**2. Materials and Methods**

#### *2.1. Site Description, Streambank Sampling for Legacy Sediments, and Sediment Analysis*

Detailed description of the streambank sampling sites, coordinates, and methods is provided in [21,22,33]. Briefly, legacy sediment sampling was performed at 15 streambank sites across five streams in northern Delaware (DE), eastern Maryland (MD), and southeast Pennsylvania (PA) (Figure 2). These included Big Elk Creek and its first order tributary Gramies Run and Christina River and its two major tributaries the White Clay Creek and the Brandywine Creek. Big Elk Creek with a drainage area of 205 km<sup>2</sup> (empties into the Chesapeake Bay) and Christina River with a drainage area of 1463 km2 (drains into Delaware Bay) straddle the fall line with upper portions of the watersheds extending into the Piedmont and Appalachian regions and the lower portions in the Coastal Plain. The drainage areas for Gramies Run, White Clay Creek, and the Brandywine Creek were 8 km2, 277 km2, and 854 km2, respectively [22].

**Figure 2.** Map showing locations of 15 stream bank sampling sites for legacy sediments across Delaware, Maryland, and Pennsylvania. These sites included: BEB = Big Elk Bridge (Agriculture), CB = Camp Bonsul (Agriculture), NCB = Nature Center Beach (Agriculture), SM2 = Scott's Mill 2 (Agriculture), SM3 = Scott's Mill 3 (Agriculture), TM = Tweed's Mill (Agriculture), RH = Cottage Mill (Suburban), CM = Casho Mill (Suburban), CDM = Cider Mill (Suburban), BZ = Brandywine Zoo (Urban), BYR = Byrnes Mill (Urban), COB = Cooch's Bridge (Urban), GMT = Gramies Run (Forested) and MR = Middle Run (Forested). Inset: Location of study sites in the mid-Atlantic tristate area.

All of the streambank legacy sediment sampling sites were located upstream of formerly breached or existing milldam locations. These sites spanned four different contemporary land use and land covers (LULC) – urban, suburban, forested, and agriculture (see Figure 1 and Table 1 in [22]). Streambank samples were collected for multiple depths (recorded from the top; Table 1 in [22]) by scraping off the surface sediment and collecting a sample using an auger or a hand trowel (where the sediments were too hard to auger). All samples were placed in sterile Ziploc bags and put on ice until they were brought back to the lab. A total of 67 sediment samples were collected across all 15 sites and all sediment sampling was performed in October–November, 2017. Sediments were ground with a mortar and pestle and sieved through a 2-mm mesh to remove small rocks and organic matter. Sediments were then sieved into coarse (>63 μm) and fine (<63 μm) fractions using an RX-29 RoTap®sieve shaker. Percent sand, silt, and clay for the samples were also determined using Beckmann Coulter LS 13 320 Particle Size Analyzer ®(Indianapolis, IN) [22].

Sediment samples were analyzed for Mehlich-3 [27] extractable elements (M3P, M3Fe, and M3Al) and microwave digestion (EPA method 3051) for total P. Using M3P, M3Al, and M3Fe values, the %DPS was computed using the equation:

$$\% \text{DPS} = \left(\frac{\left[\left\{\frac{\text{M3P}}{(\text{M3Al} + \text{M3Fe})}\right\} + 0.019\right]}{0.0042}\right) \times 100$$

where all the M3 extractable values are in molar concentrations [27]. Representative stream water samples were also collected at the time of bank sediment sampling at all 15 sites to compare stream water PO4 <sup>3</sup><sup>−</sup> against sediment EPC0 values (to assess source-sink behavior). In addition, stream water PO4 <sup>3</sup><sup>−</sup> data at bi-monthly intervals was also available for Big Elk Creek sampling site (BEB) for a temporal comparison with EPC0 [28].

#### *2.2. Phosphorus Sorption Index (PSI)*

The intent of this experiment was to determine the maximum sorption capacity of streambank legacy sediments and compare them against literature values for other sediments. Sediments from all 15 streambank sites and depths (total samples = 67) were used. Fine and coarse fractions were replicated twice for a total of 268 samples (67 samples × 2 size fractions × 2 replicates). Soils were treated with a 75 mg P L−<sup>1</sup> solution created by dissolving 0.3295 g of monobasic potassium phosphate (KH2PO4) in 1L of deionized water following the protocol of [34] (pages 20-21) based on [35]. About one gram of sediment (the exact amount was recorded) was placed in a 50 mL centrifuge tube along with 20 mL of the 75 mg P L−<sup>1</sup> sorption solution. This provides a ratio of added P to soil of 1.5 g P kg−<sup>1</sup> soil. Two drops of chloroform were added to each solution to kill and inhibit microbial activity. Tubes were placed in an end-over-end shaker and shaken for 18 h. After 18 h, samples were centrifuged at 2000 rpm for 30 min and then filtered through 0.45 μm filters using a Millipore filtration unit into 40 mL amber glass vials. The filtered solution samples were analyzed for PO4 <sup>3</sup><sup>−</sup> (mg P L−1) using EPA-118-A Rev 5 method on an AQ2 Discrete Analyzer (Seal Analytical, Mequon, Wisconsin). The PSI (mg kg<sup>−</sup>1) value was determined using the equation:

$$\text{PSI} = \frac{\left(\text{75mg} \frac{P}{L} - \text{C}\right) \times \left(0.020 L\right)}{\left(0.001 \text{ kg soil}\right)}$$

where *C* is the solution equilibrium P concentration after 18 h (mg L<sup>−</sup>1). Differences in PSI values with particle size class (fine versus coarse fraction) were determined using t tests. Pearson correlations (r) were determined to investigate relationships between PSI values and Mehlich-3 extracted Al and Fe contents of legacy sediments (M3Fe and M3Al). All statistical analyses were performed in JMP (Version 14.0).
