Abstract: Based on geostatistical method, three algorithms of spatial interpolat ion with elevation as a secondary variable, i.e., simple kriging with varying lo cal means (SKlm), kriging with an external drift (KED), and cokriging (COK), wer e used to calculate the precision of spatial interpolation for the forest duff l ayer depth, and cross validation was conducted. The results showed that among th e three algorithms, KED gave the highest precision because of its taking into ac count both the spatial variation among variables and the factors affecting local spatial change, SKlm did not yield expected precision because of the weaker corre lation between elevation and forest duff layer depth, while COK directly used th e variable elevation to estimate forest duff layer depth but many unexpected res ults yielded for the boundary area due to insufficient samplings. Comparing with the method of inverse distance weighting (IDW), only KED had a higher precision of interpolation, while for SKlm and COK, their interpolation precision was lower, suggesting that when a secondary variable was used for geostatistical interpola tion, the correlation between primary and secondary variables was of significanc e in increasing the precision of interpolation.

Key words: photosynthetic fractionation, CO₂ concentration, post-photosynthetic fractionation., carbon isotope, soil water content