*3.2. Determination of Total Phosphorus in Water Samples*

In the laboratory, total phosphorus (TP) was determined according to DIN EN ISO 6878: 2004-09 (2019). Prior to analysis, water samples were unfrozen, brought to an ambient temperature, and mixed with 1 ml 17.97 M sulfuric acid (H2SO4) to adjust pH to ~1. Samples were then oxidized with potassium persulfate (K2S2O8), while heating to 90–100 ◦C for 90 minutes in closed vessels [49]. Afterwards, extracts were filtered with ashless blue ribbon filters (2 μm pore size). The extracts' concentration of phosphate was determined on a spectrophotometer (Genesys 10S; Thermo Fisher Scientific; Bremen, Germany) via the molybdenum-blue method at 700 nm [50,51]. Samples were measured three times and averaged. Phosphate concentrations were arithmetically converted into mg TP/l.

We calculated the relative standard deviations of the method (RSDM) and the detection limits for colorimetric TP measurement [52–54]. Data below the detection limit were excluded from data evaluation. In our 14 measurements, the maximum RSDM was 4.07%. Hence, we interpret our TP concentrations with a rounded uncertainty range of ± 5.0%.

#### *3.3. Sediment Analyses*

We took sediment samples from the water courses in the Antrift catchment to assess potential correlations between water quality parameters and sediment features, especially with regard to P mobilization from sediments. The sediment sampling sites correspond to the water sampling sites (Figure 1). Sediments were sampled about 5–10 cm deep from the bottom of the rivers and the reservoir with a plastic receptacle attached to an extensible stick. The first sampling date (5 November 2018) was in the end phase of a summerly algae bloom after a prolonged period with warm, dry weather and low flow conditions (average in the week before sampling: 0.22 m3/s). The second sampling date (15 March 2019) was after an expected regeneration of water quality during winter. Even though the winter of 2018/2019 was relatively mild and dry, some precipitation occurred prior to our second sampling. Thus, the second date marks relatively moist weather and higher flow conditions (average in the week before sampling: 1.03 m3/s). Sediment samples were stored airtight in plastic bags for about one day until further processing in the laboratory. The sediments were then dried at 70 ◦C in a drying furnace for eight days. Afterwards, they were ground in a mortar and sieved (2 mm mesh).

We determined the content of organic matter (OM; via loss on ignition; DIN 19684–3:2000–08) and pH (with 0.01 M CaCl2; m:V = 1:2.5; DIN ISO 10390:1997-05) of the prepared samples. Furthermore, we assessed the sediments' texture with the integral suspension pressure method [55] after samples were prepared according to DIN ISO 11277:2002–08. We used three aliquots of each sediment sample for a P fractionation. We determined the following: (1) easily soluble P with 0.1 M hydrochloric acid (PdHCl) [56]); (2) pedogenic oxide-bound P soluble in an ammonium oxalate–oxalic acid solution (Pox; DIN 19684–6:1997–12 [56]); and (3) pseudo-total P (PAR) after extraction with *aqua regia* (12.1 M HCl and 14.4 M HNO3 in a ratio of 1:3; [56]). PdHCl was measured on the spectrophotometer according to Murphy and Riley (1962) [48]. Pox and PAR were quantified with an ICP-MS (X Series 2; Thermo Fisher Scientific; Bremen, Germany), as well as Feox, Alox, Mnox, FeAR, AlAR, MnAR, NaAR, MgAR, KAR, and CaAR. All data were converted into the unit mg element/kg. The results of the three aliquots were averaged for further evaluation.

We calculated RSDM and the detection limits for ICP measurement of each element [52–54]. Data below the detection limits were excluded from evaluation. Data measured with a relative standard deviation (RSD) of ≥20% were also excluded [57,58]. To quantify measurement uncertainty, we added RSDM (as a standard parameter of calibration-based measurement) with RSD (as a parameter depicting data reproduction, and reflecting effects of heterogeneous matrixes typical for environmental samples [59]). As our sediments were measured in one calibration, only one RSDM resulted for each element. From the several RSD calculated during this measurement, we chose the median RSD for each element, because it is insensitive to outliers. In sum, we calculated rounded-up measurement uncertainty ranges of ± 2.0% (Pox, PAR, Alox, AlAR, Feox, FeAR, CaAR, KAR, NaAR, MgAR), ± 3.0% (Mnox, MnAR), and ± 9.0% (PdHCl).

#### *3.4. Discharge and Precipitation Datasets*

At the inlet of the Antrift reservoir, there is a gauging station operated by the Wasserverband Schwalm e.V. (Schwalmstadt, Germany), which measured daily means of flow (m3/s) during our investigation period. Because there is no climate station in the Antrift catchment, average daily precipitation and air temperature were calculated based on data from three climate stations close to the catchment (7–16 km off). These stations were Neustadt (Station-ID: DWD3516, 257 m a.s.l.), Alsfeld-Eifa (Station-ID: DWD91, 300 m a.s.l.), and Meiches (Station-ID: HLNUG4288360, 467 m a.s.l.). The stations are operated by the German Weather Service (DWD) and the Hessian Agency for Nature Conservation, Environment and Geology (HLNUG) [60,61]. All climate data are freely available online.

