Applying linear discriminant analysis to predict groundwater redox conditions conducive to denitrification

• Published in: Journal of hydrology (Amsterdam) 2018-01, Vol.556, p.611-624
• Main Authors: Wilson, S.R., Close, M.E., Abraham, P.
• Format: Article
• Language: English
• Subjects: New Zealand; Nitrate; Redox status; Statistical modelling
• ISSN: 0022-1694; 1879-2707
• DOI: 10.1016/j.jhydrol.2017.11.045

•LDA provides insights into the physical conditions associated with denitrification.•Denitrification occurs on coastal plains where poorly drained soils are found.•Other indicators are soil carbon content, water table depth, and sediment type.•Hydrological bypassing of denitrifying areas occurs via artificial drainage. Diffuse nitrate losses from agricultural land pollute groundwater resources worldwide, but can be attenuated under reducing subsurface conditions. In New Zealand, the ability to predict where groundwater denitrification occurs is important for understanding the linkage between land use and discharges of nitrate-bearing groundwater to streams. This study assesses the application of linear discriminant analysis (LDA) for predicting groundwater redox status for Southland, a major dairy farming region in New Zealand. Data cases were developed by assigning a redox status to samples derived from a regional groundwater quality database. Pre-existing regional-scale geospatial databases were used as training variables for the discriminant functions. The predictive accuracy of the discriminant functions was slightly improved by optimising the thresholds between sample depth classes. The models predict 23% of the region as being reducing at shallow depths (