Representing a large region with few sites: The Quality Index approach for field studies

Madeline S Rosamond; Christopher Wellen; Meguel A Yousif; Georgina Kaltenecker; Janis L Thomas; Pamela J Joosse; Natalie C Feisthauer; William D Taylor; Mohamed N Mohamed

doi:10.1016/j.scitotenv.2018.03.113

Representing a large region with few sites: The Quality Index approach for field studies

Sci Total Environ. 2018 Aug 15:633:600-607. doi: 10.1016/j.scitotenv.2018.03.113. Epub 2018 Mar 28.

Authors

Madeline S Rosamond¹, Christopher Wellen², Meguel A Yousif³, Georgina Kaltenecker⁴, Janis L Thomas⁴, Pamela J Joosse⁵, Natalie C Feisthauer⁵, William D Taylor⁶, Mohamed N Mohamed⁴

Affiliations

¹ Department of Biology, University of Waterloo, 200 University Avenue West, Waterloo, Ontario N2L 3G1, Canada. Electronic address: msrosamo@uwaterloo.ca.
² Department of Geography and Environmental Studies, Ryerson University, 350 Victoria Street, Toronto, Ontario M5B 2K3, Canada.
³ Department of Chemical Engineering, University of Waterloo, 200 University Avenue West, Waterloo, Ontario N2L 3G1, Canada.
⁴ Environmental Monitoring and Reporting Branch, Ontario Ministry of the Environment and Climate Change, 125 Resources Rd., Toronto, Ontario M9P 3V6, Canada.
⁵ Science & Technology Branch, Agriculture and Agri-Food Canada, 174 Stone Road West, Guelph, Ontario, N1G 4S9, Canada.
⁶ Department of Biology, University of Waterloo, 200 University Avenue West, Waterloo, Ontario N2L 3G1, Canada.

PMID: 29587229
DOI: 10.1016/j.scitotenv.2018.03.113

Abstract

Many environmental studies require the characterization of a large geographical region using a range of representative sites amenable to intensive study. A systematic approach to selecting study areas can help ensure that an adequate range of the variables of interest is captured. We present a novel method of selecting study sites representing a larger region, in which the region is divided into subregions, which are characterized with relevant independent variables, and displayed in mathematical variable space. Potential study sites are also displayed this way, and selected to cover the range in variables present in the region. The coverage of sites is assessed with the Quality Index, which compares the range and standard deviation of variables among the sites to that of the larger region, and prioritizes sites that are well-distributed (i.e. not clumped) in variable space. We illustrate the method with a case study examining relationships between agricultural land use, physiography and stream phosphorus (P) export, in which we selected several variables representing agricultural P inputs and landscape susceptibility to P loss. A geographic area of 110,000km² was represented with 11 study sites with good coverage of four variables representing agricultural P inputs and transport mechanisms taken from commonly-available geospatial datasets. We use a genetic algorithm to select 11 sites with the highest possible QI and compare these, post-hoc, to our sites. This approach reduces subjectivity in site selection, considers practical constraints and easily allows for site reselection if necessary. This site selection approach can easily be adapted to different landscapes and study goals, as we provide an algorithm and computer code to reproduce our approach elsewhere.

Keywords: Field studies; Optimization; Site selection; Study design.