Mitigating the impacts of agricultural nutrients (nitrogen and phosphorus) on water quality requires a clear understanding of their transport pathways and transformation processes from land to receiving waters. For nitrate, which is subject to subsurface denitrification, it is therefore important to assess the spatial variability and temporal stability of groundwater redox conditions, as nitrate reduction typically occurs in reducing conditions. This paper presents a robust assessment of a large groundwater quality data set collected across New Zealand landscapes, develops methods to impute missing groundwater redox-sensitive variables and characterises the spatial variability and temporal stability of groundwater redox conditions against relevant landscape hydrogeochemical characteristics. Random forest and extreme gradient boosting (XGBoost) outperformed linear regression in predicting missing Mn2+ values, achieving higher accuracy (R2 > 0.8) and lower error (RMSE < 0.2mg/L). Analysis of groundwater redox conditions highlights considerable spatial variability, particularly influenced by subsurface geology (rock types) and soil characteristics such as soil carbon and drainage across various hydrogeological settings. Our findings also reveal a higher prevalence of oxidised redox status in shallower groundwater and greater temporal stability in oxidised conditions across New Zealand landscapes. These insights have significant implications for targeted management strategies to reduce nitrate losses from farming activities, particularly in oxidised, shallow groundwater across different hydrogeological land units.
Read full abstract