12 PRE-DEVELOPMENT AND POST-CLOSURE LANDSCAPE DYNAMICS IN OIL-SANDS MINING – IMPLICATIONS FOR ASSESSMENT OF RECLAMATION AND EQUIVALENT CAPABILITY J. Straker, M.Sc.1 J.B. Stelfox, Ph.D.2 1Forest ecologist, soil scientist, Integral Ecology Group, Duncan, B.C., Canada 2Landscape ecologist, ALCES Landscape and Land Use, Calgary, Canada ABSTRACT Site-level evaluation of reclamation performance and residual impacts (i.e., post-reclamation effects) of large mining projects has conventionally used an approach in which landscape conditions (e.g., amounts of certain vegetation types or wildlife habitats) immediately prior to development are characterized and used as a baseline against which to assess post-reclamation conditions. This approach is static, in that it evaluates post-reclamation conditions, usually at a single time in the future, against a “snapshot” of conditions at a single pre-development time. However, many of the characteristics typically evaluated through this approach are both spatially and temporally dynamic, on both the “pre-disturbance” and post-reclamation landscape: • “pre-disturbance” – mining is almost always not the first disturbance of a landscape, which has generally been repeatedly subjected to substantial and recent perturbations such as wildfire, and thus has undergone many pre-mining cycles of disturbance and recovery; • post-reclamation – reclaimed landscapes evolve over time, and their characteristics and utility (to humans and other species) change with this evolution, which may be affected by new future cycles of disturbance and recovery. This conventional assessment conceptual model then is highly dependent on both the specific conditions existing in a project area immediately prior to industrial development, and on the future time selected for post-reclamation assessment. In this paper we argue that the broad application of this conceptual model has resulted in a simplistic understanding of landscape responses to industrial disturbances, and has contributed an arbitrary component to reclamation planning and assessment. We propose that alternative approaches (e.g., use of ecological simulation modelling) in which pre- and post-development landscape dynamics are explicitly acknowledged and accounted for are necessary to gain a more sophisticated and useful understanding of the role of mining and reclamation in the cycle of ecosystem disturbance and recovery. An example of such an approach is presented using a case study from oil-sands mining and reclamation in northeast Alberta, Canada. KEY WORDS equivalent capability, oil sands, mine reclamation, effects assessments. 13 INTRODUCTION In many jurisdictions, prospective mining operators are required to submit Environmental Impact Assessments (EIAs, or equivalent) to obtain regulatory approval for mining activities. These assessments typically contain conceptual reclamation plans, and assess the anticipated effects of mining by comparing pre-development conditions with projected post-closure conditions. The expected differences (e.g., in wildlife habitat) between these two compared time periods (pre-development and post-mitigation) are termed residual impacts. In both Alberta and British Columbia, one of the key concepts used in these assessments is that of “equivalent land capability” (“land capability” in British Columbia). Alberta defines equivalent land capability as meaning that the ability of the land to support various land uses after reclamation is similar to the ability that existed prior to an activity being conducted (although the land uses may not be identical), and defines land capability as “the ability of land to support a given land use, based on an evaluation of the physical, chemical and biological characteristics of the land, including topography, drainage, hydrology, soils, and vegetation” (Province of Alberta 2016). Once operating, mines are required through various regulatory instruments to provide periodic updates to their conceptual/long-range and detailed/short-range reclamation activities and reclamation plans, including current information on progress and planning towards achievement of equivalent land capability. METHODOLOGY In examining the methods used for assessment of residual impacts and achievement of equivalent land capability in EIAs and reclamation planning as described above, there are three key elements to recognize in the western/northern Canadian setting (or in any relatively remote mining area with similar regulatory approaches): 1. the concepts of capability used in both Alberta and British Columbia require comparisons between the pre-development and post-closure environments, as part of both the application and operating phases of mine life; 2. because land uses are a central concept in both regulatory regimes (see capability definitions above), the pre- and post-mining comparisons focus on the ability of the land to support designated post-closure uses, and compare this ability to pre-development conditions; and 3. due to the fact that much mining in the last 50 years in Alberta and British Columbia has been conducted areas with relatively low human populations, the primary land use considered in these assessments is wildlife habitat. These elements thus require that mining operators provide information on pre-development wildlife-habitat conditions, project the provision of wildlife-habitat requisites in the post-closure 14 landscape, and assess the differences as residual impacts and/or