Habitat selection and spatial distribution of white eared-pheasant Crossoptilon crossoptilon during early breeding period

摘要

In order to characterize the habitat selection and spatial distribution of white eared-pheasant Crossoptilon crossoptilon breeding pairs, we surveyed the species around Zhujie Monastery in Daocheng C ounty, Sichuan P rovince, China from January to June 2003 and April to June 2004. Line transects of systematic sampling were used to characterize several environmental variables of the whole study area before the occurrence of breeding pairs. The distance between any two nearest transects was 100 m and grids (100 m × 100 m) were obtained every 100 m along transects. Random transects were then applied to count the number of the breeding pairs and record the actual positions they occurred after the emergence of the breeding pairs. Random transects as a whole traversed all the grids. Moran’s I coefficient of dependent variable’s residuals was the lowest when the lag was 200 m. Therefore, the grid size was 200 m × 200 m and 99 grids were obtained by uniting four coterminous 100 m × 100 m grids into one. The grids with presence of the breeding pairs were defined as detected grids and valued 1, the grids with absence as undetected ones and valued 0. Distance to the nearest water, shrub cover, shrub height and herb height showed significant differences between detected and undetected grids. Univariate analysis of logistic regression was derived with the above variables and their first-order interaction as independent variables. In univariate analysis the variables with probability less than 0.30 were remained. Forward elimination stepwise logistic regression was conducted with the remained variables as independent variables. Finally, regression equation with the lowest AIC_C value was regarded as the optimal model. The model could be formally expressed as: π ( x ) = e ^{g ( x )} / 1 + e ^{g ( x ) ,} g ( x ) = 2.473 + 0.223 × shrub cover – 0.011 × distance to nearest water. The model suggested that habitat selection of breeding pairs was negatively related to distance to nearest water and positively related to shrub cover. The value of CT that suggested the predictive accuracy of the whole model was 78.79%, the value of CP suggesting the predictive accuracy of the detected grid 77.55% and the value of CA suggesting the predictive accuracy of the undetected grid 80.00%. The observed points with the presence of the breeding pairs were mostly located in the grids with high probability of occurrence predicted by the model. We surveyed the study area again in 2004 to test the goodness-of–fit of the model. The observed detected and undetected grids in this year didn’t show significant differences with what the model predicted . The model could well predict the habitat selection of the breeding pairs. Variograms of the breeding pairs in 0 ^o (east-west), 45^o (northeast-southwest), 90^o (north-south) and 135^o (northwest-southeast) directions were calculated. Results showed that the directional variograms could all be regressed by spherical model. Spherical model variograms in 4 directions revealed the breeding pairs had distinctive spatial properties that were effectively quantified by the parameters nugget ( C ₀), sill ( C ₀ + C ), rang ( a ) and fractual ( D ). The lowest nugget, sill and rang existed in 0^o direction, the highest nugget in 90^o, the highest sill and rang in 45^o. The Kriging distribution of the breeding pairs was obviously not homogenous but rather clustered [ A cta Zoologica Sinica 51(3): 383 –392, 2005].