Appendix F — Boreal Landscape Guide: Re-estimating SRNVs

2019 Boreal Landscape Guide: Re-estimating SRNVs (Draft Report)

Phil Elkie

September 2019

F.1 Summary

This report provides a summary of potential issues with simulated ranges of natural variation (SRNVs) used to support the Forest Management Guide for Boreal Landscapes including a discussion of new science and information that may be used if SRNVs are re-estimated, a discussion of potential tools and an approximation of time to re-estimate SRNVs in a Landscape Guide Region.

F.2 Issues with Simulated Ranges of Natural Variation

F.2.1 Landscape Guide Region 3E (lowland/upland confusion)

In the initial implementation of the Boreal Landscape Guide, specifically the Hearst Forest, issues were identified with the SRNVs. The inventory used in the first round of simulations used to generate the SRNVs had differences that resulted in several prescriptive indicators being difficult to achieve. 

When we first ran the simulations, we conducted sensitivity analyses to determine how long to run the simulations with the objective to estimate the condition of a forest without recent human influence. In other words, remove the human footprint. In the sensitivity analyses we observed the changes in variance of indicators over simulation time – both within and among simulation runs. The conclusion for the Boreal forest was to run the simulations for 100 years before beginning to make measurements used to estimate SRNVs (i.e. box and whisker). The idea was that regardless of the vintage of the initiating inventory used to estimate the SRNVs, the resulting box and whisker SRNV would be the same providing the inventories used to estimate the SRNVs were accurate.

In the case of much of the Boreal East – 3E, and in some cases 3W, lowlands were classified as uplands and uplands as lowlands. The succession rules we used ensured that these functional groupings remained the same - a lowland community could not succeed into an upland community and vise versa.

Misclassification/confusion between uplands and lowlands in the initial inventories versus the current more accurate inventories, has resulted, in several cases, in SRNVs that are problematic and consequently difficult to achieve.

F.2.2 Landscape Guide Region 3S/4S (borders and parks data missing)

Landscape Guide Region 3S/4S is on the western side of the NW region in an area that has a relatively short fire cycle. Fires occur more often and can be larger than in other parts of Ontario. Fires in this area move in SW to NE - W to E directions often igniting in Manitoba and moving across the provincial border.

In the initial simulations the movement of fires from Manitoba across the Ontario border was not explicitly dealt with resulting in SRNV results biased to a partial or incomplete model. We do not fully understand the implications of this problem and there are differing opinions of how this has biased the SRNV.

Data was missing or not available including forest cover and lightning strikes from Manitoba.

To exacerbate the border problem, Woodland Caribou Provincial Park (WCPP) is in the Region and is adjacent to several management units. At the time WCPP did not have a reasonable inventory to use in the simulations. Consequently, the lack of simulated fire behaviour within the park and its effect on adjacent management units may have biased the SRNVS in the Red Lake, White Feather, Kenora management units.

The park inventory problem was not unique to WCPP. A very coarse classification biased to conifer mixedwoods was applied in Landscape Guide Region 3W - Wabikimi Provincial Park. This may have skewed results in adjacent areas particularly when the indicator apportioning is based on fire behaviour and post disturbance transitions.

Quetico Provincial Park in the 4W simulations presented a similar border problem, (e.g. southern data missing) although because much of the border is water, we expect that the problem is not as significant. Also, although a Quetico inventory was used in 4W, as with all parks (i.e. Quetico, WCPP and Wabikimi), more current inventories are now available and can be used to strengthen the SRNV estimates.

F.2.3 Landscape Guide Region 4W (red and white pine SRNVs missing)

BFOLDS simulations, inputs and calibration in 4W was like the rest of the NW Region. Work that I have done in analyzing historic survey notes suggest that the proportion of the 4W Landscape Region at one time (e.g. estimate of natural) had a significant component, at the stand level, of red and/or white pine (+/- 19% of the entire landscape). The BFOLDS simulations did not account for PR and PW appropriately. Although they were included in initial model runs, the results were not considered trustworthy and as a result were blended into the coarser conifer results.

This has become a problem for planning teams in determining what is an acceptable amount of PR and PW to use as a target and inversely how much of the spruce/jack pine conifer targets currently being used would have been PR and PW. ROD has reported that this is a contentious and time-consuming issue for planning teams.

F.2.4 Landscape Guide Region 3W (all ages conifer indicator suspect)

The all ages conifer indicator (i.e. conifer forest units (non-hardwood or balsam fir mixed conifer)) is used in the BLG to ensure future, not just current, conifer purity is accounted for. In most forest management units, the magnitude of difference between the current amount of all ages conifer and the SRNV is considerable. Practitioners have challenged the differences and achieving the lower quartile is often not possible over 100 years. In re-estimating the SRNVs this issue could be checked and focused on to ensure that the estimates are valid and defendable. I would expect the evidence-based results reports (see reference below) and potential new fire information would help to support a revised result.

