HeightModel
This
sub command object, which resides under SIMULATION.MODEL provides access to the
OSM height model interface. Variants have the option of 1) implementing their
own OSM Height Model or 2) inheriting a predefined base Height Model base type
within the OSM. Due to this flexibility, properties and methods defined in the
height model may vary widely from one OSM Variant to another. Some may not be
able to locally calibrate heights. Consult Variant documentation on the height
model used and whether there are properties in it that can be adjusted through
OSM commands.
Local height
calibration
The
abstract HeightModel
available to developers through the OSM API provides methods to locally
calibrate height predictions. If Variant developers choose to inherit methods
from this pre-defined abstract height model, the following height calibration
interface will be exposed through the HeightModel property Calibration.
The
calibration method compares any available height observations imported from
tree records against predicted heights by species. Ordinary least squares are
used to solve for a modifier that would best align predictions to observations
by species.
Any
trees with missing heights on import will have heights predicted. If during
prediction a species modifier is available, it will be multiplied against the
predicted value before assigning the height to the tree. Use of the modifier is
constrained by the species DBH range (+/- 5 cm) of height observations used to
develop the modifier; e.g., if all white spruce height observations come from
DBH between 15 and 20 cm, then OSM will
only allow the height modifier to be applied to 10 – 25 cm DBH white spruce
trees. This DBH constraint is especially important when all height trees are
small, but one or more large trees with missing heights also exist in the plot.
We would not want to apply the modifier developed from very small trees to very
large trees.
The
calibration only remains in effect while estimating tree heights before
the simulation begins. Ingrowth will not be affected by this calibration.
Properties
1)
Enabled True
| False
By default, local height
calibration is on if available. Set to FALSE if you want to turn off automatic
local calibration.
2)
MaxAbsoluteModifier
Double [Default = 0.5]
Maximum
absolute multiplier permitted. In other words, modifiers that alter predictions
by more than +/-50% should be ignored.
3)
MinObservations
Int32 [Default = 5; Must be
≥1]
Minimum
observations per species required before a species modifier will be calculated.
Reducing below 5 is not recommended unless you are averaging results for 10 or
more stands. Note that with few observations, individual stand projections may
become worse. It is important that heights come from a random tree sub-sample
to avoid correction bias.
4)
WeightMode
Text Code [Default = BA]
Controls tree record
weighting approach used when calculating height equation modifier value.
Modes:
·
BA Tree record weight = basal area per hectare of tree
record.
·
Stems Tree record weight = density per hectare of tree record.
Manually Setting
Weights on Individual Tree Height Samples
Height observation sample ‘weights’
can be specified for each tree record manually in the OSM_TreeList
table using the optional column ‘Weight’. User-defined weights can be assigned to
individual trees in OSM_TreeList in order to 1) properly
calculate stand-level Lorey’s height and 2) properly weight
self-calibration of the height model for a range of plot designs (fixed area
vs. angle gauge). If
any trees in the stand have a weight assigned > 0, then the user specified
weights in the TreeList.Weight column will override
the OSM generated weights set from WeightMode. When the
Weight column is used, any trees with a weight set to null or zero will be
ignored when counting height samples and will also be ignored when setting the
min-max DBH range over which the height modifier can operate. Setting weight to
null or zero is recommended in cases when the tree top is broken or dead,
unless the majority of tops are known to be broken or dead.
Guidance
on setting Weights for different plot designs
When localizing height predictions through OSM’s automatic self-calibration
routine, reducing height bias in large trees is usually more important than
small trees, given preference toward accurate volume estimates. Therefore, most
users weight an individual height observation by the basal-area per hectare that
each tree height observation represents (i.e., BAF = basal area factor). In the
case of a simple variable-radius plot, weight = 1 for all trees, as trees are
already selected with probability proportional to basal area. In the case of
fixed-radius plots, weight = 0.00007854*DBH2. If the sample is a
tally, then the number of trees tallied per DBH class should be multiplied by
the weight.
In the case of complex plot designs, such as the New Brunswick
Continuous Forest Inventory (NB-CLI), where tree heights are collected from multiple
sampling strategies nested together, calculation of tree weights becomes slightly
more involved. For example, the NB-CLI plot design contains the following
samples:
1) A fixed-area large tree plot (400 m2) used to record Species
and DBH ≥ 7cm.
2) A variable-radius (BAF = 2M) plot used to subsample heights ≥ 7 cm, including heights falling outside the boundary of the fixed 400 m2
large tree plot.
3) A fixed-radius small tree plot (50 m2) used to record Species
and DBH between 1 and < 7 cm, with height sampled on every tenth tree.
In the case of NB-CLI plots, weight for individual height observations ≥ 7 cm DBH would always be 2 (2M BAF), while weight (W) for individual
height observations < 7 cm DBH would vary by DBH:
, where BAFi is the basal area per hectare expansion factor for height tree i in the small tree fixed-area plot with consideration that only 1 in
every 10 trees is measured for height.
For example, weights are calculated for the following DBH values for the
NB-CLI plots as follows:
|
DBH |
Weight (BAF) |
|
1.00 |
0.16 |
|
3.00 |
1.41 |
|
6.00 |
5.65 |
|
6.99 |
7.67 |
|
7.00 |
2.00 |
|
10.00 |
2.00 |
|
25.00 |
2.00 |
|
100.00 |
2.00 |
In the OSM Acadian model (version 1.19.6.2) a Lorey’s
height (LHT) estimate for the entire plot is calculated in OSM to boost
accuracy of the Variant’s height predictions using only the observed height (H)
trees as:
If weight is not explicitly assigned in the OSM_TreeList,
then OSM will assume the plot is a fixed-radius plot and weight height samples
using Stems (tree factor) * DBH2 * 0.00007854. This default
weighting will yield incorrect LHT estimates for the CLI plots, and hence
incorrect height estimates for the Acadian Variant model. Knowing how your OSM
variant predicts tree size, growth, and survival is obviously important here.
Consult variant documentation for calibration details and calibration updates.
What if I have
heights collected outside the plot that could be useful for height calibration?
If you have a few height
samples representative of the stand or plot that fall outside the sample frame,
they can be included with the plot’s tree records in OSM_TreeList
by setting the records ‘Stems’ column to a very small number (e.g., 0.000001;
must be > 0) and manually specifying weights for both plot records and
external height sample records. In this way, the tree is considered in the plot
by OSM, but has such a small tree factor that it will not make a difference in
volume summaries or simulation results.
If you have a very large number of height tree samples from a local area that you want to use to calibrate the height model, then a better approach would be to calibrate local species-height modifiers using the OSM.Simulation.Model.Calibrator.
A Note about Tree Agitation
OSM takes into consideration the possibility that tree
records may have been agitated
when calculating number of user height observations in the tree list. If five
height observations are required for local calibration, and partitioning
increases observed height observations from say 3 to 9, local calibration will
still not occur as the actual number of tree height observations is less than five.
Also, since tree Weight property values are divided by partitions, weight
integrity is preserved across all tree records. Therefore there should be no
adverse effects of tree agitation on local height calibration.
Example
HEIGHTMODEL
CALIBRATION
Enabled TRUE
MaxAbsoluteModifier 0.5
MinObservations 5