Mapping of seasonal inundation in the Congo River basin using PALSAR ScanSAR
(1) European Commission JRC, GEM Unit, TP 440, 210 27 Ispra (VA), Italy
*** Note: This paper is submitted to the "ALOS K&C Special Session" ***
A major aim of the ALOS PALSAR applications program, within the framework of the Kyoto & Carbon Initiative Wetlands Theme, is to map and monitor seasonal inundation patters in major wetlands and river basins around the world using time series of data acquired in ScanSAR mode. Compared to the fine resolution mode, the wide swath width (360 km) ScanSAR mode is more suitable for repetitive monitoring over extensive regions, as a significantly lesser number of scenes are required to complete a full wall-to-wall coverage, and due to the coarser resolution and lower data rate, which from an operational aspect accommodate more frequent observations.
The objective of this PI study is to investigate the utility of PALSAR ScanSAR data to map the temporal and spatial properties of seasonal inundation patterns in tropical river basins. In the first part of this study, presented in the first PI report (Nov. 2007) and at the 7th ALOS Cal/Val meeting (June 2008), the temporal radiometric stability of the ScanSAR products as well as internal calibration between the individual sub-beams were investigated. Using three independent time series of data acquired over homogeneous forest areas in the Amazon and Congo basins, the absolute calibration accuracy over 10 cycles (~15 months) was measured to vary between 0.55 ~ 1.00 dB. The variations appeared to be independent of processing date or date of acquisition. The relative (within-scene) calibration varied between 0.25~0.45 dB, with a systematic difference between the ScanSAR beams observed. In all 24 scenes investigated, Beam#1 (near range) was consistently 0.2~0.7 dB darker than the scene average. A similar consistent trail-off effect (~0.4 dB) was observed in the far range part of Beam#5.
In the second part of the study, which is to be presented here, an algorithm for flood duration mapping will be developed. The prototype area is the central part of the Congo river basin, over which a time series of 6 consecutive ScanSAR passes acquired in 2007/2008 are available. To accommodate unambiguous distinction between flooded and non-flooded forest across the ScanSAR scenes, absolute radiometric normalisation and relative post-calibration will be applied, based on the results above. Given successful outcome of this trial, the methodology is to be applied on a regional scale – within the framework of the K&C Initaitve – to the full Congo river basin.