The high-latitude HF channel has been measured and characterized in terms of Doppler spread, delay spread, and signal-to-noise ratio. The performance of data modems has been determined over a comprehensive range of simulated channel conditions. A comparison is made between the channel measurements and the modem characterizations. Thereby the percentage of time that the modem works satisfactorily (availability), during the measured channel conditions, is determined. Two 75 bits/s modems (compliant with STANAG 4415 and STANAG 4285), a 2400 bits/s modem (compliant with STANAG 4285), Morse code, and voice have been selected for analysis. The two 75 bits/s modems show 60--75% higher availability than the 2400 bits/s modem on the measured paths. The 2400 bits/s modem is degraded in performance during a geomagnetic disturbance whereas the 75 bits/s modems maintain approximately the same performance level. The paper also addresses the necessary frequency pool (at a specific time) required for an HF circuit to achieve the best possible communications availability. For the 75 bits/s modems, independent of path length, a one-frequency pool is all that is required during summer and winter when the ionospheric conditions are quiet. A four frequency pool is necessary during disturbed ionospheric conditions. For the 2400 bits/s waveform, three to four frequencies are needed on the 200 km west-east path, and four to six frequencies are required on the north-south 2000 km path. This applies to all seasons and geomagnetic conditions analyzed. ¿ 2001 American Geophysical Union