Prediction of the sediment generated within a catchment basin is a crucial input in the management and design of water resources projects. Due to the unavailability and complexity of the precipitation and hydrological process, reliable sediment concentration is hardly predicted by applying linear and nonlinear regression methods. In the present study, an attempt has been made to explore the use of Tropical Rainfall Measuring Mission (TRMM-3B42) dataset for modeling suspended sediment using neural networks (NNs) with different training functions, i.e., Levenberg-Marquardt (LM), scaled conjugated gradient (SCG), and Bayesian regulation (BR) for the Kopili River basin, India. The input vector to the various models using different algorithms were derived considering the statistical properties such as autocorrelation function, partial autocorrelation function, and cross-correlation function of the time series. The daily rainfall data from 2000 to 2010 (4,018 days) were considered for the training (70%) and validation (30%) of the models. The model ANNLM6 performed better than other models during calibration (ENS = 99.49%, CORR = 0.997, RSR = 0.049) and validation (ENS = 98.11%, CORR = 0.990, RSR = 0.085), and the optimum structure of the artificial neural network (ANN) model was found to be six neurons in the hidden layer. Moreover, the analysis also concludes that ANN model using the LM algorithm is able to predict suspended sediment at the outlet fairly well. The overall analysis reveals that TRMM rainfall is promising for suspended sediment prediction in the absence of ground-based measurements of rainfall. It is inferred that, the TRMM-3B42 rainfall estimates can be used for the conceptual rainfall-runoff-sediment modeling where the gauges are sparsely distributed and the radar measurements are rarely available. © 2014 American Society of Civil Engineers.