This dataset is constructed through the systematic integration and collaborative processing of multi-source satellite remote sensing data. The specific data sources and their characteristics are as follows:GIMMS NDVI3g v1: This is a third-generation global vegetation index dataset based on the AVHRR sensor, providing biweekly NDVI data from 1982 to 2015 with a spatial resolution of approximately 8 km. The dataset has undergone radiometric calibration, orbital drift correction, and atmospheric effect processing, making it the most widely used data foundation for long-term vegetation dynamic studies in the Arctic.MODIS NDVI product (MOD13C2): Based on the Moderate Resolution Imaging Spectroradiometer (MODIS) data from the Terra satellite, this product provides monthly NDVI data from 2000 to 2015 with a spatial resolution of 0.05° (approximately 5.6 km). The product employs a compositing algorithm to minimize the effects of clouds and aerosols, offering high temporal consistency and spatial accuracy.FLUXNET2015 Arctic flux site data: This includes vegetation observation records from multiple Arctic tundra, shrubland, and wetland ecosystems, and is used for the accuracy validation of NDVI products and the calibration of phenological parameters.

 This dataset is constructed through a systematic multi-source remote sensing data fusion and vegetation dynamic reconstruction process. First, radiometric calibration, sensor calibration, and spatial registration are performed on GIMMS NDVI3g v1 and MODIS MOD13C2 NDVI data. BRDF correction is applied to eliminate observation geometric effects, and the data are uniformly resampled to the 10 km EASE-Grid 2.0 grid system. A segmented fusion strategy is employed to establish a continuous time series from 1982 to 2015: for the period 1982–1999, GIMMS data serve as the foundation, with MODIS data used to construct a cross-calibration model; for 2000–2015, the two data sources are fused, and data quality is optimized based on a Bayesian framework. Adaptive Savitzky-Golay filtering combined with the HANTS algorithm effectively removes cloud contamination, atmospheric interference, and snow/ice noise, while a time-series autoregressive model is used to fill in missing values. For the vegetation peak growth periods in July and August, the maximum value composite method is applied to generate monthly NDVI products. Combined with an improved mixed-pixel decomposition model and CAVM vegetation classification information, vegetation signals are extracted and optimized, ultimately forming a spatially continuous and temporally consistent Arctic vegetation index dataset.

 This dataset ensures the reliability and scientific applicability of vegetation index data through a multi-tiered quality control system. In terms of spatiotemporal consistency, multi-source data fusion effectively addresses monitoring blind spots caused by cloud cover, snow/ice interference, and polar night phenomena in single-sensor data in the high-latitude Arctic region. The spatial completeness of July–August data from 1982 to 2015 reaches 97.3%, and temporal continuity exceeds 98.8%. Comparisons based on FLUXNET2015 Arctic flux station vegetation observation data show that the mean absolute error between July–August NDVI and ground-measured values is 0.03, the root mean square error is controlled within 0.04, and the coefficient of determination (R²) reaches 0.86. In comparisons with high-resolution Landsat NDVI products (30-meter resolution) in typical tundra-shrub ecological transition zones, the spatial correlation coefficients for July and August are 0.89 and 0.87, respectively.