Climate Change Informed Species Selection (CCISS) Tool

Selection/Filter pane

In the left pane select the BGC subzone/variant and then the site series of interest. The edatopic space of the selected site series is displayed in the graphic below for reference. By default the report will show all species that are predicted to be feasible in at least one model and time period. Choose the “Feasible Only” option to limit the list to species that meet the threshold for classification as feasible across the global climate model ensemble in any of the time periods.

Detailed report

This report shows modelled feasibility ratios for each species in the selected site series for each time period. The colour legend for feasibility ratings is on the left hand pane. The mapped biogeoclimatic unit represents the historical equilibrium climate approximated by the climatic conditions of the 1961-1990 period. The recent time period (2001-2020) has two bars: one for the observed climate (measured by weather stations), and one for the climates simulated by the ensemble of global climate models. These two bars aren’t necessarily expected to agree: The modelled climates sample a large range of possible recent conditions, of which the observed recent climate is only one.

The report then summarizes the baseline suitability ratings and forecasted feasibility ratings for each species in the following order:

1. The CFRG suitability rating: Suitability ratings taken directly from the Chief foresters reference guide

2. The CFRG P/A value: Preferred/Acceptable ratings taken directly from the Chief foresters reference guide

3. The Historical Environmental Suitability: Expert baseline environmental suitability rating for site series.

4. CCISS Establishment Feasibility: The feasibility rating based on the baseline (1961-1990), recent observed (2001-2020), and 2021-2040 future projected feasibilities. This indicates the likely level of constraints for successful establishment of the species in the present climate. Default model settings equally weight time periods.

5. CCISS Maturation Feasibility: The mean feasibility rating across the 20-year normal periods (2021-2100). This indicates the inferred feasibility of successfully growing an established to maturity (80 years). Default model settings equally weight time periods

6. P/A: Assessed preferred/acceptable rating based on establishment and maturation feasibility.

7. Trend: the proportion of the GCM simulations indicating improving/stable feasibility (numerator) vs. declining feasibility or remaining unfeasible.

The weights used to calculate Establishment Feasibility and Maturation Feasibility can be modified using the “Adjust Parameters” dialog box in the “Select Sites” tab.

Summary report

This report compares CCISS maturation feasibility with the Chief Forester’s Reference Guide for Stocking Standards. Species codes are coloured according to trends in their future feasibility using the legend at the bottom of the selection pane: improving (green), decreasing (red). Species added to the CCISS stocking standard are coloured purple, and species dropped from the CCISS stocking standard are strike-through.

Chart

The chart mode displays a stacked bar chart showing the ratio of future BGC analogs by time period predicted across range of global climate model-scenario climate projections. Hovering over a stacked bar will display these proportions numerically.

The recent time period (2001-2020) has two bars: one for the observed climate (measured by weather stations), and one for the modelled climates (simulated by global climate models). These two aren’t expected to agree: The modelled climates sample a large range of possible recent conditions, of which the observed recent climate is only one.

The default mode of this plot simply shows the BGC analog labels. Specifying a site series of interest will display the site series in the BGC analog that overlap with the edatopic position of the historic site series, along with the proportion of the edatopic overlap. The “minimum site series overlap” slider allows the user to include or exclude site series with small edatopic overlaps.

Map

Select the site or area of interest from the drop down menu and then select a future time period. The map will show the historical BGC unit in yellow and the projected BGC analogs in grey. Darker greys indicate a higher proportion of global climate model projections matched that BGC analog.

A. Silvics and Ecology

This module summarizes the silvics of each species with historic or future feasibility in the selected BGC unit and site series. This information is drawn from Klinka et al. (1999). This information can be explored further at this link.

Klinka, K., J. Worrall, L. Skoda, and P. Varga. 2000. The Distribution and Synopsis of Ecological and Silvical Characteristics of Tree Species of British Columbia’s Forests. Canadian Cartographics Ltd., Coquitlam, B.C.

