贵州省日尺度NDVI时间序列重建法的差异对比OA

Daily-scale NDVI time series data can more precisely reflect the temporal dynamics of vegetation.However,the combination of persistent cloud cover and fog in Guizhou province with factors such as satellite revisit cycles has severely compromised its daily-scale NDVI time series data.In this study,daily-scale surface reflectance products from MOD09GQ and MYD09GQ were used to calculate the daily-scale NDVI time series data of Guizhou province in 2023,followed by maximum value compositing.6 models belonging to three categories of time series reconstruction methods—linear interpolation,Savitzky-Golay(SG)filtering,and Whittaker filtering—were selected.Among them,the SG filter was configured with three different combinations of sliding time window sizes and polynomial fitting coefficients,while the Whittaker filtering was set with two different roughness parameters.The NDVI simulation effects of the 6 models at the single-scene and monthly scales were compared and verified to evaluate their performance in Guizhou province,including gap-filling capability,fidelity,correlation,fitting performance and reconstruction accuracy.The results showed that the 6 models exhibit differences in their performance on the daily-scale NDVI time series data of Guizhou province in 2023.The Linear interpolation performed the best in gap-filling capability.Both Linear interpolation and SG_2_3 demonstrated optimal fidelity with their R² values reaching 1.000.SG_3_5 showed the strongest correlation with the existing 2023 250m×250m monthly NDVI dataset for Guizhou province on a monthly scale,with a correlation coefficient(R)of 0.557.In terms of fitting performance and reconstruction accuracy,the SG_2_15 was the best-performing method,with simulated NDVI annual average RMSE and MAE being 0.0138 and 0.0871,respectively.These findings indicate that each model has its own strengths,and the most suitable technical approach can be flexibly selected based on specific application requirements and scenario characteristics.