- Methodology
- Open Access
Model errors in tree biomass estimates computed with an approximation to a missing covariance matrix
- Steen Magnussen^{1}Email author and
- Oswaldo Ismael Carillo Negrete^{2}
https://doi.org/10.1186/s13021-015-0031-8
© Magnussen and Carillo Negrete. 2015
- Received: 16 January 2015
- Accepted: 3 August 2015
- Published: 17 September 2015
Abstract
Background
Biomass and carbon estimation has become a priority in national and regional forest inventories. Biomass of individual trees is estimated using biomass equations. A covariance matrix for the parameters in a biomass equation is needed for the computation of an estimate of the model error in a tree level estimate of biomass. Unfortunately, many biomass equations do not provide key statistics for a direct estimation of model errors. This study proposes three new procedures for recovering missing statistics from available estimates of a coefficient of determination and sample size. They are complementary to a recently published study using a computationally intensive Monte Carlo approach.
Results
Our recovery approach use survey data from the population targeted for an estimation of tree biomass. Examples from Germany and Mexico illustrate and validate the methods. Applications with biomass estimation and robust recovered fit statistics gave reasonable estimates of model errors in tree level estimates of biomass.
Conclusions
It is good practice to provide estimates of uncertainty to any model-dependent estimate of above ground biomass. When a direct approach to estimate uncertainty is impossible due to missing model statistics, the proposed robust procedure is a first step to good practice. Our recommended approach offers protection against inflated estimates of precision.
Keywords
- Linear regression
- Nonlinear regression
- Weighted regression
- Residual variance
- Robust estimation
- Parametric bootstrap
Background
The importance of forest biomass for the global carbon cycle is widely recognized [1–4]. The imperative of maintaining global levels of forest biomass and slowing regional rates of decline [5] has fostered international cooperation, initiatives, and projects to this end [6–8].
A large number of countries have agreed to implement an accounting system for forest carbon and to report on national-level annual gains and losses [9–11].
With few exceptions, the forest carbon accounting system has a national forest inventory at its core, and a suite of models to expand and transform inventory data to forest carbon [12–14]. Carbon components not fully covered by an inventory are typically estimated from activity data (e.g. harvest, disturbance, and erosion) and models fitted to data from research studies of, for examples: litter-fall; litter-decomposition; fine-root turnover; seed production; and dead and downed-woody debris.
An estimate of the uncertainty in a carbon balance has become a routine requirement [15, 16]. When the core inventory data comes from a probability sample, the uncertainty arises from three sources: observational and measurement errors [17–19], sampling errors, and errors in model parameters [12, 20]. The live above-ground forest tree biomass (AGB) accounts for the largest contribution to the forest carbon balance [21, 22].
In situ determination of AGB is extremely costly and destructive. A model‒dependent approach with prediction of biomass from a biomass equation, with easy-to-measure explanatory variables, is the only practically feasible alternative [21, 23].
The development of a complete set of regional, species- and stratum-specific biomass models constitutes a heavy financial outlay that cannot be met in many parts of the world. As a substitute for models fitted to local data, an analyst may decide to use the most suitable off-the-shelf biomass equation [21, 23–36].
It is, of course, very difficult to ascertain whether an off-the-shelf model is suitable for a particular application or not [37]. It remains a risky proposition to use externally fitted models without any form of validation or re-calibration to local conditions [38]. An adopted model generates the desired predictions of above-ground biomass but a valid estimate of the associated covariance of model-parameters is needed to compute an estimate of the uncertainty in a prediction [12, 39, p. 73, 40]. A model-bias can only be quantified in a validation with actual observations of above-ground biomass and the predictors in a model [41, pp. 172 and 232, 42].
Although we have a plethora of equations for above-ground biomass as a function of, for example, stem diameter at a reference height of 1.3 m above ground level [21, 26, 31, 43], information regarding the covariance matrix of model parameters is often missing. Available fit statistic is generally limited to one or more of the following: standard errors of estimated parameters, the coefficient of determination, the standard deviation of lack-of-fit residuals, and sample size [44].
This study demonstrates methods for recovering a covariance matrix for model parameters in a biomass equation from fit statistics restricted to: sample size (n) and the coefficient of determination (R^{2}) [44]. Our non-use of a possibly available estimate of the standard deviation of empirical residuals rests with its sensitivity to outliers [45], a strong dependency on the sampling design [39, p. 55], the distribution of the response and explanatory variables in the study that gave us the equation of interest. Wayson et al. [44] proposed a Monte-Carlo approach to recover missing estimates of the covariance among parameters in a biomass equation. The key idea in their approach is to generate a distribution of pseudo-data that mirrors, to the extent possible, a known or assumed distribution of explanatory variables in the sample trees behind an equation. The tenet behind our approach is different. It is rooted in survey sampling [39]. Hence, the recovered estimates of uncertainty are assumed compatible with estimates that could have been obtained from a sample taken from the population, for which we desire estimates of biomass. It is fully recognized that our recovery is neither perfect nor unbiased. However, supported by our results, we argue that our approach is consistent with the main objective of any recovery procedure: to estimate model errors in population estimates of biomass as opposed to a rediscovery of ‘lost’ estimates of model errors.
Our demonstrations include examples with equations and data from the first German national forest inventory in 1987 (BWI-1) [40, 46] and the 2004–2009 Mexican National Forest Inventory [47–49]. We discuss limitations to our approach, and recommend a robust recovery method. We also emphasize the need to develop new and fully documented biomass equations for important species in regions where they are currently lacking.
