Heterogeneous and patchy landscapes where vegetation and abiotic factors vary at small spatial scale (fine-grained landscapes) represent a challenge for habitat diversity mapping using remote sensing imagery. In this context, techniques of spectral mixture analysis may have an advantage over traditional methods of land cover classification because they allow to decompose the spectral signature of a mixed pixel into several endmembers and their respective abundances. In this work, we present the application of Multiple Endmember Spectral Mixture Analysis (MESMA) to quantify habitat diversity and assess the compositional turnover at different spatial scales in the fine-grained landscapes of the Cantabrian Mountains (northwestern Iberian Peninsula). A Landsat-8 OLI scene and high-resolution orthophotographs (25 cm) were used to build a region-specific spectral library of the main types of habitats in this region arboreal vegetation; shrubby vegetation; herbaceous vegetation; rocks–soil and water bodies). We optimized the spectral library with the Iterative Endmember Selection (IES) method and we applied MESMA to unmix the Landsat scene into five fraction images representing the five defined habitats (root mean square error, RMSE 0.025 in 99.45% of the pixels). The fraction images were validated by linear regressions using 250 reference plots from the orthophotographs and then used to calculate habitat diversity at the pixel (-diversity: 30 30 m), landscape (-diversity: 1 1 km) and regional ("-diversity: 110 33 km) scales and the compositional turnover (- and -diversity) according to Simpson’s diversity index. Richness and evenness were also computed. Results showed that fraction images were highly related to reference data (R2 0.73 and RMSE 0.18). In general, our findings indicated that habitat diversity was highly dependent on the spatial scale, with values for the Simpson index ranging from 0.20 0.22 for -diversity to 0.60 0.09 for -diversity and 0.72 0.11 for "-diversity. Accordingly, we found -diversity to be higher than -diversity. This work contributes to advance in the estimation of ecological diversity in complex landscapes, showing the potential of MESMA to quantify habitat diversity in a comprehensive way using Landsat imagery.