Niche-based models (NM), which are based on the correlation between species occurrence records and environmental predictors, have been generally used to project and assess the effects of climate change on species distribution range. However, most of NMs do not produce consistent prediction results of species distribution, with significant projection differences in either magnitude or direction of a species range change in response to climate change. One of the recommended solutions is to use ensembles of forecasts by simulating from more than one set of model training data, model techniques, model parameters and environmental scenarios combinations to identify a range of uncertainties. However, ensemble forecasting is in its infancy in assessing the effects of climate change on species distribution, as little is known about the sources of variations and its contributions to uncertainties in ensemble forecasting. The objective of this research is to partition and map uncertainties in ensemble forecasting of species range change under climate change. In our study, the eight niche-based models (Random forest, RF; Generalized boosted method, GBT; Generalized linear models, GLM; Generalized additive models, GAM; Classification tree analysis, CTA; Artificial neural network, ANN; Mixture discriminant analysis, MDA; multivariate adaptive regression splines, MARS) and the nine sets of model training dataset (initial data were randomly divided into two parts: mode training data and testing data), three global circulation models (GCM) (MIROC32_medres, JP; CCCMA_CGCM3; CA; BCCR-BCM2.0, NW) and one pessimistic SRES emissions scenarios (A2) were used to simulate current potential distribution(Baseline, 1961-1990) and to project future potential distributions at three time-slices (2010-2039, 2040-2069, 2070-2099) of Chinese pine (Pinus tabulaeformis). Totally, we obtained 72 predictions for the current distribution and 216 projections for the future distribution (herein termed ensemble forecasting). The area under the curve (AUC) of receiver operator characteristic (ROC) curve, true skill statistic (TSS) and Kappa were employed to objectively assess the performance accuracies of model projections. We developed ClimateChina software to downscale current and future climate data (GCM) and to calculate seasonal and annual climate variables for specific locations based on latitude, longitude and elevation. We performed a three-way analysis of variance (ANOVA) to partition the sources of uncertainties for each grid cell, with model training data, model techniques and GCMs as factor and species occurrence probability as response variable. We then obtained the sum of squares which can be attributed to training data, model techniques and GCMs and their interactions (training data×model techniques, GCMs×training data, model techniques×GCMs, training data×model techniques×GCMs). Results indicated that RF, GAM, GBM, GLM and MARS achieved successful performances in simulating the current distribution of Chinese pine with less sensitive responses to the differences from the 9 set of model training data in comparison to MDA. All projections showed different changes of the species distribution range in response to climate change. Difference among current predictions mainly located surrounding area of current distribution of Chinese pine. Differences among the future species distribution projections would increase with increasing time horizon. The distribution area projections with the great variations would expand while the distribution area projections with little variations would reduce. Model technique contributed to the largest variation in projections, but GCM and model training data had little influence on the variability of projections. This practice would reinforce our understanding of the sources of uncertainties in modeling species distribution under climate change.