I'm not quite sure I fully understand your question -- maybe this is not obvious to you, but using markers for selective sequencing and random whole genome shotgun sequencing will not inherently capture the same sequences (or sequence diversity). One is specific (16S) and the other is random (metagenomic sequencing) -- they might have some overlap of signal with regards to 16S, but this only comes (from my experience) with great metagenomic sequencing depth AND accurate and strain unbiased assembly.
how different are the species abundance simply using 16s rrna data in 16s rrna amplicon sequencing data and shotgun metagenome data (by selecting 16s rrna reads) based on current technology?
They are different -- this depends on a lot of factors: how complex a system you are studying, how deeply your sequence your environmental sample, etc. etc.
I want to know if current technology could provide comparable result in species abundance profiling simply using 16s rrna reads from these two data types?
It's possible to compare amplicons to metagenomic data, but as mentioned above this depends on the complexity of the environmental sample and the depth to which you sequence. How you interpret your data analysis is also a factor here. Obviously, there will not be a 1 to 1 ration of your 16S amplicon reads to a corresponding metagenomic 16S reads on the basis of the type of sequencing, so the answer is they are not directly comparable, but observing both might provide you with information regarding your taxonomy. 16S primers are not universal, so by looking at metagenomic data you might observe taxa not amplified by 16S primers -- does this make sense?
If not, what are the possible reason? and is there other ways to supplement?
I don't understand the "possible reason" part -- you're comparing a PCR amplified marker sequence with a random sample from a genome -- yes, you might have an overlap, but it's unlikely, especially with a metagenomic sample. It's this reason that it's hard to compare amplicon sequencing with metagenomic sequencing -- more depth increases the chances you will have overlapping information. You can supplement by choosing other marker regions other than 16S, but they all will have these inherent problems.
Also, by the way, this is a pet peeve of mine, it's 16S, the 'S' is capital and stands for Svedberg.