B. Trends

This module provides provincial-scale maps for user-selected species and site types. Three example site series representing different edatopic conditions are currently available: the B2 edatope (nutrient-poor, moisture-subxeric sites), C4 edatope (nutrient-medium, moisture-mesic sites), and D6 edatope (nutrient-rich, moisture-hygric sites).

The figure provides three maps: (left) The historical environmental feasibility rating for each species; (center) the proportion of the GCM ensemble projecting the retreat (historically feasible but projected unfeasible) or expansion (historically unfeasible but projected to be feasible) of the species in the 2041–2060 period; and (right) the mean change in feasibility across the GCM ensemble by the 2041–2060 period.

Export Report

The report is designed for documentation and off-line use of a CCISS tool report session, e.g. as an appendix to a site plan. We recommend including only the site series of interest to the user. Choose the “Feasible Only” option to limit list to species that meet the threshold for inclusion as feasible in any of the time periods.

Export data

Feasibility data can be exported for conducting further analysis such as calculating summary statistics for your area of interest. Feasibility data exported from the CCISS tool is in “long form” with each row showing projected feasibility by site of interest, site series, and tree species. Data is exported as a .CSV file; a PDF metadata report is included in the export folder.

Composite BEC for western North America

Creating species feasibility projections for the future climates of British Columbia requires finding climate analogs in Alberta and the Western US. For Alberta, we adapted the Ecological Classification of Alberta (e.g., Archibald et al. 1996), with 21 natural subregions (Natural Regions Committee 2006) as the biogeoclimatic map units and 167 ecological sites as the site series units. For Washington, Idaho, Montana, Oregon, northern California, and northwestern Wyoming, we use a draft biogeoclimatic ecosystem classification for the Western US developed by Del Meidinger and Will MacKenzie. The resulting composite biogeoclimatic units are shown at the zone level in Figure 1.

Figure 1: Baseline biogeoclimatic units of British Columbia and adjacent jurisdictions, inferred from 20th-century ecosystem observations. Excerpted from Mackenzie & Mahony (2021). (a) Biogeoclimatic zones are the highest level of the biogeoclimatic classification. (b) Each zone comprises several subzones and subzone/variants. Draft biogeoclimatic ecosystem classifications for jurisdictions adjacent to British Columbia have recently been developed to support climate change adaptation with cross-border climate analogs. Zone names are provided in Table 1.

Table 1: Names and codes of biogeoclimatic zones of western North America

Climate modeling uncertainty

Climate models are simplifications of the earth system; they involve many compromises in modeling complex processes. Consequently, an ensemble of independent climate models is required to represent modeling uncertainties about climate change outcomes over large regions. CCISS uses an ensemble of 8 global climate models (GCMs), selected by Mahony et al. (2022), for independent modeling methods that are consistent with historical climate changes and the IPCC assessed range of very likely climate sensitivity. This ensemble is described and visualized in the cmip6-BC app.

Natural variability

Global climate models, and the Earth system itself, have internal variability—weather at time scales of hours to decades. At any point in time, the climatic conditions in different GCMs can differ not only because of differences in how they model climate, but also due to internal variability (weather). Even 20-year averages can differ significantly in different runs of the same model (Figure 1). For this reason, we include three independent simulation runs of each climate model in the CCISS ensemble.

Figure 1: Trajectories of simulated and observed climate change for southern Vancouver Island, illustrating uncertainty due to natural variability (weather) and structural differences among models. Small points are the changes in climate from 1961-1990 to 2001-2020 in up to ten independent simulations for each of eight global climate models (SSP2-4.5 scenario). Larger labelled points indicate the single-model ensemble mean change in 2001-2020. Lines indicate the trajectory of each single-model ensemble mean through 2100, with dots on each line indicating the ensemble mean during the five 20-year periods of the 21st century. The large grey square is the change in observed climate from 1961-1990 to 2001-2020 averaged across weather stations for the region. Trajectories further from the dotted grey lines (no change) indicate larger projected changes in summer precipitation (y-axis) and mean temperature (x-axis) or both. Model uncertainty is driven by differences in the global climate models (i.e. different colors) as well as differences in the individual runs of the same model (i.e. small circles of same color),