Results
Examples from Germany
Species group above-ground forest tree biomass (AGB kg tree^{−1}) equations
Species | # | Equation \(\widehat{{\text{AGB}}} =\) | \(\hat{R}^{2}\) | \(\hat{\sigma }_{e}^{2}\) 10^{3} kg | Wts |
---|---|---|---|---|---|
BEECH^{a} | 1 | \(0.901{\text{DBH}}^{2} - 6.382\sqrt {\text{DBH} \times {\text{HT}}}\) | 0.96 | 75.3 | n.a. |
2 | \(\left( {9.645{\text{DBH}} - 0.648\sqrt {\text{HT}} } \right)^{2}\) | 0.95 | 79.2 | n.a. | |
SPRUCE^{a} | 5 | \(0.447{\text{DBH}}^{2} - 1.189\sqrt {\text{DBH} \times {\text{HT}}}\) | 0.98 | 18.7 | n.a. |
6 | \(\left( {0.634{\text{DBH}} - 0.426\sqrt {\text{HT}} } \right)^{2}\) | 0.98 | 18.2 | n.a. | |
PINE^{a} | 9 | \(0.450{\text{DBH}}^{2} - 0.014\sqrt {\text{DBH} \times {\text{HT}}}\) | 0.98 | 14.5 | n.a. |
10 | \(0.658{\text{DBH}}^{2} - 0.124\sqrt {\text{DBH} \times {\text{HT}}}\) | 0.98 | 14.5 | n.a. | |
BEECH | 3 | \(0.887{\text{DBH}}^{2} - 6.487\sqrt {\text{DBH} \times {\text{HT}}}\) | 0.96 | 74.4 | DBH^{−2} |
4 | \(\left( {10.127{\text{DBH}} - 1.167\sqrt {\text{HT}} } \right)^{2}\) | 0.95 | 75.3 | DBH^{−2} | |
SPRUCE | 7 | \(0.492{\text{DBH}}^{2} - 1.409\sqrt {\text{DBH} \times {\text{HT}}}\) | 0.97 | 19.0 | DBH^{−2} |
8 | \(\left( {0.689{\text{DBH}} - 0.064\sqrt {\text{HT}} } \right)^{2}\) | 0.97 | 18.7 | DBH^{−2} | |
PINE | 11 | \(0.479{\text{DBH}}^{2} - 1.927\sqrt {\text{DBH} \times {\text{HT}}}\) | 0.97 | 14.7 | DBH^{−2} |
12 | \(0.690{\text{DBH}}^{2} - 0.205\sqrt {\text{DBH} \times {\text{HT}}}\) | 0.97 | 14.5 | DBH^{−2} | |
BEECH^{b} | 13 | \(Exp\left[ {0.006 + 10.933\,\frac{\text{DBH}}{\text {DBH} + 21.216}} \right]\) | 0.99^{c} | n.a. | n.a. |
SPRUCE^{b} | 14 | \(Exp\left[ { - 1.694 + 10.825\frac{\text{DBH}}{\text{DBH} + 11.816}} \right]\) | 0.99^{c} | n.a. | n.a. |
PINE^{b} | 15 | \(Exp\left[ { - 2.688 + 10.745\,\frac{\text{DBH}}{\text{DBH} + 8.062}} \right]\) | 0.99^{c} | n.a. | n.a. |
Actual, refitted, and recovered covariance matrices of non-weighted regression coefficients in equations
Eq # | Actual | Refitted | Recovered | Recovered (robust) | P _{ 1 } _{×100 } | P _{ 2×100 } | P _{ 3 } _{×100 } |
---|---|---|---|---|---|---|---|
1 | \(\left( {\begin{array}{*{20}c} {2.62 \times 10^{ - 3} } & { - 0.132} \\ { - 0.132} & {8.67} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {3.30 \times 10^{ - 3} } & { - 0.179} \\ { - 0.179} & {11.4} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {2.75 \times 10^{ - 3} } & { - 0.154} \\ { - 0.154} & {10.2} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {3.01 \times 10^{ - 3} } & { - 0.167} \\ { - 0.167} & {10.9} \\ \end{array} } \right)\) | 0.5 | 64 | 68 |
2 | \(\left( {\begin{array}{*{20}c} {0.267} & { - 0.746} \\ { - 0.746} & {2.25} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {0.316} & { - 0.871} \\ { - 0.871} & {2.53} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {0.240} & { - 0.677} \\ { - 0.677} & {2.33} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {0.264} & { - 0.740} \\ { - 0.740} & {2.19} \\ \end{array} } \right)\) | 99 | 95 | 98 |
5 | \(\left( {\begin{array}{*{20}c} {1.14 \times 10^{ - 3} } & { - 0.605 \times 10^{ - 1} } \\ { - 0.605 \times 10^{ - 1} } & {3.56} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {1.39 \times 10^{ - 3} } & { - 0.598 \times 10^{ - 1} } \\ { - 0.598 \times 10^{ - 1} } & {3.00} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {1.02 \times 10^{ - 3} } & { - 0.436 \times 10^{ - 1} } \\ { - 0.436 \times 10^{ - 1} } & {2.21} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {1.08 \times 10^{ - 3} } & {0.465 \times 10^{ - 1} } \\ {0.465 \times 10^{ - 1} } & {2.35} \\ \end{array} } \right)\) | 4 | 6 | 7 |
6 | \(\left( {\begin{array}{*{20}c} {1.47 \times 10^{ - 1} } & { - 0.383} \\ { - 0.383} & {1.04} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {2.00 \times 10^{ - 1} } & { - 0.469} \\ { - 0.469} & {1.15} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {1.38 \times 10^{ - 1} } & { - 0.319} \\ { - 0.319} & {0.785} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {1.46 \times 10^{ - 1} } & { - 0.338} \\ { - 0.338} & {0.831} \\ \end{array} } \right)\) | 5 | 36 | 36 |
9 | \(\left( {\begin{array}{*{20}c} {5.53 \times 10^{ - 4} } & { - 0.322 \times 10^{ - 1} } \\ { - 0.322 \times 10^{ - 1} } & {2.19} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {1.33 \times 10^{ - 3} } & { - 0.636 \times 10^{ - 1} } \\ { - 0.636 \times 10^{ - 1} } & {3.44} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {5.61 \times 10^{ - 4} } & { - 0.273 \times 10^{ - 1} } \\ { - 0.273 \times 10^{ - 1} } & {1.50} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {6.05 \times 10^{ - 4} } & { - 0.293 \times 10^{ - 1} } \\ { - 0.293 \times 10^{ - 1} } & {1.61} \\ \end{array} } \right)\) | 4 | 3 | 5 |
10 | \(\left( {\begin{array}{*{20}c} {7.84 \times 10^{ - 2} } & { - 0.227} \\ { - 0.227} & {0.697} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {1.27 \times 10^{ - 1} } & { - 0.499} \\ { - 0.499} & {0.86} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {8.20 \times 10^{ - 2} } & { - 0.210} \\ { - 0.210} & {0.561} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {8.82 \times 10^{ - 2} } & { - 0.225} \\ { - 0.225} & {0.599} \\ \end{array} } \right)\) | 12 | 15 | 19 |
Taking into consideration that the relative error in the regression coefficient to diameter DBH or DBH^{2} is six to twelve times smaller than the relative error in the regression coefficient to \(\sqrt {\text{DBH} \times {\text{HT}}} \;{\text{or}}\;\sqrt {\text{HT}}\), a bias in the former is much more serious than in the latter. For un-weighted linear and nonlinear equations, the robust procedure appears as the most attractive. As well, the strong impact of errors in the first regression coefficient on a tree-level estimate of AGB amplifies concerns surrounding the overestimation of model-errors encountered with the refitting procedure.
