Abstract: Based on the observation data and the export data of microclimatic gradient system during growing and non-growing seasons of Stipa krylovii grassland in Inner Mongolia, this paper investigated the accuracy of variational method (VT) on estimating the surface heat fluxes of grassland. The results showed that the changes of sensible and latent heat fluxes estimated by VT were consistent with those obtained by open-path eddy covariance method (EC), and the fluxes estimated by VT could better satisfy the surface energy balance equation. The sensible heat fluxes (H) from VT were mostly higher than the corrected fluxes from EC, and the values and their difference from these two methods all got the maximum at twelve o’clock in daytime, and were close in nighttime. The variations of latent heat fluxes from VT lagged behind those from EC. VT could solve the problems that the conventional Bowen ratio energy balance method (BREB) was computationally unstable and produced spurious large values when the Bowen ratio was in vicinity of -1.Compared with EC and BREB, VT was more of rationality and stability.

Key words: Wetland ecosystem health, Temporal-spatial scale, Health indicator