Although this issue exists in the entire NW region it is most pronounced in 3W.

In summary the following are current issues/needs (ordered in priority/magnitude) for re-estimating Boreal SRNVS;

  • Robust SRNV estimates of red and white pine (esp. 4W and 3S/4S),

  • New starting inventories with confirmation of upland / lowland ecosites

  • Complete inventories for each Landscape Guide Region that include new park inventories (esp. 3W 3S/4S and 4W)

  • All ages conifer SRNV confirmation (esp. 3W)

F.3 New Science and Information and Science Projects

F.3.1 Natural Succession

Ken Lennon and colleagues recently published a report (see citation below). The report is a summary of successive remeasurements of growth and yield plots with a quantitative summary and description of natural succession pathways for northeastern forest units used in forest management planning.

Much of the first run BFOLDS succession was opinion based and used information from current (at the time) forest management plans as a guide.

From my discussions with NW Regional staff, one of the key findings is that several of the more dominant forest units don’t succeed at the ages that were used in original BFOLDS simulations. Specifically, the time they would remain in the mature and old landscape classes is longer than first thought.

Presumably rerunning simulations may have unintended consequences in forest management planning with revised simulations resulting in higher levels of more “mature and old” and “old growth” forest.

Report citation: Lennon. K., J. Parton, K. Major and G.J. Kayahara. 2016. Evidence-based natural succession pathways for forest management planning in northeastern Ontario. Ontario Ministry of Natural Resources and Forestry. Science and Research Branch, Peterborough, ON. Science and Research Technical Report TR-15. 47 p. + appendices.

Currently a similar draft report for the Northwest region is being reviewed and is expected to be published and available early in 2020. This work is being led by Chris Stratton, Forest Productivity Specialist – Science and Research Branch – Thunder Bay.

This new science and information is available and should be used in new revised simulations. However, each simulation model/tool we’ve used to date to generate SRNVs (e.g. BFOLDS, TELSA/VDDT and ST-Sim) grows and succeeds the forest differently. Each model has its own unique set(s) of algorithms.  Consequently, depending on the model chosen to re-estimate the SRNVs, the results of the new reports would have to be translated to be compatible with the chosen model. 

F.3.2 Post Disturbance Transitions

Like natural succession, much of the information we used in the first round of SRNVs was expert opinion. In discussions with Gordon Kayahara and others there is new evidence-based information from both Ontario and Quebec that would be useful if rebuilding models. Gordon has indicated that this would require someone to do a literature review etc.

F.3.3 Science/Information Projects

  1. Translate new natural succession from Ken Lennon’s reports into selected simulation model algorithms.

  2. Review new literature focusing on forest composition - post disturbance (e.g. fire and insect infestation) transitions and provide report focusing on quantifying results.

  3. Translate post disturbance findings into selected model algorithms.

  4. If simulation model other BFOLDS is chosen, review literature and report on historic fire (crown and surface) in 4W and 3S/4S focusing on amount burned year-to-year and disturbance sizes. This would be used to calibrate simulation model.

F.4 Simulation and Assessment Tools

In the past we used several simulation tools to estimate SRNVS. BFOLDS was used in the entire Boreal Forest Region (e.g. Landscape Guide Regions 3S/4S, 4W, 3W and 3E). VDDT/TELSA was used in initial SRNV estimates in the GLSL and recently re-estimated using ST-Sim.

F.4.1 BFOLDS

BFOLDS is a process model where simulations use geography specific data (e.g. soils layers and lightning strikes) including daily weather, succession and post disturbance pathways combined with scientific knowledge of Canadian Forest Fire Behavior Prediction System and Fire Weather Index system.

In recent discussions with Marc Ouellette a spruce budworm module/add-on using LANDIS-II has been completed and is ready to be used. The lack of spruce budworm attacks/outbreaks has been raised as an issue in the Boreal simulations (GLSL simulations include budworm). Acknowledging spruce budworm in BFOLDS would support the objective of more robust estimations of the SRNV.

Ajith Perera (BFOLDS lead) has recently retired from the MNRF and my understanding is that Marc Ouellette is working on other projects although he has indicated that he is able and willing to rerun SRNVs. BFOLDS support will be important if a full scale rerun of Boreal SRNVS. In the initial SRNVs Marc had several computers running BFOLDS with results and data being exchanged between himself and others over several years.

F.4.2 ST-Sim

ST-Sim is an adaptation of the VDDT/TELSA tools we originally used in the GLSL.

