Sampling areas and River Basins
Samples were taken from Friemar (Cambisol-Tschernosem from Loess) and Großenstein (Luvisol from Loess), both in Thuringia (Central Germany) (Fig. 1). The selected samples represent the most common agricultural soil types in this region which contribute considerably to P introduction into surface waters. This allows both clayey (Cambisol-Tschernosem) and sandy (Luvisol) soils to be studied and compared. From each study site, six samples were taken: three topsoil samples with different levels of fertilisation (no fertilisation, optimum and -30% of optimum); two topsoil samples from the area next to a nearby stream (“Hillside” and “Riparian strip”); and one sediment sample from the streambed. All topsoil samples (0–25 cm) were taken randomly at each study site and mixed on site. Samples Soil 1–4 from both study sites are quadruplicates from four sections (a, b, c, and d) of the test fields run by Thuringian State Office for Agriculture and Rural Areas (TLLLR). The samples from study site Großenstein were taken on 23/03/2020 and from study site Friemar on 26/03/2020. The last time the test fields were fertilised before sampling was 08/08/2019 in Friemar and 15/11/2019 in Großenstein (Supplementary material (SM), Table S 1).
Method validation for P was carried out using inductively coupled plasma-optical emission spectroscopy (ICP-OES) SPECTRO ARCOS (SPECTRO Analytical Instruments GmbH, Germany) as total P (SM, Table S 2) and for flow injection analysis (FIA) QuikChem 8500 (Lachat Instruments, USA) as inorganic PO43- (SM, Table S 3). The scope of the validation included the following eight parameters: (1) Selectivity, specificity, (2) Limit of Detection/Quantification, (3) Working range / linear range, (4) Recovery, (5) Sensitivity, (6) Measurement precision, (7) Repeatability/reproducibility, and (8) Robustness. All results are listed in the supplementary material (S 1). The conclusion of the method validation was that the methods used are adequate for the questions to be answered.
Sample Preparation and Analysis
All samples were dried (40 °C), sieved (< 2 mm) (JEHMLICH BSM 53,870, Nossen, Germany), and used for all experiments to ensure comparable conditions (grain sizes and soil/sediment characterisation: Table 1‒Table 3). For standards and working solutions, deionised high purity water (> 18.2 MΩ ∙ cm, UV lamp, pH = 6.0, T = 26 °C) (GenPure, TKA Wasseraufbereitungssysteme GmbH, Germany) was used throughout. Soil composition (Köhn-pipette (DIN ISO 11277:2020–04)) and some basic parameters such as pH (Type: senTix® 81, WTW GmbH, Weilheim, Germany), element composition, and P content (Totals: X-Ray Fluorescence (XRF) spectroscopy from ash, S8 TIGER, Bruker AXS GmbH, Karlsruhe, Germany; Pseudo totals: Aqua regia extraction, ICP-OES) were analysed by TLLLR (Table 1‒Table 2). XRF spectroscopy represents the total element content while the pseudo total element content does not include silicate-bonded P which is indeed not as important from an ecological point of view. Therefore, the term “P-pool” is used for the P pseudo total content in the following text. The plant available P was determined using the Calcium-Acetate-Lactate (CAL) method. In addition, the most likely elements to associate with P (Al, Ca, Fe, and Mn) were measured in all experiments using ICP-OES (Blume et al. 2016; Echterhoff and Meißner 2015; Holtan et al. 1988; Zorn 1998) (SM, Table S 2).
Three different experiments (described in detail in subsections 2.4.1‒2.4.3) were performed to investigate the dynamics of P input processes into receiving waters. In all soil experiments, deionised water (simulating rain) was used, and for the sediment samples synthetic water, representing surface water, such as in rivers/streams, was used (SM, Table S 4). In Thuringia, the water salinity varies greatly, so medium-hard synthetic water was prepared and used for sediment samples. As recommended by Smith et al. (2002), for CaCO3 containing solutions, CO2 gas was bubbled through for ca. 1 h until the poorly soluble CaCO3 precipitate was dissolved. All samples, reference soils, and blanks were extracted in duplicate due to the many samples. Soil 2 from each study site (Friemar and Großenstein) was extracted in quadruplicate to get information about the uncertainty of each experiment. The limits of our laboratory simulations are described in SM S 2.
