Since sediment concentration is an effective factor on increasing debris flood’s peak flow and damages from floods, developing new models to predict the sediment concentration of debris floods has crucial importance. In this study, a hybrid SVR-PSO model was proposed to predict the concentration of sediment in typical and debris floods, and it was examined in three basins located in Gilan, Mazandaran, and Tehran Provinces, Iran. Mean elevation and slope of the basin, the area of the basin, current day’s rainfall, the rainfall of previous days (1–3 days before flood) for all rain-gauge stations of the basins, as well as the discharge of the previous day, were used as the input variables of the model. Then, various combinations of variables were tested to assess the factors influencing the concentration of sediment in typical and debris floods in order to find the best variable combination with a high performance in predicting the concentration of sediment in the studied floods. The results showed that basin elevation, current day’s rainfall, previous day’s discharge, rainfall of the previous day, basin area, rainfall of the previous two days, basin slope, and rainfall of the previous three days were the key factors influencing the concentration of sediment in typical and debris floods, respectively. Coefficient of determination, root mean square error, and mean absolute percentage error were estimated 0.96, 0.003, and 14.38% for the proposed model at the testing phase, respectively. This implies model’s good performance for predicting the concentration of sediment in typical and debris floods so that the present model can provide reliable predictions of flood character in basins.

由于泥沙浓度是增加泥石流洪峰流量和洪灾破坏的有效因素，因此开发新的模型来预测泥石流泥沙浓度至关重要。在这项研究中，提出了一种混合SVR-PSO模型来预测典型洪水和泥石流中的沉积物浓度，并在位于伊朗吉兰，马赞丹兰和德黑兰省的三个盆地中进行了研究。流域平均高度和坡度，流域面积，当日降雨量，流域所有雨量计站前几天（洪水发生前1-3天）的降雨量以及前一天的流量用作模型的输入变量。然后，测试了各种变量组合以评估影响典型洪水和碎石洪水中沉积物浓度的因素，从而找到最佳的变量组合，并具有较高的预测洪水中沉积物浓度的性能。结果表明，流域海拔，当日降水量，前天流量，前一天降水量，流域面积，前两天降水量，流域坡度和前三天降水量是影响泥沙浓度的关键因素。在典型的洪水和泥石流中。在测试阶段，该模型的估计系数，均方根误差和绝对绝对百分比误差分别估计为0.96、0.003和14.38％。