Actual, refitted, and recovered covariance matrices of regression coefficients in weighted least squares equations in Table 1
Eq. # | Actual | Refitted | Recovered | Recovered (robust) | P _{ 1 } _{×100 } | P _{ 2 } _{×100 } | P _{ 3 } _{×100 } |
---|---|---|---|---|---|---|---|
3 | \(\left( {\begin{array}{*{20}c} {2.62 \times 10^{ - 3} } & { - 0.132} \\ { - 0.132} & {8.67} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {2.15 \times 10^{ - 3} } & { - 0.249} \\ { - 0.249} & {4.11} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {2.66 \times 10^{ - 3} } & { - 0.0995} \\ { - 0.0995} & {5.00} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {3.54 \times 10^{ - 3} } & { - 0.175} \\ { - 0.175} & {10.4} \\ \end{array} } \right)\) | 0.1 | 0.2 | 0.0 |
4 | \(\left( {\begin{array}{*{20}c} {0.268} & { - 0.747} \\ { - 0.747} & {2.25} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {0.203} & { - 1.60} \\ { - 1.60} & {1.26} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {0.300} & { - 0.723} \\ { - 0.723} & {1.90} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {0.493} & { - 1.28} \\ { - 1.28} & {3.51} \\ \end{array} } \right)\) | 13 | 8 | 0.0 |
7 | \(\left( {\begin{array}{*{20}c} {0.816 \times 10^{ - 3} } & { - 0.0336} \\ { - 0.0336} & {1.66} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {3.78 \times 10^{ - 3} } & { - 0.115} \\ { - 0.115} & {4.37} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {1.36 \times 10^{ - 3} } & { - 0.0417} \\ { - 0.0417} & {1.61} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {0.578 \times 10^{ - 3} } & { - 0.0251} \\ { - 0.0251} & {1.27} \\ \end{array} } \right)\) | 1 | 0.8 | 29 |
8 | \(\left( {\begin{array}{*{20}c} {0.139} & { - 0.328} \\ { - 0.328} & {0.813} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {0.521} & { - 1.01} \\ { - 1.01} & {2.11} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {0.192} & { - 0.381} \\ { - 0.381} & {0.82} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {0.297} & { - 0.630} \\ { - 0.630} & {1.41} \\ \end{array} } \right)\) | 0.0 | 7 | 4 |
11 | \(\left( {\begin{array}{*{20}c} {5.53 \times 10^{ - 4} } & { - 0.0322} \\ { - 0.0322} & {2.19} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {2.13 \times 10^{ - 3} } & { - 0.0778} \\ { - 0.0778} & {3.46} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {5.78 \times 10^{ - 4} } & { - 0.0213} \\ { - 0.0213} & {0.959} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {1.40 \times 10^{ - 3} } & { - 0.0549} \\ { - 0.0549} & {2.30} \\ \end{array} } \right)\) | 0.0 | 16 | 0.5 |
12 | \(\left( {\begin{array}{*{20}c} {7.84 \times 10^{ - 2} } & { - 0.227} \\ { - 0.227} & {0.698} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {0.153} & { - 0.342} \\ { - 0.342} & {0.813} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {8.96 \times 10^{ - 2} } & { - 0.202} \\ { - 0.202} & {0.485} \\ \end{array} } \right)\) | \(\left( {\begin{array}{*{20}c} {0.202} & { - 0.457} \\ { - 0.457} & {1.06} \\ \end{array} } \right)\) | 0.0 | 38 | 4 |
Recovering an estimate of the residual variance was, as expected, easier than recovering a covariance matrix. The relative error in recovered estimates of the residual standard error varied from approximately −20 to +35 %. Two of eight estimates were significantly different from the actual values (F-ratio test, P = 0.02), for the remaining six, the level of significance was 0.10 or greater.
Attempts at a recovery of the covariance matrices for the generalized above-ground biomass Eqs. 13–15 in Table 1 [26] failed, regardless of method. With the recovery methods, the estimated standard deviations of the three regression parameters were 2–8 times greater than those listed in Table 3 of Muukkonen [26]. Had we used the tabled values of the root mean squared error in lieu of the recovered substitute, the estimated errors would have been approximately 30–70 times too small. The failure is easy to explain: the fit-statistics of the generalized model apply to the set of models that are generalized. Footnotes to Table 3 in Muukkonen [26] carefully explain the constrained interpretation of the table entries. Due to the poor accuracy of the recovered generalized covariance matrices they were not used to gauge the error-propagation to estimates of tree-level AGB.
All recovery procedures are fraught with numerical problems due to co-linearity among regression coefficients (correlations coefficients varied between −0.87 and −0.97), and large differences in accuracy of parameter estimates. For example, the matrix condition number varied between 14.1 and 14.9, and determinants were less than 10^{−5} suggesting a serious potential of amplified estimation errors when inverting a covariance matrix [50, 51]. Challenges of this nature will also be encountered in applications of the proposed procedures.
Relative model errors (%) in estimates of the mean per tree above-ground tree biomass with actual (ACT), refitted (REFIT), recovered (RECOV), and robustly recovered (RREC) covariance matrices for the parameters in the biomass equations in Table 1
Species | Model | Weights? | ACT | REFIT | RECOV | RREC |
---|---|---|---|---|---|---|
BEECH | LIN | No | 11.5 | 14.0 | 12.3 | 12.8 |
NLIN | No | 8.7 | 9.1 | 8.0 | 8.4 | |
LIN | Yes | 4.7 | 8.8 | 7.8 | 11.6 | |
NLIN | Yes | 4.4 | 6.6 | 6.9 | 9.4 | |
SPRUCE | LIN | No | 9.8 | 8.4 | 7.2 | 7.4 |
NLIN | No | 7.0 | 6.6 | 5.4 | 5.6 | |
LIN | Yes | 6.2 | 10.0 | 5.4 | 7.4 | |
NLIN | Yes | 5.6 | 8.5 | 4.8 | 6.6 | |
PINE | LIN | No | 6.9 | 5.9 | 4.9 | 5.1 |
NLIN | No | 5.3 | 4.7 | 3.8 | 4.0 | |
LIN | Yes | 4.8 | 22.3 | 3.5 | 4.3 | |
NLIN | Yes | 4.2 | 19.0 | 3.0 | 3.8 |
Weighted regressions were uniformly superior with the lowest relative errors. Results with the substitute covariance matrices followed—by and large—these trends with estimates within one to 6 % points from results with the actual estimate of covariance. In the case of weighted regressions: two poor results with the refitting procedure with PINE data, and two for the robust recovery with BEECH data, stands out as examples of inflated estimates of model-error. The remaining estimates of error appear reasonable; yet do not indicate that one recovery procedure is substantially and consistently better than the presented alternatives.
Examples from Mexico
Actual, refitted, and recovered variances of regression coefficients in Eqs. 1–7 in Table 8
Eq # | Actual | Refitted | Recovered | Recovered (robust) | P _{ 1 } _{×100 } | P _{ 2 } _{×100 } | P _{ 3 } _{×100 } |
---|---|---|---|---|---|---|---|
1 | (0.04, 0.01) | (0.08, 0.01) | (0.07, 0.01) | (0.10, 0.02) | 56 | 74 | 49 |
2 | (0.55, 0.42, 0.020) | (2.64, 1.39, 0.044) | (2.36, 1.25, 0.039) | (2.88, 1.52, 0.048) | 0.2 | 0.1 | 0.1 |
3 | (0.27, 0.048) | (0.29, 0.041) | (0.26, 0.039) | (0.35, 0.053) | 99 | 98 | 96 |
4 | (0.084, 0.017) | (0.33, 0.049) | (0.29, 0.043) | (0.37, 0.06) | 2 | 5 | 1 |
5 | n.a. | (0.32, 7.1)10^{−3} | (0.20, 9.4)10^{−3} | (0.23, 10.6)10^{−3} | n.a. | n.a. | n.a. |
6 | n.a. | (0.61, 4.7)10^{−1} | (0.42, 32.2)10^{−2} | (0.12, 9.0)10^{−1} | n.a. | n.a. | n.a. |
7 | n.a. | (0.98, 16.8)10^{−3} | (0.29, 19.4)10^{−3} | (0.74, 29.4)10^{−3} | n.a. | n.a. | n.a. |
Substitute estimates of the residual standard error were considerably and statistically significantly smaller (30–240 %) than the published values. These results paired with the inflation of the variance of regression coefficients suggest a much smaller variation of the explanatory variables in the samples from the national inventory than in the sample used for fitting. A uniform distribution of the explanatory variables in the model fitting sample [53] could explain our results.
