Skip to content

Growth, Removals, and Mortality

Growth, removals, and mortality (GRM) refers to a set of estimation procedures used by the FIA to quantify forest change. Broadly, trees are assigned to change classes, which we refer to as transitions. Then, estimates are constructed based on the transition and desired output of the user. Typical outputs include estimates of mortality volume, removals, and others. Transitions are dependent on a pre-specified tree_basis and land_basis. For example, ingrowth with respect to growing stock is more stringent than ingrowth with respect to all live trees, because growing stock requires a larger minimum diameter and more strict defect constraints. Thus, users must specify the tree and land bases, which will facilitate fiaplyr in correctly applying the appropriate transition rules, which is done when the handler is initialized.

Generally, users produce estimates using a series of facilitating functions called targets, which are called within aggregate and estimate contexts. The target names should be familiar, and are all prefixed with grm_. Common examples include:

  • grm_mortality
  • grm_removal
  • grm_ingrowth
  • grm_net_change

and others. Users supply the desired variable to the target, whether or not to annualize the estimate, and any other relevant arguments. Aside from the targets, analysis of GRM data is largely similar to that of status analysis. We animate GRM analysis with an example from Vermont.

To begin with GRM analyses, specify a handler as always. However, this handler must use an evaluation ending in 3 which indicates a GRM-specific evaluation. We also pass the GRMAnalysis specification.

library(fiaplyr)
library(DBI)
library(duckdb)
library(dplyr)
con <- dbConnect(duckdb(), fiadb_vt_mini_path())
handler <- eval_handler(con, 501103, spec = grm_analysis())
handler
EvalHandler
----------
EVALID: 501103
Description: VERMONT 2011: 2003-2007 to 2008-2011: AREA CHANGE, GROWTH,
REMOVALS, MORTALITY
Plots: 770
GRM Spec
Tree basis: all_live
Land basis: forest_land
Rules: 8

Here, we see the handler is initialized with a GRM evaluation, and the summary statement gives particulars. The specification defaults to a tree basis of all_live and a land basis of forest_land, but these can be changed by setting the tree_basis and land_basis arguments in the grm_analysis() function. Options for the tree basis include all_live, growing_stock, and sawtimber and options for the land basis include forest_land and timberland.

The core table involved with GRM analyses is the tree_history table, which contains the tree-level histories for each tree in the remeasurement period, and relies on the GRM paradigm of beginning, midpoint, and end values. Users can view the lazy query for the tree_history table with

handler@tables$tree_history |>
select(DIA_begin, DIA_midpt, DIA) |>
head()
# A query: ?? x 3
# Database: DuckDB 1.5.5
DIA_begin DIA_midpt DIA
<dbl> <dbl> <dbl>
1 11.7 12.6 13.6
2 8.6 9.5 10.4
3 9.3 10.1 10.9
4 3.9 4.25 4.6
5 3.2 3.4 3.6
6 2.6 3.3 4

Notice that the beginning and midpoint values have suffixes, while the end does not. This complies with typical FIA conventions.

In the GRM paradigm, trees undergo transitions, primarily based on their life history. For example, a tree that is measured and alive at the beginning of the period may survive to the endpoint, wherein it is classified as a survivor tree. Tree transitions are applied when the handler is initialized, using transition rules that comply with the selected tree and land bases specified in the grm_analysis() function. Users should interpret transitions as a column in the tree_history table that contains the transition type, for which we use the conventional names from the FIA database, but in lower case to differentiate from the original values. While it is important to be aware of the transitions, they are typically abstracted away when using GRM targets. Still, some inspection is useful, such as a count of trees by transition

handler@tables$tree_history |>
group_by(transition) |>
summarise(n = n()) |>
arrange(transition) |>
print(n = 10)
# A query: ?? x 2
# Database: DuckDB 1.5.5
# Ordered by: transition
transition n
<chr> <dbl>
1 cut1 413
2 cut2 3
3 diversion1 96
4 ingrowth 1456
5 mortality1 959
6 not_used 7082
7 reversion1 16
8 reversion2 182
9 survivor 15369
10 unknown 80

Estimation of Change Components with Targets

Section titled “Estimation of Change Components with Targets”

Change components refer to the various outputs that users may be interested in, such as mortality volume, removals, and others. For most use cases, it is sufficient to use the GRM targets to specify the desired change components, which abstract away complex transition rules and other accounting methods.

Beginning with a simple example, let’s assume we want to estimate the total mortality volume. The target grm_mortality is used, which generates mortality for the specified variable. The annualize argument is used to produce annual mortality rates, as is done in most standard FIA estimates.

handler |>
estimate(
tree_history(mort = grm_mortality(VOLCFSND, annualize = TRUE)),
output = "total"
)
# A query: ?? x 3
# Database: DuckDB 1.5.5
var estimate se
<chr> <dbl> <dbl>
1 mort 101032246. 6741699.

Likewise, ingrowth can also be estimated with the grm_ingrowth target

handler |>
estimate(
tree_history(ingrowth = grm_ingrowth(VOLCFSND, annualize = TRUE)),
output = "total"
)
# A query: ?? x 3
# Database: DuckDB 1.5.5
var estimate se
<chr> <dbl> <dbl>
1 ingrowth 31024427. 1391141.

Higher order change components are grm_net_change, grm_net_growth and grm_gross_growth, all of which use convenient targets.

handler |>
estimate(
tree_history(net_change = grm_net_change(VOLCFSND, annualize = TRUE)),
output = "total"
)
# A query: ?? x 3
# Database: DuckDB 1.5.5
var estimate se
<chr> <dbl> <dbl>
1 net_change 160580318. 20799837.

Change estimates also need to be expressed as ratios in many settings, especially when normalizing to forested area. Estimates of this kind proceed along the same lines as ratio estimates. The only modification, of course, is the specification of the GRM estimates themselves.

handler |>
subset(cond(COND_STATUS_CD == 1)) |>
estimate(
ratio(
tree_history(mort = grm_mortality(VOLCFSND, annualize = TRUE)),
cond()
)
)
# A query: ?? x 4
# Database: DuckDB 1.5.5
var_n var_d estimate se
<chr> <chr> <dbl> <dbl>
1 mort prop 22.0 1.46