Overview

Welcome to the CBGC ALCES homepage. CBGC ALCES is a population dynamics model for the Bathurst, Bluenose-East, Bluenose-West, Cape Bathurst, and Tuktoyatuk Peninsula central barren-ground caribou herds in the Northwest Territories and Nunavut. The model is a customized version of the ALCES landscape and population dynamics (PopDyn) models. ALCES PopDyn simulates wildlife population dynamics in response to habitat, fecundity, and mortality. It is a cell-based spatial model, with each cell defined as a Leslie-matrix population model with a carrying capacity dictated by the cell’s habitat. The model is linked to ALCES landscape simulations so that habitat, fecundity, and mortality risk respond to landscape and climate dynamics. To represent unique habitat and mortality risks associated with seasons, CBGC ALCES applies 5 seasonal models that run in succession each simulation year.

Population dynamics simulation outputs can be viewed using publicly accessible scenario results viewer.

Scenario Results Viewer - used to view maps and charts of scenario outcomes, including population dynamics and habitat. Two version of the results viewer are available: 1) an analyst version, which is available to users with analyst privileges; and 2) a public version that can be used by anyone to view outcomes of publicly shared simulations.

Model Description

CBGC ALCES integrates a range of data and model logic, including a spatial representation of current landscape composition, a 40-year forecast of landscape and climate dynamics, a default future development scenario, and a seasonal barren-ground population dynamics model. These components are now described.

Current Land Cover

A coverage of land cover and climate data for the full extent of NWT as well as the western portion of Nunavut occurring above (to the north of) NWT was created for the model to be used for habitat modelling. This dataset captures the extent of the five caribou herds studied in this project (Bluenose East, Bluenose West, Cape Bathurst, Tuktoyatuk Peninsula, Bathurst). Geospatial data sets were prepared in ALCES Online as described below.

A landscape composition data set was developed to provide proportional coverage of each cell in the study area by each land cover and human footprint type. Table 1 provides a prioritized list of the cover types and a summary of the source data sets. The unity data set was prepared by intersecting the datasets with the 100 m x 100 m (1 ha) cell grid, and assigning priorities to source data sets during the intersection so that unity (i.e., no more or less than 100% coverage) is respected. The source data sets were selected based on input from experts within the Government of the Northwest Territories. Land Cover of Canada was selected as the primary natural land cover datasets because it was the most up-to-date inventory with complete coverage. The Human Disturbance dataset was the primary development footprint dataset. A 2020 version of the human disturbance dataset provided to the project included an expanded extent to include the Nunavut portion of the study area. The Human Disturbance dataset was augmented by CanVec datasets to achieve more comprehensive representation of footprint.
Digital elevation model (DEM) characteristics – aspect, slope, mean elevation, minimum elevation, and maximum elevation – were assigned for each 1 ha spatial unit within the study area (100 m x 100 m cell).
Forest age was assigned to forested spatial units based on estimated time since disturbance, which was derived from information on time since the most recent fire or timber harvest event. Forest age was estimated from three data sources: the NWT fire history dataset (1965 to 2020), the National Burn Area Composite for fires in Nunavut between 1986 and 2019, Canada Landsat Disturbance 2017 for timber harvest between 1984 and 2015. Where harvest and fire disturbance did not occur, forest age was established based on a national stand age data layer (circa 2011 and adjusted to 2019). Where harvest and fire disturbance overlapped, the most recent disturbance type and age was applied.

Landscape and Climate Forecast

CBGC ALCES simulates habitat and population response to a RCP 8.5 climate scenario that incorporates projected changes in climate variables as well as landscape dynamics caused by ecoregional shifts and fire.

Climate data were downscaled from CanESM2 (https://climate-scenarios.canada.ca/?page=pred-canesm2) using DEM, baseline and anomaly grids based on methods presented in Wang et al. (2016) . Climate data include monthly and annual temperature (min, max, mean), precipitation, precipitation as snow, shortwave radiation, and evaporation, downscaled to 1 km2.