ST-Sim is different from BFOLDS in that historic landscape attributes (e.g. fire size, fire amount etc.) are used as inputs. ST-Sim uses these as targets when simulating landscape dynamics. We have been successful with ST-Sim due to its ability to include generic inputs and easily modify. ST-Sim is available with support, at a cost, from Apex-RMS in Ottawa (contact name Colin Daniel).

Currently the ST-Sim models used in the GLSL could be modified and used in 4W and 3S/4S where we were unable to model the big pines with BFOLDS.

F.4.3 LANDIS-II

Landis-II is a widely used generic landscape simulation model which may be used as a stand-alone model in Ontario Boreal regions. Landis-II has many users world wide including users groups and includes modules and extensions that may be of use in climate change scenarios and when simulating other forest stressors. I am not aware if LANDIS II (generally stand level dynamics) is able to be parameterised on large landscapes. This would have to be tested.

F.4.4 ALCES

Alces is a tool that Rob Rempel has used here in Ontario for investigating cumulative effects among varying predicted future scenarios. ALCES may be used for SRNV estimating but with limited capabilities. In discussions with Rob Rempel, he has advised that ALCES is not a tool to re-estimate SRNVs

F.4.5 Landscape Scripting Language (LSL)

LSL is a landscape assessment tool developed by Rob Rempel and Rob Kushneriuk at CNFER. We have used and continue to use LSL in estimating all pattern indicators in the Landscape Guides – both in the GLSL and Boreal regions.

LSL is also used in Ontario’s Landscape Tool (OLT). OLT uses LSL as a spatial engine to assess surfaces and create maps and charts.

LSL is a relatively stable application. Rob Rempel is retiring in December 2019 and Rob Kushneriuk retired several years ago. We have not experienced issues (e.g. bugs) with LSL since Rob Kushneriuk left. However, we continue to depend on LSL due to its ability to measure landscape indicators using sampling techniques dependant on hexagons layers. We (OLT and future simulation assessments) will be vulnerable with LSL support being unavailable.

LSL has been used by other scientists in the OPS but nobody is currently maintaining.

F.5 Re-estimating SRNVS

When we estimated the first round of SRNVs we assembled science and modelling teams in each region consisting of MNR staff from Regional science units and ROD with some industry consultation.

Considering that at the time many of the inputs were mined from existing forest management plans and much was opinion based - many people contributed to the development of the simulation models. In contrast, in re-running the simulations, much of the same information may be used, and revisions made if and where new science and information is available (e.g. Lennon et al. reports).

It is difficult to estimate the capacity and resources required to re-estimate the SRNVs as it will be dependent on the tool used and experience of people available.

Instead I’ve listed steps to re-estimate which may help if/when scoping out project.

F.5.1 Steps in re-estimating SRNVs

  1. Create year 0 simulation start surface

    a. Gather most current inventories

    b. Classify inventories into modelling currency (forest units)

    c. Join inventories at an appropriate scale (e.g. originally done at the landscape guide region with buffers)

  2. Develop succession and post disturbance rules

    a. Translate Lennon et al.’s work into succession rules (dependent on simulation model used)

    b. Search, review and report on new literature for post disturbance succession

    c. Translate new post disturbance rules (dependent on simulation model)

  3. Search and review fire and wind literature and translate into selected simulation model (only applicable if BFOLDS not used)

  4. Search and review insect infestation literature and translate into selected simulation model (only applicable if BFOLDS not used)

  5. Build simulation model – this will require experienced working knowledge of model being used. In the past this was done by model developers,

  6. Run simulations

    1. If BFOLDS is used – in the past this was done at OFRI – Marc Ouellette ran the simulations on several high-level computers.

    2. If ST-Sim is used – in 2018 we (Apex-RMS) uploaded the ST-Sim models to external Amazon servers where the models were run. Other models may be uploaded to external servers as well.

  7. Process simulation results

    1. Use LSL to summarize the resulting simulation surfaces (n= 40…….?) including estimating the box and whisker values for prescriptive and evaluative indicators and all spatial indicators (texture and patch based). This will include using existing LSL scripts where possible but will require some coding/script writing (done in the past by Julie Elliott).

  8. Summarize simulation results and rebuild Science and Information Packages

    1. LSL will create raw results in various file formats. This step includes creating box and whisker diagrams and texture maps/charts – and creating new Science and Information Packages.

  9. Populate Ontario’s Landscape Tool with new estimates

    1. This is relatively straight forward and not time consuming but will require working knowledge of OLT from a developer’s perspective.

F.5.2 Estimated time to re-estimate SRNVS for a Landscape Guide Region

Note: Steps 2,3 and 4 can occur at same time. Total time to re-estimate 53 weeks.