OSM
Acadian Version History
See also change history for OSM
No Variant-specific changes to report.
See change history
for OSM for this major version upgrade to the OSM API source code.
· Added Ingrowth model ‘commandable’ properties to allow ingrowth probability and abundance to be modified using a multiplier. Two new commands are:
o
Simulation.Model.IngrowthModel.IngrowthProbabilityMultiplier [value >= 0]
o
Simulation.Model.IngrowthModel.IngrowthAbundanceMultiplier [value >= 0]
o Where a value of 1.5 would result in a 150% increase in probability or abundance of ingrowth, and 0.5 would reduce them by half.
· Acadian diameter growth model – new property/command: LockAnnualPredictions
o Command: OSM.Model.DBHI.Model.LockAnnualPredcitions TRUE
o Default = FALSE
o This is a new property restricted to the current Acadian default DBH increment model. If TRUE and years per cycle = 1, then tree DBH increments are only updated every 5th cycle (i.e., once every five years). Increments are always updated if harvesting occurs, or when DBH increment is not yet set.
o This property is only active when years per cycle = 1 (SIMULATION.YPC 1).
o When set to TRUE, this property reduces the difference between model outcomes when run with annual cycles compared to 5-year cycles. Using periodic 5-year cycles are preferred because most calibration data is measured in 5-year intervals, but annual cycles are sometimes desired for growing the inventory to a standard inventory projection start year, or for simulations that require treatments or outputs with annual resolution.
o For young stands, diameter and height growth will often be overpredicted during annual simulations if this property remains set to FALSE (default), and sometimes under-predicted for old stand conditions, compared to periodic (5-year) cycle simulations. It is therefore recommended to set this property to TRUE whenever you are running annual OSM cycles with the Acadian variant. Because height growth is currently connected to DBH growth in this variant, height growth will also be ‘corrected’. Background mortality and self-thinning will continue to change each year during annual simulations, so some difference between annual and periodic 5-year simulations will still be noticeable when this property is set to TRUE, but the difference should be substantially smaller than when set to FALSE (default).
·
Spruce Budworm
Impact Extension
o Provides means to run spruce budworm (SBW) impact and protection simulations with a few simple OSM commands. The impact estimates are derived from the SBW DSS (MacLean et al. 2001) with significant upgrades over the past 20 years.
· Managed stand calibration changes for current default growth and mortality models:
o PCT and Planted types have always had faster growth and lower mortality in OSM, compared to unmanaged stand types, and these higher rates are based on a lot of field data. As managed stands age and trees grow larger, managed growth rates are slowly diluted back to natural growth rates because our data for managed stand growth is limited beyond 35 to 40 years old.
o Age and DBH breakpoints governing this calibration transition are influential on the timing of stand decline and peak volumes estimated in OSM for managed stands. Data to support when this transition should occur in the model is limited and was probably set too late in PCT and for fir and jack-pine plantations.
o Before
August 2021:
§ Managed stand (PCT and Planted) tree growth & mortality rates were set to transition back to unmanaged growth rates at stand age 50.
§ Before age 50, managed growth and mortality rate modifiers are linearly decreased starting at DBH 25 cm down to unmanaged rates by DBH 50 cm; e.g., at 37.cm, growth rate = average of managed and unmanaged rate.
§ These rules do not vary by species, but they should, especially for species that have been observed to decline quickly in these conditions like balsam fir and jack pine.
o After
August 2021:
§ PCT stands now transition back to unmanaged growth in OSM at age 40
§ Before age 40, for Fir and JP, managed growth and mortality rate modifiers are linearly decreased starting at DBH 20 cm to unmanaged rates by DBH 30 cm
§ This change results in an earlier onset of fir and JP decline and reduced stand peak volume, compared to previous OSM-Acadian versions.
· Added general documentation on current height prediction model and height prediction procedures used by this Variant.
· Added documentation on the current self-thinning line equation.
· Lorey’s height (LHT) boosting is set to occur only if > 5 height samples are present; however, tree heights with weights of zero were being counted toward sample size. Now fixed.
· Restricted use of Lorey’s height (LHT) prediction boosting in tree height imputation models when LHT < 4 m or > 20 m, or when DBH < 5 cm & > 50 cm. LHT adjustment can result in very odd height predictions in most of these extreme cases, under some circumstances. These extremes are also at the upper and lower end of data used to fit the model.
· If tree agitation occurred height samples could be artificially increased, and this was influencing calculation of Lorey’s height during height imputation boosting. This issue is now fixed.
