The upper Esopus Creek is located in the Catskill Mountains of New York State and is the primary tributary to the Ashokan Reservoir, with a 192 square mile (mi²) watershed from the source, Winnisook Lake, to the Ashokan Reservoir near Boiceville, New York (Smith and others, 2008; fig. 1). The area of the Ashokan Reservoir watershed is 255 mi², and it is one of six reservoirs in the West-of-Hudson Catskill-Delaware system. Suspended-sediment and turbidity are water-quality concerns in New York City’s water-supply system (NYSDOH, 2017) and have been the subject of extensive monitoring studies in the upper Esopus Creek watershed (McHale and Siemion, 2014; Siemion and others, 2016; NYCDEP, 2017). Numerous stream-management projects have been implemented in the upper Esopus Creek watershed in recent years (Catskill Streams, 2020), and more are planned through at least 2026 to increase stream stability and decrease suspended-sediment and turbidity transport (NYCDEP, 2019).

Suspended sediment is an important component of sediment transport, and turbidity is the constituent of regulation for the New York City drinking-water supply; however, sediment transport also includes the material transported as bedload. Measurements of bedload were not included in previous sediment-monitoring projects in the upper Esopus Creek watershed. Bedload transport, however, is an important design component of stream-channel stabilization projects. Stream-project designs in the upper Esopus Creek watershed have generally used bedload estimates from published sediment-transport equations. The use of bedload-transport equations, however, can be problematic (Gray and others, 2010) and can yield results that are inaccurate (Hinton and others, 2018) because they are often based on data collected in gravel or sand-bed systems. Although there is no evidence to suggest that inaccurate bedload measurements have adversely influenced previous stream projects, the use of inaccurate estimates, particularly if large, may influence project design, cost, and success. Calibration of bedload-transport equations with physical measurements can greatly improve the accuracy of predictive results (Hinton and others, 2018); however, physical measurements of bedload are often difficult and expensive, and may be subject to large temporal and spatial variability (Gray and others, 2010, Itoh and others, 2018).

The U.S. Geological Survey, in cooperation with the Ashokan Watershed Stream Management Program in July 1, 2017–September 30, 2020, conducted a pilot study to determine the feasibility of measuring bedload in the upper Esopus Creek watershed at two existing streamgages on watersheds that include differing geomorphic conditions. The two sampling locations were at Stony Clove Creek at Jansen Road at Lanesville (station 01362336) and at the Lasher Road bridge downstream from Birch Creek at Big Indian (station 013621955). Multiple bedload-measuring techniques were tested. These techniques included the use of traditional bedload samplers, active and passive tracers, and hydrophones to measure the transport of bed material through a range of streamflow conditions.

The objectives of this report are to describe the field-data collection methods used during the study and to summarize the resulting data. Supporting data are available in Siemion and others (2021).

The distribution of bed-sediment particle sizes was estimated by Wolman pebble counts (Wolman, 1954). Two 30-meter longitudinal transects were measured at each monitoring site at the initiation of the project, one immediately upstream and one downstream from the two bedload-sampling locations. Particle sizes were measured by using a gravelometer along ten evenly spaced stream cross sections in each longitudinal transect.

Bedloads were sampled by using the equal-width increment method and a pressure-difference bedload sampler (Edwards and Glysson, 1999). An Elwha bedload sampler with a 203-millimeter (mm) by 102-mm opening was deployed from a type-E truck and crane with a power-adapted B-56 reel. Sampling times, depending on streamflow conditions, ranged from 30 seconds to one minute. Four passes were made across the stream, and 10 equally spaced locations were sampled during each pass. Samples from each pass were combined. The sample from each pass was analyzed individually at the Cascade Volcano Observatory Sediment Laboratory for total-mass and particle-size distribution. With this method, the results from the four passes could be averaged if streamflow conditions were steady or could be used individually if conditions were rapidly changing. Bedload-transport rates were calculated by following methods in Edwards and Glysson (1999).

Passive tracers of bed-material movement were constructed by using radio-frequency identification (RFID) technology that uniquely identified each rock (Olinde and Johnson, 2015; Phillips and Jerolmack, 2014). Rocks from the study reach that were representative of the size distribution found in the channel, but greater than the opening of the Elwha sampler deployed for traditional sampling, were selected. Each rock had a ¼-inch hole drilled in it, a passive RFID tag inserted, and the hole sealed with epoxy. The rocks were then placed in a 1-meter-wide swath across the stream channel at Stony Clove Creek at Jansen Road, and the coordinates of the location were recorded by using a real-time kinematic (RTK) Global Positioning System (GPS). Rocks that were too large to be transported were prepared in place. Following the occurrence of potential bedload-transporting streamflows, a backpack RFID detector was used to locate and identify the tagged rocks, and the coordinates of the new location were recorded again with the RTK GPS. The passive tracers were left in place and relocated after subsequent events.

