Within the parlance of Hydrology, forecasting streamflow plays a vital role in water resource management. Based on the temporal scale, streamflow forecasting can be broadly classified into two categories: short-term and longterm forecasts. Short-term forecasting at hourly and daily scale is critical for flood warning, mitigation (Gül et al., 2010;Rogelis & Werner, 2018), and disaster management (Roulin, 2007). Long-term forecasting at monthly and seasonal scales is extensively utilized in reservoir monitoring (Rezaie-Balf et al., 2019), design of hydraulic structures (Lowe, 2006), and irrigation practices (Droogers & Bastiaanssen, 2002). However, uncertainty in hydrological predictions is still a serious concern that needs to be addressed before planning and management activities. These uncertainties may arise due to the error in the forcing variables (e.g., rainfall and temperature data), the model's inherent structural error (parameters and boundary conditions), and improper initial condition (Alvarado-Montero et al., 2017). While the quality of the input data such as the rainfall rate and temperature can be improved by deploying better observation systems, errors in model control variables consisting of the initial and boundary conditions and parameters can be corrected using the well-established tools from the theory of dynamic data assimilation (DDA, Lakshmivarahan et al., 2017;Lewis et al., 2006). In this paper, the goal is to demonstrate the power of a new class of method called forward sensitivity method (FSM, Lakshmivarahan & Lewis, 2010) for assimilating data into a simple conceptual two parameter model (TPM, Xiong & Guo, 1999) in the analysis of streamflow.DDA is the process of combining a model of a process of interest in the analysis with a finite set of relevant but noisy observations of the same process. Existing literature on DDA can be broadly classified into three classes. First is a class of sequential methods known as Kalman filtering (Kalman, 1960b) and many if its extensions (Evensen, 1994;Puente & Bras, 1987;Sakov et al., 2012;Whitaker & Hamill, 2002). This class of methods rests on the basic principle of best, linear unbiased estimation (BLUE), where best is in the sense minimum variance. While Kalman filter-based methods (KF) provide a natural framework for sequential dynamic data assimilation,