Socioeconomic uncertainty

The third major category of climate change uncertainty relates to future concentrations of greenhouse gas concentrations in the atmosphere that result from global emissions policies and socioeconomic development. The climate model projections used by CCISS follow scenarios of future greenhouse gas concentrations commonly referred to as Shared Socioeconomic Pathways (SSPs). CCISS includes projections for three major SSP scenarios: SSP1-2.6, SSP2-4.5, and SSP3-7.0 (Figure 2). SSP1-2.6 assumes strong emissions reductions (mitigation) roughly consistent with the goal of the Paris Climate Accords to limit global warming to 2^oC above pre-industrial temperatures. SSP2-4.5 assumes moderate mitigation and is roughly consistent with current emissions policies and economic trends. SSP3-7.0 is representative of a broader range of “baseline” scenarios that assume the absence of mitigation policies and is characterized by a linear increase in the rate of greenhouse gas emissions. Collectively, SSP1-2.6, SSP2-4.5, and SSP3-7.0 provide a reasonable representation of optimistic, neutral, and pessimistic outlooks (respectively) on global GHG emissions reduction efforts (Hausfather and Peters 2020).

Figure 2: Projected change in summer mean temperature for the Southern Interior Ecoprovince of BC, showing the ensemble mean and range of the 8-model climate ensemble for the three greenhouse gas concentration scenarios used as a default setting in the CCISS tool.

The CCISS BGC model

CCISS uses a Random Forest machine learning model trained on a balanced set of training points for the BGC units of western North America. The climate predictors in this model are a set of seasonal bioclimate variables selected for low correlation, ecological relevance, and predictive importance. Biogeoclimatic modeling methods are a current focus of the CCISS team and the biogeoclimatic model is expected to evolve over the next year.

Overview of BGC projection trends

Biogeoclimatic projections are illustrated in Figure 1, excerpted from Mackenzie & Mahony (2021). While there is considerable variability in the pace and character of biogeoclimatic projections driven by different climate models, there are some common zone-level trends: the expansion of the IDF analogs northward into the current SBPS zone and into higher elevations in the current MS zone; the expansion of ICH analogs northward into the current SBS zone and into higher elevations in the current ESSF zone; the expansion of CWH analogs into higher elevations in the current MH zone and eastward into the current SBS zone; and the displacement of the current SWB zone by the ESSF analogs. Inter-model differences in precipitation changes are reflected in the biogeoclimatic projections, such as in the expansion of ICH analogs into the central interior in the wetter CanESM2 model instead of IDF analogs in the drier CESM1-CAM5 model. However, the tight relationship between warming and climatic displacement (Figure 1f) suggests that inter-model differences in the rate of displacement of historical climates are primarily driven by the amount and regional pattern of warming.

The biogeoclimatic projections for the 2011-2040 period include the incursion of exotic biogeoclimatic zone analogs into the province; e.g., the Sub-Boreal Aspen Parkland (SBAP) zone from Southeastern Alberta and the Interior Grand Fir (IGF) zone from Northwestern Oregon. In addition to these exotic biogeoclimatic zones, analogs for the projected future climates of BC also include exotic subzone/variants of familiar zones. The largest area of projected exotic subzone/variant analogs is in the boreal northeast of the province, where Albertan subzone/variants of the BWBS zone dominate. In later periods (2050s and 2080s), exotic analogs are also projected in the major valley systems of southern BC and on the south coast.

