Comparative study on NDVI time series fitting methods based on GF-1 WFV data

The domestically developed high-resolution satellite GF-1 WFV serves as a vital data source for regional vegetation dynamics monitoring. However, its NDVI time series frequently suffers from noise and missing values, making the selection of an appropriate fitting method crucial for ensuring the accuracy of subsequent phenological parameter extraction. Taking Aihui District, Heihe City as the case study area, this research systematically evaluated the performance of three commonly used methods—double logistic, Savitzky-Golay filtering, and Fourier series fitting—in reconstructing NDVI time series. The analysis was based on GF-1 WFV data from 2024 that underwent radiance calibration, atmospheric correction, and quality control processing. It should be noted that the findings are derived from a single growing season (2024), and interannual variability may affect method performance. Within this observational context, the double logistic function fitting exhibited the lowest RMSE (0.085) among the three methods and yielded fitting parameters with clear phenological interpretability. Savitzky-Golay filtering effectively preserves local sequence details with high computational efficiency, while Fourier series fitting excels at capturing stable interannual periodic signals. Based on these numerical results, the double logistic function fitting appears preferable for temperate-region vegetation monitoring aimed at explicit phenological parameter extraction, whereas Savitzky–Golay filtering may be considered when rapid smoothing and preservation of growing-season curve morphology are prioritized. This study provides concrete evidence and practical reference for selecting temporal reconstruction methods when conducting vegetation dynamics analysis based on domestically produced high-resolution satellite data, though further validation with independent ground-truth data is warranted to confirm these observational findings.