This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution license (http://creativecommons.org/licenses/by/3.0/).

This study aims to estimate forest above-ground biomass and biomass components in a stand of ^{2} values of 0.748, 0.749, and 0.727, respectively, root mean squared error (RMSE) values of 9.876, 1.520, and 15.237 Mg·ha^{−1}, respectively, and relative RMSE values of 12.783%, 12.423%, and 14.163%, respectively. Moreover, fruit and crown biomass may be estimated with relatively high accuracies; estimates have adjusted ^{2} values of 0.578 and 0.648, respectively, RMSE values of 1.022 and 5.963 Mg·ha^{−1}, respectively, and relative RMSE values of 23.273% and 19.665%, respectively. In contrast, foliage biomass estimates have relatively low accuracies; they had an adjusted ^{2} value of 0.356, an RMSE of 3.691 Mg·ha^{−1}, and a relative RMSE of 26.953%. Finally, above-ground biomass and biomass component spatial maps were established using stepwise multiple regression equations. These maps are very useful for updating and modifying forest base maps and registries.

Forest biomass is an essential factor in environmental and climate modeling. Also, standing forest biomass is an essential, active participant in the global carbon cycle. Quantifying the amount of biomass within a forest stand is necessary for property managers to make informed decisions about the value and use of their forested land. Light Detection and Ranging (LiDAR) is one of the most promising remote sensing technologies for estimating various biophysical properties of forests. LiDAR provides the most accurate measurements of terrain elevation and vegetation height; this accuracy holds even on sloped terrain or in dense forests. LiDAR data are well suited to biomass estimation, as point clouds generated from forest canopies can accurately depict the physical characteristics of the canopy surface [

Though above-ground biomass estimates can be extracted from LiDAR data with high accuracies, little was known about biomass component estimates. Forest biomass can be sub-divided into its components, such as stem, branch, and foliage (

In this paper, above-ground biomass and biomass components, including stem, branch, foliage, fruit, and crown biomass were estimated in a stand of

The study site of Dayekou is situated in the Qilian Mountain area, with its geographic coordinates ranging from N38°29’ to 38°35’ in latitude and from E100°12’ to 100°20’ in longitude. The site is situated within Gansu province, western China (

LiDAR data were acquired on 23 June 2008 using a Riegl LMS-Q560 laser scanner and the Litemapper 5600 system. The scanner operated at a flight altitude of 800 m, and was configured to acquire data using a narrow scan angle of <0.5 mrad (with respect to either side of the NADIR) and with a point density of about 1 point/m^{2}. The

The location of the study area (

In order to investigate the capability of LiDAR data for forest biomass estimation, ground truth data were collected through field work from July to August of 2007, and from June to July of 2008. A high resolution airborne CCD image mosaic was used to identify each forest stand through manual interpretation. Some forest stands were selected for field plot data measurement. The size of sample plots was limited to either 20 m × 20 m or 25 m × 25 m. The height (H) and diameter at the breast height of 1.3 m (DBH) of each individual tree within each plot was measured. The DBH was defined as the ground truth data on the plot level. For the plots, the plots mean DBH ranged from 5.56 cm to 26.66 cm, with an average value of 15.55 cm; the mean tree height ranged from 3.16 m to 16.72 m, with an average value of 9.59 m. A set of total 83 forest sample plots was selected from the measurement database by the following criteria: plot location fixed with a differential GPS system; prevalence of forest plots dominated by

Using relative allometric Equations (1)–(4) [^{2}H to estimate forest biomass and biomass components could be found in other studies [^{2} × ^{0.8665}^{2} × ^{0.8905}^{2} × ^{0.4701}^{2} × ^{0.5779}

The main LiDAR processing steps in this study are illustrated in a flow chart.

