Dividing your query into "instant" and "later" deaths makes the speculation more difficult; frankly, it may require taking a wide sample of answers of coming up with your own theories, averages, etc. and then be able to show how and why you came up with those figures, which will only strengthen the impact of your project in my opinion, giving you a boost in the teacher's impression of the work, thought and synthesis or your efforts.
For example, you may find several, and conflicting, numbers about the number of deaths and you and theorize a "range" (median) that may have been killed instantly based upon the research of "x" number of studies and list them in a bibliography of sorts (where you got your data from), and then from other research, studies, and numbers provided on deaths through a certain date period, like the UCLA research project http://www.aasc.ucla.edu/cab/200708230009.html that tracked deaths through December 1945, and theorize and show a number of deaths taking place after the initial blasts (instant) and the number that later died from radiation exposure and other related nuclear diseases.
A few good starting point for you hopefully
Between 60,000 and 80,000 people were killed instantly.
The final death toll was calculated at 135,000
About 40,000 people were killed instantly and a third of the city was destroyed. The final death toll was calculated as at least 50,000.
Good luck and great learning!