This letter presents a modification to a very recent iterative microphone removal array design process, called the Array Reduction Method (ARM). The ARM iteratively reduces an array based on the product of sidelobe and main lobe criteria, yet the main lobe may become distorted and the beamformer output can disproportionately favor one of the criteria. The Adaptive Array Reduction Method presented here uses the derivative estimates of the maximum sidelobe level and main lobe width with respect to the number of microphones removed, and the distortion of the main lobe during iteration, to produce an improved source image map.