Nonground hits, designated as vegetation hits, were normalized for varying terrain elevations, thereby enabling volume and biomass models to incorporate actual LiDAR point heights [

As above-ground biomass estimation is most sensitive to tree height. The major independent variables for stepwise multiple regression in this study were the mean height and height quantiles from LiDAR points [_{p}_{p}_{p}

A total of 19 vegetation height quantiles were calculated by sorting the vegetation points in ascending order at each plot, and classifying them into classes at 5% intervals, ranging from

Crown cover (CC) was selected as another statistical variable. To generate equivalent CC estimates from LiDAR data, returns greater than 1.3 m in height were considered as tree crown elements. All points were interpolated into a raster image. When the grid unit had multiple echoes, the maximum value was selected as the interpolation value. According to the point cloud density, the digital surface model (DSM) was interpolated into a resolution of 0.5 m. The canopy height model (CHM) was obtained to indicate the difference between the DSM and DTM. The 1.3 m height threshold was used to conform to definitions of forest cover, with all 0.5 × 0.5 pixels above this threshold coded as either 1 or 0. For each field plot, crown cover percentage (CC %) was calculated as the sum of all cells with a value of 1 as a percentage of the total. The result was assessed in [^{2} between the retrieved CC% values and those measured by Hemiview is 0.3901.

Finally, these variables were selected for forest biomass estimation. To select the most significant variables for biomass estimation from these variables including mean height, height quantiles, and CC, stepwise multiple linear regressions were used.

Parametric estimation via the multiple linear regression method was conducted for LiDAR data. The metrics follow a normal distribution, and from the scatter plot, we determined that a linear regression was suitable. The literature has been consistent with these conclusions [

The conventional multivariate regression model can be expressed as follows:
_{0} + _{1}_{1} + … + _{i}X_{i}_{0} is the intercept, _{1…i} and _{1…i} are the regression coefficients and values of independent variables, respectively. In this study, _{1…i} are the mean heights, height quantiles, and CC values. Collinearity was diagnosed through the Variance Inflation Factor (VIF). Generally, if the VIF is less than 10, collinearity is not serious [

Some sample data were used for estimation, while the rest were used for validation. The estimation accuracy was calculated as follows:

Ground-based above-ground biomass components are calculated using field-measured heights and DBH values; the component values are presented in ^{−1}, with a mean value of 104.53 Mg·ha^{−1}, and a standard deviation of 33.86 Mg.ha^{−1}. From ^{−1} for young plantation stands, to 174.88 Mg·ha^{−1} for mature, highly stocked stands. On average, 72% of the total above-ground biomass of a tree is contained within the stem, 11% in branches, 13% in the foliage, and 4% in fruit.

Stepwise multiple regression analysis is used to establish the above-ground biomass and biomass components equation, and the results are shown in ^{2}^{−1}, and relative RMSE values of 12.783%, 12.423%, and 14.163%, respectively. Also, the VIF for mean height and CC is 1.003, thus indicating an absence of collinearity. ^{2}^{−1}, and a relative RMSE of 26.953%. ^{2} of 0.578, an RMSE of 1.022 Mg ha^{−1}, and a relative RMSE of 23.273%. The VIF for the mean and H85 is 7.961, which means that there is no collinearity. ^{2}^{−1}, and a relative RMSE of 19.665%. The VIFs for the mean, H85, and CC are 8.577, 8.561, and 1.079, respectively. Thus, none of these variables are collinear.

First of all, the research area is segmented using a 20 m × 20 m grid. The LiDAR parameters were then measured in each segmented cell; parameters include mean height, height quantiles, and CC. Finally, above-ground biomass and biomass components in each segmented cell can be obtained based on the establishment of the above-ground biomass and biomass components estimation equation. The result is shown in ^{−1}, 0–27 Mg·ha^{−1}, 0–26 Mg·ha^{−1}, 0–9 Mg·ha^{−1}, 0–61 Mg·ha^{−1}, and 0–230 Mg·ha^{−1}, respectively.

