Quantification of anthropogenic and natural changes in oil sands mining infrastructure land based on RapidEye and SPOT5

https://doi.org/10.1016/j.jag.2013.11.013Get rights and content

Highlights

  • A hybrid method is proposed to monitor anthropogenic change associated with oil sands developments.

  • A new landscape index, the Re-growth Index, has been formulated for monitoring reclamation of transient land disturbances.

  • SPOT5 and RapidEye images have been utilized to create consistent time series of change.

  • Accuracies of 80% and 90–95% are achieved for pixel and object level processing.

Abstract

Natural resources development, spanning exploration, production and transportation activities, alters local land surface at various spatial scales. Quantification of these anthropogenic changes, both permanent and reversible, is needed for compliance assessment and for development of effective sustainable management strategies. Multi-spectral high resolution imagery data from SPOT5 and RapidEye were used for extraction and quantification of the anthropogenic and natural changes for a case study of Alberta bitumen (oil sands) mining located in the Western Boreal Plains near Fort McMurray, Canada. Two test sites representative of the major Alberta bitumen production extraction processes, open pit and in situ extraction, were selected. A hybrid change detection approach, combining pixel- and object-based target detection and extraction, is proposed based on Change Vector Analysis (CVA). The extraction results indicate that the changed infrastructure landscapes of these two sites have different footprints linked with their differing oil sands production processes. Pixel- and object-based accuracy assessments have been applied for validation of the change detection results. For manmade disturbances, except for those fine linear features such as the seismic lines, accuracies of about 80% have been achieved at the pixel level while, at the object level, these rise to 90–95%.

Since many disturbance features are transient, a new landscape index, entitled the Re-growth Index, has been formulated at single object level specifically to monitor restoration of these features to their natural state. It is found that the temporal behaviour of the Re-growth Index in an individual patch varies depending on the type of natural land cover. In addition, the Re-growth Index is also useful for assessing the detectability of disturbed sites.

Graphical abstract

Growth of an oil sands development project near Kearl Lake, Alberta, from 2007 to 2011 based on a temporal sequence of SPOT5 images.

  1. Download : Download high-res image (197KB)
  2. Download : Download full-size image

Introduction

Land cover and land uses in natural resources development regions are altered by human activities associated with the exploration, production and transportation processes. Quantification of these landscape changes is a critical information component effective compliance monitoring and for the development of effective sustainable management practices (Jordaan et al., 2009, Alberta Environment, 2009, Alberta Environment, 2012, ERCB, 2012, CAPP, 2012). Since land changes related to natural resources development are primarily of fine spatial scale, high resolution remote sensing data and associated techniques hold good promise as a cost-efficient approach for this land change monitoring.

The ultimate goal of the research described here is to enhance operational compliance monitoring of bitumen (oil sands) development by provincial regulator agencies. Our approach involves assessing the utility of high resolution earth observation (EO) data to map permanent infrastructures and transient land disturbances associated with resource exploration and extraction at oil sands sites. The former group of includes permanent roads, well pads, pipeline corridors and processing facilities while the latter includes clearings created for core hole drilling, access corridors to core hole sites and seismic lines. Current compliance activities involve site visits by agency personnel. This is both costly and time consuming. There is a need for frequent, synoptic reconnaissance monitoring to identify development areas of specific interest for in-depth ground-based study. Satellite earth observation data is a potential contributor to this solution. Our research therefore involves assessment of the relevant information content of a suite of high resolution (1–10 m) satellite image sources as it relates to the requirements noted above and evaluation of the cost effectiveness of various satellite image sources.