Tabled estimates of the residual standard errors for the three Quercus spp. Were three to four times smaller than recovered estimates. We noted that even a small reduction of 1–2 % in the published value of \(\hat{R}^{2}\) would bring the two sets of estimates within approximately 20 % of each other. Power functions are notorious in this regard.
Estimates of mean AGB kg tree^{−1} and relative errors (%) in estimates in mean AGB for seven Mexican species
Species | \(\hat{\text{AGB}}\) | REFIT | RECOV | RREC |
---|---|---|---|---|
Guazuma ulmifolia | 54 | 5.0 | 4.7 | 5.4 |
Inga vera | 198 | 8.6 | 8.2 | 9.0 |
Ochroma pyramidale | 50 | 11.8 | 11.2 | 12.8 |
Trichospernum mexicanum | 51 | 13.9 | 13.1 | 15.1 |
Quercus castenea | 75 | 23.2 | 28.6 | 33.9 |
Quercus latea | 129 | 101.7 | 23.9 | 30.7 |
Quercus spp. | 135 | 53.0 | 13.1 | 13.8 |
Discussion
The need for forest biomass equations has increased sharply over the past decades in response to efforts directed at quantifying stock and stock-changes in forest carbon and the potential for bioenergy extraction [21, 42, 54]. Ideally there would be an equation for each tree species and region with distinct growth forms and management regimes [55, 56]. We are still far from this ideal. Even the equations we have are generally based on very limited sampling within a relatively small area and range of tree sizes [21]. This is understandable in light of the high costs of producing a biomass equation [21, 26, 57]. Biomass estimates for large trees are therefore fraught with problems of applicability of available biomass equations.
In the computation of forest biomass in a large region, country, or even a continent, it is common practice to use a suitable biomass equation for a particular species and growth region [58–60]. In most cases, there is no separate calibration of chosen biomass equations.
On this background, national and regional estimates of above-ground biomass should be regarded as no more than first-order approximations [16]. The requirement [11] to quantify or at least assess uncertainty in a national or regional estimate of forest biomass has precipitated a need for estimates of errors in the parameters of employed biomass equations. For a large number of equations, this information is partially or entirely missing [31, 44].
In a context of model-dependent estimation of forest tree biomass and model-errors in these estimates, a covariance matrix of the model parameters is needed [12, 40]. When this statistic is missing a substitute is needed. Wayson et al. [44] proposed a computationally intensive method for generating a large number of pseudo data of the dependent and independent variables in a biomass equation. Samples are then drawn repeatedly and the model is refitted each time. The sampling aims at mimicking the actual sampling process (if known) of the original data behind a biomass equation.
Our proposed procedures for computing a substitute for a missing covariance matrix are computationally faster and make direct use of data of the explanatory variables sampled from the population targeted for an estimation of biomass. The distribution of the explanatory variables used to compute (recover) a covariance matrix plays a pivotal role in both approaches. If the actual distribution behind an equation differs from the distribution in the recovery process, a covariance matrix different from the actual (but unknown) will emerge from a recovery procedure. We saw several examples of this in our examples, but an equal number of examples where a substitute matrix was not statistically different from the target matrix. Wayson et al. [44] do not report at this level of details, but we surmise that they encountered similar issues. It is now a question of whether these differences are relevant or not. We argue, that sampling the explanatory variables from the target population vouch for estimates adapted to the application domain rather than to a small sample of trees with unknown representation in the target population.
The most intuitive approach to recover a missing covariance matrix is a variant of the parametric bootstrap [61]. In the textbook version of a parametric bootstrap, n pseudo observations of Y are generated a large number of times (say B) by adding a random draw from the observed empirical regression residuals to the n model predictions obtained from the original regression model and the observed explanatory variables. The regression model is then refitted B times to the pseudo observations of Y. At the end, the analyst has B replications of the covariance matrix of the model parameters. Without observed residuals, this approach is not feasible. Instead, our recovery by refitting resorted to random sampling of the explanatory variables from the target population for biomass estimation, and residuals from a distribution deemed realistic to the case at hand (e.g. a gamma distribution for multiplicative residuals). Although this method in many cases was as good as with alternative approaches, it was equally clear that it entails a considerable risk of poor results. A risk traced to random interactions between residuals and the explanatory variables. For that reason we do not recommend a recovery by model refitting with simulated residuals.
The examples from Germany confirmed that in presence of heteroscedasticity in the model residuals, a weighting with the inverse to the presumed residual variance can be effective [62, ch. 2.11, 63, ch. 2.1]. To carry this efficiency through to a recovered covariance matrix, a weighting scheme applied to the original biomass equation should be replicated in a recovery procedure.
A matrix recovery based on the average (vector) gradient of the model parameters with respect to the explanatory variables was, in the balance, better suited for the purpose of estimation of model errors in tree-level biomass estimates. A robust variant of the recovery is easy to compute and—despite expected and observed larger estimates of model-errors—we recommend this procedure as a prudent choice. For the purpose of reasonable estimates model-errors in tree-level estimates of biomass, it is not a strict requirement that a recovered covariance matrix is close to the actual but missing matrix. Most of our estimates, but especially those obtained with the robust recovery procedure, seem reasonable [13, 16, 20, 57, 64]. Our resampling of explanatory variables from inventory data representing the population targeted for an estimation of biomass, ensures that the mean of the explanatory variables will be close to the mean in the target population. Ceteris paribus, this will counter the aforementioned inflation of model-parameter variances [39, ch. 5.4].
An attempt to recover a covariance matrix can end in failure. A failure was demonstrated with the generalized biomass equations for beech, pine, and spruce in the temperate zone [26, Table 1]. A failure is pre-ordained when estimates of R^{2} and a root mean squared errors are incompatible with the biomass equation applied to actual data. Our experience should raise awareness of potential pitfalls in published fit-statistics for a generalized equation, unless they reflect a proper meta-analysis [65].
