Abstract:River systems, particularly plain river systems, are rich in fine-grained sediments, which exhibit a significant adsorption capacity for phosphorus. The fine-grained sediments play a significant role in river water pollution, serving as both sources and sinks for pollutants. The confluence zone is a crucial node for phosphorus transport and deposition in river systems. Due to the complex flow structure in this area, it is difficult to reveal the distribution characteristics of phosphorus on bed sediments at river confluences.An experimental system and detection method were designed to investigate the distribution of phosphorus on bed sediments within the confluence zone. The three-dimensional flow field of the confluence and the total phosphorus content of the bed sediments were measured. Subsequently, the relationship between the flow zone and the spatial distribution of total phosphorus on the bed sediments was analyzed. Based on 204 sets of hydrodynamic parameters and total phosphorus content data obtained from the confluence experiment, an artificial neural network model was established to quantitatively analyze the influence of hydrodynamic parameters within the confluence zone on the total phosphorus content of bed sediments, thereby identifying the key hydrodynamic parameters. The results show that: (1) Under the condition of large confluence ratio (Qr= 0.6), the total phosphorus content of bed sediments in the separation zone and the shear layer zone was high, indicating zones with high occurrences of phosphorus. The total phosphorus content of bed sediments in the acceleration zone was 21.4% lower than that in the separation zone, indicating a zone with low phosphorus occurrence. Under the condition of small confluence ratio (Qr= 0.4), except for sediments in the shear layer zone, which exhibited a high phosphorus content, the other areas exhibited a comparatively low phosphorus content. (2) The low horizontal flow velocity in the separation zone facilitated the migration of phosphorus from the water body to the bed sediments, resulting in a high phosphorus content of the bed sediments. Conversely, the high horizontal flow velocity in the acceleration zone outside the separation zone resulted in a low phosphorus content of the bed sediments. However, there was a significantly high phosphorus content in the local area (shear layer zone) within the acceleration zone, which was attributed to the downward flow (primarily generated by spiral flow) in this area. The downward flow promoted the exchange of surface water and pore water, increasing the contact probability between phosphorus and the bed sediments. (3) An artificial neural network model, formulated using hydrodynamic parameters, has accurately predicted the phosphorus content of the bed sediments at the confluence. Remarkably, the average relative error between the model predictions and the experimental results was only 5.71%. Through parameter sensitivity analysis of the model, it was found that w1 and w2 exerted the most significant influence on the model prediction accuracy, indicating that the vertical velocities in the lower and middle layers were crucial hydrodynamic parameters impacting the total phosphorus content of the bed sediments at the confluence. The horizontal flow velocity is low in the separation zone, resulting in high phosphorus content of the bed sediments. Conversely, in the acceleration zone outside the separation zone, the horizontal flow velocity is higher, leading to lower phosphorus content of the bed sediments. However, there is a notable exception in a specific region of the acceleration zone known as the shear layer zone, where a large amount of phosphorus is present. This phenomenon is related to the downward flow caused by spiral flow in this area, which promotes the exchange of surface water and pore water, increasing the contact probability between phosphorus and bed sediments.