A GPGPU accelerated modeling environment for quantitatively characterizing karst systems

The ability to derive quantitative information on the geometry of karst aquifer systems is highly desirable. Knowing the geometric makeup of a karst aquifer system enables quantitative characterization of the systems response to hydraulic events. However, the relationship between flow path geometry and karst aquifer response is not well understood. One method to improve this understanding is the use of high speed modeling environments. High speed modeling environments offer great potential in this regard as they allow researchers to improve their understanding of the modeled karst aquifer through fast quantitative characterization. To that end, we have implemented a finite difference model using General Purpose Graphics Processing Units (GPGPUs). GPGPUs are special purpose accelerators which are capable of high speed and highly parallel computation. The GPGPU architecture is a grid like structure, making it is a natural fit for structured systems like finite difference models. To characterize the highly complex nature of karst aquifer systems our modeling environment is designed to use an inverse method to conduct the parameter tuning. Using an inverse method reduces the total amount of parameter space needed to produce a set of parameters describing a system of good fit. Systems of good fit are determined with a comparison to reference storm responses. To obtain reference storm responses we have collected data from a series of data-loggers measuring water depth, temperature, and conductivity at locations along a cave stream with a known geometry in southeastern Minnesota. By comparing the modeled response to those of the reference responses the model parameters can be tuned to quantitatively characterize geometry, and thus, the response of the karst system.