Rather than me regurgitate others' work, have a look at page 7 (93),
Direction, of chapter six,
Air Movement, in the
Plant Growth Chamber Handbook 
ETA: Here it is
DIRECTION
Air in controlled-environment rooms can flow from bottom to top, top to bottom (often called downward or reverse air flow), or horizontal. The chief argument for moving the air from bottom to top is that this method provides a laminar flow of air through the plant growing area. Unfortunately, the laminar flow is quickly disrupted once plants are placed in the chamber, and uniform air flow cannot be maintained unless the plant containers are spaced evenly over the growing area. Moreover, the air movement across the leaf will change considerably depending upon the number and size of the plant containers per unit area of the growing space. For example, a plant growth chamber with an area of 2.97 m² might have an air flow of 0.5 m/sec when empty. If the space were filled with 15-cm pots and about 112 could be used, the reduction in free area conceivably could result in a three-fold increase in air velocity as the air passed between the containers.
From the point of view of both engineering and plant scientists, downward air flow is preferable to upward air flow because temperature gradients are smaller (Morse, 1963). Dimock (1963) noted that the results of intensive study at Cornell University showed that the vertical temperature gradient with downward air flow was only half as great as with upward flow. Matsui et al. (1980) also concluded from their studies on humidity distribution that downward air flow gave the most consistent pattern of air movement. In general, downward movement of air will more closely mimic humidity and temperature profiles found under field conditions. Humidity levels should always increase with depth in the canopy (Allen, 1975). Morse and Evans (1962) reported that plant growth of tomato, lucerne, and subterranean clover was slightly greater when the air flow was downward (Fig. 5).
The argument for using horizontal air flow in a growth chamber is that canopy turbulence more closely resembles natural conditions than with an upward or downward air flow (Monteith, 1964; Doorenbos, 1972), although no data have been presented to verify this assumption. The obvious disadvantage of a horizontal air flow is that in the presence of the radiant lamp load, air temperature and humidity will rise as the air crosses the chamber. The greater the lamp load, the greater the horizontal gradient. Forrester (1979) claims that the plants create turbulence that reduces the gradient to a negligible level. Smeets (1978) claims that the temperature rise is not significant when the radiant heat is reduced by cold water running over the glass barrier that separates the lamps from the growing area. If the distance across a growth chamber is very great, however, this gradient in temperature and humidity may be quite large (Allen, L H. Jr., Gainsville, Florida, Personal communication, 1989). On the other hand, even though a horizontal gradient is present, rooms can be used effectively for research studies by assigning blocks (replicates) along the horizontal gradient. Allen indicates that if humidity is kept low and air temperatures are high enough, the temperature will drop because of evaporative cooling by the plants. However, a horizontal humidity gradient is still produced.
