Hmm, well, it's not ideal, obviously, but it sounds like you had no choice. I think you'll just have to be clear in arguing why you've done it this way, and replicate the experiment. Since it's on the same field and location, but blocked temporally, one concern might be that certain abiotic conditions change over time. Thus, I would think it's best to alternate/intersperse the within-line and between-line crosses temporally so that you have, over time: within, between, within, between, etc. You don't want to do all of the, for example, within-line crosses, replicated several times, and then move on to doing all the between-line crosses, since one might suspect that seasonal changes or changes in the soil conditions might confound your contrast of within vs. between (which, depending on your question, could be an integral contrast to have for the analysis). Also, you could probably help avoid criticism by taking all the abiotic measures you can...  So, (I'm not a plant person) soil salinity, pH, light, temperature, etc. to help you argue that the conditions were stable over time. If they were not stable over time, they you can include these measures as covariates in your analysis (so for example, mean pH for each block, or if you have different dates within blocks then you can use the different measures for each date) to help you identify environmentally induced variance and reveal the different forms of genetic variance that you're probably looking for.  

The short answer is no, but Karl Grieshop's comments are generally very appropriate, although it sounds like he is referring to spatial rather than temporal confounding with the difference between parental and hybrid lines as two classes.  The idea of trying to find and measure environmental covariates to correct for biases in comparisons is appropriate in principle, the problem being to ascertain what are appropriate covariates to use.  Spatial analysis, designed to detect and correct for trends in environmental effects (without attempting to identify specific environmental factors) might be appropriate, but it could be difficult to use effectively in your situation, and you would have to see what trends can be detected.

But, with the confounding you have, you may just need to admit to the problem and make the best of a bad job.

Comparing parental vs. hybrid performance may be interesting. However, it is usually not the goal of diallel analysis. In my opinion, the performance of a parent is much less important than its GCA. You don't need any data from the parents to estimate their GCA.

I would advise you NOT to combine your data for parents and hybrids for analysis since the experimental design is not appropriate for such a comparison. Assuming that the hybrids were properly randomized and replicated, you can use the data for your hybrids to estimate GCA, SCA, and heritability. Then, if you'd like, you can analyze the parental performance separately. You shouldn't make inferences from one experiment to the other. However, you could potentially compare ranks of parental performance vs. parental GCA, which might be interesting.

Best of luck with your analysis!

Well, I was referring to temporal blocks, since he said he's done the parents and hybids on the same field/location. But actually, Jack is right... I got caught up thinking about my own work; sorry. You won't need the parent lines (the diagonal) to analyze the diallel, except for some very specific questions. 

Thank you to Karl and Jack for their contribution but am still waiting for more response,idea and suggestion for way out

You may try to find out homogenity of error variance of both trial. If homogenity exist, then you can use all the data together and proceed for diallel.

answer is no, since add no value to aim of the design

this helps in find out the linkage between high-GCA parents and superior per se of a trait.  this not so in many crops but there is chance to find out right parental combination and the same parent can be use with superior tester for hybrid production. i suggest you to  follow the Dr Arunachalam (1974) paper dealing with the L x T mating design including the parents called modified method of kempthorne paper.   Arunachalam, V., 1974. The fallacy behind the use of a modified linex tester design. Indian J. Genet. 34:280-287.

Govind