Contrasting physical controls on phosphorus transport to shallow 1 groundwater at the hillslope scale 2

Abstract. In well-drained agricultural catchments transport of phosphorus (P) to groundwater (GW) can be controlled by static and dynamic factors and where surface water is GW fed this can lead to elevated P concentrations at the catchment outlet. In order to better control P transport along hillslopes a spatial and temporal conceptual view of P loss to GW must be developed. Initially in the present study, hillslope GW quality and rainfall data were examined for 2017 utilising a transect of piezometers at upslope (US), midslope (MS) and downslope (DS) locations. Two dominant scenarios emerged where GW P concentrations at DS and MS were simultaneously low or at other times DS became elevated and MS remained low. To examine the potential reasons for such scenarios, a one-dimensional hydrological transport model was developed for the unsaturated zone at DS and MS using rainfall and depth specific soil physical and hydraulic data. Results indicated that the DS zone facilitated transport (higher sand content, soil saturated hydraulic conductivity (Ks) and lower soil compaction) with higher modelled concentration peaks towards higher GW P concentrations whereas the MS zone had more potential to attenuate transport (lower soil Ks and higher soil compaction). Moreover, inter-annual variations of GW P concentrations at DS were related to rainfall and GW level. Hence, mitigation strategies should particularly (but not exclusively) focus on reducing P sources in the DS zone. This also indicates a need to identify hotspots of facilitated transport to shallow GW using finer scale soil properties surveys. Here, this is defined by low soil compaction, high sand content and soil Ks. However, challenges arise as soil properties can vary in time with soil management and with the difficulty of assessing the transport potential of deeper soil.



Field methods -meteorological and soil data
For the purposes of the present study meteorological data taken from a Campbell Scientific 151 BWS-200 weather station ( Fig. 1) from January 2017 to December 2017 were examined.

152
Absence of rainfall for at least 12 hours was used to separate one rainfall event from another 153 https://doi.org/10.5194/hess-2020-248 Preprint. Discussion started: 21 July 2020 c Author(s) 2020. CC BY 4.0 License. depending on the total rainfall amount (A = 5.0-9.9 mm, B = 10.0-19.9 mm, C = 20.0-29.9 156 mm, D = 30.0-39.9 mm, E = ≥40 mm). Using the hybrid soil moisture deficit (SMD) model of resulting breakthrough curves were used to assess P hydrological transport to GW at DS and 203 MS (Fig. 3).   [%] data from 2017 were used as input parameters. Solute dispersivity was set at 10 % of soil 225 profile depth (Fetter, 2008;Šimůnek et al., 2013). A conservative solute was used in order to 226 examine the role of soil hydraulic properties on the potential for P transport to GW. Thus, no 227 soil chemical input data were used in the models and chemical P attenuation processes in soil 228 are not considered here. Conservative solute initial concentration at the soil surface was set at 229 10 mmol cm −1 and 1 cm precipitation was applied with no evaporation, in order to initiate

234
For objective 1, analysis of variance (ANOVAs) was used to investigate significant (P < 0.05) 235 difference of soil properties between depths within each site and between sites for each depth.

236
Residuals plots were used to assess the normal distribution of the residuals and the equal 237 variance of the data; data were log transformed before statistical analyses when those 238 conditions were not met. Trends were studied when the variation between replicates was very 239 high (e.g. K s ). Pearson r correlations were used to measure the degree of relationship between 240 soil parameters. Statistical analysis was carried out using R Studio 3.5.2.  Table S3 and Table S4, respectively. Below is a description of the 246 overall (at the scale of the sampling area, including the four replicates) variations observed 247 between sites and depths. The SWRC shape parameters α 1 and α 2 , n 1 and n 2 , ω 2 , as well as θ s 248 and θ r are not presented as they are not considered to be the main parameters controlling 249 hydrological transport to GW. A detailed description of hydraulic parameters is presented in 250 S5. Soil at DS is a Sandy Loam whereas MS soil has a Loamy texture.

252
Average soil ρ b was higher (not significantly) at MS than DS for both shallow and deeper soil 253 cores. Average soil ρ b increased with depth (not significantly) in each site: from 0.89 ± 0.05 g 254 cm -3 to 0.95 ± 0.03 g cm -3 at DS, and from 1.20 ± 0.05 g cm -3 to 1.27 ± 0.05 g cm -3 at MS. Soil 255 organic matter (OM %) was higher at DS (8.3 %) than at MS (4.6 %).

257
At both sites and for both depths, soil K s were variable. Average K s was higher (not 258 significantly) at MS than DS for both shallow and deeper soil cores. Average K s decreased 259 with depth (not significantly) at each site: from 1 648 ± 791 to 829 ± 600 cm d -1 at DS, and 260 from 2 981 ± 1 417 to 2 242 ± 1 248 cm d -1 at MS.

262
Average ϕ was higher (not significantly) at DS than MS for both shallow and deeper soil 263 cores. Average ϕ decreased with depth (not significantly) at each site: from 66 ± 2 % to 64 ± 1 264 % at DS, and from 54 ± 2 % to 51 ± 2 % at MS. Average ε was higher (not significantly) at 265 DS than MS for both shallow and deeper soil cores. Average ε increased with depth (not 266 significantly) at each site: from 22 ± 5 % to 24 ± 2 % at DS, and from 14 ± 1 % to 19 ± 2 % 267 at MS.    Over the year 2017, concentrations in TDP were higher at DS than at MS with a higher 332 variability in concentrations at DS (Fig. 4a). In particular, TDP concentrations at DS were in Figure 6 for each rainfall event R1 [B; long duration with low total rainfall] (Fig. 6a), R2

339
[D; short duration with high total rainfall] (Fig. 6b) and R3 [E; long duration with high total 340 rainfall] (Fig. 6c). Tracer first and last occurrences, concentration peak and total transport 341 duration are shown in Table 3.

373
This study highlighted the spatial variability of P hydrological transport through the soil 374 profile to GW within a hillslope of contrasting GW P concentrations, and examined the inter-375 annual variability of GW P concentrations. A range of modelled hydraulic properties and 376 transport capacities were identified to 1) determine static soil hydraulic properties controlling 377 hydrological transport to GW along the hillslope, 2) examine variations in GW P 378 concentrations in relation to dynamic physical controls and 3) reveal contrasting physical 379 controls on the potential for P transport to GW at the hillslope scale. The combined analysis 380 of meteorological data, high resolution soil physical/hydraulic data and GW chemical data 381 revealed contrasting spatial and temporal P hydrological transport potential to GW along the 382 hillslope in relation to the existence of a static system (soil) and a dynamic system (rainfall, 383 GWL, soil moisture), respectively. The DS zone showed a higher hydrological transport