as contributions to achievement of equivalent land capability. Although this approach has been effective in encouraging and supporting thoughtful closure planning, there are some limitations to its application. A primary limitation is that conventional methods do not address or even explicitly acknowledge the dynamic nature of both pre-development and post-closure landscapes. In the pre-development landscape – often referred to as the “pre-disturbance” landscape – mining is almost invariably not the first ecosystem disturbance. There are many potential natural disturbance agents such as insect outbreaks, wind throw, and mass movements, but in particular in western and northern Canada, most mining jurisdictions have been repeatedly subjected to frequent and often large fire events which are easily capable of burning areas larger than the largest mining footprints.1 In these areas, ecosystems experience many pre-mining cycles of disturbance – often severe – and recovery. Similarly, post-reclamation landscapes will evolve over time, and thus their vegetation conditions, and the suitability of these conditions as habitat for different wildlife species, changes with this evolution. These changes continue until a new disturbance – natural or anthropogenic – initiates a new recovery trajectory. Results of this conventional assessment approach are highly dependent on both the specific conditions existing in a project area immediately prior to industrial development, and on the future time selected for post-reclamation assessment. Broad application of the EIA residual-impact conceptual model has resulted in a simplistic understanding of landscape responses to industrial disturbances, and has contributed an arbitrary – and potentially skewed – component to reclamation planning and assessment, whereby the pre-development conditions are used as targets for reclamation planning, regardless of the particular context of those conditions. We propose that alternative approaches (e.g., use of ecological simulation modelling) in which pre- and post-development landscape dynamics are explicitly acknowledged and accounted for are necessary to gain a more sophisticated and useful understanding of the role of mining and reclamation in the cycle of ecosystem disturbance and recovery. APPLICATION OF AN ECOSYSTEM SIMULATION MODELLING APPROACH Ecosystem simulation models have been used to provide increased understanding of fluctuating conditions under pre-industrial natural disturbance regimes, and of conditions over time in ecosystems recovering from both natural and anthropogenic disturbances. The explicit temporal aspect of these models addresses some of the shortcomings of the current EIA model noted above, and in some cases is the only way to produce desired information. For example, in western/northern Canada, the temporal span of datasets on ecosystem conditions is typically short, and thus it is difficult to understand historic variation over longer time periods without application of a modelling component that simulates historic disturbance regimes. Similarly, 1 For instance, in 2011, a single fire event, the Richardson fire originating north of McClelland Lake, burned over 575,000 ha, or an area almost 25 times as large as the Mildred Lake lease area modelled in this study. 15 projection of conditions into the future requires a simulation framework in which to conduct such projections. In this study we used the ALCES model (A Landscape Cumulative Effects Simulator) to assess pre-development and post-reclamation conditions on the Syncrude Canada Ltd. (Syncrude) Mildred Lake lease in the Athabasca oil-sands region of northeast Alberta. ALCES was used to track presence and evolution of ecosystem types and associated attributes (e.g., wildlife habitat suitability) on an annual time step. Range of Natural Variation A key concept and metric modelled for this study is the range of natural variation (RNV). RNV can be considered the normal variation of a specific ecological indicator (e.g., moose habitat suitability2) that occurs in response to the full suite of natural and episodic disturbances that characterize an ecological system. In the study area for this project, the primary natural disturbance agent is fire. RNV based on fire was simulated stochastically using a random draw from a lognormal distribution with an average annual burn rate of 1.25% (i.e., an 80-year mean fire-return interval), a maximum annual burn rate of 10%, and corresponding fire size-class distributions, based on fire research in the region (Andison 2005). Lognormal distributions are appropriate in areas such as Canada’s boreal forest where burn rates are highly variable across years, and typically low annual burn rates are punctuated by occasional large fire years (Armstrong 1999). Information on ecosystem occurrence in pre-development conditions was derived from retrospective Ecosystem Land Classification and Alberta Vegetation Inventory mapping conducted using pre-clearing aerial photography and supplied by Syncrude. RNV simulations were conducted over 1,000 simulation years, and were conducted for both the Mildred Lake study area (24,621 ha) and a much larger study area – the 6.6-million-ha Regional Municipality of Wood Buffalo (RMWB) – in which the Mildred Lake study area occurs. The RMWB was modelled to illustrate differences between the conditions particular to the Mildred Lake lease and those found within the larger region. An example of derivation of