A new NSF-sponsored Engineering Research Center, known as the Center for Collaborative Adaptive Sensing of the Atmosphere (CASA), is developing a dynamic, collaborative, adaptive system (DCAS) radar network capable of detecting and tracking hazardous weather in real-time. Data from a network of radars are collected at a central processing node, where data are then quality controlled, analyzed, and new tasking commands are generated and sent back to the radars. This paper describes the design of this central processing software, collectively known as the Meteorological Command and Control (MC&C), as well as a brief overview of the system architecture.

As a proof-of-concept study, CASA is building its first test bed in central Oklahoma. Known as NetRad, the network will consist of four radars and a central processing facility located at the University of Oklahoma. The radars in this test bed will be spaced an average 25 km apart, each with a maximum range of 30 km. This network configuration allows for the beams to overlap, thus maximizing collaborative sensing of the atmosphere among the networked radars. It is the goal of this test bed to demonstrate that the presence of tornado can be verified within 60 seconds of touchdown and that its centroid can be located with a spatial resolution of 100 meters. To achieve this goal, collaborative, simultaneous scanning from several radars is necessary. Thus, one of the challenges for CASA is to build an MC&C system that processes input from multiple radars in parallel and retasks these radars in a very short amount of time.

The MC&C is the key software component that makes the NetRad system collaborative and adaptive. It ingests data as input from the sensing components, applies quality control (QC) on that data, and invokes detection and prediction algorithms on that data. All output from these detection algorithms is then organized within the Feature Repository, a 2-dimensional grid that can be used to visualize all relevant, 'real-time' features. Furthermore, the Feature Repository will store past and predictive information. This collection of feature information is then passed to a "Resource Allocation" module. This module will apply numerical techniques to optimize and generate the next radar scanning strategies, based on meteorological data, signal-to-noise ratio data (radar attenuation), and end-user priorities. It is expected that this entire feedback loop - from radar to MC&C to Resource Allocation to radar - will be completed within 30 seconds.