Adhesion failure is a significant reliability concern and a major barrier to commercializing MEMS devices. In this paper, a beam-peel-based experimental setup was improved to measure the adhesion energy between a microcantilever and a substrate at different humidity levels. The microcantilever arrays were separated from the substrate and their fixed ends were directly attached to a piezoactuator to control the beam displacement with sub-nanometer accuracy. The experimental setup was sealed in a chamber with precise humidity control. An in situ interferometer was used to measure the beam deflection and crack length during the peel test. To examine the effects of surface roughness and relative humidity on adhesion energy, four different surface pairs were measured at humidity levels ranging from 40 to 92%. Before testing, the microcantilevers and substrates were scanned using an Atomic Force Microscope. Surface roughness parameters and the exact probability density function of the asperity heights were extracted and directly entered into a statistics-based model. An Extended-Maugis-Dugdale single asperity meniscus model considering both asperity deformation and solid surface interaction was coupled with the Pearson surface statistical model to develop an improved elastic asymmetrical surface meniscus model. The model compared favorably with the experimental data. (C) Koninklijke Brill NV, Leiden, 2008.