Method for land use and land cover identification in a tropical area using multisensor optical and radar imageries

Synergystic use of Landsat-TM and JERS-I SAR imageries were used for land use/land cover identification in a tropical area. There are several approaches carried out to investigate an optimum result, including : geometric factor on radar backscattering coefficient through the influence of local slope correction ; speckle noise reduction using wavelet transform ; investigating relationship between radar backscattering coefficient (σ o ) and Normalized Difference Vegetation Index (NDVI); spectral analysis derived from Landsat-TM then combined with JERS-I SAR using IHS (Intensity, Hue, Saturation) and PCA (Principal Component Analysis) techniques. Significantly local slope from 2° to 4° degrees increased radar backscattering coefficient (σ o ) amount 0.5-2 dB. Correction of local slope on σ o was effective in mountainous areas, while was not effective in flat areas, such as commercial area and settlement area. In mixed crops area, small significant non-linear relationships were established between NDVI and JERS-I SAR data in 1994/06/22 (r 2 = 0.393) and 1994/09/18 (r 2 = 0.3303), respectively. On the other hand, there was no correlation between NDVI and JERS-I SAR data in paddy field, settlement, and commercial areas, respectively. The use of multiresolution analysis wavelet transforms (WT) for speckle noise reduction of JERS-I SAR resulted a smooth image without obscuring edge information, especially in level-2 for airport and level-3 for mountainous area. Land use/land cover classification were performed using Landsat-TM & JERS-1 SAR through image fusion techniques. The combination of 5 channels, consists of PC-I (principal component-1), PC-2 (principal component-2), PC-3 (principal component-3), NDVI, and WT-filtered JERS-I SAR (94/09/18, which is acquired in dry season) produced the best classification.