There are ways to convert tetraploids to diploids.
One way is to tissue culture either the pollen or the ovule (gametes) since they are already 'diploid' in a tetraploid. In other words a tetraploid has 44 chromosomes and the gametes have 22 chromosomes.
Producing diploids from tetraploids is not extremely unusual - look for research describing the production of haploids from diploids. In practice a very large percentage of all plant species have in the distant past naturally doubled all their chromosomes one or more times. The end result is that many diploid species are ancient tetraploids that now act like perfect diploids (most of the time). So creating a haploid (half of the normal number of chromosomes) can effectively create a diploid form.
There is also another way that researchers can produce 'diploids' from tetraploids and that is by pollinating the tetraploid with pollen from a completely different species. The pollen must be basically almost completely incompatible (you do not want to produce hybrids) and it acts to stimulate the growth of the ovule but the pollen chromosomes are either destroyed or otherwise lost from the developing embryo. Again look for research on creating haploids by using intergeneric crosses or incompatible pollen, etc.
Below are two starting points - there will be many research papers for many different species but not that I know of specifically for tissue culture of daylily pollen.
Barley haploids from intergeneric crosses with maize (Zea mays L.)
http://wheat.pw.usda.gov/ggpag...
Anther and Pollen Culture (Production of Haploid Plants)
http://www.eplantscience.com/i...
You might try pollinating the tetraploid daylily with Lilium pollen or with the pollen from other members of the Hemerocallidoideae (older - Hemerocallidaceae), such as Dianella or Stypandra or Simethis or perhaps Phormium, etc.
However, when a tetraploid is converted to a diploid any unusual phenotypes that are based on four copies of each gene may be lost when only two copies are present. Daylilies are somewhat unusual in that hybridizers have continued to select in both the diploid and tetraploid gene pools. The tetraploids have developed some phenotypes that are not obviously present in the diploids. Whether there are genotypes present in the tetraploids that do not have corresponding genotypes in the diploids (other than those based on four alleles versus two alleles - if any) is not known.
Any diploid produced from a tetraploid will have a random sample of two of the four chromosomes for each of the 11 sets.
When looking for new mutations it is always best to do so in diploids rather than tetraploids (far easier in diploids). That is also the case for selecting for new phenotypes (genetically based).
