Spring 2022
A well-developed land classification model can accurately assess the risks associated with variable exposures to soil contamination.
Exposure assessment is a key step in assessing the risk of potential adverse human health effects from environmental hazards. In the case of contaminated soil, an important consideration for exposure potential for a given population is the extent of uncovered or bare soil within an area of interest. Several studies have addressed the impact of ground cover on exposure to soil contaminants (e.g., Dixon et al., 2006). These studies demonstrate that different land cover types such as vegetation (e.g., grass, bushes, trees) or impervious surfaces (e.g., buildings, structures, pavement) are generally associated with less exposure to soil contamination. The importance of land cover as a determinant for potential human health risks is considered by agencies in a regulatory context. For example, both United States Environmental Protection Agency (US EPA) and Centers for Disease Control and Prevention (CDC) recognize that vegetative cover, such as grass, can be used as an interim measure to reduce exposure to lead in soil (US EPA, 2004; CDC, 2014). In addition, US EPA’s definition of a soil-lead hazard is specific to bare soil on residential properties (US EPA, 2014).
The location and extent of different land cover types can be an important factor to consider when conducting both human health and ecological risk assessments in the context of making site remediation decisions. For many conceptual site models, bare soil can be considered a potential source of contamination to surface water via erosion due to overland flow or to groundwater via infiltration. In addition, while impervious surfaces can inhibit infiltration of stormwater into groundwater systems, it can also convey contaminant-laden runoff towards ecological receptors. Direct ingestion of soil is also a potentially relevant pathway for both human and ecological receptors.
Land cover plays an important role in how contaminants move through the environment.”
Furthermore, it is important to consider the variability in the types of land cover when assessing the merits of a class certification, such as in the context of class action litigation conducted under Rule 23, which requires a class to pass a test of commonality, among other requirements. Properties within a large proposed class area can vary significantly based on their uses (e.g., commercial, industrial, residential), their use densities (i.e., the number of properties with a particular use over the area covered by that use), and land cover (i.e., the amount of impervious surface, vegetative, or bare soil cover). This variability can be characterized both among the properties and over time. It is therefore important to accurately and efficiently quantify this variability (both spatially and temporally) to help determine if the proposed class satisfies the requirements of Rule 23(a).
One effective approach to evaluating the land cover for large areas is to leverage aerial and satellite imagery and Geographic Information Systems (GIS) data to develop a land cover classification model. These models are based on the statistical pattern recognition of an image’s pixels. These patterns are tagged to specific land cover types using ground reference information (e.g., assessors’ data, zoning information, ground-based photography) and/or first-hand knowledge of the area. In addition to evaluating variability in potential exposures to contaminants in the environment, a land cover classification model is a powerful tool that when properly developed and validated has many different applications (e.g., quantifying changes in shorelines due to erosion, assessing vegetative health and wildfire risk). Finally, since high quality aerial and satellite imagery dating back several decades is readily obtained through publicly available sources and private vendors, these models can be used to evaluate remote areas where first-hand observation of land cover may be unavailable.
As with any scientific model, it is important to assess the accuracy of a land cover classification model. This typically involves systematically comparing the modeled land cover results to known land cover types using well-distributed and randomly selected ground reference points from the area of interest. In addition, the validity of the model results must be weighed against the quality of the information on which the model is based, as well as the reliability of the assumptions made when developing and conducting the analysis. For example, one should carefully consider the characteristics of the imagery being modeled (e.g., pixel resolution, acquisition methods) and the data used to train the model and confirm the results (e.g., ground reference information). Land cover plays an important role in how contaminants move through the environment. Quantifying the types of land cover and how it varies across different areas using land cover classification models helps scientists assess the risks from contaminants to human and ecological populations.
The authors can be reached at mmayo@gradientcorp.com and cmarsh@gradientcorp.com.