an RNV simulation for a wildlife-habitat-suitability metric for the Mildred Lake Study Area (MLSA) is shown in Figure 1. Reclamation For modelling of development of reclaimed landscapes and associated attributes, simulations were based on projected ecosystem occurrence from Syncrude’s 2011 closure plan. Wildlife Habitat Suitability Coefficients for wildlife indicators were derived from habitat-suitability-index (HSI) models developed for northeastern Alberta. HSI models are primarily knowledge-based (as opposed to 2 This paper primarily discusses wildlife habitat suitability (i.e., “the ability of the habitat in its current condition to provide the life-requisites of a species”) and not capability (i.e., “the ability of the habitat under optimal natural [seral] conditions for a species to provide its life-requisites, irrespective of the current condition of the habitat”) (MELP 1999). 16 empirical) models that can incorporate information from both empirical studies and expert knowledge (USFWS 1980). The HSI models used in this study were based on a review of published literature as well as expert opinion; they were initially developed for application in the Cumulative Environmental Management Association’s (CEMA) Terrestrial Ecosystem Management Framework for the RMWB, and subsequently revised through the modelling conducted in support of the Government of Alberta’s Lower Athabasca Regional Plan. Figure 1 RNV model results for moose habitat suitability index using two 500-year simulations. Individual blue and red lines show simulated indicator performance over each 500-year simulation. The blue band shows the observed range of indicator performance for both simulations, while the blue circle shows the median of that range. Results The following discussion presents examples of application and interpretation of the simulation-modelling approach, using moose habitat suitability as an indicator. These examples are provided to illustrate the central theses of this paper, and not to provide specific data or conclusions on reclamation and equivalent capability in the project study area. A comparison of different methods for assessing pre-development conditions is presented in Figure 2. This comparison shows that relative to the larger region, conditions in the lease-specific study area are relatively favourable for moose habitat, due to a higher proportion of upland deciduous forests on the pre-development lease area than is typically found throughout the larger C 17 region.3 Moose HSI based on the static pre-development inventory data – analogous to the typical pre-development approach used in EIAs – is near the top of the simulated lease RNV, reflecting unusually favourable conditions for moose at this particular point in time. This is due to the fire history in this area prior to development, and the abundance of young forests from recent fires, which generally have a higher suitability as moose habitat than older forests. The observed difference between the static pre-development conditions and the simulated RNV for this indicator is illustrative of a primary thesis of this paper: that assessment of equivalent capability versus immediately pre-development conditions and stewarding to these conditions in reclamation planning can lead to skewed outcomes due to incomplete understanding of temporal dynamics. In the case of reclamation planning, targeting the return of an abundance of very high-value moose habitat similar to that occurring at the onset of lease development is unnecessary given a more complete understanding of fluxes in pre-development conditions. Stewarding to this target may induce other stresses/risks in the reclaimed system, such as replacement of an over-abundance of deciduous-dominated forests with high water use, which will have effects on overall water balances and downstream ecosystems in the reclaimed landscape. In the case of assessment of equivalent capability, the static pre-development HSI value sets a benchmark that is abnormally high in the context of longer-term ecosystem dynamics, and may result in a conclusion that reclamation has underachieved in capability for this land use, simply because of unusual conditions at the time of the pre-development inventory. Figure 2 RNVs and pre-development conditions for moose HSI for the Mildred Lake Study Area. The blue and red bands show RNV for the Mildred Lake Study Area (MLSA) and RMWB Study 3 The RNV for the lease study area is larger than that of the regional study area due to the ability of larger simulated fire events to affect a higher proportion of the smaller lease area than of the larger regional area. 18 Area (CEMA). The blue and red circles show the median of these ranges. The yellow circle shows the single-time HSI value immediately prior to development. In Figure 3, the RNV information discussed above is extended to a simulation of historic and future conditions including development of the lease, reclamation, and recovery dynamics of the post-closure landscape. This figure emphasizes the contrasts between a dynamic and static assessment of equivalent capability for a wildlife-habitat end land use. Because the deciduous forests of the pre-development landscape are replaced to a large extent by mixed wood forests in the reclamation plan (due to the understandable desire to mitigate risk by planting multiple overstory species rather than single species), the HSI simulations show reduced values in the post-closure landscape in