Geostatistical analysis of the integration of spatial information on nitrate-N observations and surface land uses for identifying groundwater pollution zones

Abstract

Groundwater nitrate-N pollution mainly originates from surface agricultural activities. Integrating spatial information on nitrate-N observations and agricultural land-use data is crucial for identifying groundwater pollution zones. This study used regression kriging (RK) to determine groundwater pollution zones in the Choushui River alluvial fan in Taiwan according to nitrate-N observations and agricultural land uses. Areal ratios of agricultural land-use types within buffering zones were first characterized using geographical information systems. A multivariate linear regression (MLR) model was employed to explore the relationship between groundwater nitrate-N pollution and agricultural land-use types. Then, simple kriging (SK) was adopted to analyze residuals obtained from gaps between nitrate-N observations and MLR predictions; the SK estimates of the residuals with the addition of the MLR predictions served as the RK estimates for groundwater nitrate-N pollution. Finally, groundwater pollution zones were determined according to a specific anthropogenic nitrate-N pollution level. The study results revealed that the “orchard” land-use type positively contributed to groundwater nitrate-N in contract to the “livestock house” and “agricultural facility” land-use types, which were negatively related to groundwater nitrate-N. Moreover, the RK estimates had the ability to characterize the potential pollution source of the orchard land-use type and were suitable for identifying groundwater pollution zones. Therefore, the amount of fertilizer used in the orchards located in groundwater pollution zones must be reduced considerably.