Figure 1: An example of biogeoclimatic projections for British Columbia, excerpted from Mackenzie & Mahony (2021). (a) Mapped biogeoclimatic zones, which encompass the 211 biogeoclimatic subzone/variants used to model tree species feasibility. (b) Biogeoclimatic projection of the recent period (1991-2018). (c-e) Biogeoclimatic projections of the 2041-2070 period (RCP4.5) for two GCMs with medium (CESM1-CAM5), low (MRI-CGCM3) and high (CanESM2) regional climate sensitivity. (f) Biogeoclimatic displacement relative to the change in the BC-mean temperature change for each of 90 model projections. Biogeoclimatic displacement is the proportion of grid cells across BC that have a different projected biogeoclimatic unit than their model-predicted biogeoclimatic unit of the 1961-1990 reference period. Boxplots show the full range and 25th-75th percentile range of the temperature change projected by the 15-GCM ensemble in each RCP/time period combination.

Guidance for interpretation of biogeoclimatic projections

Although the visual effect of biogeoclimatic projections is of BGC zones and subzone/variants shifting across the map, these spatial shifts should not be taken literally. No analog is perfect, and analogs may be highly imperfect for a number of reasons. The actual future climate at any location will be a hybrid of (1) the characteristics of the analog climate, (2) novel climatic characteristics (e.g., extremes) that are not represented by the analog, and (3) enduring features of the local climate such as frost pooling, lake effects, and wind patterns. The likely proportions of these categories in any location and time period deserve careful consideration by the end users of CCISS products.

The misinterpretation of biogeoclimatic projections as literal spatial shifts in BGC units has led to a common perception that climate change is rendering biogeoclimatic mapping obsolete. This is not the case. The linework of biogeoclimatic subzone/variants in many cases will remain useful as units of relative climatic variation across landscapes. The terminology we use for biogeoclimatic projections can help to emphasize that biogeoclimatic analogs are only approximations and that the biogeoclimatic units themselves are not undergoing spatial shifts. Rather than saying “this location is becoming IDFxh1”, it is more correct to say “the future climate at this location is predicted to be similar to the historical climate of the IDFxh1.” Rather than “the IDF is moving north into the SBS”, it is better to say “the SBS is transitioning into more IDF-like climates”.

The edatopic grid

In the BEC system, variations in site conditions within each climate type (i.e., biogeoclimatic subzone/variant) are represented by an edatopic grid. An edatopic grid has 8 relative soil moisture regimes (SMRs) and 5 soil nutrient regimes (SNRs). Each cell in this grid, called an edatope, is the combination of one SMR and one SNR, and represents the finest scale of site differentiation in the BEC system. A site series is a group of edatopes over which a classified BEC Plant Association has been observed to occur. The distribution of site series across the edatopic grid is unique to each biogeoclimatic subzone/variant.

Site series misalignment

In the CCISS analysis, the edatope of each location of interest remains constant as the climate changes. If species feasibility ratings were specified for each edatope, and if the edatope of the location of interest were known, then there could be a direct transfer of species feasiblity from the future BGC analog to the location of interest. However, species feasiblity ratings are at the site series level, and typically the management decision is also being made at the site series level. This creates a problem for the CCISS analysis: since the edatopic distributions of corresponding site series at any edatope do not align among the BGC subzone/variants, there is ambiguity about which site series of the projected future BGC subzone/variant is the correct one to choose for any site series in the historical climate.

The problem of site series alignment is illustrated in Figure 1. In this example, the historical site series of interest is the 09 site series, which occupies four edatopes: D5, D6, E5, and E6 (left panel). The BGC subzone/variant analog for the future climate has three site series that overlap with this edatopic space: 07, 08, and 09 (right panel). Tree species feasibility ratings from each of these three site series are applicable to the historical site series 09.

Figure 1: Example of the misalignment of site series of a historical BGC subzone/variant (left) and the BGC subzone/variant analog for its projected future climate (right).

Weighting site series contributions to CCISS feasibility projections

Instead of choosing a single “best match” analog site series, the CCISS tool uses all analog site series weighted by their overlaps with the historical site series (Figure 2). There are two directions of overlap: forward overlap (how much of the historical site series is covered by the analog site series), and reverse overlap (how much of the analog site series is covered by the historical site series). Overlaps are measured at the edatope level: partial occupancy of a site series in an edatope is counted as a full edatope. In the example from Figure 1, the historical 09 and analog 07 site series have a forward overlap of 25% and a reverse overlap of 50%. These overlaps are multiplied to produce an overlap agreement of 12.5%. After rescaling the overlap agreements by 1.273 so that they sum to 100%, the contribution of the 07 site series to the projected tree species feasibility is 16%.