A number of reviews on change detection methodologies can be found in Singh (1989), Macleod and Congalton (1998), Mas (1999), Lunetta and Elvidge (1999), Lu et al. (2004). Change detection has a broad range of applications in remote sensing (Chen et al., 2009, Jensen et al., 1995, Lambin and Ehrlich, 1997, Larsson, 2002, Miller et al., 1998, Petit et al., 2001, Yang and Lo, 2000, Yang and Lo, 2002, Zhang et al., 2002), mostly for the low (AVHRR, MODIS) and moderate resolution satellite data (MSS, TM, ETM+, SPOT before SPOT5). In recent years, high and very high resolution sensors have emerged and are being used by a growing number of user groups. There is a need for modification of conventional change detection techniques, especially in the incorporation of object-based technologies such as image segmentation and classification based on both object spectral and shape attributes (Al-Khudhairy et al., 2005, Blaschke, 2010, Chen et al., 2012, Jyothi et al., 2008, McDermid et al., 2008, Myint et al., 2011, Van der Werff and van der Meer, 2008, Wulder et al., 2008).

There is a limited literature on remote sensing of the Alberta oil sands mining. Efforts have been made to study spectral characteristics of bitumen (e.g. Rivard et al., 2010) and to evaluate the utility of hyper-spectral imaging technology for monitoring purposes (e.g. Lyder et al., 2010). In addition, preliminary assessment has been undertaken of a range of satellite broad-band sensors for land cover land use mapping (e.g. Aronoff et al., 1982, Dean et al., 2007, Gillanders et al., 2008).

Bitumen (oil sands) mining, as with many other resource development activities, involves complex landscape alterations that have important implications for the application of change detection techniques to operational monitoring. Firstly, some disturbances, for example, infrastructure such as road and process facilities as well as pipelines can be considered as permanent change (PC). On the other hand, there will also be disturbances that are transient in the sense that attempts are made, following the disturbance, to return these lands to some form of ‘natural’ state. This latter class of change in turn can be subdivided into two categories. Many disturbances associated with exploration activities involve an abrupt disturbance followed by a period of gradual recovery. Examples of this category include core hole drilling sites, access corridors to reach core sites and seismic lines. We will refer to these as transient short term (TST) disturbances. The other category involves sites that undergo long term human activities. A principal example would be open pit sites where land characteristics undergo continual change during the course of oil sands extraction including removal of overlying soils, exposure and extraction of the sands themselves, creation and use of settling ponds as well as land in-filling prior to re-vegetation activities associated with reclamation. We will refer to these as transient long term (TLT) disturbances.

There are a number of implications of the above observations for the creation of monitoring information products (e.g. time series of change) based on application of change detection methods to multi-temporal earth observation image datasets.

  • (a)

    Conventional image change detection methods typically involving searching for differences between two co-registered images. Long term monitoring requires the integration of changes between temporally consecutive pairs of images. Consistency checking is required to account for possible false alarms.

  • (b)

    Reclamation implies that transient features such as core hole sites may be easily detected when they are initially created. On the other hand in subsequent images, where reclamation is underway, these same sites may appear similar to surrounding, undisturbed land, and will be more difficult to detect. In addition, reclamation monitoring must include spatial context measures, i.e. how similar is a previously disturbed area to its surrounding, undisturbed lands.

The objectives of the research include (a) detection and precise mapping of infrastructure and transient disturbances and (b) assessment of regrowth of vegetation at TST sites. In this work, a change detection methodology is described that has been tailored to the landscape of northern Alberta of Canada and oil sands extraction activities.

Section snippets

Test sites

Two test sites near Fort McMurray, Alberta, Canada, were selected which are representative of the two principal bitumen extraction processes, namely, open pit mining (Kearl Lake site) and in situ extraction (Christina Lake site). Open pit mining is a surface mining technique that is used when oil sand deposits lie near the surface while in situ mining is used for deeper deposits of oil sands. The latter type of mining production is expected to be required for 97% of Alberta's bitumen resources

Data sets

Image data selection for operational monitoring must be based on trade-offs among cost, information potential and expected performance of machine processing algorithms. As noted above, infrastructure and other disturbances related to oil sands development exhibit a range of spatial dimension. On the one hand, while overall site extent can be large, typically tens of kilometres, complex processing facilities tend to be concentrated in much smaller areas. Given the fact that satellite data costs

