Remote Sensing Data Fusion to Fill Spatiotemporal Gaps in Harmful Algal Bloom Monitoring

Lakes suffering from harmful algal blooms demand effective monitoring data with high-frequency and high-resolution. However, such data are not consistently available from any single satellite sensor due to cloud coverage and revisit period. To address this limitation, this study evaluates the performance of two widely used spatiotemporal fusion algorithms—STARFM and STNLFFM—in constructing daily high-resolution observations over the third largest lake in China. We assessed fusion accuracies under varying temporal intervals, bloom dynamics, and spatial heterogeneity, using Normalized Difference Vegetation Index (NDVI) derived from cloud-free Gaofen-4 and GOCI-II imagery. Two conventional error-based indices (MAPE and SMAPE) and two novel algal-specific indices (algal area difference, Areadiff, and opposite algal pixels, Pixeldiff) were proposed to evaluate the fusion results. The results showed that remote sensing data fusion can accurately estimate algal bloom extent and distribution with an average Areadiff of 2.9% and an average Pixeldiff of 1.5%, indicating good spatial consistency between fused and true images. STARFM generally performs better for NDVI extraction under short time gaps and for delineating algal bloom extent. This study demonstrates that remote sensing data fusion can enhance bloom monitoring by improving spatial resolution to 50 meters and temporal resolution to one day.