To do a "perfect job" LEDs MUST be current driven.
In this case it happens that 16 in series driven by a 48V fixed voltage supply will probably work OK BUT power level will be uncertain and may vary widely with different LEDs. In some cases, doing what you propose would destroy LEDs in some cases, but in this case they should always be "in spec".
You do not say what light output you want or what you expect or what you are replacing. All these matter.
The hlh-120h-48V is specified at 2.5A. This MAY mean it limits sharply at 2.5A but it MAY provide 3A or 3.5A either at 48V or at a reduced voltage. If we assume that it makes 2.5A max then the LEDs will make ABOUT 2.5A x 3 = 7.5W each IF they are driven to the full 2.5A. They are rated at 3A max so 2.5A would be acceptable if they were properly heatsunk. The LED will very roughly radiate 1/3 of its power as light so 16 on one heatsink will dissipate 2.5A x 3V x 16 x 2/3 dissipated = 80 Watts.
Each LED will have an internal temperature rise at 2.5 C/W of about
dT = 7.5W x 2.5 C/W =~ 20C less any reduction due to light radiation.
Design operating temperature is Tj=85C to Tc (or Tsolder) is about 85-20 = 65C (too hot to touch.) Use Tambient of 20C (coolroom but be safish) so Tca = 65-20 = 45C. Heatsink must dissipate 80W so C/W of heatsink needs to be < 45/80 or around 0.5 C/W. That is mnot a minor heatsink!. You need either a VERY substantial Al molding, or bolt to some large metal surface or use blown air (or flowing liquid) cooling. If the lights obly ever operate when the coolroom is cool you can relax that spec BUT you must design for worst case.
The Xm-L is not the most efficient LEd available from Cree or other sources. At the time the data sheet was printed it was the most efficient SINGLE DIE LED that Cree made but you do not need a single die LED (used to ensure precise optical source), and you can get Cree LEDs that give approaching 200 l/W for top flux bin parts. The XM-L2 is specified at about 105 l/W so you should be able to get say 150+ l/W from some other LEDS. Note that l/W efficiency falls as I_:ED approaches maximum. The datasheet shows that at 3A the XM-L2 is only 88% as efficient as at 1A.
The XM-L2 has a typical Vf (forward voltage) of about 3.2V at 2.5A (datasheet page 2 interpolated and page 5 on graph).
AT 3V Vf I_LED is about 1.3A (datasheet page 5)
So if you used a 48V supply and placed 16 in series for mean Vf = 3.0V then I_LED typical = 1.3A. This is about half of what your supply can provide and about 40% of LED maximum. To get say the full 3A you need 3.3 V/LED typical and perhaps 3.6V worst case. LEDs from one batch tend to 'clump' and you would tend to get most high or most low or whatever. To operate 16 x 3.3V needs ~= 53V or if 48V is used you can operate <= Vsupply/_V_LED = 48/3.3 = 14.5 LEDs = 14 LEDs. If VLED =3.6V in all cases you can operate <= 48V/3.6 = 13 LEDs.
The A type supplies are current adjustable and MAY be suitable as CC sources for LED driving as they stand. If not, you can make a N Amp constant current source using eg an LM350 and a few resistors and caps and suitable heatsinking. Vheadroom ~= 2.5V say so with a 48V supply you can operate (48-2.5)/3.3 = 13 LEDs down to (48-2.5)/3.6 = 12 LEDs.
If you have 12 LEDs and get a batch that have mean Vf of 3.3V then CC voltage drop is (48-(3.3 x 12)) = 48V - 39.6V = 8.4V.
Dissipation at 2.5A in CC is 8.4 x 2.5 = 21 Watts - An LM350 would need good heatsinking to handle this.
A CC supply which regulates the supply current but does not use a linear dropper resistor is desirable.
Rather more could be written but sleep calls. More anon if wanted.
Ask questions.
