Seriously, there has to be an operating temperature range specification. If you can't cover the ends of the range with actual outdoor exposure at both ends, and/or don't have time or the logistics are too formidable, you would then buy/ build/ rent an environmental chamber to allow controlled testing at the ends of that range as a minimum. Similarly for humidity. You also have to cover precipitation of all kinds, freezing, condensation, dust, etc.
The first step then, would be read the specification defining what the product is supposed to survive, and figure out how to validate that it will survive. If you have to write the specification yourself, it takes a little longer.
Testing quickly gets expensive, in money and/or time, so you might face a business decision balancing the actual cost of the testing you are willing to do against the potential cost of warranty failures or other liabilities associated with the intended use. I personally would try to foist that off on a business person, and just try to estimate and present the alternatives to be considered.
Cheap, dirty, fast: Place a freebie in service with the Army; if they can't break it, it must be pretty good. Offered with tongue not entirely in cheek.
It really depends on what causes reliability to be low. There are many ways to accelerate particular failures, otherwise, a 10-year reliability would require 10 years of real time. You can Google "accelerated life" for some hits. Temperature is one way, using the Arrhenius equation, but that assumes you know what the activation energies of the failures are.