Throughout we have treated published fit statistics as known entities. It would have been preferable to consider an empirical Bayesian recovery procedure [66]. The coefficient of determination is pivotal in our proposed procedures. Its sampling variance can only be estimated from the data supporting a biomass equation [67]. To recognize sampling variance in R ^{2}, a recovery is repeated a large number of times, each with a random draw from an anticipated distribution of R ^{2}, to create an empirical Bayes posterior distribution of the recovered statistic. The recovery procedure by Wayson et al. [44] contains elements of a Bayesian approach.
Although a recovered covariance matrix affords an estimate of the model error in a tree level biomass estimate, the model-error is conditional on a correctly specified model. If a published biomass equation is the result of an intensive model and variable screening process, we must expect optimism in published statistics and model-bias [68].
We have demonstrated the recovery of a missing covariance matrix without too much concern about sample size. Clearly, a biomass equation derived from a small sample size has a relatively high risk of model bias due to a high influence of individual observations [62, p. 170]. It is not possible to give a definite recommendation about the minimum sample size for our robust recovery procedure. However, a first approximation can be gained from the following example: If we have fitted a linear regression model with three parameters, and we wish to declare a standardized regression residual of 3 as significant at the 5 % level (an indication that the model is unduly influenced by residuals of this magnitude), we need a sample size of approximately 55 [69]. Thus an application of our recovery procedure for regression models supported by less than 55 trees should proceed with caution and attention to robustness.
In large sample inventories the model errors in point estimates of biomass will often dominate sampling errors [12, 40]. Fortunately, when estimating a temporal change in biomass and carbon stock between two inventories, model errors in a difference all but cancel [Ibid]. Thus applying recovered conservative (robust) estimates of a missing covariance matrix will have little impact on the estimate of model errors in a difference.
We have demonstrated that reasonable (robust) estimates of model-errors in estimates of tree-level biomass can be derived from a minimum of two available fit statistics for a biomass equation: the coefficient of determination, and sample size. To complete an estimation of model-errors an analyst need access to forest inventory sample data of the explanatory variables from the population targeted for biomass estimation.
Conclusions
It is good practice to provide estimates of uncertainty to any model-dependent estimate of above ground biomass. When a direct approach to estimate uncertainty is impossible due to missing model statistics, the proposed robust procedure is a first step to good practice. Our recommended approach offers protection against inflated estimates of precision.
Methods
The biomass model
The estimation problem
It is clear from (2) that we cannot estimate the error in an out-of-sample estimate of the AGB in a single tree unless we have reasonable estimates of \(\hat{\sigma }_{e}^{2}\) and \(\hat{\text{cov}}\left( {{\hat{\mathbf{b}}}} \right)\). Note, when we wish to estimate the error in an average of AGB in a large number (m) of trees, the contribution to the error from the residual variance can be ignored as it declines at a rate of m ^{−1}, the second term, however, is only averaged over m [62, pp. 28–30].
When we are tasked with estimating the error variance in (2) but do not have estimates of \(\hat{\sigma }_{e}^{2}\) or \(\hat{\text{cov}}\left( {{\hat{\mathbf{b}}}} \right)\) we have to recover reasonable substitutes. Equations (2) and (3) implicitly suggest how to obtain substitutes \(\tilde{\sigma }_{e}^{2}\) for \(\hat{\sigma }_{e}^{2}\) and \(\text{c} \widetilde{\text{o}}v\left( {{\hat{\mathbf{b}}}} \right)\) for \({\text{c}}\widetilde{\text{o}}{\text{v}}\left( {{\hat{\mathbf{b}}}} \right)\) when we at least know the sample size n used to estimate the parameters in the biomass model in (1), and the coefficient of determination \(\hat{R}^{2}\) or, preferably, the adjusted coefficient of determination [62, p. 91].
Recovery of missing fit statistics
It is clear from (4) that the estimate \(\tilde{\sigma }_{e}^{2}\) depends on the sampling distribution of \(f\left( {{\mathbf{x}}_{b,j} |{\hat{\mathbf{b}}}} \right)\) which may be quite different from the distribution in the original sample used in model fitting. Most biomass functions are fitted to an approximate uniform distribution of the explanatory variables, as it achieves large-sample optimality for model fitting [39, ch. 7.5]. However, for typically small sample sizes in biomass studies, this no longer holds. Our repeated sampling from the target population assuage more robust and realistic estimates of the desired covariance matrix. Albeit under the proviso that the reported coefficient of determination has not been maximized by a combination of model- and variable-selection procedures, and a sampling design that C. paribus favors a linear model.
Recovery via refitting
A recovered estimate of the residual variance (see (4)) can be used in a parametric bootstrap [71] to recover a substitute for a missing covariance matrix \(\text{c} \widetilde{\text{o}}v\left( {{\hat{\mathbf{b}}}} \right).\) The refitting begins with n random draws of residuals (e _{ j } ^{*} , j = 1, …, n) from a t-distribution with n − q degrees of freedom. Pseudo data \(y_{j}^{*} = f\left( {{\mathbf{x}}_{j} |{\hat{\mathbf{b}}}} \right) + e_{j}^{*}\) is then used to re-estimate the parameters \({\hat{\mathbf{b}}}_{{}}^{*}\) and the associated covariance matrix \(\text{c} \widetilde{\text{o}}v\left( {{\mathbf{b}}^{*} } \right)\). This process is repeated B times; the mean of the covariance matrices is now the substitute to use in computing an error of AGB via (2).
Recovery of off-diagonal elements in \(\text{c} \widetilde{\text{o}}v\left( {{\hat{\mathbf{b}}}} \right)\)
Robust recovery
A weighted recovery
In regressions with a positively valued dependent variable (y), it is not uncommon to observe an increase in the variance of regression residuals with an increase in y [73, ch. 5.1]. A weighted least squares (WLS) approach to model-fitting would be appropriate. If f (x _{ i }; b) was fitted using WLS the recovery of a substitute for \({\text{c}}\hat{\text{o}}{\text{v}}\left( {{\hat{\mathbf{b}}}} \right)\) should also employ a weighting scheme. Equation [8] provides an example.
A weighting scheme is also needed when trees for model-fitting were selected by an unequal probability selection scheme. Weights should then be proportional to the inverse of the sample inclusion probability [73, p. 41].
The number B of resampling replications
The value of B was determined adaptively by monitoring the Monte Carlo error as a function of B [74]. In our examples we fixed B to 800. With this value of B, the Monte Carlo error in the determinant of \({\text{c}}\widetilde{\text{o}}{\text{v}}\left( {{\hat{\mathbf{b}}}} \right)\) was less than 4 %.
Comparing recovered and actual covariance matrices
A recovered substitute for a covariance matrix may vary considerably from an unknown target estimate when the joint distribution of the explanatory variables in the sample used for fitting differs from the joint-distribution in the target population for model application. In our demonstrations we knew, in most cases, the actual estimates of the missing covariance matrix. It is therefore of interest to test the hypothesis of equality between a recovered substitute and the actual estimate. We use Box’s M-test to obtain a Chi square test-statistic and the probability of this test statistic under the null hypothesis of no difference [75, p. 281]. The same test was applied in examples where only the covariance in \({\hat{\mathbf{b}}}\) are unknown.
