There are a number of choices, though it depends on many factors. I'll propose several ideas, but they may not be workable depending on your conditions. Also depends a bit on if this is qualitative sampling or quantitative.

1) Sweep net sampling. If the majority of the canopy is under 15' this will work well. I would sweep for a fixed time interval for each tree.

2) Visual searching. This is good for scale insects that insects that live on braches or bark. Such insects are not well sampled using a sweep net.

3) Fogging. Use a fogger to spray the tree with insecticide. Place a white sheet on the ground below the tree before spraying to catch the falling insects. This might be a bad idea if you need to rear caterpillars to adulthood in order to identify them to species.

4) For a canopy that is 20' or more I might try a malaise trap set in the branches.

5) D-vac might work. Here is one, though there are others: http://rinconvitova.com/bulletins_product_htm/D-Vac%20model%2024.htm

6) Apply something like tanglefoot to the trunk to catch insects that might migrate up and down the trunk. http://www.amazon.com/s/?ie=UTF8&keywords=tanglefoot&tag=mh0b-20&index=aps&hvadid=3527195770&hvqmt=e&hvbmt=be&hvdev=c&ref=pd_sl_97r2mek6dt_e

7) You might also look in the soil. There are insects that specialize on root feeding. Cicadas, beetles, and a few aphids come to mind.

8) Grow small trees of relevant species (or find saplings) and place them in the forest to act as trap species.

We find ligt traps and fogging most effective for Homoptera - although inevitably some are missed (soil fauna etc).  Sweeping and beating will also work but are harder to quantify.

Caterpillars , as the last responses says, are more of a challenge.  Endophages can only be found by hand searching - for a fixed period, distance on a transact or whatever.  Free-living caterpillars will turn up in fogging samples - but many end up simple dangling on their silk in mid drop!  Beating and sweeping remain effective and have the advantage of turning up live material.

Thank you so much Timothy and Roger!!!

I very much appreciate your kind and thorough answers!

The truth is for these groups in the canopy you have to go up there and get them. For Sternorrhyncha (the Order Homoptera is no longer valid - these insects are all grouped in Hemiptera) and caterpillars, fogging will not be an efficient way to capture these insects. They are killed by the insecticide fog but caterpillars get hung up on silk or in the foliage and rarely make it to the collecting sheets on the ground. Aphids and scales die with their mouthparts inserted into their food plant and do not fall to the collecting sheets. In any case, with fogging you cannot know what the insects were feeding on or doing in the canopy; they just die and fall to the ground. Similarly light traps, even those hoisted into the canopy will not catch these groups efficiently; caterpillars are not attracted to light and only male scales fly and are not reliably attracted to lights. Similarly only alate aphids will come to lights but you have no idea what they are feeding on. For specific feeding studies on lianas, you are going to have to go up the tree with ropes and ascenders. and collect the insects feeding on the lianas you are interested in by hand. Even if you have access to a canopy crane or tower, you are going to have to rope over to the plants you are interested in. This is not trivial as many lianas position their leaves at the very top of the canopy in the sun and all you see from below is the woody parts anchored on the tree or on the ground. This illustrated the major problem in working with the biota in the canopy - a great area of undiscovered biodiversity and ecological marvels. Good luck with your studies.

Dear Clifford, thank you so much for your critical, yet extremely invaluable answer. I think we need to somehow standardise sampling by perhaps the number of caterpillars we collect not the area or volume. We have a canopy crane established in our site and more cranes are planned to be built across southern China so accessibility is not an issue (but the number of spatial replication is still a big problem). Thanks again for your comments!! Best :)