IFM: A Scalable High Resolution Flood Modeling Framework Swati Singhal 1 , Sandhya Aneja 2 , Frank Liu 1 , Lucas Villa Real 1 , and Thomas George 1 1 IBM Research 2 Universiti Brunei Darussalam, Brunei Darussalam Abstract. Accurate and timely flood forecasts are essential for effective management of flood disasters, which has become increasingly frequent over the last decade. Obtaining such forecasts requires high resolution in- tegrated weather and flood models with computational costs optimized to provide sufficient lead time. Existing overland flood modeling soft- ware packages do not readily scale to topography grids of large size and only permit coarse resolution modeling of large regions. In this paper, we present a highly scalable, integrated flood forecasting system called IFM that runs on both shared and distributed memory architectures, effec- tively allowing the computation of domains with billions of cells. In order to optimize IFM for large areas, we focus on the computationally expen- sive overland routing engine. We describe a parallelization scheme and novel strategies to partition irregular domains to minimize load imbal- ance in the presence of memory constraints that results in 40% reduction in time compared to best uniform partitioning. We demonstrate the scal- ability of the proposed approach for up to 8192 processors on large scale real-world domains. Our model can provide a 48-hour flood forecast on a watershed of 656 million cells in under 5 minutes. 1 Introduction Operational flood forecasting is becoming increasingly important due to the changing global climate and frequent incidence of flood disasters [1]. The most common causes for flooding are sudden precipitation in urban areas with poor drainage or seasonal storms resulting in persistent rainfall, which results in over- flowing water bodies. Hence, in recent years there has been a strong focus on two stage mechanisms to predict flooding events. The first stage employs a weather model to predict precipitation. The second stage uses these predictions as input to an overland flood model, which computes surface runoff and routes the flow taking into account surface characteristics such as variation in land use type and topography. In such a system, the weather forecasting is performed using fine resolution atmospheric models that discretize the partial differential equations representing evolution of atmospheric flows in time [12], while the overland flows are simulated via equations based on conservation of mass and momentum with the vertical effects simplified to yield the 2-D shallow water equation [14]. F. Silva et al. (Eds.): Euro-Par 2014, LNCS 8632, pp. 692–703, 2014. c  Springer International Publishing Switzerland 2014