· Updated the Hennigar 2018 background mortality model (current default Acadian mortality model) to operate stochastically when trees/ha are < 0.5 (kills individual trees randomly according to annual mortality probability), rather than deterministically (proportion of tree record stems/hectare/year killed equals annual mortality probability).
o In past versions, deterministic mortality was always used for all tree records, unless Simulation.Model.IsRandom = TRUE, in which case individual tree stochastic mortality was always used.
o Deterministic mortality is desirable for repeatability of results, but deterministic mortality can result in accumulation of thousands of tree and snag records with extremely small trees/ha (<0.0001) over long simulation time horizons, which can bloat memory and slow model performance. As a hybrid approach, the mortality procedure was modified to operate stochastically when trees/ha are < 0.5 per tree record, and deterministically in all other situations. Repeated stochastic simulations ought to converge on deterministic results. As the majority of tree records will have > 0.5 trees/ha, this change will have very minor effects on user results; especially during the first 25-50 years of the simulation. Even over long-term (100-year simulations), volume/ha fluctuations resulting from this stochastic behaviour are likely to be small (< ± 1 m3/ha compared to deterministic runs).
o For these reasons (performance improvement, minimal effect on results), this hybrid approach will be used in this version of this mortality model, even when Simulation.Model.IsRandom = FALSE.
o When Simulation.Model.IsRandom = TRUE, mortality of individual trees in every tree record will be set stochastically.
o To ensure results are repeatable from run to run, for testing, benchmarking, and validation work. Use command Simulation.Model.SetRandomSeed #, where # is a fixed positive integer. Do this for each stand simulation to force the random number generator to provide exactly the same series of random numbers during simulation. For example, random seed value could be the plot number.
·
Dr.
Chris Hennigar (ERD) presented at the ECANUSA conference in October 2018 in
Fredericton on accuracy of the latest OSM-ACD calibration (version 1.18.5.2)
compared to the University of Maine FVS-ACD model (calibrated by Dr. Aaron
Weiskittel and Dr. John Kershaw from 2010-2015) in terms of volume accrual (MB,
RMSE) across all plots used to calibrate these models. Accuracy was assessed by
forest and management type, spatially, by jurisdiction (NS, NB, Maine, PEI),
and overall. Recent calibration in OSM-ACD has improved accuracy over FVS-ACD
in all Acadian jurisdictions for nearly all forest and management types;
however, some OSM-ACD residual spatial bias was evident, which seemed to be
visually correlated with site (over-prediction on some poor sites, under-prediction
on some rich sites). Improvements to site productivity mapping are needed
before recalibration of growth and survival models used by OSM-ACD should be
revisited because current model accuracy is acceptable (best in class) for
management planning purposes.
·
Dr.
Martin Béland (U de Moncton) also presented at the same ECANUSA conference on
using OSM-ACD to model hardwood crop-tree release vs. traditional
pre-commercial thinning and on simulated treatment results.
·
Northern Hardwood Research Institute (NHRI)
noted some over-prediction of mortality and under prediction of ingrowth
probability in tolerant hardwood unmanaged and partial cut stands. Diameter
increment predictions on the other hand were found to be not
statistically different than observed.
·
University de Moncton (U de M) have highlighted
the need for a more comprehensive look into sapling regeneration modeling
(ingrowth rates), as there validation work has provided some evidence the
current ingrowth model in OSM-ACD (version 1.18.5.2, based on the Li et al.
Acadian ingrowth models) is under-predicting softwood ingrowth in mature and
partial-cut hardwood-mixedwood stand conditions in northwestern NB. Currently,
U de M, NHRI, and ERD are working toward revising the ingrowth models used by
OSM-ACD; first in NB, and then for the entire Acadian region. This work is
expected to be complete sometime in 2020.
·
Removed corrections ‘tweaks’ to DBH growth
models (introduced in version 1.0.3.2 as a result of previous validation work)
for clearcut, PCT, and planted stands. It seems these adjustments are not
needed/justified anymore following revision of mortality model and further
validation comparative work over the summer. Removing these adjustments
slightly reduced DBH and volume accrual rates in stand simulations for these
regenerating stands.
·
Revised
mortality and height model described in version 1.0.3.5 following validation
and testing against plot observations in NB, NS, PEI and Maine. Current models
are described here.
·
Adjustments to self-thinning line and ingrowth
for certain management types/species. Model was self-thinning too early in red
pine and too much ingrowth was forecast in well managed PCT and planted stands.
·
Addition of ‘PoorSite’ (true/false)
attribute in OSM_StandList for
Acadian Variant inputs to reduce growth, ingrowth, survival, and stocking on
sites with severe growth restrictions. Survival and growth will be
substantially over-predicted for such sites if PoorSite is not set to TRUE.
Many poor site barren, wet, and high elevation sites in NS (Atlantic Coastal,
Western Barrens, Taiga, Highlands) are still over predicted by OSM even when
‘PoorSite’ is set to TRUE; however, these sites represent marginal forest land
with severe regeneration and stocking limitations.
·
Make
sure to include ‘Zone’ (state / province code) and ‘BGI’ (Biomass Growth Index) in your
OSM_StandList table. Default Zone = NB.
·
‘NbSite’
and ‘NbClimate’ attributes in OSM_StandList
input table no longer required
in any default OSM models. These attributes are only used if you switch growth
or mortality back to STAMAN models.
