Synthesis of Habitat Models Used in the Oil Sands Region
Welcome to the Cumulative Environmental Management Association (CEMA) database of species habitat models used in the Athabasca Oil Sands Region (AOSR). The Wildlife Task Group (WTG) of the Reclamation Working Group (RWG) of CEMA completed CEMA Contract 2010-0034 - Synthesis of Habitat Models Used in the Oil Sands Region in 2011. The deliverables of this project included a final report documenting the results of the habitat models synthesis (Muir et al. 2011) and a database summarizing attributes of each model reviewed as part of the synthesis.
PURPOSE OF DATABASE
The purpose of this database is to compile and summarize the state of the knowledge of the species habitat models used in the AOSR during the preparation of Environmental Impact Assessment (EIA), Supplementary Information Requests (SIR), and Closure Planning phases of oil sands projects from 1990 to the present.
The database structure is hierarchical; users can assess information by navigating through linked tables by clicking on primary or secondary headings to access information on specific projects, species, model types or proponents.
Primary headings indicate major search categories (e.g., Species, Model Type, Project Name, Proponent) and secondary headings provide specific information (e.g., project details, species data, closure planning, etc.).
Each secondary heading is comprised of a table with a comprehensive list of data fields that attempts to capture all relevant aspects of each species habitat model developed for the various projects. Tables have been transformed into forms for viewing (data has already been entered) and/or easy data entry by the user.
- Search by Species
- Search by Model Type
- Search by Project
- Search by Proponent
This project was initiated following the publication of the Guidelines for Reclamation to Forest Vegetation in the Athabasca Oil Sands Region (AENV 2010) to address uncertainty regarding the validation status of habitat models used for predicting environmental effects in Environmental Impact Assessments (EIAs) and for closure planning, and for the reclamation guidance developed based on habitat suitability indices derived from unvalidated models (see pp. 103-104 in AENV 2010). This project assessed the current state of habitat models used in oil sands region EIA and closure planning to meet the following objectives:
1. Determine which habitat models are used in EIAs and closure planning, and how these models were used;
2. Determine what linkages exist between the habitat model predictions in the EIAs and closure plans;
3. Determine which habitat models have been validated, and of these, describe and evaluate the validation procedures that were used on each model with recommendations for improvement if needed; and
4. Recommend procedures to validate non‐validated models.
Twenty two oil sands operations EIAs and 228 species habitat models in the EIA documents were reviewed. The number of models in an EIA varied from 4 to 15, with a median of 12 models. The majority of these models (82.0%) were reused from previous EIAs, often with some degree of modification although this was frequently not clearly specified in the model documentation. Three types of species habitat models were mainly used. Habitat suitability index (HSI) models were used the majority of the time (74.1%), with resource selection functions (RSFs) and habitat capability and/or suitability (CAPSU) models used 7.9% and 16.7% respectively. RSF models were solely used for Moose (Alces americanus), Canada Lynx (Lynx canadensis), Fisher (Martes pennanti), Barred Owl (Strix varia) and Black-throated Green Warbler (Dendroica virens) and were only included in three EIAs (Kirby In-situ, Christina Lake and Jackpine Mine Expansion).
Of the 228 models, 101 (44.3 %) had some validation documented in an EIA, with RSF models most likely to be validated (66.7 %) followed by HSI models (52.1 %). Only one of the CAPSU models was validated. Data gaps were the main reason given for performing limited or no model validation. Three main data gaps were identified:
i) lack of species observations within the study area [e.g., American Black Bear (Ursus americanus)];
ii) species observations were available, but had been made during a different season than that targeted by the species habitat model (e.g. fall surveys of waterfowl when models applied to nesting habitat suitability); and
iii) species observations were available, but these observations did not sample all the seasons represented by the model [e.g., only breeding season data for an all-year Canadian Toad (Bufo hemiophrys] model).
There appeared to be an increasing trend of documented model validation with EIAs submitted from 2005 and later. An average of 54.8% of the models in these EIAs had some level of validation documented. Only 32.4% of the models were apparently validated in EIAs from 1996-2003 (no EIAs were reviewed for the year 2004). The percentage of validated models was similar for new models and reused models. When species/species groups were considered, Canada Lynx models were validated most often (75.0%). Other frequently modelled species/species groups (i.e., Fisher/Marten, Canadian Toad, Moose, Ruffed Grouse, Snowshoe Hare and Woodland Caribou) had 62.5 to 70.0% of their models validated. No Barred Owl, non-treed wetland/riparian birds or Black-throated Green Warbler models were validated. Statistical methods were used in 48.5% of all validated models when HSI and RSF model types were pooled.