Ten active tracers of bedload movement were constructed, deployed, and recovered by using methods established in Olinde and Johnson (2015). The tracer containers were produced by injection molding that used a copper-filled alloy at 3.3 grams per cubic centimeter with a B-axis measurement of 175 mm. The containers duplicated the density of sandstone when the hollow interior that housed the electronics was considered. An accelerometer (Onset, 2013) and RFID tag were placed in the center of the tracer container. The accelerometer measured the static acceleration of gravity along three axes, even when the tracer was sitting motionless. After each of the first three surveys, the active tracers were excavated, their data were downloaded, and they were replaced in their observed positions in the streambed. After the fourth survey, the active tracers were excavated, their data were downloaded, and they were replaced in the initial transect. The rationale for the change in the replacement method was that a minimal quantity of useful data was gathered by leaving the active tracers buried in the streambed.

The distance traveled for each active and passive tracer was determined by using the Near and the Locate Feature Along Routes tools in ArcGIS, a geographic information system (Esri, 2019). The tracers were represented as a scattered-point dataset after obtaining the latitudes and longitudes of the tracers locations, all within the stream channel. All tracers were snapped to the center line of the stream by using the Near tool for standardizing the measurement process. The tool had several methods by which to assign a new coordinate pair. The preferred method was geodesic, which relocated the tracers along the shortest possible distance between their original locations and the streamline center. The streamline was converted to a route feature and the Locate Feature Along Routes tool was then used to compute the distances along the streamline once all tracers were snapped to the streamline center. The Locate Feature Along Routes tool calculated the distance along the line on which the tracer was located from the starting point of the streamline. Comparisons of the calculated distances of the tracers for each survey indicated how far the tracers traveled between each survey period.

Four hydrophone-based underwater sound-recording systems (Marineau and others, 2016) were constructed and programmed, and two systems were installed at each monitoring site (one system on each bank). Each system comprised two hydrophones, one liquid-detection sensor, and one recorder. At each study site, two systems were installed to monitor spatial variability in transport and to provide redundancy in data collection in case equipment was damaged by flood debris. The hydrophones were mounted on bridge abutments or boulders on the stream bank. The recorders were programmed to periodically check the sensor to determine if the water level had risen. When water levels were low, the recorders remained in standby mode to preserve power. Once activated, the recorders recorded 1 minute of audio data at 15-minute intervals. The data were stored on a removable memory card. The recorders were serviced periodically to replace batteries and the memory cards.

The data stored on the removable memory card were postprocessed following procedures described in Marineau and others (2019) with a fast-Fourier transform to calculate sound level at various frequency ranges. The sound levels over specific frequency ranges that correlated well with coarse-bedload transport were intended to be averaged together. The acoustic data and physical bedload samples were intended to be used to develop an acoustic-based sediment-rating curve and, if possible, a time series of estimated bedload-transport rates for one or more size ranges.

The results of bed-sediment measurements, traditional bedload measurements, and active and passive tracer surveys are given in the following sections. The hydrophone data are not presented because rock-generated noise was not distinguishable from background streamflow noise at the monitoring sites; instead, the hydrophone results are discussed in context of other recent studies.

Two sets of bedload samples were collected at Birch Creek at Big Indian, New York (station 013621955), and Stony Clove Creek at Jansen Road at Lanesville, New York (station 01362336), during the 2018 water year. One of the sample sets from Stony Clove Creek at Jansen Road was collected during bankfull-streamflow conditions. Measured bedload-transport rates ranged from less than 1 to 37.2 tons per day during the study. Active tracers with a B-axis measurement of 175 millimeter were loosely placed on the streambed and were generally mobilized at streamflows of 130 to 375 cubic feet per second. Passive-tracer data provided a range of peak streamflows that mobilized specific size classes and indicated that particles larger than the opening of the Elwha sampler may have been in motion during sampling. Sediment-generated noise collected in water year 2017 was not distinguishable from background stream noise in hydrophone recordings.