No single band - or even two bands - will guarantee coverage over the long term.
Over the last few years we've had to drop down to 160m to cover paths out to
400 miles on many occasions. It all depends on the conditions in the ionosphere,
and that changes throughout the day, year, and with the sunspot cycle: current
conditions aren't the same as those 2 years ago, or 2 years from now.
So you really need an antenna system for the two paths over multiple bands. For
a long term solution that probably means 160m / 80m / 40m and 20m at least.
(If you are providing a data link then 30m might be a useful option.)
One useful approach is to look at some of the propagation prediction tools to see
what they forecast for coverage. I like the tools from the Australian Ionospheric
Prediction Service because you can run them online here:http://www.ips.gov.au/HF_Systems/7/1
The "Combined HF Prediction Tool" gives you a number of options: the "Local
Area Mobile Prediction" (LAMP) tool shows what bands you can use for
distances out to 1000km (600 miles), which is easier to read than the more
detailed GRAFEX tool for point-to-point circuits. The latter shows which
paths may be open, and includes the "Optimum Working Frequency" (OWF)
value - the frequency that is most likely to give reliable coverage for the path -
for each hour of the day. (Currently it varies from about 5.5. MHz to 12.1 MHz
for the specific 800 mile path that I tried.) For the LAMP display it has a preset
option for the ham bands, and color-codes the display based on the best band
vs. distance for each hour.
These will give you a prediction for the current conditions - you can manually
enter values for other conditions to see how it would affect the frequencies
In terms of antenna design, the optimum vertical radiation angle for a 400 mile
path is about 40 to 50 degrees, while for an 800 mile path it is 20 - 30 degrees.
(Approximations from scaling off a chart - exact value depends on the height
of the ionosphere.) That means the optimum height for horizontal antennas will
be about 3/8 to 1/2 wavelengths. Lower heights will work: at 1/4 wavelength
the signal at 20 degrees is only down 8dB or so. Higher antennas may not work
as well: at 3/4 wavelength there is a null at 45 degrees, and the null drops to
30 degrees at 1 wavelength over ground. (If the antenna is on a large metal
roof the calculation gets more complicated.)
For these angles a vertical antenna isn't a bad choice, as long as it has an
adequate ground plane, otherwise a simple dipole may out perform it.
Since both other sites are to the West you could plan some sort of directional
antenna, at least on some bands, to optimize coverage in that direction.