Case Studies in High Throughput Computing (HTC) using Condor for Physics-Based Remote Sensing Algorithm Development and Application

The use of physics-based numerical models has been an emerging paradigm in the analysis of remote sensing data and the development of exploitation algorithms. Understanding the key parameters that affect the complex imaging chain to exploit a scene and associated phenomenology is facilitated by these numerical models. As computing power to perform these calculations improve, the application of algorithms become operationally viable. This gain, however, is offset by the need to address complex scenarios through higher fidelity modeling. This involves computing additional model factors at many more levels. While the iterative process of executing these computations is conceptually simple, the implementation to make them operational requires special methods. Unfortunately, the mechanics and infrastructure to the solution of realizing these large number of calculations are often glossed over in the remote sensing literature. This leaves investigators without the technical details to address these class of problems. We present our experiences in addressing these types of problems through the use of the Condor High Throughput Computing (HTC) system. We present several remote sensing research case studies conducted by the Digital Imaging and Remote Sensing Laboratory at RIT over the past decade and highlight the computational gains realized by these HTC frameworks.