Use anemometer for measuring the air velocity. There are digital ones which are very accurate. Temperature measurements are usually done by thermocouples. There are different types of thermocouples depending on the temperature range you are interested to measure.
We use a very good and cheap anemometer (around 50 pounds) from Audon Electronics. Performance is very good. There are cheaper solutions around 20 pounds, you can find on ebay.
About air temperature we use the Sensirion sensor for humidity, light and temperature. Good accuracy, around 20USD. This is a chip however, you need to have it on some board.
We have our own boards that we connect to our wireless transceivers; I can show them to you reporting online if you wish.
You've asked two questions: "how to measure the accuracy of" and "how to accurately measure" air flow. Important distinction. The comments thus far address your second question.
To better answer this second question, however, we would benefit from knowing where the measurement will occur. Is your goal to simply measure air flow, or to measure airflow acting on some body or test article? Is it in open air (where there might be plenty of room for an anemometer, as Nikos suggested) without altering your test environment); is it in a pipe or duct, where an instrument will likely interfere with the flow (although to a lesser extent with John's hot wire); or is this in a wind tunnel where you need to measure the specific flow over the test article (in which a surface static pressure port connected to a transducer might be better).
Veera's resources are a good start down this path, assuming you know what these parameters and limitations are.
if you're actually interested in knowing or measuring the accuracy of such a measurement (your first question)...
You can start with manufacturers specifications. However duplicate instruments will likely give different answers (perhaps within specs). If you want to prove the accuracy, it may be necessary to design an experiment to measure the instrument(s) in a controlled environment before using it in your actual test. This isn't too hard to do, but the design of such a calibration experiment depends heavily on the instrument and environment.