Simulation of Total P Re-dissolution Following Erosion Caused by Two Heavy Rain Events
The heavy rain event simulation was designed to determine the total water-extractable content of P, simulating the situation following erosion and transport processes from soil to receiving water caused by heavy rain events. A heavy rain event was defined as ≥ 50 mm day-1 based on precipitation data from TLLLR weather stations (SM, Fig. S 1a + b). At both study sites, two heavy rain events were recorded in the months with special risk for erosion (May‒October), in Großenstein from 2005 to 2019 and in Friemar from 1994 to 2019. One event at each study site had intensities between 50 and 60 mm day-1 (2010: Friemar, 2007: Großenstein) and the other between 70 and 80 mm day-1 (2017: Friemar, 2019: Großenstein). To ensure comparability of both study sites, and to simulate extreme events in terms of increasing number and intensity of heavy rain events in future induced by climate change, an extreme value of two heavy rain events per year was chosen for the upscaling process of this simulation. Applied to sediments, it shows the total P re-dissolution from disturbed sediment into water. The experimental set-up is modified from the S 4-method (DIN 38, 414–4:1984–10). The suggested ratio of extract solution and soil (10:1) and extraction time (24 h) described in the S 4-method were used throughout. 4 ± 0.04 g of each sample was weighed into 50 mL centrifuge tubes and 40 ± 0.12 mL deionised/synthetic water was added to each sample. The suspensions were shaken in an overhead shaker ELU (Edmund Bühler GmbH) at 31 rpm for 24 h. Following extraction, the suspensions were centrifuged (4,000 rpm, 10 min; Centrifuge 5810, Eppendorf AG, Germany) and the supernatant decanted into new 50 mL centrifuge tubes. Approximately 37.5 mL was transferred to the new centrifuge tube and 2.5 mL remained, so 375 µL saturated MgSO4 solution (490 g L-1) was added to the supernatant to cause aggregation of particles, e.g. suspended matter. The supernatant was centrifuged again, filled into 13 mL tubes, and measured using ICP-OES and FIA. Additionally, organic carbon (Corg.) (centrifuged) was measured in all samples (multi N/C 2100, Analytik Jena AG, Germany; (DIN EN 1484:2019–04)). The extraction process was repeated three more times as sequential extraction. In addition, a similar laboratory simulation within 24 h was performed using Soil 2 (Großenstein) to determine the dissolution velocity of P, Al, Ca, Fe, and Mn as described previously, with the only difference being that 14 separate 50 mL centrifuge tubes were used and two tubes were sampled at every sampling timepoint. Samples were taken every 1, 2, 3, 4, 6, 8, and 24 h.
Simulation of P Re-dissolution via Diffusion Following Erosion
The P re-dissolution simulation investigated the P re-dissolution process from sediment via diffusion. The experimental set-up was designed to investigate P re-dissolution in lentic areas of waterbodies such as stagnant water areas (SM, Fig. S 2). The P re-dissolution from sediment under fluvial conditions such as in a river was also investigated (SM, Fig. S 3). 80 ± 0.8 g of each sample was placed into 1 L glass beakers, moistened with 30–40 mL deionised/synthetic water and a 90 mm glass fibre round filter (LABSOLUTE, particle retention 1.60 µm) was placed on top. A quartz sand–filled tube and three 1 cm diameter polytetrafluoroethylene (PTFE) balls were used to fix the filter in place. After that, 800 mL deionised/synthetic water was slowly filled into each beaker which was then covered with watch-glasses. A sample of 15 mL each was taken after 1, 2, 3, 4, 5, 6, 7, 10–11, 14–15, 21, and 28 days. One replica from samples Soil 1 and Soil 2 (Friemar) was continued over a period of 49 days. In addition to the main experiment, the diffusion speed of Ca3(PO4)2 was investigated by adding 0.45–0.46 g Ca3(PO4)2, based on the Ca concentration of the optimal fertilisation (Soil 2, Friemar), under 1 to 2 cm layer of quartz sand (131–136 g). The sampled volume was replaced with fresh extraction solution after each sampling and a correction was applied to account for this. The unmoved set-up, designed to simulate stagnant water bodies, was carried out in a dark cabinet to reduce biological activity. The moved set-up, designed to simulate flowing water bodies, was placed on a shaking machine SM 30 CONTROL (Edmund Bühler GmbH, 31 rpm). The centrifuged extracts (4,000 rpm, 10 min) were measured using ICP-OES and FIA. In addition, the redox potential (senTix® ORP900), oxygen concentration (FDO® 925), and conductivity (TetraCon® 325) were measured using electrodes (all: WTW GmbH, Germany).
