Global climate change is closely linked with forest ecosystem carbon (C) cycling. Quantitative assessment of forest C budgets is of importance in ecosystem and global change sciences. Studies on forest ecosystem C cycling have been advanced considerably during the past 30 years, but there are still large uncertainties in the C budgets at both global and regional scales. These uncertainties can be attributed not only to the complexity of forest ecosystems but also to the methodology applied. In this paper, we reviewed the fundamental concepts and major field measurement methods of C cycling in forest ecosystems. First, we the concepts of concentration, density, flux, allocation and turnover. Carbon concentration and stock are static attributes of the C cycling, while C flux and turnover are dynamic ones. In meteorology, CO₂ concentration can be expressed in three ways: mass/molar density, molar fraction to the moist air, and mass/molar fraction to the dry air, among which the dry molar fraction is the most convenient and conservative. In ecology, the C concentrations in biomass, necromass and soil are mostly expressed as mass fraction to the dry weight. Carbon density is defined as the C stock per unit forest area. Carbon flux is the mass flow per unit time through unit forest area (or particular organ of an individual tree). Carbon allocation pertains to standing biomass distribution, the absolute and relative partitioning of gross primary production to C flux components. Net primary production, the most frequently investigated C flux, is often underestimated due to missing components. Net ecosystem production is often higher than the net C accumulation rate of the ecosystem. After clarifying the related concepts and terminologies, we modified the conceptual framework of forest C cycling. In the second part, we discussed field measurement methods of forest C cycling, focusing on principles, pros and cons, and potential uncertainties in measuring forest C fluxes with biometry (or inventory), chamber and eddy covariance methods. The biometry method is the most widely-used method for measuring forest C pools particularly for aboveground biomass, which requires little equipment or technology but long-term commitment, and lacks of fine temporal resolutions. The chamber method (e.g. ecophysiological approach) often focuses on specific ecological processes such as photosynthesis and respiration, which provides process-based parameters for forest C cycling modeling with fine temporal resolutions, but needs sophisticated expensive instrumentation and is difficult to up scale. The eddy covariance method directly measures C and water vapor between forest canopy and the atmosphere at ecosystem level. It provides continuous automated measurements with high frequency and reduces potential human errors, but it required expensive instrumentation and sophisticated data-processing and is limited by spatially-representative sampling locations. In the last part of this paper, we recommend four ways to reduce uncertainties in forest C accounting: (1) Appropriately use biomass allometric models to estimate tree biomass and its increment; (2) measure all C flux components in the forest ecosystem, or at least report the missing components; (3) estimate the uncertainty of C fluxes rather than only report the mean value of estimation; and (4) cross-validate the C fluxes with independent methods.