It is like my colleague Hongliang states. The best soil depth for sampling depends on the ecosystem on that soil. Moreover, one does have to take samples at different soil depths to have a good idea of changes in carbon or other variables in the soil profile. Hence, a 'best' or ideal soil depth doesn't exist. You will have to sample the full depth of a profile, with at least 5 depth intervals so that you can plot the changes i.f.o. depth of the variables considered. Moreover, a depth profile model requires a minimum of at least 5 intervals, to enable a fit of a depth model. Only then one can understand the profiles of different variables in a soil. To estimate the total carbon content in the soil, one can the numerically integrate the regression model of a soil depth profile for all soil variables!
One measurement in a sol profile is definitely insufficient.