Conclusions

Characterization of the structure of the ionosphere during auroral roar observations is critical as the proposed generation mechanisms all rely on the presence of structure and/or gradients in the electron density at F region altitudes. Several theories rely on density cavities, such as auroral ionospheric cavities, to enhance wave growth by reflecting the waves multiple times through a source region, to provide for conversion of trapped Z mode waves to escaping O mode radiation, or to duct O or X mode radiation. Other generation mechanisms rely on theory that predicts excitation of upper hybrid Z mode waves where $f_{uh}$ = 2$f_{ce}$ or 3$f_{ce}$, a condition that can be tested using density profiles obtained from the Sondrestrom ISR.

The ISR data show that no particular density feature persists in all the cases for which auroral roar emissions are observed, at least at scale sizes >~30 km probed by the radar. Table 1 summarizes the results. Five of the six cases show strong electron density gradients, and three (possibly four) of these events are characterized by the observation of an auroral ionospheric cavity. One case has no apparent density structure and remains laminar during the auroral roar observations. It is possible that the spatial/temporal resolution and spatial coverage of the ISR, as configured for these particular runs, was simply too limited to address density features critical to the evaluation of cavity theories. In five of six cases, $\vert\nabla$$N_e$/$N_e$ $\vert _{min}^{-1} <$ 120 km (measured with 23-137 km resolution). An auroral roar radiowave imager, planned for installation at the Sondrestrom site in the Summer of 1998, will be used to identify the location of the source region and narrow the ISR beam coverage, minimizing the possibility of missing the source region.

In 16 of 18 radar scans on 5 of the 6 days studied the upper hybrid matching condition $f_{uh}$ = 2$f_{ce}$ or 3$f_{ce}$ is satisfied in the F region ionosphere over nearly the entire frequency range of the observed 2$f_{ce}$ or 3$f_{ce}$ auroral roar emissions. The few cases in which the matching condition is not observed may result because the radar does not sample the entire ionosphere and has limited spatial and temporal resolution as configured for other experiments. These data are consistent with a theory of auroral roar in which an electrostatic upper hybrid or Z mode wave is generated at F region altitudes and subsequently mode converts to an escaping EM wave which propagates to the ground.

This research was supported by National Science Foundation grants ATM-9713119 to Dartmouth College and ATM-9616136 to the University of Maryland. The authors would like to thank Mike Trimpi for his work in designing, building, and deploying the Dartmouth College radio receiver, and John and Tommy Jørgensen for their help maintaining and operating this receiver. Operation of the Sondrestrom ISR and support for analysis of ISR products was provided by NSF cooperative agreement ATM-9317167.

The Editor thanks J. Minow and another referee for their assistance in evaluating this paper.