Simulation of P Re-dissolution from Interstitial Sites
The percolation simulation shows the P leaching process of rain with respect to the flow of rain through the soil interstitial sites. This simulation was also applied to sediment samples, where it simulates the flow of surface water through the interstitial sites of a streambed or river sediment. Disturbed topsoils and sediments were used for this simulation, which indeed does not describe the original pore system but ensures direct comparability to the previous simulations to evaluate the P input due to re-dissolution from interstitial sites. The experimental set-up (SM, Fig. S 4) was as follows: A 25 mm glass fibre filter (1.0 µm particle retention rate, GE Water & Process Technologies; e.g. available at Fisher Scientific GmbH (2021)) was put into a 50 mL syringe and a small ceramic sieve with fine holes was placed on top of the filter for a constant flow. 65 \(\pm\) 0.65 g of each sample was added and then filled up to 1.5 cm from the top with quartz sand (1–2 cm layer) to avoid dead-volume. Another 25 mm glass fibre filter was added and the top of the syringe was closed with a plunger and silicon plug, both with a hole in the centre. Each plug was fixed using two cable ties. A tube connected the syringe to the volumetric flask for sampling and de-ionised/synthetic river water was pumped from the bottom to the top of the syringe. The main experiment (SM, Fig. S 5) was run for 28 days using two peristaltic pumps (type: 205U and 302SL, Watson-Marlow GmbH, 2.5 rpm). One replicate of Soils 1 and 2 from both study sites was continued over a period of 153 days. The percolation volume flow rate was calculated (SM, Eq. S 1) using the soil hydraulic conductivity data (10–40 cm day-1) for a comparable soil type (Eckelmann et al. 2005). Therefore, the percolation simulation was carried out using volume flow rates from 66 to 118 mL day-1. For the first 14 days, a sample was taken daily from 500 mL measuring flasks and the percolated volume noted. From day 15, a sample was taken every 2 days. The extracts were measured using the ICP-OES and FIA. In addition, dissolved Corg. (< 1.0 µm) and pH of the extracts from the first week were measured.
Upscaling to a Realistic Scenario
Upscaling from the laboratory simulations to a realistic scenario illustrates the P input processes from soil/sediment into receiving waters mainly via erosion (SM, Fig. S 6). It predicts a realistic scenario and shows the order of magnitude of the P input processes. Each experiment mimics a specific scenario which is possible in nature. The total Pload per year from soils/sediments (kg P (ha a)-1) into receiving waters at each site was calculated based on extracted P concentrations from each experiment. The Pshift, simulating water percolation through the soil interstitial sites, was extrapolated to real leaching of P into deeper regions of the soil beneath the scaling up the experimental conditions to reality using lysimeter data from TLLLR. Oehl et al. (2002) observed that 5‒11 kg P ha-1 a-1 were shifted from topsoils (0‒20 cm) to deeper subsurface layers (30‒50 cm), highlighting the need to consider shifted P. Khan et al. (2021) came to the same conclusion: available P-pools are translocated and accumulated to deeper layers in the soil profile. In addition, the resulting P concentrations (mg L-1) in the nearby surface water (river/stream) caused by soil erosion were calculated and compared to real water P concentrations measured by Thüringer Landesanstalt für Umwelt und Geologie (TLUG) (2014) to determine if the calculated P concentrations were within the real existing range. If the real values are much higher, this may be due to other point sources such as sewage waste and the P input through atmospheric deposition. They were further compared to the LAWA orientation values for total P (≤ 0.10 mg L-1) (Bund/Länderarbeitsgemeinschaft Wasser (LAWA) 2015) to judge environmental impact and evaluate eutrophication risk. The calculations used are listed in SM, Eq. S 2a–S 4c and all values reported in SM, Table S 5‒6.