Statistical methods were used in 42.0% of the validated HSI models and in all instances of validated RSF models. The HSI models were validated using various methods; however, there was no evidence of validating the model structure or calibrating the model as recommended in the reviewed literature. Most of the documented HSI model validation consisted of a portion of the recommended methods, i.e., testing species observations against model predictions. There was also no indication that an iterative validation process was carried out whereby model validation results were used to improve the model, which was then subjected to further validation until satisfactory results were achieved, or no more improvements could be made to the model given the limitations of current knowledge.
Habitat variables and species observations collected as part of EIA survey work are often not appropriate for HSI model validation. Observations may not provide the correct type of samples, adequate sample size, stratification across the full range of predicted HSI values, or in the case of species observations, span sufficient time to adequately represent patterns in species habitat use. None of the reviewed models provided any design or implementation of specific studies to collect validation data for a model. Typically data from baseline studies for the project and occasionally other projects in the area were used instead. This approach limited HSI model validation because of data gaps within these baseline data.
The RSFs presented in the reviewed EIAs modelled habitat suitability for Moose, Canada Lynx, Fisher, Barred Owl and Black-throated Green Warbler. Models for Moose, Canada Lynx, Fisher, Barred Owl were all of the “used/available” RSF form. Each of these models was validated using the methods recommended by Boyce et al. (2002), i.e., k-fold validation and use of the Spearman rank correlation to assess the relationship of area corrected frequencies of presence to bins of predicted RSF scores.
The review of EIA species habitat models revealed that while RSF models tended to be appropriately validated, the HSI models were not. Only 52% percent of HSI models were validated and this validation did not appear to conduct all four steps recommended by standards documents (e.g., USFWS 1981, RISC 1999). Furthermore, descriptive comparisons of species observations in habitat suitability classes used in several cases of HSI model validation did not correct for the available area of the suitability classes, and so may have misinterpreted the validation results.
Based on the findings of the report, the Guide to Preparing Environmental Impact Assessment Reports in Alberta Updated March 2013 (Government of Alberta 2013) was updated to state that wildlife habitat models should be appropriately validated (see Muir et al 2011). CEMA conducted a follow-up study (Muir et al 2012; CEMA Contract 2011-0034) to:
1. Review existing data sources to determine if these data can be used to validate existing species habitat models identified in the Phase 1 report.
2. Develop a data collection program to address data gaps limiting procedures to validate remaining models.
3. Describe validation procedures for existing species habitat models based on available data and the data collection program.
The Phase 1 and Phase 2 reports (Muir et al 2011, 2012) provide substantial guidance for improving the validation procedures and methods for wildlife habitat models used in the AOSR. The database is available as a resource to consultants to review and prepare wildlife habitat models for future studies.
Alberta Environment. 2010. Guidelines for Reclamation to Forest Vegetation in the Athabasca Oil Sands Region, 2nd Edition. Prepared by the Terrestrial Subgroup of the Reclamation Working Group of the Cumulative Environmental Management Association, Fort McMurray, AB. December 2009.
Boyce, M. S., Vernier, P. R., Nielsen, S. E. and F. K.A. Schmiegelow. 2002. Evaluating resource selection functions. Ecological Modelling 157(2‐3):281‐300.
Government of Alberta. 2013. Environmental Assessment Program Guide to Preparing Environmental Impact Assessment Reports in Alberta Updated March 2013.
Muir, J.E, V.C. Hawkes, K.N. Tuttle, and T. Mochizuki. 2011. Synthesis of Habitat Models used in the Oil Sands Region. LGL Report EA3259. Unpublished report by LGL Limited environmental research associates, Sidney, BC, for the Cumulative Environmental Management Association (CEMA) – The Reclamation Working Group (RWG), Fort McMurray, AB. 30 pp + Appendices.
Muir, J.E, M. D’entremont, J. Gatten, L. Ainsworth and D. Robichard. 2012. Validation Procedures for Habitat Models in the Oil Sands Region. LGL Report EA3354. Unpublished report by LGL Limited environmental research associates, Sidney, BC, for the Cumulative Environmental Management Association (CEMA) – The Reclamation Working Group (RWG), Fort McMurray, AB. 95 pp + Appendices.
RISC. 1999. British Columbia Wildlife Habitat Rating Standards Version 2.0 Prepared by Ministry of Environment, Lands and Parks Resources Inventory Branch for the Terrestrial Ecosystems Task Force Resources Inventory Committee.
USFWS 1981. Standards for the development of habitat suitability index models (103 ESM). Division of Ecological Services U.S. Fish and Wildlife Service. Department of the Interior. Washington, D.C.