Accounting for spatial autocorrelation is needed to avoid misidentifying trade-offs and bundles among ecosystem services. (October 2021)

• Record Type: Journal Article
• Authors: Shaikh, Shaikh Fairul Edros Ahmad; See, Sin Ching; Richards, Daniel; Belcher, Richard N.; Grêt-Regamey, Adrienne; Galleguillos Torres, Marcelo; Carrasco, L. Roman
• Is Part Of: Ecological indicators. Volume 129(2021)
• Publication Date: 2021-10
• Subjects: Ecosystem services bundles -- Spatial analysis -- Trade-offs -- Synergies -- Nature's contributions to people -- Nature's benefits; Environmental monitoring -- Periodicals; Environmental management -- Periodicals; Environmental impact analysis -- Periodicals; Environmental risk assessment -- Periodicals; Sustainable development -- Periodicals; 333.71405
• DOI: 10.1016/j.ecolind.2021.107992
• Languages: English
• ISSNs: 1470-160X

Abstract

Highlights: We assessed how spatial autocorrelation affects the identification of ES relationships such as trade-offs, synergies and bundles. We determined ES relationships using common methods: correlation, regression and principal component analysis. For each method, we conducted a comparison between one that accounted for spatial autocorrelation with another that did not. Not accounting for spatial autocorrelation is likely to result in misidentification of ES trade-off, synergies and bundles. Abstract: The identification of relationships between multiple ecosystem services (ES) (i.e. trade-offs, synergies and bundles) is essential for ES management. However, the identification of ES relationships may be susceptible to spatial autocorrelation — a statistical bias due to ES observations being related to each other across space. Spatial autocorrelation remains largely overlooked in the literature on ES relationships and its implications are not clear. Here we assess the implication of not accounting for spatial autocorrelation when determining ES relationships using four ES found in the city-state of Singapore. We quantify the ES relationships using some of the most common methods of determining relationships between ES: correlation, regression and principal component analysis. We then compare each method with the corresponding method that accounts for spatial autocorrelation. We found that accounting for spatial autocorrelation resulted in less statistically significant ES relationships, especially at finer resolutions, in correlations (33.3% less significant relationships) and regressions (50% less relationships). Depending on the spatial resolution, different ES were bundled when accounting for spatial autocorrelation when using principal component analysis. Our results suggest that not accounting for spatial autocorrelation in ES relationship studies is likely to result in the misidentification of ES trade-off, synergies and bundles. We thus recommend that future ES relationship studies consider the effects of spatial autocorrelation in their analyses.