Conceptual Flood Inundation Modelling: Computationally Efficient Methods for Large Data-scarce River Basins

India is one of the world's most flood-prone countries, with 113 million people exposed to floods. Large-scale hydrological models integrated with complicated Navier–Stokes based hydraulic, and inundation models traditionally address flood preparedness, control, and mitigation. In addition to being highly data-intensive at the fine spatial and temporal resolution, this approach has a considerable computational cost that limits real-time applications. We employ the parameter-free Dynamic Budyko (DB) hydrological model to map observed precipitation with gridded runoff to overcome data scarcity. We propose a time-efficient Slope-corrected, Calibration-free, Iterative Flood Routing and Inundation Model (SCI-FRIM) framework that can be used with any hydrological model to generate a probability map of inundation. To model the catastrophic flood extents that the state of Kerala in India experienced during August 2018, we use gridded 0.25 deg × 0.25 deg IMD precipitation data. We use a parameter-free iterative approach to update flood velocity by assuming that river velocity does not fluctuate geographically across a particular river network at a given time instant. We pre-compute the iterative velocity and model the relationship between flood velocity-discharge and discharge-inundation height for each reach by combining the globally available SRTM/ASTER DEMs with empirically obtained river-reach geometry data (JPL). We compute the reach slope from the absolute vertical error-prone DEM by segmenting the river network into a series of independent channels and extracting the relationship between the channel pixel's elevation and the pixel's distance to the pour point. We use the Height Above Nearest Drainage (HAND) to map the probabilistic spatial extent corresponding to an ensemble of derived reach inundation heights. We then compare the proposed model with observed flood data points provided by the Kerala State Disaster Management Authority (KSDMA). The model captures up to 52% of 370,000 flood data points in a single run for the peak flood day within 15 minutes on a desktop computer. With reliable estimates of empirical bankfull discharge, the proposed model can achieve higher accuracy in lesser time.