From an electronics viewpoint, of course it's perfectly possible to drive your pumps with transistors or relays; and of course the relays are likely to need a transistor to drive them.
From a scientific experiment point of view, consider the following:
- How easy is it to determine the system is working correctly?
- Which of your team can do this work?
- How much time/effort is this portion of your research worth?
- What happens if you have a system failure? (repeat experiment?)
- Are field failures easy to detect?
Taking those into account:
- Consider using high-reputation suppliers with modular interfacing, such Mikroe Click and Digilen PMOD. Mikroe for example has 4-relay and 16-relay cards with all the interfacing already done, which might save you an enormous amount of wiring. (As do many other manufacturers, good and bad.)
- Consider redundancy: would it make sense to have two systems to switch the pumps on?
- Consider feedback: is there any way to detect that a pump has operated?
- Consider logging: if you're driving this with an Arduino, what is doing the logging?
- Consider testing: are there buttons you can easily press to manually drive a pump (for testing the pump, relay, etc). Are there LEDs so you can easily see which pump is supposed to be activated?
This sort of manufacturer provides good circuit diagrams, good repeatability etc. If you get economy products or hobby items you may well have a source of poor repeatability. And you may well find the product ranges and approach are helpful for other or future automation that you might have.
This even before you get to software, which is of course the usual cause of field failures. For reliability and reproducibility, I recommend microcontrollers with gcc-avr or similar. This gives you good makefile-type documentation and eliminates GUI compilation, a notorious source of variation; of course at the price of additional learning curve if you're not familiar with it.