Drought tolerance is a complex phenomenon. In the field, irrigation control that creates a differential moisture gradient can easily discriminate genotypes. However, such work can be taken up in suitable rain free season. You need one two plots for genotypes, one for the stressed environment and the other for non-stressed. Stress levels and stage of exposure can be modulated according to your needs.
Almost all drought environments have some rainfall. When people create a open sided rain barrier the precipitation can be largely eliminated and then water can be added back forming different drought stresses based on timing severity and duration. Part sheltered system simulates the open field conditions and levels of water addition can be achieved giving drought levels.
Not perfect but much better than not having any comparisons at all.
In all field experiments it is important to provide as uniform base conditions to reduce the uncontrolled variation which will confound the study of real management factors because of other variation. On that comes immediately to mind to field uniformity and the depth of soils for the tests. Shallow variable top soils come in drought areas will have huge soil variability working to confounding varietal or other differences.
I prefere to use TDR instrument for evaluation of drought tolerance. Accurate results is too much hard in droght tolerance investigation but TDR could be helpful in field experiment.