The benefit of synthetically generated RapidEye and Landsat 8 data fusion time series for riparian forest disturbance monitoring

Insect defoliation causes forest disturbances with complex spatial dynamics. In order to monitor affected areas, decision makers seek but often lack information with high spatial and temporal precision. Within the context of a riparian Tugai forest disturbed by the insect Apocheima cinerarius, this study examines whether the analysis of a RapidEye time series would benefit from the availability of synthetically generated images at the spatial resolution of RapidEye and the additional temporal resolution of Landsat 8. We applied the Enhanced Spatial and Temporal Adaptive Reflectance Fusion Model (ESTARFM) to downscale Landsat 8 Normalized Difference Vegetation Index (NDVI) scenes to concurrent RapidEye NDVI scenes. We a) performed a pixel-based regression analyses in order to evaluate the quality of the synthetically created NDVI products and b) examined if forest disturbance maps produced with synthetic images improve the accuracy of disturbance detection. The results show that the ESTARFM predictions have a sufficiently good accuracy, with a correlation coefficient between 0.878 < r < 0.919 (p < 0.001) and an average root mean square error 0.015 < RMSE < 0.024. The overall accuracy of forest disturbance detection with added synthetic images increased from 42.8% to 61.1 & 65.7% compared to the original data set. Forest recovery detection accuracy improved from 59.5% to 80.9%. The main source of error in the disturbance analysis occurs during the temporal interweaving between foliation and defoliation in spring.