The effect of rock fragments on the water retention properties of New Zealand stony soils : A thesis submitted in partial fulfilment of the requirements for the Degree of Doctor of Philosophy at Lincoln University

Abstract

Globally there is increasing evidence of nutrient enrichment and water depletion in surface and groundwater systems due to agricultural practices. To mitigate nutrient enrichment, it is necessary to quantify nutrient discharge within catchments so that robust and effective land management practices and regulations can be developed. To provide this information, several countries and global projects have developed soil databases and information systems to supply maps of soil spatial variability and estimates of crucial soil water retention properties such as field capacity (FC) and available water holding capacity (AWC). These data are used with models to quantify nutrient losses so that nutrient discharge within catchments can be managed. However, when estimating soil water retention properties in stony soils, it is common practice to assume that rock fragments (RFs) have no effect and that FC is the water content (WC) at defined matric potential criteria such as -10 kPa. To test these assumptions and their impacts, I conducted four experiments as part of this PhD. Experiment 1 involved characterising the depth variability of matric potential and WC (fines and RFs independently) at FC in 52 pits excavated to 60 cm depth in stony soils across the Canterbury Plains. For Experiment 2, matric potential after four to five days of drainage following a saturation event was measured to a depth of 1.5 m at five of the sites used in Experiment 1. Experiments 1 and 2 showed that matric potential was generally higher than the default -10 kPa typically assumed for New Zealand soils at FC. The matric potential-depth profile of the pits could be characterised into one of five modes. The most common mode was hydrostatic equilibrium, which generally develops when a shallow (~<2 m) water table establishes a zero matric potential boundary condition near the soil surface. The groundwater tables at all the sites studied (except one) were deep (>2 m), and instead, a coarse sandy gravel layer at ~1 m depth established a near-zero but finite matric potential boundary condition. Very slow unsaturated hydraulic conductivity in this layer allowed the near-zero matric potential to be maintained, above which hydrostatic equilibrium could evolve. This condition, referred to as a capillary break, corresponded to either a layer of open framework gravels or fine earth with a specific surface area <15 m2g-1. Experiment 1 WC results indicated that RFs could influence the fine earth bulk density, porosity, and soil chemistry within an in situ stony soil. RFs could also retain water: 2-20 mm RFs retained twice as much water (0.07 m3 m-3) as >20 mm RFs (0.03 m3 m-3). The water retention of the hard sandstone is low compared to other lithologies, but the volumetric abundance of RFs in the sampled stony soils meant that they accounted for ~10% of the water retained to a depth of 60 cm at FC. The results demonstrate that ~13 mm of water retained by RFs at FC is not currently considered in water budgets and nutrient leaching predictions, which may be relevant to best practice land management. To understand the effect of including or excluding RF water storage on soil water retention predictive models, I developed two pedotransfer functions (PtFs) using data from Experiment 1. Results showed it was possible to accurately predict the WC at FC in stony soils using only explanatory variables that could be easily measured or estimated from a minimalistic field survey. An existing PtF calibrated on NZ soils (the logit PtF), which was constructed on the assumption that RFs had no effect on WC at FC other than reducing the fine earth volume, performed worse than the models developed in this study. By modifying the logit PtF, it was concluded that its poorer performance stems from its inability to account for deviations from 1) the matric potential it assumes for FC (-10 kPa), 2) water held by RFs, and 3) the effect of RFs on the water retention characteristics of the fine earth. The results demonstrate that even the low porosity RFs measured in this study can significantly affect model performance, but by including two variables (depth and volumetric proportion of RFs) that are routinely measured or estimated in most soil sampling projects, it is possible to improve prediction accuracy in established models such as the logit PtF. Experiment 3 required developing a novel repacked soil core experiment to measure the water retention curve (WRC) of low porosity, greywacke RFs. The new method was necessary to account for the low water retention properties of greywacke RFs and the effect RFs have on fine earth porosity and bulk density (which is not considered in most repacked soil studies). A new measurement set up was developed to accomplish this, which allowed the use of large cores with repacked soils that incorporated RFs, glass fragments and fine earth. The method was accurate enough to measure the WRC of the greywacke RFs, which had an AWC of 0.03 ± 0.02 m3 m-3 released between tensions of -10 kPa and -1500 kPa. In an average Canterbury stony soil to a depth of 60 cm, the AWC of just the RFs alone could release 6.4 ± 4.7 mm of water, a significant amount considering most New Zealand stony soils have low AWC. Results from Experiment 1 and 3 were then used in OVERSEER® simulations to determine the effect of the RF plant available water on nutrient loss predictions for a simulated dairy farm in Canterbury. Three soil types were tested (an average stony Brown soil, an average stony Recent soil and a very stony Brown soil) in simulations that included the WC of the RFs, compared to simulations that used only fine earth WC. The inclusion of RF WC had little to no effect on P and GHG losses but could reduce predicted N losses by 1-6 kg N ha-1 yr-1 depending on soil type. Variation in N losses was equivalent to a relative change of 4-19% for the simulated soils, which indicates that farmers on stony soils may be subject to N-leaching overestimates. A caveat to this conclusion is that the OVERSEER® model does not account for bypass flow, a common phenomenon in stony soils. Any N-leaching overestimates indicated by the present research should be treated as a desirable buffer for potential underestimates generated by N-loss processes unaccounted for in the current version of OVERSEER®. Rock fragments are demonstrably able to affect not only the structural and chemical properties of stony soil but also the water retention properties. This project has indicated that the standard assumptions are prone to error when measuring stony soils, namely that: 1. De facto matric potential criteria can define FC in undisturbed stony soils 2. RFs do not retain water 3. RFs do not affect fine earth properties This thesis has shown that current soil information systems and modelling practices regarding stony soils could be inaccurate if these assumptions are made. The work also justifies the need to 1) quantify the WC of RF lithologies at varying states of weathering; 2) explore the generality of findings to stony soils in other sedimentary facies or under differing land uses; 3) develop a method of efficiently identifying the depth of open framework gravels at the paddock and farm-scale and to fully characterise the capillary break, including what conditions and matric potentials it occurs.