| Abstract: | Numerical simulation of propagation through atmospheric turbulence of an initially spherical wave is used to calculate irradiance variance σ(2)(I), variance of log irradiance σ(2)(ln I), and mean of log irradiance ?In I? for 13 values of l(0)/R(f) (i.e., of turbulence inner scale l(0) normalized by Fresnel scale R(F)) and 10 values of Rytov variance σ(2)(Rytov), which is the irradiance variance, including the inner-scale effect, predicted by perturbation methods; l(0)/R(f) was varied from 0 to 2.5 and σ(2)(Rytov) from 0.06 to 5.0. The irradiance probability distribution function (PDF) and, hence, σ(2)(I), σ(2)(In I), and ?ln I? are shown to depend on only two dimensionless parameters, such as l(o)/R(F) and σ(2)(Rytov). Thus the effects of the onset of strong scintillation on the three statistics are characterized completely. Excellent agreement is obtained with previous simulations that calculated σ(2)(I). We find that σ(2)(I), σ(2)(In I), and ?ln I? are larger than their weak-scintillation asymptotes (namely, σ(2)(Rytov), σ(2)(Rytov), and - σ(2)(Rytov)/2, respectively) for the onset of strong scintillation for all l(0)/R(f). An exception is that for the largest l(0)/R(f), the onset of strong scintillation causes σ(2)(ln I) to decrease relative to its weak-scintillation limit, σ(2)(Rytov). We determine the efficacy of each of the three statistics for measurement of l(0), taking into account the relative difficulties of measuring each statistic. We find that measuring σ(2)(I) is most advantageous, although it is not the most sensitive to l(0) of the three statistics. All three statistics and, hence, the PDF become insensitive to l(0) for roughly 1 < β0(2) < 3 (where β0(2) is σ (2)(Rytov) for l(0) = 0); this is a condition for which retrieval of l(0) is problematic. |
