Non-sampling error in ozone biomonitoring: the role of operator training

Two groups assessed ozone symptoms on tobacco leaves: one was represented by young students and the other by scientists with experience in plant biology, but not experienced scorers. In the first case, results demonstrate that in the first week of exposure the extent of injury is almost always overestimated, but in the second week it is correctly evaluated or slightly underestimated: this can be due to the variable ambient ozone levels. In the second case, the average accuracy levels ranged from 40 to 82%, with an average repeatability of 95.2%. Central classes of damage are more difficult to evaluate: this may depend on the fact that two leaves may have similar total injured area, but substantially different number and spatial distribution of the lesions. Some practical suggestions in order to reduce non-sampling errors and to improve operator training are given.