Methods

There are two principal approaches to scene interpretation in support of change detection, namely, data-driven (e.g. pixel level classification) and object-based (aggregation of contiguous groups of pixels into ‘image objects’ followed by assignment of a physical label to image objects based on object attributes such as grey level, size, shape, context etc.). Data-driven approaches have typically been applied to cases of low spatial resolution imagery, i.e. where resolution is coarse relative

Compatibility assessment of image data

The principal results of application of the above processing tools are illustrated using representative example sub-images of the Christina Lake area. Fig. 4 illustrates (A) a portion of a SPOT5 scene of the area and (B) a visual interpretation of the same area aided by inspection of very high resolution imagery (Worldview2). It should be noted that the interpretation does not include seismic lines that are visible on the imagery as their anticipated width is much less than the sensor

Anthropogenic changes

The image data set available for the Christina Lake site is more limited than is the case for the Kearl site as it includes only two SPOT5 (2006, 2008) images and one RapidEye (2011) image. Fig. 5 shows a sub-area from (A) the SPOT5 2008 image, (B) the RapidEye 2011 and (C) the change between the two image dates. This figure clearly illustrates the small scale nature of most anthropogenic changes in the in situ test site. Although this change product covers a three-year time period, there is

Summary and discussions

Resource extraction efforts from Alberta's oil sands are both extensive in area and of long term duration. Related exploration and exploitation activities involve significant changes to the landscape both in terms of the construction of permanent infrastructure as well as transitory disturbances. Provincial regulatory agencies require reliable geospatial information on these developments to ensure compliance with approved development and reclamation programmes. The principal objective of the

Acknowledgement

The work was supported partially by the GRIP funding from the Canadian Space Agency.

References (38)

  • D.H.A. Al-Khudhairy et al.

    Structural damage assessments from Ikonos data using change detection, object-oriented segmentation, and classification techniques

    Photogrammetric Engineering & Remote Sensing

    (2005)
  • CAPP (Canadian Association of Petroleum Producers)

    Land Use in Canada's Old Sands

    (2012)
  • J. Chen et al.

    Land use land cover change detection using improved Change-Vector Analysis

    Photogrammetric Engineering and Remote Sensing

    (2003)
  • X. Chen et al.

    A spatial geostatistical analysis of impact of land use development on groundwater resources in the Sangong oasis region using remote sensing imagery and data

    Journal of Arid Land

    (2009)
  • G. Chen et al.

    Object-based change detection

    International Journal of Remote Sensing

    (2012)
  • A.M. Dean et al.

    Monitoring sustainable development in the oil sands region of Alberta, Canada, using SPOT-5 and Envisat ASAR and MERIS imagery

  • ERCB (Energy Resources Conservation Board)

    Regulating Unconventional Oils & Gas in Alberta. A Discussion Paper

    (2012)
  • S. Gillanders et al.

    Application of Landsat satellite imagery to monitor land-cover changes at the Athabasca oil sands, Alberta, Canada

    The Canadian Geographer

    (2008)
  • J. Jensen et al.

    Inland wetland change detection in the Everglades Water Conservation Area2A using a time series of normalized remotely sensed data

    Photogrammetric Engineering and Remote Sensing

    (1995)
  • Cited by (11)

    • Estimation of land-use/land-cover changes associated with energy footprints and other disturbance agents in the Upper Peace Region of Alberta Canada from 1985 to 2015 using Landsat data

      2021, International Journal of Applied Earth Observation and Geoinformation
      Citation Excerpt :

      Conversely, non-energy footprints can result in a permanent change in land cover and land use (e.g., urban, transportation corridors, industrial infrastructure). Disturbance related to energy footprints are usually transient and go through reclamation process at the end of their lifecycle (Chowdhury et al., 2017; Zhang et al., 2014). These transitions can take on various forms, and can occur over differing periods, thus necessitating a time series approach and a suitable temporal resolution to capture these intermediate and final (typically reclaimed) stages.

    View all citing articles on Scopus
    View full text