Sedimentary contamination study in the Great Chao Phraya River watershed, Thailand, using geographically weighted regression
Issued Date
2024-01-01
Resource Type
ISSN
17538947
eISSN
17538955
Scopus ID
2-s2.0-85193813713
Journal Title
International Journal of Digital Earth
Volume
17
Issue
1
Rights Holder(s)
SCOPUS
Bibliographic Citation
International Journal of Digital Earth Vol.17 No.1 (2024)
Suggested Citation
Bhandari R., Xue W., Yu S., Virdis S.G.P., Winijkul E., Tabucanon A.S., Kurniawan T.A., Sriratana M. Sedimentary contamination study in the Great Chao Phraya River watershed, Thailand, using geographically weighted regression. International Journal of Digital Earth Vol.17 No.1 (2024). doi:10.1080/17538947.2024.2356113 Retrieved from: https://repository.li.mahidol.ac.th/handle/20.500.14594/98523
Title
Sedimentary contamination study in the Great Chao Phraya River watershed, Thailand, using geographically weighted regression
Corresponding Author(s)
Other Contributor(s)
Abstract
This research investigated the spatial variations of heavy metals and nutrients in the sediment of the Great Chao Phraya River network of Thailand using geographically weighted regression (GWR). The performance of GWR in explaining the spatiality of selected contaminants using single and multiple predictors was compared and discussed, and a suitable approach was recommended for sediment study. Extensive explanatory variables, including land use type and topographic, socio-economic, and meteorological indicators were utilized in the geospatial regressions of river sediment contaminants at the sub-catchment scale. In the case of simple regression, GWR outperformed the ordinary least square (OLS) regressions with global R2 (range 0.34–0.72 for GWR and 0–0.24 for OLS, respectively), over 2-fold of which was derived by OLS. Similar observations were obtained in multiple regressions, where GWR offered global R2 values between 0.27 and 0.6, higher than those that were provided by OLS (i.e. 0.12–0.44). Furthermore, multiple GWR model presented higher local explanatory power and performance with significantly increased local R2 (up to 51%) compared with simple GWRs (p ≤ 0.05). It is recommended to combine the applications of simple and multiple GWR models to understand the effects of selected explanatory predictors on sediment contaminations from different perspectives.