Forest carbon sink is important for mitigating the climate change. Monitoring and quantifying the aboveground forest carbon sink with remote sensing has become a hotspot in the research field of forest remote sensing. Wuyi Mountain National Reserve in Fujian province was chosen as the case study area, while 24 permanent sampling plot data in 2003 and Landsat TM remote sensing images in the same year were collected as the main sources of information to estimate the forest carbon stocks in the study area in 2003 by means of four methods of multiple linear regression, K-nearest neighbor classification, artificial neural networks, and land cover classification, followed by geographically weighed regression and spatial pattern analysis of forest carbon density. Study results show that: ① Among the four models, artificial neural network outperformed others with the highest correlation coefficient, the lowest standard error and the minimum average relative error;② The difference of carbon density was not very obvious among three functional areas of the reserve, average carbon density was 52.40 t/hm~2, totaly 2 785 423 t; ③ Carbon density was negatively correlated with the elevation and aspect, positively correlated with slope, hydrological conditions of forest soil and plant growth conditions; ④ With the decrease of altitude and enhancement of human disturbance, from the core zone, buffer zone to the experimental area,the spatial aggregation of forest carbon density tended to become more weakened and more fragmented.