Abstract. The identification and monitoring of the status of all fishponds with Fishpond Lease Agreements (FLAs) in the Philippines is limited mainly to site visits. However, this approach is tedious, time-consuming, and costly due to the sheer number and locations of fishponds. This study explores the use of radar remote sensing in expediting the mapping of fishpond status (i.e., active, abandoned). Historical Sentinel-1A SAR images acquired from 2022–2023 covering identified fishpond areas were analysed. Several temporal statistics of their sigma nought backscatter values were calculated per polarization (i.e., VV and VH). Non-fishpond areas were masked out, and then Principal Component Analysis was applied to the stacked temporal statistics images. Afterwards, K-Means Cluster classification was applied to the first 7 components of the resulting PCA images to generate 4 classes. The resulting class with relatively low backscatter values is identified as water; “intermittent” water if their backscatter values vary greatly; and non-water or vegetated if their backscatter values are relatively high. Status was determined based mainly on the dominant class in each plot. Water class dominated plots are considered active; non-water or vegetated class dominated plots and plots that are covered mostly by “intermittent” water class are labelled as abandoned. Based on field-validated data, active fishponds, and old, abandoned fishponds (e.g., already have mangroves, or have no water control structures) were easier to correctly map, while abandoned fishpond with certain characteristics were harder to correctly identify. The approach is promising and can help BFAR map and monitor the fishponds in the country.
Read full abstract