A background field-based framework for reconstructing spatially and temporally continuous daily NDVI (1982–2020) from the AVHRR LTDR

• Developed a background field-based framework (BFF) for daily NDVI reconstruction. • Incorporated background field information to exploit the full potential of LTDR NDVI. • Generated seamless daily NDVI (1982–2020) for mid- to high-latitude forests. • Provides a solid data foundation for ecological indicator development and applications. High-quality and long-term Normalized Difference Vegetation Index (NDVI) time series are essential for investigating vegetation dynamics and their responses to climate change. The Advanced Very High Resolution Radiometer (AVHRR) provides one of the longest NDVI records available. However, these data are often affected by artifacts caused by clouds, fog, sensor, and other factors, which significantly limit their usability. Moreover, existing NDVI reconstruction methods struggle to address large temporal gaps and cannot generate seamless daily NDVI time series effectively. In this study, we develop a background field-based framework (BFF) to reconstruct spatially and temporally continuous daily NDVI time series over forested regions in the mid- to high-latitudes of the Northern Hemisphere from 1982 to 2020, fully leveraging the long-term AVHRR Long-Term Data Record (LTDR) dataset. The BFF method significantly outperforms five widely used approaches (Savitzky–Golay, Double Logistic, Harmonic Analysis, Iterative Interpolation, and Whittaker Smoother). Based on experiments in which subsets of high-quality observations were randomly selected for training and testing (train sample ratios from 80% to 10%), BFF consistently achieved the lowest reconstruction errors (e.g., RMSE: 0.0124–0.0314) and the highest R 2 (0.949–0.993). Compared with the existing methods, BFF reduced MAE and RMSE by up to 85.5% and 73.7%, respectively, and increased R 2 by up to 11.3%. Visual assessments further confirm the superior performance of BFF in bridging both temporal and spatial discontinuities. This study provides an effective solution for producing high-quality daily NDVI time series over long periods, which is fundamental for developing ecological indicators, monitoring ecosystem dynamics, and investigating climate–vegetation interactions.