CFT

The CFT package for R

• The CFT package for R
  • IMPORTANT
  • References
  • Introduction
  • Requirements
  • How to install the CFT package
  • Examples
    • Estimating beta diversity
    • Variance estimators for greenhouse gas emissions in national reporting
    • A model of merchantable volume for 26 tree species
• Version history

IMPORTANT

The dissimilarity indices have migrated to an independent package called betadiv. All the information is available here.

References

Fortin, M. 2020. Package CFT. Version 0.1.6. PDF documentation.
Fortin, M. 2021. Package CFT. Version 0.2.0. PDF documentation.

Introduction

The CFT package is meant to provide a connection to a Java library that implements forest-related models in the Canadian context. At the moment, two features are available through the CFT package:

• Variance estimators for greenhouse gas emissions (Fortin 2021)
• A model of merchantable volume for 26 tree species in Québec, Canada (Fortin et al. 2007 (doi.org))

Requirements

To properly work, the CFT package requires the installation of the J4R package, which ensures the connection to Java. Please see the J4R webpage for more information about the installation of J4R and its requirements.

How to install the CFT package

The CFT package relies on the J4R package. To make it work you will have to go through the following steps:

1- Install the J4R package by copying this line of code in your R console:

install.packages("https://sourceforge.net/projects/repiceasource/files/latest", repos = NULL,  type="source")

2- Then download and install the CFT package by copying this line of code in your R console:

install.packages("https://sourceforge.net/projects/mrnfforesttools/files/latest", repos = NULL,  type="source")

Examples

Estimating beta diversity

The dissimilarity indices have migrated to an independent package called betadiv. All the information is available here.

Variance estimators for greenhouse gas emissions in national reporting

The estimation of greenhouse gas emissions often involves the product of two or more estimates. Several estimators of the variance of this product exist. The CFT contains functions and classes that implement these estimators. The createEstimate function requires two arguments, one for the mean and the other for the variance. An optional argument is the distribution of the estimate, which can be either dist="normal", dist="uniform" or dist="lognormal". By default, this distribution is assumed to be normal. The function returns an object of the class estimate as follows:

> library(CFT)
> est1 <- createEstimate(mean = 5, variance = 20)
> class(est1)
[1] "estimate"
attr(,"package")
[1] "CFT"
> summary(est1)
[1] "Mean = 5 ; Variance = 20 ; Distribution = normal"
>

In this example, the object est1 is a normally-distirbuted estimate with mean 5 and variance 20. The estimate can be plotted using the plotEstimate function:

> plotEstimate(est1)
>

which yields a graph similar to this one:

If we create another estimate called est2 and we are interested in the product of est1 by est2, the function multiplyEstimate can be used to obtain an estimate of the product and its variance:

> est2 <- createEstimate(mean = 4, variance = 8)
> product <- multiplyEstimates(est1,est2)
> summary(product)
[1] "Mean = 20 ; Variance = 360 ; Distribution = unknown"
>

The multiplyEstimate function implements five variance estimators

1. a naive variance estimator
2. a propagation-based estimator
3. Goodman's estimator
4. a Monte Carlo estimator
5. a rescaled Monte Carlo estimator

By default, the function relies on Goodman's estimator. However, the user can decide to use the naive variance estimator as follows:

> product.naive <- multiplyEstimates(est1, est2, method="Naive")
> summary(product.naive)
[1] "Mean = 20 ; Variance = 680 ; Distribution = unknown"
>

The Monte Carlo and rescaled Monte Carlo estimators are both based on resampling. By default the number of realizations is set to 5000. The user can change the number of realizations through the realMC argument as in the following example:

> product.mc <- multiplyEstimates(est1, est2, method = "MonteCarlo", realMC = 10000)
> summary(product.mc)
[1] "Mean = 19.6712569018852 ; Variance = 657.241031072833 ; Distribution = empirical"
> 

The product.mc object is an instance of the monteCarloEstimate which is derived from the estimate class. This estimate can be plotted through the plotEstimate function which displays the realizations of the product in a histogram:

> plotEstimate(product.mc)
>

If the product of interest implies more than two variables, there can be entered in the multiplyEstimate function. For instance, if we create a third estimate, the function can estimate the product of these three estimates as follows:

> est3 <- createEstimate(mean = 3, variance = 3)
> product.goodman <- multiplyEstimates(est1,est2,est3)
> summary(product.goodman)
[1] "Mean = 60 ; Variance = 3360 ; Distribution = unknown"
> 

Theoretical developments and some simulation studies have shown that Goodman's and the rescaled Monte Carlo variance estimators are unbiased whereas the others tend to overestimate the true variance.

A model of merchantable volume for 26 tree species

The model is implemented through a single function called getUnderbarkMerchantableVolumeDm3. The function requires three arguments: a species code, tree DBH (cm) and height (m). For instance, the species code "SAB" stands for balsam fir and the following line

getUnderbarkMerchantableVolumeDm3("SAB", 18, 15)

will return the underbark merchantable volume in dm3 of a balsam fir tree with 18 cm in DBH and 15 m in height.

The species codes are the usual codes in the province of Québec. Those recognized by the model are shown in the following table.

Latin name	Common name	Species code
Abies balsamea	Balsam fir	SAB
Acer rubrum	Red maple	ERR
Acer saccharum	Sugar maple	ERS
Betula populifolia	Grey birch	BOG
Betula alleghaniensis	Yellow birch	BOJ
Betula papyrifera	White birch	BOP
Fagus grandifolia	American beech	HEG
Fraxinus americana	American ash	FRA
Fraxinus nigra	Black ash	FRN
Larix laricina	American larch	MEL
Ostrya virginiana	American hophornbeam	OSV
Picea glauca	White spruce	EPB
Picea mariana	Black spruce	EPN
Picea rubens	Red spruce	EPR
Pinus banksiana	Jack pine	PIG
Pinus resinosa	Red pine	PIR
Pinus strobus	White pine	PIB
Populus balsamifera	Balsam poplar	PEB
Populus grandidentata	Large-tooth aspen	PEG
Populus tremuloides	Trembling aspen	PET
Prunus serotina	Black cherry	CET
Quercus rubra	Red oak	CHR
Thuja occidentalis	Eastern white cedar	THO
Tilia americana	American basswood	TIL
Tsuga canadensis	Eastern hemlock	PRU
Ulmus americana	American elm	ORA

Version history

Date	Version released	Features and bugs fixed	Comments
2021-08-13	0.2.0		The volume model of Fortin et al. (2007) (pubs.cif-ifc.org) has been integrated into the package.
2020-07-30	0.1.6		The dissimilarity indices have migrated in a new package called betadiv.
2020-04-26	0.1.5		The BioSim client has migrated into a subproject.
2020-03-17	0.1.4	Feature #3 and upgrade to J4R 1.0.5	Implementation of climate models in BioSim client
2020-01-17	0.1.3	Feature #2 and upgrade to J4R 1.0.4
2019-11-04	0.1.2	Upgrade to J4R 1.0.2	Refactoring in the dissimilarity functions, adding the variance estimators for greenhouse gas emissions
2019-07-24	0.1.1	Upgrade to J4R 1.0.1
2019-04-16	0.1.0		First beta version