Field Site Field experiments were performed to examine arsenic transport in irrigation water flowing through constructed distribution channels. The work was conducted at a well-characterized field site near Sreenagar in the Munshiganj province, approximately 30 km south of Dhaka, Bangladesh (Harvey et al., 2002; Neumann et al., 2009). The site is located approximately 7 km from the Ganges River and has typical rice paddy soils and management practices of the region (Dittmar et al., 2007). During the dry season, a network of soil-based distribution channels is used to convey irrigation water from a central well to 11 individual rice fields (Roberts et al., 2007). Arsenic concentrations of well water at the site are ~400 mg L-1 (Harvey et al., 2002; Polizzotto et al., 2013), and heterogeneous arsenic removal from irrigation water has been observed over space and time in rice fields (Roberts et al., 2007; Dittmar et al., 2007; Polizzotto et al., 2013). Experimental Design Channel flow experiments were conducted in December 2011, after the recession of floodwaters but before planting of the Boro rice crop. Within experimental channels, samples were obtained from flowing irrigation water to quantify the impacts of channel physical properties on arsenic removal from solution. Four sets of experiments were run: control channels (45 m long, 45 cm wide), soil-free channels (continuous plastic tarp covering the soil), channels with increased widths (2× and 3× width channels), and channels with increased length (200-m-length channels) (Fig. 1; Supplemental Fig. S1). These experiments altered residence times, water depths, flow lengths, and soil–water contact. All experiments were run in triplicate except the soil-free channels, which were run in duplicate. Volumetric flow rates for all experiments, as measured from good discharge, averaged 18.4 L s-1 and ranged from 16.0 to 20.5 L s-1 . Channel Construction Five irrigation channels were initially constructed in a bare field for use in the experiments. Each experimental channel was linked to an existing manifold channel that brought irrigation water from the central tubewell. Channels were constructed by hand, mounding and shaping mucky soil into channel walls, as is customary in the area. Three channels had widths of 45 cm (the approximate width of existing irrigation channels at the site, 1× width), one channel had a width of 90 cm (2× width), and the final channel had a width of 135 cm (3× width). Drainage was directed by a terminal channel, located perpendicular to the five experimental channels. For experiments involving channel lengths greater than 45 m, the 2× and 3× width channels were rebuilt to widths of 45 cm, and all five channels were linked to make one winding channel 200 m in length.Flow Experiments Before experimentation, irrigation water was allowed to flow down the manifold channel for 10 min to flush out water in residual pools along the channel. Experimental trials were then initiated by removing the earthen dam separating the experimental channel from the manifold channel and creating a new earthen dam that diverted all the water into the experimental channel. Water rushing into the experimental channel mixed with loose soil and percolated along the length of the channel bottom. This initial water flowing through the channels, which included a high quantity of suspended soil particles, is termed the wetting front for the purpose of these experiments. Over time, the water levels naturally rose in the channels until a stable hydraulic profile was achieved, termed full flow. Samples were obtained at different points in space and time to examine the evolution of irrigation water chemistry. For the 45-m channels, sampling times were at 0 (wetting front), 5, 20, and 35 min, and for the 200-m channel, sampling times were at 0 (wetting front), 12, 35, and 45 min. Wetting-front samples and the 12-min, 200-m-channel samples were collected over several minutes, depending on the amount of time required for the water stream to flow along the lengths of channels. Samples were collected every 15 m in the 45-m channels and every 50 m in the 200-m channel. Sampling was conducted from a movable sampling station situated near the center of each channel. Water samples were collected into field-rinsed 250-mL bottles and immediately transferred to the sampling station. Samples were filtered to 0.2 mm using vacuum pumps or syringes; in most cases, filtration was complete within seconds, but the high content of suspended solids in the wetting-front samples frequently resulted in longer filtration times of ~1 to 2 min. After filtration, a portion of select samples was speciated with arsenic-speciation cartridges (MetalSoft Center) attached to the ends of syringes to quantify the percent total arsenic as As(III). Filtered and speciated samples were acidified with concentrated nitric acid, placed on ice, transported, and stored at 4°C until analysis. After the last sampling time point for each experimental trial, dissolved oxygen, pH, conductivity, and redox potential were measured with a Hanna HI-9828 multiprobe at each sample location.