·
New
default mortality and height models for the entire Acadian region developed in
2017 by Dr. Chris R. Hennigar at the NB Forest Stewardship and Planning Branch
have been set in this version. These models are beta and have had limited
testing, but given their improvement over previous models, especially in NS and
Maine, it was decided to release them for broader testing now. The height model
is fairly well tested and in use, but the mortality model is still being
tweaked a bit and will undergo more comprehensive testing, so expect updates in
2018.
·
The
mortality model includes modifiers for Acadian Zone and managed stands, while
the height model includes BGI and Zone modifiers. The height model also uses a
prediction boosting equation if Lorey’s height of the plot or stand can be
internally estimated from sample height trees in the plot.
·
Make
sure to include ‘Zone’ and ‘BGI’ in your StandList table. Default Zone = NB.
·
New snag modelling capabilities; see
also: OSM Snag Model.
·
Minor tweaks to Acadian NB calibration to
correct bias in regenerating stand conditions.
Acadian Variant
· 14 new species and new species groups.
· OSM_StandList table required column changes, additions:
o ‘NbManagement’ column changed to ‘Management’
o ‘District’ column changed to ‘Zone’.
o BGI column added. Most default tree equations use BGI as a measure of site productivity in this version.
· New calibration for many tree models. Main change was the introduction of a new non-linear DBH growth regression model(s) as a function of Zone, BGI, Species, DBH, basal area, QMD, and basal area of larger trees. Each species model was fit independently and not all variables were used depending on available data across NS, NB, PEI, and Maine. Species group (genus) models were also developed. Each species model was fit independently and not all variables were used in each model depending on available data. Fit by C.R. Hennigar, NB DNR 2016.
·
Updated
Variant to take advantage of the new OSM tree grade object model. See extensive
documentation in OSM on building custom tree grade models
(predictions/transitions) for OSM.
·
Fixed
some bugs with NHRI tree grade (form) model.
o
Replaced
this command “SIMULATION.MODEL.FormPredictionOn”, introduced in previous
version for testing, with “SIMULATION.MODEL.GradeModels.IsEnabled True”.
·
Tree
grade related commands are now less subject to change; however, calibration of
the tree grade model is likely to change significantly over 2015 as NHRI
finalizes their models. For this reason, there is no Variant documentation on
the new tree grade model; however, there are good examples in the Demo files
shipped with this build.
·
See
OSM Version changes for tree grade and documentation links.
·
Added
‘Beta’ tree Form prediction and transition models for tolerant hardwood
according to the Northern Hardwood Research Institute.
·
To
turn form prediction on use the following command:
o
SIMULATION.MODEL.FormPredictionOn
True
·
Form
commands and form model are subject to major change as they are currently under
development. No documentation is available, and this functionality is only
exposed for testing purposes.
·
Added
site tree calculations based on Carmean et al. (1989) for most Acadian species.
These routines do not influence Variant predictions and can only be used
through direct application linking with OSM.Acadian.dll.
·
Minor
code updates to make this version compatible with new OSM calibration features.
·
No
changes.
·
None
· Added ability to turn on STAMAN calibration using the following command:
o SIMULATION.MODEL.Switch STAMAN
o Use of STAMAN calibration is recommended at this time.
o This model release has known issues with the default Acadian diameter growth model, which will hopefully be resolved by the time of the next release.
· Local height calibration is ignored if the calibrated modifier results in prediction changes by more than >50%.
· Switched maximum-size density prediction from C. Woodall (2005; FEM) to predictions developed by C. Hennigar using a similar predictive approach as Woodall, but using New Brunswick 13,000+ temporary stand samples with DBH resolution down to 1cm.
o A-line set to 80%
o Included a maximum basal area constraint, fit using the same stand specific gravity approach down to 1cm DBH.
o Included two max SDI constraints, one fit to all trees down to 1cm, and the other down to only 9cm DBH.
o All three of these maximum density constraints need to be satisfied at the end of each cycle.
o Validation of these new max size-density relationships is currently underway with all Acadian PSPs.
· Changed species codes.
o PB changed to WB
o TA changed to TL
· Acadian ‘ClimateIndex’ default changed from 25 to 13.5 (approximate average in Maine).
· STAMAN ‘NbSite’ default changed from Good (2) to Moderate (3); more representative of entire Acadian forest.
· STAMAN ‘NbClimate’ default changed from Northern (2) to Southern (3); more representative of entire Acadian forest.
· Default mortality model changed from’JAK2012’ to ‘JAK2013’.
· Local height calibration is automatically applied if greater than 5 heights are present in the input data.
· Added new tree species background mortality models (JAK2013).
· Added support for generic USDA plant codes to ensure rare or unknown species in inventory data will not cause the model to fail.
o 2TB - Hardwood
o 2TE - Evergreen (conifer, includes Tamarack)
o 2TREE – Unknown tree.
· Added local height calibration modelling capabilities.