comparison to pre-development conditions. Depending on approach, one could interpret these results in different ways: Static pre-development inventory approach: reclamation replaces habitat equivalent to approximately 50-70% of that that existed in the pre-development inventory. Development and subsequent reclamation has resulted in a substantial reduction in moose habitat. Dynamic assessment approach: reclamation replaces habitat values that are often reduced in comparison to the median value of RNV, but are still within this range. Therefore there is no meaningful reduction in moose habitat resulting from development and reclamation. Further, post-closure habitat within the lease is still in the upper half of or above the regional RNV, meaning that in the regional context the post-closure landscape is still relatively favourable for moose. A related observation is that the timeframe selected for the post-closure assessment of conditions influences the outcome of these assessments. Due to forests aging on the post-closure landscape, some time periods (e.g., early and late in the post-closure assessment period) provide higher moose habitat values than others. In addition, assumptions on management of the post-closure landscape (e.g., fire suppression4) also influence these outcomes, as in general re-introduction of a natural fire regime results in higher HSI values for moose, due to its maintenance of higher proportions of younger forests. Figure 4 shows an analysis equivalent to that presented in Figure 3 for a different wildlife species, fisher. Because fisher, a mustelid furbearer, requires conifer-dominated upland forest, it is simulated to perform well in the post-closure environment, due the conversion of relatively less-favoured ecosystems – deciduous forests and organic wetlands – to these conifer types. All assessment methods show fisher HSI outperforming pre-development conditions in the post-closure landscape. This example is provided to illustrate the temporal dynamics of habitat evolution in the post-closure environment, and specifically the fact that different wildlife species have different response to these dynamics: there are times at which the post-closure landscape is simulated to have relatively high habitat suitability for moose and low suitability for fisher, and other times at which these patterns are reversed. These divergent responses are attributable 4 The example of fire suppression is provided here, but other external management actions and land-use decisions potentially also have impacts on post-closure HSI, such as the level of human access that would be expected on and adjacent to the post-closure landscape. 19 partially to specific preferences for certain forest age classes, and partially to similarly divergent responses to external management factors: because fisher favour older forests as habitat, their habitat is best in the absence of a simulated post-closure fire regime, as these conditions allow development of higher proportions of mature forest, where the same development is detrimental to moose habitat. Figure 3 Moose HSI in two simulations through four phases: 1) RNV; 2) development to current conditions (backcast); 3) ongoing development and reclamation to mine closure (reclamation); and 4) post-closure dynamics. RNV bands and the immediately pre-development conditions are depicted as described for Figure 2. The two simulations include one in which fires are not modelled in the post-closure landscape (blue line), and one in which they are (red line). Assumptions The model results discussed above incorporate the assumption that reclamation of mining landforms will be at least partially effective at replacing functional ecosystems that provide habitat requisites present in the pre-development environment. This is a substantial uncertainty that cannot be tested through simulation modelling, and will only be resolved through diligent and creative reclamation that aims to restore functional components of the pre-mining landscape, and through long-term monitoring of reclamation outcomes. However, simulation modelling does provide a framework to evaluate uncertainties like these. The ability of reclamation to effectively replace wildlife habitat requisites was evaluated by modelling habitat-suitability indices with varying levels of habitat discounting applied to the reclaimed landscape. This analysis is presented for moose in Figure 5, and shows that habitat values are relatively insensitive to HSI 20 discounting versus other over-riding factors such as proportions of different ecosystem types in the post-closure landscape. In other words, this analysis suggests that moose are more sensitive to coarse ecosystem differences such as relative abundance of deciduous upland forests, and less sensitive to the relative quality of the habitats within these ecosystem types. Clearly there are limitations to this observation, but it is provided to illustrate both the importance of the assumption and its associated uncertainties, and that sensitivity of assessments of equivalent capability to such assumptions can be at least partially tested in a simulation-modelling framework, whereas they can generally only be addressed qualitatively in a conventional EIA model. Figure 4 Fisher HSI in two simulations through four phases: 1) RNV; 2) development to current conditions (backcast); 3) ongoing development and reclamation to mine closure (reclamation); and 4) post-closure dynamics. RNV bands and the immediately pre-development conditions are depicted as described for Figure 2. The two simulations include one in which fires are not modelled in the post-closure landscape (blue line), and one in which they are (red line). CONCLUSION The primary point advanced by this study is that assessment of the equivalency of reclaimed ecosystems to pre-development conditions is complex, due to the dynamic nature of both pre-development and post-closure landscapes. 