Over the past decades, tropical storms and hurricanes in the Southeast United States have become more frequent and intense, mainly due to the effects of climate change. They often produce torrential rains that may result in catastrophic floods depending on hydrologic, geomorphologic and orographic characteristics of the region. Although hydrological models are widely used to provide estimates of such floods, their predictions most often are not perfect as the models suffer either from inadequate conceptualization of underlying physics or non-uniqueness of model parameters or inaccurate initialization. Data Assimilation (DA) based on Particle Filtering (PF) has been recognized as an effective and reliable mean to integrate the hydrometeorological observations from in-situ stations and remotely sensed sensors into hydrological models for enhancing their prediction skills while accounting for the associated uncertainties. Although recent developments in DA theory and remote sensing technologies have made significant progress in enhancing the performance of the hydrologic models, their usefulness are subject to some inherent limitations that may result in inaccurate and imprecise model predictions, especially in the case of an extreme event such as flooding. This dissertation is an attempt to identify these limitations and address those by conducting four studies. The first tackles a fundamental problem associated with the utilization of remotely sensed observations in hydrologic data assimilation applications. The two and third are progressive studies that address two conceptual/theoretical problems of using particle filtering approach in hydrologic studies. As a result, the fourth study demonstrates the effectiveness and usefulness of the developments in all three studies in improving the hyper-resolution hydrologic model predictions over a region in the Southeast Texas where heavy rainfall from Hurricane Harvey caused deadly flooding.