Characteristics of above-ground biomass (Mg·ha^{−1}) and biomass components for

Statistic | Minimum | Maximum | Mean | Standard deviation |
---|---|---|---|---|

Stem biomass | 9.80 | 130.82 | 75.19 | 26.02 |

Branch biomass | 1.41 | 21.00 | 11.92 | 4.21 |

Foliage biomass | 2.92 | 22.87 | 13.19 | 3.47 |

Fruit biomass | 0.87 | 6.85 | 4.23 | 1.17 |

Crown biomass | 5.68 | 44.42 | 29.34 | 8.23 |

Above-ground biomass | 17.89 | 174.88 | 104.53 | 33.86 |

Stepwise multiple regression results for above-ground biomass and individual biomass components.

Variables | Std |
^{2} |
Adjusted
^{2} |
Model | |
---|---|---|---|---|---|

Stem biomass | Mean, CC | 13.823 | 0.756 | 0.748 | −13.595 + 8.446Mean + 20.378CC |

Branch biomass | Mean, CC | 2.229 | 0.757 | 0.749 | −2.447 + 1.367Mean + 3.300CC |

Foliage biomass | H5 | 2.692 | 0.366 | 0.356 | 7.767 + 0.861H5 |

Fruit biomass | Mean, H85 | 9.507 | 0.591 | 0.578 | 1.726 + 0.541Mean − 0.210H85 |

Crown biomass | Mean, H85, CC | 5.025 | 0.664 | 0.648 | 8.017 + 4.038Mean − 1.502H85 + 7.287CC |

Above-ground biomass | Mean, CC | 18.640 | 0.736 | 0.727 | −9.013 + 10.812Mean + 25.105CC |

Scatter plots of between above-ground biomass and biomass components and their significant variables. (

Estimated accuracy of above-ground biomass and individual biomass components; RMSE: root mean squared error.

Biomass | Stem | Branch | Foliage | Fruit | Crown | Above-ground |
---|---|---|---|---|---|---|

RMSE(Mg·ha^{−1}) |
9.876 | 1.520 | 3.691 | 1.022 | 5.963 | 15.237 |

Relative RMSE(%) | 12.783 | 12.423 | 26.953 | 23.273 | 19.665 | 14.163 |

Accuracy(%) | 87.45 | 87.60 | 80.00 | 80.15 | 82.59 | 87.08 |

Spatial distributions of above-ground biomass and biomass components. (

^{−1}. Compared with forests that exist in relatively shaded areas, forests existing in relatively sunnier areas receive more solar radiation, which causes dehydration via intense evapotranspiration. Consequently, there is very little biomass in relatively sunnier areas (

CCD image together with the DTM, Slope, and Aspect images. (

Scatter plots plots comparing above-ground biomass against elevation, slope, and aspect. (

For biomass estimation using small footprint LiDAR, distributional metrics—such as the mean canopy height and the standard deviation of the canopy height—are taken from either an interpolated grid corresponding to the height of the canopy (

In [^{2}^{2}^{2}^{2} than the study in [^{2}^{2}^{2}^{2}^{2}

From the CCD image, together with the DTM, Slope, and Aspect images, we observed that the forest comprises altitudes ranging from about 2600 m to 3600 m, at slopes varying from 10 to 50 degrees. The shapes of scatter plots plotting above-ground biomass against elevation and slope are similar to normal distribution curves. Most forests exist at azimuths range from 0 to 100 degrees, from 250 to 360 degrees, and from 100 to 350 degrees. The shape of the scatter plots plotting above-ground biomass against azimuth is also similar to a normal distribution. This means that

From ^{2} is used. In another’s study [^{2} was sufficient for estimating forest inventory variables at the plot and stand levels for various forest types.

In this paper, we investigated a

This paper has been supported jointly by the Natural Science Foundation of China (Grant No. 41101308, 41101374 and 41271361), the Fundamental Research Funds for the Central Universities (Grant No. 2011B06714), and the National Science and Technology Support Plan During the 12th Five-year Plan Period of China (2012BAC19B03; 2013BAC10B01). The authors also wish to thank all people who participated in the field experiment. The authors also wish to thank all the people that have given helped prepare this paper. The authors are sincerely grateful to the four anonymous reviewers for the constructive and insightful comments which improved this study.

The authors declare no conflict of interest.