Since no one has given you an answer I'll try to help although I'm not completely familiar with the subject.
Like any other balance equation you can probably assume that the total water consumption of any system is given by:
total water consumption = mesquite plant consumption + sum(factor "i" consumption)
Now in a laboratory you could easily measure the input of water and the output of water in any given time for some samples of mesquite plant and calculate the water consumption as long as the the second term on the right side of the equation is zero (no other consumption factors besides mesquite plant).
I'm guessing however (from the question topics) that you intend to do spatial study meaning you'll have a much more complex system to study. With such complexity (and assuming that lab studies may not be feasible, I'll talk later about the possibility of that being feasible) I would advise calculating a final number (and an uncertainty) by steps:
0) Define what is your system (it usually is an area given by a shape).
1) Estimate, in your system and for a controlled time range (1 year for instance to avoid seasonal bias), the total input of water and the total output. Input is every water that come in, output every water that comes out. This is usually a number with a big uncertainty but at the very least you can come out with a global consumption value (input - output = consumption).
2) Having that number the next stage would be characterizing the consumption factor by populations (being mesquite plant class 1, for instance, plant A being class 2, plant B being class 3, plant C and D being class4, and so on...). The criteria here would be to specifically individualize the populations which you know you can estimate (even if roughly) the consumption.
3) Since you have the populations you can start building a map (there are many methods available) of those populations and calculate the are occupied by each one.
4) So, for example, if you know that population B has a water consumption X for every m2, multiplying that value for the total area covered by population B is going to give a rough estimation of the consumption by that specific class. You can now subtract that value to the result of point 1 being one step closer from the real number.
5) By repeating point 4) for every class that is not mesquite plant you'll get a final value.
An important remark, however, is that usually plants do not organize themselves neatly in a given area. So in 100 m2, for example, you can have probability 0.5 of having plant A, 0.3 of plant C and 0.2 of plant B. Meaning that if you want to know the consumption of plant C for that area you must multiply the area by the probability since only a part of the area is actually plant C (it would be: 100m2*0.2*X = Y).
If by any chance you can measure mesquite plant consumption in a controlled system you'll get estimates with far less uncertainty since your only preoccupation will be to calculate the percentage of coverage of mesquite plant in the study area.
thanks Pedro Correia for your valuable suggestions, I definitely try the possible means from your suggested methods
My Idea is to Map the area of mesquite plant using satellite image, then measure that this much area can consume this much of groundwater.
To share the information suggested by scholars, that we can measure the root pressure of any plant using auxanometer and transpiration by Potometer these can helps to calculate the water usage. I also got some literature related to water usage of this particular plant "http://ac.els-cdn.com/S0378377400001141/1-s2.0-S0378377400001141-main.pdf?_tid=dd647daa-e49e-11e3-8885-00000aab0f27&acdnat=1401085852_6a763fff60b34050826f425c52caf07c".