Refining the estimates of subsurface nitrate attenuation in New Zealand agricultural landscapes : a thesis presented in partial fulfilment of the requirements for the degree of Doctor of Philosophy in Soil Science at Massey University, Palmerston North, New Zealand

Abstract

Nutrient losses from intensive land use are degrading freshwater quality in agricultural catchments. A robust understanding of nutrient sources, pathways and their potential attenuation is needed to design effective water quality mitigation measures. However, estimates of subsurface nitrate attenuation dynamics have typically relied on sub-catchment-scale mass balance investigations, inferred from the difference between the observed (attenuated) river nitrogen loads and the estimated (unattenuated) root zone nitrogen losses from different land use typologies in a catchment. These nitrogen mass balance calculations yield the nitrogen attenuation factor (𝘈𝘍𝘕), representing the proportion of nitrogen removed during transport processes from the root zone to receiving waters, with estimates often ranging from 0.1 to 0.9 at sub-catchment scales. While the mass balance approach highlights spatial variability in subsurface nitrate attenuation capacities at sub-catchment scale, it obscures modelling and mapping of finer-scale variability of subsurface nitrate attenuation dynamics within a sub-catchment and catchment.

This study developed the Landscape Subsurface Nitrate-Attenuation Index (𝘓𝘚𝘕𝘈𝘐), a novel approach for modelling and mapping spatially variable subsurface nitrate attenuation capacity at landscape unit scale across the Horizons region. The 𝘓𝘚𝘕𝘈𝘐 refines 𝘈𝘍𝘕 estimation from sub-catchments to landscape units, a more realistic scale for representing subsurface nitrate attenuation in agricultural landscapes. The study compiled and analysed a national data set for assessment of spatial and temporal variability of groundwater redox conditions across New Zealand landscapes, and an XGBoost machine learning model was trained to predict subsurface redox probabilities for each landscape unit in the Horizons region. The subsurface redox probability predictions, combined with the rock type and soil drainage information, were classified into five 𝘓𝘚𝘕𝘈𝘐 categories (from 𝘝𝘦𝘳𝘺 𝘭𝘰𝘸 𝘵𝘰 𝘝𝘦𝘳𝘺 𝘩𝘪𝘨𝘩) representing subsurface nitrate attenuation of different landscape units across the region. This 𝘓𝘚𝘕𝘈𝘐 classification was validated with a data set of independent groundwater redox and denitrification measurements in shallow groundwater, and with continuous soil redox probe measurements in a hill country farm, which confirmed that the 𝘓𝘚𝘕𝘈𝘐 classification reflected the observed spatial variability in subsurface redox conditions in the study area.

The 𝘓𝘚𝘕𝘈𝘐 categories were then integrated into sub-catchment scale nitrogen export coefficient modelling (ECM) to parameterise the 𝘓𝘚𝘈𝘕𝘐’𝘴 five categories into a set of calibrated values of 𝘈𝘍𝘕 for the Horizons region’s ~11,000 landscape units. The sub-catchment scale ECM modelling accounted for the root zone nitrogen losses from different land use typologies, any point-source (wastewater) nitrogen discharges, and estimates of river instream nitrogen uptake/attenuation in the study region. The initially assigned 𝘈𝘍𝘕 for the 𝘓𝘚𝘈𝘕𝘐’𝘴 categories (from 𝘝𝘦𝘳𝘺 𝘭𝘰𝘸 𝘵𝘰 𝘝𝘦𝘳𝘺 𝘩𝘪𝘨𝘩) were calibrated by comparing the predicted with the observed average annual loads of soluble inorganic nitrogen (𝘚𝘐𝘕) at 38 river monitoring sites. The calibrated 𝘈𝘍𝘕 values successfully predicted average annual river 𝘚𝘐𝘕 loads in larger sub-catchments (> 500 t 𝘚𝘐𝘕 yr⁻¹), although the errors between the observed and predicted the average annual river 𝘚𝘐𝘕 loads were higher in smaller sub-catchments (< 500 t 𝘚𝘐𝘕 yr⁻¹). The ECM modelling performance improved slightly when the 𝘈𝘍𝘕 values were calibrated for individual catchments, as compared to the whole region. Further research is recommended to improve calibration of 𝘈𝘍𝘕 for the 𝘓𝘚𝘈𝘕𝘐’𝘴 categories in the study region and other similar catchments.

Compared with traditional nitrogen mass balance approaches, the 𝘓𝘚𝘈𝘕𝘐 provided a more consistent framework that better captured underlying spatial heterogeneity and aligned more closely with observed subsurface redox characteristics and landscape attributes across the study region. This research demonstrates that the newly develop approach of 𝘓𝘚𝘈𝘕𝘐 can refine estimates of subsurface nitrate attenuation from sub-catchment to landscape unit scale, improving how spatially variable subsurface nitrate attenuation dynamics are represented when predicting average annual river nitrogen loads. With further refinement, this approach offers a promising basis for targeted and effective evaluation of land use, on-farm and edge-of-field mitigations, and regulatory scenarios, supporting efforts to reduce river nitrogen loads and improve water quality in agricultural catchments.