a long-term monitoring program of forest aboveground biomass dynamics.This study explores the development of a methodological framework using historical national forest inventory plot data and Landsat TM timeseries images.This method was formulated by comparing two parametric methods:Linear Regression for Multiple Independent Variables(MLR),and Partial Least Square Regression(PLSR);and two nonparametric methods:Random Forest(RF)and Gradient Boost Regression Tree(GBRT)based on the state of forest aboveground biomass and change models.The methodological framework mapped Pinus densata aboveground biomass and its changes over time in Shangri-la,Yunnan,China.Landsat images and national forest inventory data were acquired for 1987,1992,1997,2002 and 2007.The results show that:(1)correlation and homogeneity texture measures were able to characterize forest canopy structures,aboveground biomass and its dynamics;(2)GBRT and RF predicted Pinus densata aboveground biomass and its changes better than PLSR and MLR;(3)GBRT was the most reliable approach in the estimation of aboveground biomass and its changes;and,(4)the aboveground biomass change models showed a promising improvement of prediction accuracy.This study indicates that the combination of GBRT state and change models developed using temporal Landsat and national forest inventory data provides the potential for developing a methodological framework for the long-term mapping and monitoring program of forest aboveground biomass and its changes in Southwest China.     

Inventory-based estimates of forest biomass carbon stocks in China: A comparison of three methods

Several studies have reported different estimates for forest biomass carbon (C) stocks in China. The discrepancy among these estimates may be largely attributed to the methods used. In this study, we used three methods [mean biomass density method (MBM), mean ratio method (MRM), and continuous biomass expansion factor (BEF) method (abbreviated as CBM)] applied to forest inventory data to estimate China's forest biomass C stocks and their changes from 1984 to 2003. The three methods generated various estimates of the biomass C stocks: the lowest (4.0–5.9 Pg C) from CBM and the highest (5.7–7.7 Pg C) from MBM, with an intermediate estimate (4.2–6.2 Pg C) from MRM. Forest age class is a major factor responsible for these method-induced differences. MBM overestimates biomass for young-aged forests, but underestimates biomass for old-aged forests; while the reverse is true for MRM. Further, the three methods resulted in different estimates of biomass C stocks for different forest types. For temperate/subtropical mixed forests, MBM generated a 92% higher estimate than CBM and MRM generated a 14% lower than CBM. The degree of the overestimates is closely related with the proportion of young-aged forest within total area of each forest type.