In dynamic data driven application systems, the predictions are improved based on measurements obtained in time. Predicted quantity often satisfies differential equation models with unknown initial conditions and source terms. A physical example of the problem we are attempting to solve is a major waste spill near a body of water. This can be, for example, near an aquifer, or possibly in a river or bay. Sensors can be used to measure where the contaminant was spilled, where it is, and where it will go. In this paper, we propose techniques for improving predictions by estimating initial conditions and source terms. We show how well we can solve the problem for a variety of data-driven models.

A wildland fire model based on semi-empirical relations for the spread rate of a surface fire and post-frontal heat release is coupled with the Weather Research and Forecasting atmospheric model (WRF). The propagation of the fire front is implemented by a level set method. Data is assimilated by a morphing ensemble Kalman filter, which provides amplitude as well as position corrections. Thermal images of a fire will provide the observations and will be compared to a synthetic image from the model state.

This paper describes a method to improve the cache locality of sequential programs by scheduling fine-grained threads. The algorithm relies upon hints provided at the time of thread creation to determine a thread execution order likely to reduce cache misses. This technique may be particularly valuable when compiler-directed tiling is not feasible. Experiments with several application programs, on two systems with different cache structures, show that our thread scheduling method can improve program performance by reducing second-level cache misses.

Multilevel methods have been studied extensively for solving certain partial differential equations, and such equation solvers have been successfully applied to major computational problems of physical interest and technological importance. KLEV is a collection of routines for solving the N X N system of linear equations Az = g when A and g are discretizations of some partial differential equation. The current release includes a direct method, iterative methods, and multilevel methods. A brief review of multilevel methods and an overview of the implementation are presented.