Applications
Examples from Germany
We demonstrate the above recovery procedures with 15 biomass equations (Table 1) and data (HT, DBH) from 335 plots in the first German national forest inventory (BWI-1987). Note, the data represent trees selected with probability proportional to their basal area. Their mean HT and DBH are therefore larger than the mean of trees selected with equal probability. However, for purpose of a demonstration, this fact is deemed unimportant.
Means of DBH, HT, and AGB of trees from the 1987 German National Inventory used in this study
Trees | DBH (cm) | HT (m) | AGB (kg/tree) | |
---|---|---|---|---|
BEECH | 1,595 | 35 (15) | 25 (7) | 1,111 (1,110) |
SPRUCE | 2,221 | 30 (12) | 24 (7) | 492 (460) |
PINE | 1,221 | 33 (12) | 23 (6) | 1,100 (410) |
Examples from Mexico
Above-ground forest tree biomass (AGB kg tree^{−1}) equations for five species and a species group in Mexico
Species | # | Equation | \(\hat{R}^{2}\) | \(\hat{\sigma }_{e}^{{}}\) kg | Sample size |
---|---|---|---|---|---|
Guazuma ulmifolia | 1 | \(\log \left( {\hat{\text{AGB}}} \right) = - 1.62 + 2.12\log \left( {\text{DBH}} \right)\) | 0.97 | 0.48 | 18 |
Inga vera | 2 | \(\log \left( {\hat{\text{AGB}}} \right) = - 4.04 + 4.00\log \left( {\text{DBH}} \right) - 0.29\log \left( {\text{DBH}} \right)^{2}\) | 0.97 | 0.39 | 15 |
Ochroma pyramidale | 3 | \(\log \left( {\hat{\text{AGB}}} \right) = - 2.45 + 2.30\log \left( {\text{DBH}} \right)\) | 0.90 | 0.96 | 16 |
Trichospernum mexicanum | 4 | \(\log \left( {\hat{\text{AGB}}} \right) = - 2.82 + 2.42\log \left( {\text{DBH}} \right)\) | 0.96 | 0.40 | 16 |
Quercus castenea | 5 | \(\hat{\text{AGB}} = 0.0416{\text{DBH}}^{2.7154}\) | 0.97 | 11.6 | 38 |
Quercus latea | 6 | \(\hat{\text{AGB}} = 0.0333{\text{DBH}}^{2.6648}\) | 0.92 | 12.8 | 7 |
Quercus spp. | 7 | \(\hat{\text{AGB}} = 0.0342{\text{DBH}}^{2.7590}\) | 0.93 | 15.7 | 45 |
Summary of tree size (mean DBH cm, mean HT m), stem density of species groups (N ha^{−1}), and model-dependent predictions of above-ground forest tree biomass (AGB Mg ha^{−1}) in the Mexican NFI (2004–2009) plots
Species | State | Stratum | Trees | n _{ plots } | DBH | HT | N × ha^{−1} |
---|---|---|---|---|---|---|---|
Guazuma ulmifolia | Chiapas | Mediana subperennifolia | 177 | 12 | 12.9 (5.1) | 5.5 (2.4) | 201 |
Inga vera | Chiapas | Alta pernnifolia | 37 | 12 | 17.1 (9.5) | 12.0 (4.1) | 103 |
Ochroma pyramidale | Chiapas | Alta pernnifolia | 144 | 20 | 13.9 (6.2) | 10.1 (2.6) | 49 |
Trichospernum mexicanum | Chiapas | Alta pernnifolia | 612 | 48 | 13.9 (6.5) | 9.6 (3.3) | 156 |
Quercus castenea | Michoacàn | Bosque de encino | 416 | 17 | 16.6 (7.7) | 7.1 (2.6) | 612 |
Quercus latea | Michoacàn | Bosque de pino | 37 | 4 | 17.2 (8.5) | 9.5 (3.2) | 231 |
Quercus spp. | Michoacàn | Bosque de encino | 420 | 19 | 16.6 (7.7) | 7.0 (2.6) | 553 |
Application of recovered statistics
The above recovery procedures are motivated by the need to supply inventory estimates of AGB with an estimate of sampling and model errors. The latter is not possible without published or recovered substitutes for missing values of \(\hat{\sigma }_{e}\) and \({\text{c}}\widetilde{\text{o}}{\text{v}}\left( {{\hat{\mathbf{b}}}} \right)\).
Declarations
Acknowledgements
We want to thank Heino Polley, Thünen-Institute for Forest Ecology, for making the data of the German National Forest Inventory available for this study.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
Authors’ Affiliations
References
- Kindermann GE, McCallum I, Fritz S, Obersteiner M (2008) A global forest growing stock, biomass and carbon map based on FAO statistics. Silv Fenn 42:387Google Scholar
- Pan Y, Birdsey RA, Fang J, Houghton R, Kauppi PE, Kurz WA et al (2011) A large and persistent carbon sink in the world’s forests. Science 333:988–993View ArticleGoogle Scholar
- Houghton R (2005) Aboveground forest biomass and the global carbon balance. Glob Chang Biol 11:945–958View ArticleGoogle Scholar
- Goodale CL, Apps MJ, Birdsey RA, Field CB, Heath LS, Houghton RA et al (2002) Forest carbon sinks in the Northern Hemisphere. Ecol Appl 12:891–899View ArticleGoogle Scholar
- Hansen MC, Stehman SV, Potapov PV (2010) Quantification of global gross forest cover loss. Proc Natl Acad Sci 107:8650–8655View ArticleGoogle Scholar
- Köhl M, Baldauf T, Plugge D, Krug J (2009) Reduced emissions from deforestation and forest degradation (REDD): a climate change mitigation strategy on a critical track. Carbon Balance Manag 4:10View ArticleGoogle Scholar
- Martin H, Margaret S (2011) Monitoring, reporting and verification for national REDD+ programmes: two proposals. Environ Res Lett 6:014002View ArticleGoogle Scholar
- Plugge D, Baldauf T, Köhl M (2011) Reduced emissions from deforestation and forest degradation (REDD): why a robust and transparent monitoring, reporting and verification (MRV) system is mandatory. In: Blanco J, Kheradmand H (eds) Climate change–research and technology for adaptation and mitigation, chap 9. InTech, Rijeka, pp 155–170Google Scholar
- Stinson G, Kurz WA, Smyth CE, Neilson ET, Dymond CC, Metsaranta JM et al (2011) An inventory-based analysis of Canada’s managed forest carbon dynamics, 1990 to 2008. Glob Chang Biol 17:2227–2244View ArticleGoogle Scholar
- Wertz-Kanounnikoff S, Verchot LV, Kanninen M, Murdiyarso D (2008) How can we monitor, report and verify carbon emissions from forests. In: Angelsen A (ed) Moving ahead with REDD: issues, options, and implications, chap 9. CIFOR, Bogor, Indonesia, pp 87–98Google Scholar
- Frey C, Penman J, Hanley L, Suvi MOS (2006) Uncertainties. In: Eggleston HS, Buendia L, Miwa K, Ngara T, Tanabe K (eds) 2006 IPCC guidelines for national greenhouse gas inventories. Prepared by the National Greenhouse Gas Inventories Programme, vol 1. Institute for Global Environmental Strategies (IGES), Hayama, Kanagawa, JP, p 66Google Scholar
- Ståhl G, Heikkinen J, Petersson H, Repola J, Holm S (2014) Sample-based estimation of greenhouse gas emissions from forests: a new approach to account for both sampling and model errors. For Sci 60:3–13Google Scholar