#### *3.5. Statistical Analyses*

Basic statistical operations were performed in Microsoft Excel 2013 (Microsoft; Redmond, WA, USA), in R (R Core Team, 2013) and RStudio (Version 1.1.447; RStudio Inc.; Boston, MA, USA). Data visualization, tests for normal distribution (Shapiro–Wilk test), and Spearman correlation analyses were conducted with the R-packages "graphics", "stats", and "corrplot" [62,63]. Significances were tested on different levels. We interpret significant (*p* ≤ 0.05) correlation coefficients as: weak (rSP 0.4–<0.6), clear (rSP 0.6–<0.8), and strong (rSP ≥ 0.8) [64]. For the sediment data, we performed Spearman correlation analyses for all our data, as well as separately for the data of each sampling date.

Each of our P fractions was determined for a new aliquot (i.e., 1.0 g) of the respective sediment sample. Hence, the resulting P contents are cumulative and contain P forms of different solubility (i.e., Pox includes PdHCl, and PAR includes Pox). To deduce the reactive behavior of sediment P, we calculated differential P fractions, which represent only one class of similar P solubility. Next to easily soluble PdHCl, we calculated moderately labile Pml (Pox minus PdHCl), and recalcitrant Prc (PAR minus Pox).

To depict tendencies in the reactive behavior of sediment P, we determined the degrees of P mobilization (DPM). These ratios between two differently soluble P fractions elucidate whether the sediment has a tendency for P mobilization (large DPM) or P bonding (small DPM [59]). We calculated DPM2 via PdHCl: Pox, and DPM3 via Pox: PAR.

The poor ecological status of the water courses of the Antrift catchment is largely attributed to erosion and overland flow [65–67]. Hence, there should be P inputs into the surface waters shortly (i.e., maximally within a few hours) after precipitation events with sufficient intensity [68–71]. Thus, we used our limited dataset to seek an indication of temporal relationships between TP concentrations in the water and the occurrence of precipitation events. We applied a three-step data evaluation: (1) we tested for similarities between the time series of precipitation and discharge, as a function of the increments of precipitation relative to discharge ("cross correlation"; *p* ≤ 0.05) [62], to find out if there is a temporal connection between a precipitation event and a discharge event, and to elucidate how quickly the first affects the later. (2) Next, we performed a Spearman correlation analysis between TP concentrations in the water and the number of days between each precipitation event and the next sampling date. For this, rainfall events were classified according to the daily precipitation sum into heavy (>10 mm/day), and maximal precipitation (maximal daily precipitation sum during each three-week measurement period). Moreover, we determined "higher precipitation" for events with a daily precipitation sum between the third quartile and the maximum of each three-week measurement period. Because we wanted to investigate the (short-term) effects of precipitation on erosion and P loss, we picked the "higher precipitation" event that occurred closest to our measuring date. (3) Finally, we checked the absolute number of days between each precipitation event and the next sampling date to evaluate the correlation results and exclude all events which are backdated more than two days. This two-day threshold was chosen according to the results of step (1) and the normal duration of erosion or overland flow events [68,69].

#### **4. Results**

#### *4.1. Evaluation of Climate Data: Discharge and Precipitation*

According to data from the climate station Alsfeld-Eifa (ca. 7 km northeast of the Antrift reservoir), the last six years show a positive temperature deviation between 0.29 ◦C and 1.30 ◦C from the reference period 1991–2020 (Figure 2) [72]. Between 2009 and 2019, the highest positive temperature increase was documented in 2018. A strong negative deviation in annual precipitation sums was reported. Precipitation was ca. 19% lower in 2018 (−135.75 mm), and ca. 10.5% lower in 2019 than in the reference period (1991–2020).

**Figure 2.** Temperature (**a**) and precipitation trend (**b**) in the Antrift catchment (2009–2019). (Data source: HLNUG (2019); operator: Deutscher Wetterdienst, data for the climate station Alsfeld-Eifa, station-ID: DWD91, 50.7447◦ N, 9.345◦ E.).

Between July and November 2018, monthly precipitation sums were lower by between 17.8% (August) and 38.3% (November), compared with the long-term (1979–2020) average monthly precipitation sums (Figure 3). Except for March and May 2019, monthly precipitation sums were below the long-term average precipitation sum all through our field measurement and sampling campaign. The clear increase of the precipitation sum in May 2019 (74.7% higher than in April 2019) resulted from a thunderstorm event with heavy rainfall on 20 May 2019 (Figure 4) [73]. For that day, a 24-h precipitation sum of 45.21 mm was calculated on the basis of data from three regional climate stations (see Materials and Methods). During our investigation period, eleven events with precipitation of >10 mm per day occurred.

**Figure 3.** Monthly precipitation sums in the Antrift catchment during the investigation period, compared to long-term average monthly precipitation (reference period 1991–2020). (Data source: HLNUG (2019); operator: Deutscher Wetterdienst, data for the climate station Alsfeld-Eifa, station-ID: DWD91, 50.7447◦ N, 9.345◦ E.).

Discharge remained below its long-term average (MQ), with the exception of a few increases in discharge during the winter of 2018/2019, and as a result of the precipitation events previously mentioned (Figure 4). Discharge was low from late spring to summer 2018 and 2019 (data not shown). Except for the weeks with higher rainfall, daily discharge declined below the long-term average of lowest discharge (MNQ).

**Figure 4.** Monthly precipitation and discharge trend of 2018 and 2019 in the Antrift catchment, compared to average discharge in the reference period 1991–2000 (MQ = average discharge; MNQ = average low-flow discharge). Data source for discharge: Wasserverband Schwalm e.V. (2019), water level station 42881009, 50.76245◦ N, 9.20654◦ E. Precipitation data triangulated from HLNUG (2019): climate stations DWD3561 (50.8494◦ N, 9.1253◦ E), DWD91 (50.7447◦ N, 9.345◦ E), and HLNUG4288360 (50.6281◦ N, 9.26143◦ E).
