Overal 214 41 % 65 % 54 %
MATERIALS AND METHODS
The experiments were carried out at Treborth Botanic Garden, University of Wales, Bangor, Using pitfall traps, a rhizotron previously described by Sackville-Hamilton, et al (1991) and time lapse photography. Between January 1993 and December 1995, numbers of predatory soil arthropods were recorded using 20 pitfall traps. Traps used here simply consisted of plastic jars (7.5 cm high and diameter 6.0 cm at mouth). The rhizotron contains 34 windows (90cm x 50 cm) which allow the soil profile to be observed from 5-10 cm above the surface, down to 100 cm below. A time lapse camera was set up in one of the rhizotron windows to record predator - prey interactions.
Predator activity was defined in these observations as the number of times a predator contacted a given prey. In preliminary experiments with no baits, no predators were recorded during three days of filming and this was time consuming and unacceptably expensive. A time lapse unit was set up, composed of 4 sub units : camera motor, control unit, timer unit and flash. Colour possitive fim (Kodachrome 25, 16 mm x 30 m, containing 4,000) frames was used.
Locust and stick insect eggs, mustard beetle and mealworm pupae were compared as possible baits for increasing activity. Mealworm pupae (Tenebrio molitor) were selected. A time lapse cinematographic technique was developed and using an analysing projector, it was found that an eight minute interval between frames maximised the number of significant results obtained per unit cost. The parameters measured were pupal disappearance, contacts between predators and prey and the identity of predators.
RESULTS AND DISCUSSION
This study is one of the few attempts to assess the activity and importance of subterranean predation on the resting stages of insects, and the field study has been made possible by the use of a rhizotron which allows direct observation to be made. Oksamen et al. (1985), studying small predators and their prey emphasized that experiments should be performed under as natural condition as possible and this has been a guiding principle in the present work.
The use of a rhizotron is not however without problems.
1. It is difficult to identify soil dwelling organisms to species when they are behind glass and are active, and when they cannot be caught and killed without creating unacceptable levels of disturbance. This necessitates the use of broad taxa for classification and this is never satisfactory.
2. The area of soil profile which can be viewed is limited by the size of the windows and as ut is not possible to see any distance into the soil, it is virtually two dimensional, so that the volume of habitat which can be inspected is highly restricted. As a consequence, interactions between animals are rarely observed and the observational food web published by Gunn and Cherrett (1993) had to be based on years and years of regular observation. More frequent observations are only made possible by placing resting stages next to the windows in artificially high concentrations and after considerable soil disturbance. Moreover the prey chosen were not normally found in the soil and so may have been much less resistant to predator attack than resting stages normally found there. In effect, predators were attracted by artificial baits.