- Breidenbach J, Antón-Fernández C, Petersson H, McRoberts RE, Astrup R (2014) Quantifying the model-related variability of biomass stock and change estimates in the Norwegian National Forest Inventory. For Sci 60:25–33Google Scholar
- Gasparini P, Gregori E, Pompei E, Rodeghiero M (2010) The Italian national forest inventory: survey methods for carbon pools assessment. Sherwood - Foreste ed Alberi Oggi (168):13–18Google Scholar
- Podur J, Wotton M (2010) Will climate change overwhelm fire management capacity? Ecol Model 221:1301–1309View ArticleGoogle Scholar
- Petersson H, Holm S, Ståhl G, Alger D, Fridman J, Lehtonen A et al (2012) Individual tree biomass equations or biomass expansion factors for assessment of carbon stock changes in living biomass—a comparative study. For Ecol Manag 270:78–84View ArticleGoogle Scholar
- Berger A, Gschwantner T, McRoberts RE, Schadauer K (2014) Effects of measurement errors on individual tree stem volume estimates for the Austrian National Forest Inventory. For Sci 60:14–24Google Scholar
- McRoberts RE, Westfall JA (2013) Effects of uncertainty in model predictions of individual tree volume on large area volume estimates. For Sci 60:34–42Google Scholar
- Gertner GZ, Köhl M (1992) An assessment of some nonsampling errors in a national survey using an error budget. For Sci 38:525–538Google Scholar
- Moundounga Mavouroulou Q, Ngomanda A, Engone Obiang NL, Lebamba J, Gomat H, Mankou GS et al (2014) How to improve allometric equations to estimate forest biomass stocks? Some hints from a central African forest. Can J For Res 44:685–691View ArticleGoogle Scholar
- Zianis D, Seura SM (2005) Biomass and stem volume equations for tree species in Europe. Silv Fenn Monogr 4:63Google Scholar
- Niiyama K, Kajimoto T, Matsuura Y, Yamashita T, Matsuo N, Yashiro Y, et al (2010) Estimation of root biomass based on excavation of individual root systems in a primary dipterocarp forest in Pasoh Forest Reserve, Peninsular Malaysia. J Trop Ecol 26:271–284View ArticleGoogle Scholar
- Segura M, Kanninen M (2005) Allometric models for tree volume and total aboveground biomass in a tropical humid forest in Costa Rica. Biotropica 37:2–8View ArticleGoogle Scholar
- Nogueira EM, Fearnside PM, Nelson BW, Barbosa RI, Keizer EWH (2008) Estimates of forest biomass in the Brazilian Amazon: new allometric equations and adjustments to biomass from wood-volume inventories. For Ecol Manage 256:1853–1867View ArticleGoogle Scholar
- Yarie J, Kane E, Mack M (2007) Aboveground biomass equations for the trees of interior Alaska. Agricultural and Forestry Experiment Station Bulletin, 115. University of Alaska-Fairbanks, Fairbanks, p 16Google Scholar
- Muukkonen P (2007) Generalized allometric volume and biomass equations for some tree species in Europe. Eur J For Res 126:157–166View ArticleGoogle Scholar
- Lambert MC, Ung CH, Raulier F (2005) Canadian national tree aboveground biomass equations. Can J For Res 35:1996–2018View ArticleGoogle Scholar
- Levy PE, Hale SE, Nicoll BC (2004) Biomass expansion factors and root: shoot ratios for coniferous tree species in Great Britain. Forestry (Oxford) 77:421–430View ArticleGoogle Scholar
- Lehtonen A, Mäkipää R, Heikkinen J, Sievänen R, Liski J (2004) Biomass expansion factors (BEFs) for Scots pine, Norway spruce and birch according to stand age for boreal forests. For Ecol Manag 188:211–224View ArticleGoogle Scholar
- Nelson BW, Mesquita R, Pereira JLG, de Souza SGA, Batista GT, Couto LB (1999) Allometric regressions for improved estimate of secondary forest biomass in the central Amazon. For Ecol Manag 117:149–167View ArticleGoogle Scholar
- Ter-Mikaelian MT, Korzukhin MD (1997) Biomass equations for sixty-five North American tree species. For Ecol Manag 97:1–24View ArticleGoogle Scholar
- Senelwa K, Sims REH (1997) Tree biomass equations for short rotation eucalypts grown in New Zealand. Biomass Bioenergy 13:133–140View ArticleGoogle Scholar
- Singh T (1984) Biomass equations for six major tree species of the Northwest Territories. Information Report NOR-X-257. Environment Canada, Canadian Forestry Service, Northern Forest Research Centre, Edmonton, AlbertaGoogle Scholar
- Singh T (1982) Biomass equations for ten major tree species of the prairie provinces. Information ReportNOR-X-242. Environment Canada, Canadian Forestry Service, Northern Forest Research Centre, Edmonton, AlbertaGoogle Scholar
- Ker MF (1980) Tree biomass equations for ten major species in Cumberland County, Nova Scotia. Information Report. M-X-108. Environment Canada, Canadian Forestry Service, Maritimes Forest Research Centre, Fredericton, New BrunswickGoogle Scholar
- Green DG, Grigal DF (1978) Generalized biomass equations for jack pine (Pinus banksiana Lamb.). Minnesota School of Forestry, Research Note 268Google Scholar
- Yang YQ, Monserud RA, Huang SM (2004) An evaluation of diagnostic tests and their roles in validating forest biometric models. Can J For Res 34:619–629View ArticleGoogle Scholar
- Lappi J (1991) Calibration of height and volume equations with random parameters. For Sci 37:781–801Google Scholar
- Chambers RL, Clark RG (2012) An introduction to model-based survey sampling with applications. Oxford University Press, New YorkView ArticleGoogle Scholar
- Magnussen S, Köhl M, Olschofsky K (2014) Error propagation in stock-difference and gain–loss estimates of a forest biomass carbon balance. Eur J For Res 133:1137–1155View ArticleGoogle Scholar
- Claeskens G, Hjort NL (2008) Model selection and model averaging. Cambridge University Press, CambridgeView ArticleGoogle Scholar
- Woodall CW, Heath LS, Domke DM, Nichols MC (2011) Methods and equations for estimating aboveground volume, biomass, and carbon for trees in the U.S. forest inventory, 2010. General technical report, NRS-88. Newtown Square, PA, p 30Google Scholar
- Návar J (2009) Allometric equations for tree species and carbon stocks for forests of northwestern Mexico. For Ecol Manag 257:427–434View ArticleGoogle Scholar
