In radar detection, the existence of the clutter edge in the reference samples considerably degrades the performance of the detector. Hence, clutter edge estimation not only improves the CFAR detectors, but also can be used for partitioning the various areas of the clutter in the clutter map. In this paper, we propose an adaptive algorithm for detecting the clutter edge between two Weibull clutter areas with different shape and scale parameters. First, we apply the UMPI detector as a data pre-processor which is optimal among the invariant detectors. Then, based on the results of the previous pre-detection, the reference data are partitioned by an adaptive algorithm and the position of the clutter edge is estimated. For this, we need to the probability density functions of the pre-detector output samples which are provided by the nearest neighbor method.