Figure 2: Calculation of the overlap agreement for the example shown in Figure 1. Overlap agreement is weight of each site series to contribute its tree species feasibility ratings to the CCISS projection for the historical site series of interest.

Extra-edatopic site series (e.g. floodplains) are aligned to similar types by predefined rule sets (e.g. historical high-bench floodplain site series aligned to analog high-bench floodplain site series).

Feasibility ratings

Tree species’ environmental tolerances are complex and many approximations are required to translate them into a simple feasiblity ratings metric used in CCISS. Feasibility ratings for each tree species in each site series have been assigned primarily through expert judgement with support from vegetation plot data. There are inevitably errors in this database, particularly for species that were not historically prominent in reforestation, such as deciduous species. The feasibility ratings are undergoing ongoing review by the CCISS team and regional ecologists, via the By-BEC Portal (Figure 1). Further, the CCISS team is currently conducting predictive modeling to quantitatively evaluate and refine feasibility ratings.

Figure 1: Screenshot of the By-BEC Portal, which facilitates expert review and revisions of the CCISS feasibility ratings. This example shows feasibility ratings for western redcedar on zonal sites, with editing for the CWHdm open.

Climate mapping

The reference climate maps for CCISS are the PRISM (Parameter-elevation Regressions on Independent Slopes Model) climate surfaces of temperature and precipitation at 800m spatial resolution developed by the Pacific Climate Impacts Consortium. These surfaces are best-in-class, but they have important limitations. They are interpolated from weather station data that is sparse in many regions of BC (Figure 2). Most valleys and some larger regions have no stations, so the nuances of climate in these locations may not be well represented, particularly cold air drainage and elevational gradients. Further, the PRISM method does not model microclimatic factors such as heat loading (warm vs. cold aspects), vegetation influences, and lake effects. Microclimatic factors need to be accounted for during professional interpretation of CCISS results.

Figure 2: PRISM climate map of July daily maximum temperature, in the Cariboo-Chilcotin region. Stations are concentrated at low-elevations and are absent from large areas, demonstrating the potential for errors in the climate mapping.

Biogeoclimatic modeling

Biogeoclimatic modeling involves using machine learning to classify climate conditions as biogeoclimatic units (subzones/variants). We have found that BGC projections are sensitive to many aspects of model training, especially climate variable selection. Improving BGC modeling methods and representing uncertainty in BGC projections is a current priority for the CCISS team. An example of these planned improvements is integrating climate variables that capture extremes such as heat waves and drought. Even with these future improvements, biogeoclimatic modeling will retain inherent imprecision and will continue to be a source of error in CCISS projections.

Climate data downscaling

Climate changes are modeled in CCISS by overlaying very coarse-scale global climate model projections (50-150km resolution; Figure 3) onto the 800m PRISM climate maps, a method called change-factor downscaling. This approach results in a uniform warming rate from valley bottom to mountain top. In reality, we would expect elevation-dependent differences in warming rate due to loss of snowpack, for example. Similarly, change-factor downscaling can’t represent the role of other fine-scale features like lakes, vegetation, cold-air pooling, aspect, and soil moisture in modifying the regional average climate change. The Ministry of Forests’ ClimatEx project seeks to improve the spatial resolution of climate change modeling, but will not be integrated into CCISS for at least two years. CCISS users are encouraged to consider the role of site-specific modification of regional climate changes.

Figure 3: Coarse resolution of climate changes provided by global climate model data. This panel shows the low-resolution interpolated changes in July mean temperature for the EC-Earth3 global climate model for the 2041-2060 time period, relative to the model’s 1961-1990 climate. The coastline of Vancouver is shown for scale.