Two methods were employed to find spatial regularity in a complicated mount
ain landscape of Beijing, China on the basis of functional and structural a
ffinities. The first approach applied Affinity Analysis based on species co
mposition to landscape. The mosaic diversity of the landscape was 3.5298 >3
, which means the study landscape is complex and controlled by multiple env
ironmental gradients. These landscape types were divided into 3 parts accor
ding to the mean affinity values of 0.2143 and 0.7857 (0.5 +/-1 SD). Modal
sites are the central types of the landscape, which include a zonal broad-l
eaved forest of the region and a conifer plantation replacing the former. O
utliers are found in the highest altitude and the lowest, both have few spe
cies in common with the above two modal types. The remaining landscape type
s are intermediate sites, which are transitional between modals and outlier
s, broadly distributed throughout mountain environments. Neighbor types hav
e more species in common than those more widely separated, which probably d
istributed adjacently in space or in similar quality habitat. The other met
hod employed is the new TWINSPAN analysis by substituting spatial neighbori
ng data of landscape types for species composition data. It clearly divided
the landscape types into three groups, i.e., subalpine, middle and low mou
ntain groups, which were correlated with altitude, as well as influenced by
human disturbance. The new TWINSPAN classification method is more reliable
in finding spatial gradient of patchy landscapes than affinity analysis; h
owever, affinity analysis is useful in finding species diversity pattern an
d the importance of landscape types in a region. Integrating advantages of
the two methods could supply complete and reliable information on how lands
cape types are distributed in space, which environmental gradient dominates
the spatial distribution of the landscape types, as well as where importan
t and unusual types are located.