An optimization process to reduce nonstationarity problems yet preserve the reliability of the radar echo signal distribution is developed combining the conventional χ2 test and numerical simulation techniques. As an application to the real data, we determine the optimized data length to be 128 elements per set for SOUSY data and 256 elements per set for Chung-Li data. For each of such data sets, the echo power P and the modified Rice parameter γ are calculated and their statistics are examined. When their time averages are plotted as functions of height, we frequently observe the existence of maximum P along with minimum γ in Chung-Li data. Such a coupling can be explained as an evidence of the existence of turbulence layer. The cause of the occurrence of turbulence layers in the troposphere over Chung-Li area is briefly discussed.