Kalman Assimilation Model for Retrieving Time-Series Chlorophyll Content Based on Multisource Remote Sensing Images

The chlorophyll (Chl) is closely related to vegetation respiration and photosynthesis, and its long time series data with high spatial resolution is significant for dynamics monitoring for vegetation growth status and environmental management. However, extant methodologies for long-term Chl content retrieval are encumbered with data lacunae and imbalance in high spatial resolution and continuous time series cover-age. To address these limitations, we have developed a novel Chl assimilation inversion model with Kalman filtering method (KFCAM) to synthesize high spatiotemporal resolution Chl content data in Wuhan. The KFCAM is used to I) obtain Chl time-varying rates from low-resolution (Landsat-8) images and high-resolution Chl information from Sentinel-2/SPOT5 images, simultaneously II) calculating the state error and Kalman gain; then III) this model integrates the high-resolution information into the time series by state updating model, and finally IV) synthesis continuous time series images for Chl with 10 m resolution. The results demonstrate that I) the synthetic Chl image with KFCAM exhibits high accuracy with a R2 of 0.906±0.047 and a RMSE of 1.801±0.369, and II) the RMSE values reveal a 12% and 33% reduction compare to the Spatio-Temporal Gap-Filling (STGF) and temporal adaptive reflectance fusion model (STARFM) respectively. III) Furthermore, in comparative experiments that considered different assimilation time intervals and changes in land cover, KFCAM outperformed two reference models in terms of stability and anti-interference capability. It provides a novel means for large scale and global Chl detection.