- Wayson CA, Johnson KD, Cole JA, Olguín MI, Carrillo OI, Birdsey RA (2014) Estimating uncertainty of allometric biomass equations with incomplete fit error information using a pseudo-data approach: methods. Anna For Sci. doi:10.1007/s13595-014-0436-7 Google Scholar
- Draper D (1995) Assessment and propagation of model uncertainty. J R Stat Soc Ser B 57:45–97Google Scholar
- Kublin E, Scharnagl G (1988) Biometrische Lösungen für die Berechnung des Volumens, der Sortierung. der Rindenabzüge und der Ernteverluste im Rahmen der Bundeswaldinventur, Verfahrens- und Programmbeschreibung zum BWI-Unterprogramm BDATGoogle Scholar
- Couturier S, Mas JF, López-Granados E, Benítez J, Coria-Tapia V, Vega-Guzmán Á (2010) Accuracy assessment of the Mexican National Forest Inventory map: a study in four ecogeographical areas. Singap J Trop Geogr 31:163–179View ArticleGoogle Scholar
- Couturier S, Mas J-F, Vega A, Tapia V (2007) Accuracy assessment of land cover maps in sub-tropical countries: a sampling design for the Mexican National Forest Inventory map. Online J Earth Sci 1:127–135Google Scholar
- Magnussen S, Smith B, Uribe AS (2007) National Forest Inventories in North America for monitoring forest species diversity. Plant Biosyst 141:113–122View ArticleGoogle Scholar
- Edelman A (1988) Eigenvalues and condition numbers of random matrices. SIAM J Matrix Anal Appl 9:543–560View ArticleGoogle Scholar
- Ledoit O, Wolf M (2004) A well-conditioned estimator for large-dimensional covariance matrices. J Multivar Anal 88:365–411View ArticleGoogle Scholar
- Aguilar R, Ghilardi A, Vega E, Skutsch M, Oyama K (2012) Sprouting productivity and allometric relationships of two oak species managed for traditional charcoal making in central Mexico. Biomass Bioenergy 36:192–207View ArticleGoogle Scholar
- Douterlungne D, Herrera-Gorocica AM, Ferguson BG, Siddique I, Soto-Pinto L (2013) Allometric equations used to estimate biomass and carbon in four neotropical tree species with restoration potential. Agrociencia 47:385–397Google Scholar
- Domke GM, Woodall CW, Smith JE, Westfall JA, McRoberts RE (2012) Consequences of alternative tree-level biomass estimation procedures on U.S. forest carbon stock estimates. For Ecol Manag 270:108–116View ArticleGoogle Scholar
- Alves LF, Vieira SA, Scaranello MA, Camargo PB, Santos FAM, Joly CA et al (2010) Forest structure and live aboveground biomass variation along an elevational gradient of tropical Atlantic moist forest (Brazil). For Ecol Manag 260:679–691View ArticleGoogle Scholar
- Groen TA, Verkerk PJ, Böttcher H, Grassi G, Cienciala E, Black KG et al (2013) What causes differences between national estimates of forest management carbon emissions and removals compared to estimates of large-scale models? Environ Sci Policy 33:222–232View ArticleGoogle Scholar
- Brown S, Gillespie AJR, Lugo AE (1989) Biomass estimation methods for tropical forests with applications to forest inventory data. For Sci 35:881–902Google Scholar
- Kurz WA, Dymond CC, White TM, Stinson G, Shaw CH, Rampley GJ et al (2009) CBM-CFS3: a model of carbon-dynamics in forestry and land-use change implementing IPCC standards. Ecol Model 220:480–504View ArticleGoogle Scholar
- Gonzalez P, Kroll B, Vargas CR (2014) Tropical rainforest biodiversity and aboveground carbon changes and uncertainties in the Selva Central, Peru. For Ecol Manag 312:78–91View ArticleGoogle Scholar
- Gallaun H, Zanchi G, Nabuurs GJ, Hengeveld G, Schardt M, Verkerk PJ (2010) EU-wide maps of growing stock and above-ground biomass in forests based on remote sensing and field measurements. For Ecol Manag 260:252–261View ArticleGoogle Scholar
- Efron B (1982) The jackknife, the bootstrap, and other resampling plans. Conference Board of Mathematical Science/National Science Foundation, PhiladelphiaView ArticleGoogle Scholar
- Draper NR, Smith H (2014) Applied regression analysis, 3rd edn. Wiley, New YorkGoogle Scholar
- Gallant AR (1987) Nonlinear statistical methods. Wiley, New YorkView ArticleGoogle Scholar
- Fehrmann L, Lehtonen A, Kleinn C, Tomppo E (2008) Comparison of linear and mixed-effect regression models and a k-nearest neighbour approach for estimation of single-tree biomass. Can J For Res 38:1–9View ArticleGoogle Scholar
- Wirth C, Schumacher J, Schulze ED (2004) Generic biomass functions for Norway spruce in Central Europe—a meta-analysis approach toward prediction and uncertainty estimation. Tree Physiol 24:121–139View ArticleGoogle Scholar
- Rao J, Wu C (2010) Bayesian pseudo-empirical-likelihood intervals for complex surveys. J R Stat Soc: Ser B (Stat Methodol) 72:533–544View ArticleGoogle Scholar
- Éric M (1997) On moments of beta mixtures, the noncentral beta distribution, and the coefficient of determination. J Stat Comput Simul 59:161–178View ArticleGoogle Scholar
- Efron B (2014) Estimation and accuracy after model selection. J Am Stat Assoc 109:991–1007View ArticleGoogle Scholar
- Cook RD (1977) Detection of influential observations in linear regression. Technometrics 19:15–18View ArticleGoogle Scholar
- Wolter KM (2007) Introduction to variance estimation, 2nd edn. Springer, New YorkGoogle Scholar
- Efron B, Tibshirani RJ (1993) An introduction to the bootstrap. Chapman & Hall, Boca RatonView ArticleGoogle Scholar
- Goodman LA (1960) On the exact variance of products. J Am Stat Assoc 55:708–713View ArticleGoogle Scholar
- Valliant R, Dorfman AH, Royall RM (2000) Finite population sampling and inference. A prediction approach. Wiley, New YorkGoogle Scholar
- Koehler E, Brown E, Haneuse J-PA (2009) On the assessment of Monte Carlo error in simulation-based statistical analyses. Am Stat 63:155–162View ArticleGoogle Scholar
- Rencher AC (1995) Methods of multivariate analysis. Wiley, New YorkGoogle Scholar
- Muukkonen P, Heiskanen J (2007) Biomass estimation over a large area based on standwise forest inventory data and ASTER and MODIS satellite data: a possibility to verify carbon inventories. Remote Sens Environ 107:617–624View ArticleGoogle Scholar
- Gregoire TG, Valentine HT (2008) Sampling strategies for natural resources and the environment. Chapman & Hall/CRC, Boca RatonGoogle Scholar