Time series fractional green vegetation cover (FVC) is crucial for monitoring vegetation cover status monitoring, simulating growth processes and modeling land surfaces. Through the integration of remotely sensed data and FVC estimation models, FVC can be routinely and periodically monitored using remote sensing images over large areas. However, due to frequent cloud contamination and trade-offs in satellite sensor design, the FVC estimates from remote sensing data are not continuous, either spatially or temporally, and cannot simultaneously depict details in spatio-temporal variation. Taking the seasonally inundated Zoige alpine wetland in China as a case area, the objective of this paper is to develop a practical and effective approach to quantifying the explicit vegetation FVC details with both high spatial and temporal resolution. In this approach, 30-m multi-spectral images from the Chinese HJ-1A/B (HuanJing (HJ), which means environment in Chinese) satellite constellation with a 2-day revisit time were first composited at 16-day intervals to improve spatio-temporal continuity. Then, a new adaptive endmember selection linear spectral mixture model (ASLSMM) was proposed to improve the accuracy of FVC estimation by considering the endmember dynamics for each pixel. FVC time series were finally estimated by applying the ASLSMM to the cloudless HJ composites. The performance of the model and the spatio-temporal representational capability of the FVC estimation results were comprehensively evaluated using Unmanned Aerial Vehicle (UAV) reference images and ground measurements from an integrated, multi-scale remote sensing experiment. A traditional LSMM with fixed endmembers and the Multiple Endmember Spectral Mixture Analysis (MESMA) model were also used for model performance comparison. The results showed that the R-2 and RMSE values between the FVC estimated from the proposed model and the UAV reference were 0.7315 and 0.1016 (unitless) respectively, which was better than the results from the linear spectral mixture model with a fixed number of endmembers, with R-2 of 0.5924 and RMSE of 0.3821. The R-2 and RMSE values between the FVC estimated from MESMA and the UAV reference were 0.6327 and 0.1578, which was comparable with the ASLSMM. The accuracy evaluation usingmulti-temporal in situ measurements indicated the consistently high performance of the ASLSMM. This study highlights the feasibility of using HJ satellite constellation images to generate the temporally dense and fine spatial resolution FVC estimations for wetland and wetland-like heterogeneous landscape monitoring. The proposed approach can be viewed as a reference for generating FVC datasets from the on-going HJ constellation and similar constellation missions such as Sentinel-2A/B. (C) 2017 Elsevier Inc. All rights reserved.