Because of its high level of spatial heterogeneity, the fine root biomass of forest ecosystems can be accurately estimated only by adopting an appropriate sampling strategy. In this study, 100 soil cores each 1 m in length were taken from randomly distributed locations in a Chinese fir plantation at Sanming, in Fujian Province. Then, the soil cores were each divided into ten 10 cm sections. The cores were then washed with tap water, after which the roots were removed and separated into two diameter classes (0-1 mm and 1-2 mm). Dead roots and large living roots were discarded. In order to design a sampling strategy suitable for the study of fine roots in forest ecosystems, we examined the spatial heterogeneity of the distribution of fine roots of different types (Chinese fir, undergrowth vegetation, and total fine roots). The minimum sample sizes required to estimate fine root biomass were also calculated. The coefficient of variation (CV) and the required sampling sizes for different types of fine root biomass all increased with increasing diameter class (0-1 vs. 1-2 mm). As soil depth increased, the fine root biomass per unit ground area decreased in both diameter classes, and their respective CVs and required sampling sizes increased accordingly. With the exception of the 1-2 mm fine roots of undergrowth vegetation, which had a CV of 120% for biomass per unit ground area, all fine roots from different diameter classes and different root types had CVs of biomass per unit ground area consistently between 40%-90%. Meanwhile, the biomass per unit ground area of 0-2 mm roots of different types was characterized by medium variations. With the exception of 0-1 mm fine-roots in 0-20 cm and 50-60 cm soil depth, which exhibited medium variations in biomass per unit ground area, all fine roots of different diameter classes and different root types were characterized by high levels of variation. The total fine roots (0-2 mm) exhibited high levels of variation in biomass per unit ground area at all soil depths, except for the 0-10 cm and 10-20 cm depths, where there were medium variations. The Shapiro-Wilk test showed that, among different diameter classes and root types, only the biomass per unit ground area of the 0-2 mm fine roots of Chinese fir and the total fine roots were normally distributed; the roots of other diameter classes, soil depths, and types conformed to a noticeably right-skewed distribution. The minimum sampling sizes were estimated using the Monte Carlo simulation, at a confidence leve1 of 95% and with a precision level of 80%. To determine the biomass per unit ground area of 0-1 mm, 1-2 mm, and 0-2 mm roots, 95, 96, and 32 cores were needed for Chinese fir roots, 98, 98, and 63 cores were needed for understory roots, and 93, 93, and 18 cores for the overall roots. To estimate biomass per unit ground area for roots from the top soil layer to the lowest soil layer (at 10 cm intervals), 39, 63, 117, 97, 119, 94, 613, 318, 478, and 532 cores, respectively, were required for the 0-1 mm fine roots, 97, 168, 207, 302, 219, 333, 769, 598, 722, and 3168 cores, respectively, were required for the 1-2 mm fine roots, and 38, 59, 103, 125, 120, 123, 382, 262, 432, and 628 cores, respectively, were required for the 0-2 mm fine roots.