| Abstrak/Abstract |
The monitoring of mangrove and its trend over large areas can be done using multi-temporal remote sensing technology. However, remote sensing data is often contaminated by cloud cover, and its corresponding shadow resulted in missing data. This study aims to assess the performance of the existed gap-filling techniques, such as, linear, spline, stineman, data interpolation Empirical Orthogonal Function (Dineof) and spatial downscaling strategy employing the Proba-V imagery in 100 m, when being used for estimating the missing data and depicting the trend in NDVI from Landsat 8 OLI by using Mann-Kendall. The study was conducted in the Mangrove Forests at Segara Anakan, Central Java which threatened by climate change and human activities. Our result suggested that EOF-based interpolation gave better prediction results and more accurate in predicting longer missing data. Linear interpolation, on the other hand, was accurate to predict shorter missing data. Overall, all interpolation results can reconstruct 64 (spline) to 72 % (Dineof) of missing data in NDVI with the RMSE of 0.10 (Dineof) – 0.13 (spline). A consistently decreasing trend was also found from the four interpolation methods, which showed the consistency of the interpolated values when used for deriving trends with similar patterns of overall decreasing trend and magnitude of changes of -0.0095 to -0.0099 (NDVI unit) over the mangrove areas in 2015. The result demonstrated the potential ability of gap-filling methods for simulating the value of missing data and for deriving trends.
|