21 Baseline conditions and pre-development dynamics Characterization of pre-development baseline conditions, and assessment against these conditions, should include recognition that non-industrial landscapes are inherently dynamic, and subject to repeated perturbations from natural disturbance regimes. Further, this characterization and assessment should incorporate quantification of the effects of these non-industrial dynamics, whether through ecosystem simulation modelling or other methods. In this study, pre-development ranges of natural variation for wildlife-habitat indicators were simulated. Despite the fact that average (median) values can be provided for these simulated ranges, variation around these average values is substantial (±50% of the mean value), particularly at the scale of individual mine leases. Thus we suggest that ranges of pre-development conditions, rather than precise single values, should be used as the basis for defining envelopes of equivalency for land-use indicators of interest. Figure 5 Moose HSI with varying levels of discounting of habitat suitability applied to reclaimed landscapes (0%, 20% and 40% in comparison to fire-origin stands). Levels are distinguished by the numbered legend and corresponding numbers on the lines depicting simulation results. These simulations were run with a deterministic (mean, non-random) fire rate in the pre-development landscape, and in the absence of fire in the post-closure landscape. In contrast, static characterizations of baseline conditions derived from observed or inferred pre-development inventories typically do not recognize any variation around metrics, but provide precise single values in various habitat-capability classes for different species. Presentation of this information typically does not acknowledge that actual habitat on the lease and on these hectares 22 would have varied temporally according to changes in stand ages, successional patterns, and other attributes. These static characterizations are unlikely to capture pre-development conditions in an average state, as fire history and other factors may mean that a pre-disturbance inventory actually records relatively unusual conditions. This appears to be the case for ecosystem conditions in this study, in which forest ages immediately prior to development were substantially younger than median RNV due to recent fire history. Equivalency targets based on these static and potentially non-average measures may be misleading, as they may suggest stewardship and/or comparisons to precise but inaccurate and unusual conditions. Post-closure dynamics Not only do landscape capabilities vary over time in the pre-development environment, but they also vary temporally in the post-reclamation and closure environment. The result of such variation is that at any given assessment time, some indicators may be performing well in comparison to pre-development metrics, and others less so, while this situation may be reversed at a different time of assessment. Thus, when assessing equivalency, it is critical not only to understand the variation in the pre-development environment from which targets are derived, but to also understand the temporal dynamics of recovery and variation in the post-reclamation and closure landscape. Use of pre-development RNVs as baseline envelopes of equivalent capability, and comparison to ranges of expected performance on the post-closure landscape, can provide this understanding, and would be an improvement on current methods of setting targets for reclamation planning and for assessment of achievement of equivalent capability. ACKNOWLEGEMENTS The authors wish to thank Audrey Lanoue of Syncrude Canada Ltd. for her intellectual contributions to the discussed case study, and for Syncrude’s support in this project. REFERENCES Andison, D.W. 2005. Natural Levels of Forest Age Class Variability on the RSDS Landscape of Alberta. Unpublished Report Submitted to the Cumulative Environmental Management Association by Bandaloop Landscape-Ecosystem Services, Vancouver, BC. 86 pp. December 2005. Armstrong, G.W. 1999. A stochastic characterization of the natural disturbance regime of the boreal mixed wood forest with implications for sustainable forest management. Canadian Journal of Forest Research 29: 424–433. Province of Alberta. 2016. Environmental Protection and Enhancement Act, Conservation and Reclamation Regulation. Alberta Regulation 115/1993, with amendments up to and including Alberta Regulation 103/2016. Alberta Queen’s Printer, Edmonton, AB. 23 Ministry of Environment, Lands and Parks. 1999. British Columbia Wildlife Habitat Rating Standards. Prepared for the Terrestrial Ecosystem Task Force, Resources Inventory Committee by the Resource Inventory Branch, Ministry of Environment, Lands and Parks, Victoria. 111 pp. United States Fish and Wildlife Service (USFWS) 1980. Habitat evaluation procedures. United States Fish and Wildlife Service (USFWS), ESM 102. Release 2-80, Department of the Interior, Washington, DC.