Beta diversity of geometrid moths (Lepidoptera: Geometridae) in an Andean montane rainforest

Turnover in species composition of the extremely species-rich family Geometridae (Lepidoptera) was investigated along an elevational gradient ranging from 1040 m to 2677 m above sea level. Moths were sampled using weak light traps (30 W) in three field periods in 1999 and 2000 in an Andean montane rainforest in the province of Zamora-Chinchipe in southern Ecuador. A total of 13 938 specimens representing 1010 species were analysed. Similarities of ensembles of all geometrid moths and of the subfamilies Ennominae and Larentiinae were calculated using the NESS index (with m(max)). Ordinations performed using nonmetric multidimensional scaling (NMDS) and correspondence analysis depicted a gradual change of the ensembles along the altitudinal gradient. Extracted ordination scores significantly correlate with altitude (-0.97 less than or equal to r less than or equal to -0.95, P < 0.001) and with ambient air temperature (0.93 <= r <= 0.97, P < 0.001). Temperature is therefore assumed to be the most important abiotic determinant responsible for the species turnover among the moths. Matrix correlation tests were performed in order to compare faunal matrices with matrices derived from available environmental factors. Both tree diversity and vegetation structure significantly correlate with faunal data, but tree diversity explains considerably more of the data variability (range: Mantel r = 0.81-0.83, P < 0.001) than vegetation structure (range: Mantel r = 0.35, P < 0.005 to r = 0.43, P < 0.001). Tree diversity also changes gradually and scores of the first NMDS dimension are highly significantly correlated with altitude (r = 0.98, P < 0.001). A common underlying factor such as ambient temperature might also be responsible for such vegetation changes. Additionally, simulated model data was developed that assumed a constant turnover of moth species and equal elevational ranges of all species involved. Despite the simplicity of the models, they fit empirical data very well (Mantel r > 0.80 and P < 0.001 in all models).