A landscape simulation was completed using ALCES Mapper to represent potential shifts in land cover and forest disturbance in response to the RCP 8.5 climate scenario. The simulation used an annual time step four decades into the future. Cell size for the simulation was 1 km2.

Simulated expansion and contraction of taiga and tundra cover types was informed by climate-projected distributional shifts for North American ecoregions under RCP8.5 (Stralberg 2018). Areas where tundra ecoregions are projected to transition to taiga ecoregions were identified as being eligible for shrubification (e.g., Mod and Luoto 2016), simulated here as conversion of grassland to shrub land cover. Immediate conversion of grassland to shrub is unlikely. Instead, the simulation converted 1.0% of eligible land cover per year over the next 40 years. Because only a portion of the study area is eligible for shrubification, the area affected by these scenarios is substantially lower than 1.0% of the total study area per year. 1% of eligible grassland equals about 0.11% (891 km2) of the study area.

Spatial distribution of shrubification was random with the following constraints:

Shrubification was limited to within areas that are projected to shift from a tundra ecoregion to a taiga ecoregion.
The likelihood of shrubification was inversely proportional to distance (km) to forest and shrub land cover. In other words, likelihood of conversion increased in closer proximity to forest and shrub.
The likelihood of shrubification was inversely proportional to the tundra refugia value (Stralberg 2019). In other words likelihood of conversion decreased with increasing tundra refugia value. Tundra refugia is a 0 to 1 index, with higher values indicating greater climate persistence and therefore tundra ecoregion resilience to change. The highest tundra refugia value occurring within the study area is 0.5.
Shrubification occurred within cells at levels of 0.1, 0.3, 0.5, 0.7, and 0.9 km2 based on the current distribution of cell coverage by shrubland in the study area.

Expansion of tundra was assumed to be catalyzed by fires occurring in areas where taiga ecoregions are projected to transition to tundra ecoregions. In the simulations, fire within the area of tundra ecoregion expansion caused coniferous and mixed forest to convert to deciduous forest, and caused deciduous forest and shrubland to convert to grassland. As such, conversion of coniferous forest to grassland required two fires during a simulation: the first burn to convert coniferous forest to deciduous forest and the second burn to convert deciduous forest to grassland. Locations with a tundra refugia value (Stralberg 2019) greater than 0.5 were excluded from the conversions.

Fire was simulated by applying projected changes in fire area by homogeneous fire regime zone (Boulanger et al. 2014). Baseline annual fire area for each homogeneous fire regime zone (HFRZ) was calculated as the average annual area of forest and shrub burned from historical (1965-1990) fire data for the study area . Simulated future fire area was obtained by multiplying each HFRZ’s baseline fire area by the area-weighted average projected annual area burned ratio across HFRZs under climate scenario A2 for time periods 2011-2040 and 2041-2070. The average burn ratio for the 2011-2040 period was 2.1 and for the 2041-2070 period was 4.2.

In addition to differences in fire rate by HFRZ, local scale (1 km2) differences in fire rate were incorporated in simulations using fire selection ratios that differ by forest type and age class (Bernier et al. 2016). Cover types other than forest and shrub were assumed to be nonflammable. Fire location during simulations was random but guided by a relative likelihood layer that reflected the fire selection ratios and HFRZ burn rates. The fire size class distribution used in the simulations was based on burned forest and shrub patch size distribution occurring in the study area between 2010 and 2020.

Although annual burn area tends to vary substantially from year to year, simulations excluded interannual variation so that random differences in burn area from year to year did not obscure differences between scenarios. The effect of this simplification on caribou modelling outcomes is likely small given that forest age is incorporated in caribou habitat models at a coarse level of temporal detail (i.e., forest younger than 50 years).

Development Forecast

The landscape forecast applied a number of development footprints related to potential future projects. The projects are those identified for the Increasing Development scenario that was prepared in consultation with a working group made up of Government of NWT staff, as described in the Land-use Scenarios Workbook (link to document to be provided soon). Development projects in the forecast include:

Multiple mines (Courageous Lake, Indin Lake, Kennady North, High Lake, Izok Lake, NICO, Sabina George, Sabina Goose, Ulu, Yellowknife Gold, and Nechalacho)
Izok to Grays Bay road
Slave province road and transmission corridor
Sabina to Bathurst Inlet road
Transmission lines to NICO and Nechalacho
Inuvialuit Energy Security Project

The user can select from these projects when defining a scenario using the Scenario Analysis Tool. As well, additional development projects can be added using the Footprint Catalog Tool.

Population Dynamics Model

Overview

CBGC ALCES simulates barren-ground caribou population dynamics in response to habitat, fecundity, and mortality. It is a cell-based spatial model, with each cell defined as a Leslie-matrix population model with a carrying capacity dictated by the cell’s habitat. The model is linked to ALCES landscape simulations so that habitat, fecundity, and mortality risk respond to landscape and climate dynamics. CBGC ALCES implements the ALCES PopDyn model. The PopDyn code is open-source and supporting documentation is available. The Popdyn Source code can be viewed at https://bitbucket.org/alceslanduse/popdyn/src/master/ and documentation can be found at https://popdyn.com.

Seasonality is a key characteristic of the annual life cycle for barren-ground caribou. To represent unique habitat and mortality risks associated with seasons, the caribou population dynamics model is actually 5 seasonal models that run in succession each year. The seasonal models are spring migration, calving, summer, fall, and winter. Each year, output from the spring migration model provides input to the calving model, output from the calving model provides input to the summer model, and so on. Output from the winter model at the end of the year then provides input to the spring migration model to start the next year. A simulation starts with the spring migration model because population estimates that are used to initialize the population are typically based on pre-calving population estimates.

The computational steps that are used by ALCES PopDyn are summarized below to provide an overview of how the inputs are applied to simulate barren-ground caribou population dynamics.

The initial population dictates the starting point of the simulation in terms of the spatial distribution of animals within each sex and age class.
Habitat layers for each season are prepared using landscape covariates, and each cell’s carrying capacity by season is calculated for subsequent use when applying density dependence relationships for fecundity and mortality.
Fecundity rates for each cell are calculated, adjusting for density dependence if necessary. Fecundity rates are applied to the number of females within relevant age classes to calculate the number of births per cell. Each cell’s population is adjusted accordingly.
Mortality rates for each cell are calculated for the calving season, adjusting for density dependence if necessary. Mortality rates are applied to the number of animals by sex and age class to calculate the number of deaths per cell. Each cell’s population is adjusted accordingly.
The population remaining at the end of the calving season migrates to the summer range and is distributed across cells based on habitat availability.
Mortality rates for each cell are calculated for the summer season, adjusting for density dependence if necessary, and applied to the number of animals by sex and age class to calculate the number of deaths per cell. Each cell’s population is adjusted accordingly.
The population remaining at the end of the summer season migrates to the fall range and is distributed across cells based on habitat availability.
Mortality rates for each cell are calculated for the fall season, adjusting for density dependence if necessary, and applied to the number of animals by sex and age class to calculate the number of deaths per cell. Each cell’s population is adjusted accordingly.
The population remaining at the end of the fall season migrates to the winter range and is distributed across cells based on habitat availability.
Mortality rates for each cell are calculated for the winter season, adjusting for density dependence if necessary, and applied to the number of animals by sex and age class to calculate the number of deaths per cell. Each cell’s population is adjusted accordingly.
The population remaining at the end of the winter season migrates to the calving range and is distributed across cells based on habitat availability. This provides the starting point for the next simulation year.
Steps 2 through 11 are repeated for each year of the simulation.

Population Model Assumptions

ALCES PopDyn requires a number of inputs to simulation caribou population dynamics including:

seasonal range boundaries
seasonal habitat models
initial population size and composition
maximum population density
fecundity rates and relationship with climate
mortality rates and relationship with climate
density dependent mortality
harvest

Assumptions for these inputs are described in the report: Bathurst Caribou Population Dynamics Models Inputs and Example Outputs