Research

Currently, my research is mainly focused on:

• Atmospheric Correction of hyperspectral data
• Hyperspectral data analysis and applications in land cover classification
• Derivation of optical properties and water quality parameters of sea and inland waters from remote sensing reflectance data

I am currently Principal Investigator in the following international projects:

• NASA's EO-1 Science validation team on Forest classification using Hyperion data
• ESA's ENVISAT research project on Forest fire and haze monitoring using AATS R and MERIS data
• NASDA's ADEOS2 - research project on algal bloom monitoring using GLI data

• NASDA's ALOS research project on land cover classification using multi-polar ization SAR (with M. Nakayama of the Tokyo University of Agriculture and Technology as PI)

I am Head of Research, CRISP.

Go to CRISP's Research Web

Some of my earlier projects

Texture of SAR Images

Different physical mechanisms are involved in the image formation processes of SAR and optical images. Hence, different techniques are required for interpretation of SAR and optical images. In this project, we attempt to use texture features in the interpretation of SAR images. The texture of an image region depends on the relation between grey levels of neighbouring pixels. The texture features are commonly derived from the grey level co-occurrence matrix (GLCM), which is a matrix of relative frequencies P(a,b;s) with which two neighbouring pixels separated by a distance s occur on the image, one with grey level a and the other with grey level b.

SAR Texture Image of Singapore

The RGB colour-composite map of Singapore shown here is derived from three texture features: contrast (red), entropy (green) and inverse-difference moment (blue). Three regions can be distinguished clearly in the map. Built-up areas appear in yellow, forested areas in light green and water mass in blue. Flat coastal areas can also be distinguished as dark green areas in the map. It is interesting to note that different shades of colours appear in the sea, which correspond to different sea surface features.

Remote Sensing Applications in Forestry/Agriculture

Rice Monitoring and landuse classification in the Mekong River Delta, Vietnam Rice cultivation in the Mekong River delta is largely governed by hydrology, rainfall pattern and the availability of irrigation, resulting in a large diversity of rice cropping systems practiced in this region. Our research aims to delineate the various rice cropping systems in this region and to generate a thematic map of the major rice cropping systems using multitemporal ERS and RADARSAT synthetic aperture radar imagery. Multitemporal SAR is particularly suited to this study due to the changing landcover in the rice planting areas throughout the rice growing seasons. The use of cloud penetrating SAR also overcomes the problem of cloud cover during the rainy season. SAR imagery from RADARSAT is used to complement ERS SAR in landuse classification, to take advantage of the different polarisation and incident angles of the RADARSAT SAR. Monitoring changes in rice cropping systems using space-borne SAR imagery - Paper presented at the IEEE 1999 International Geoscience and Remote Sensing Symposium, 28 June - 2 July 1999, Hamburg. Application of multitemporal ERS-2 synthetic aperture radar in delineating rice cropping systems in the Mekong River Delta, Vietnam - Published in the journal IEEE Transactions on Geoscience and Remote Sensing 36(5) 1412-1420, Part 1 Sep 1998. Delineation of Rice Cropping Systems in the Mekong River Delta, Vietnam, using Multitemporal ERS SAR Images. - presented at the 3rd. ERS Symposium, Florence, March 1997. Landcover Classification over the Mekong River Delta using ERS and RADARSAT SAR images - presented at IGARSS 97, Singapore. Rice crops monitoring in the Mekong river delta using combined ERS and RADARSAT synthetic aperture radar (pdf file, 500 kbytes) - presented at IGARSS 98, Seattle.

Forest fires The 1997/98 forest fire episode in South East Asia has attracted international attention. The fires which occurred primarily in the Sumatra and Borneo islands and aggravated by the drought due to the El Nino Southern Oscillation phenomenon, resulted in increased aerosol loading (smoke-haze) over the region. The economic damage resulting from the haze has been reported to be in the range of billions of dollars. The environmental impacts of the fires include loss of forests as carbon sinks and emission of greenhouse gases which may influence the global climatic systems. In response to this situation, daily fire monitoring operation of the region using full-resolution SPOT images has been carried out at CRISP, in collaboration with the Ministry of Environment, Singapore. CRISP has also acquired many SPOT, ERS and RADARSAT images of the areas affected by the 1997/1998 fire. Detailed analysis of the fires is being carried out using these images. Forest Fire Monitoring Operation at CRISP using SPOT Images Forest Fire Monitoring and Research Activities at the Centre for Remote Imaging, Sensing and Processing (CRISP), National University of Singapore - published in the newsletter International Forest Fires News No. 19, September 1998 Delineation of Forest Fire Burnt Scars Using Interferometric SAR - Published in the journal Geophysical Research Letters 26(16), 2409-2412, 1999. Monitoring Vegetation Cover Changes in Peat Swamp Area of Central Kalimantan Using ERS Interferometric Synthetic Aperture Radar. Paper presented at the International Conference and Workshop on Tropical Peat Swamps - "Safeguarding a Global Natural Resource", 27-29 July 1999, Penang, Malaysia. A Study of the 1997 Forest Fires in South East Asia using SPOT Quicklook Mosaics

Multisensor Study of Tropical Forests and Agricultural Regions The use of spaceborne remote sensing for forest applications has been widely demonstrated as an important tool in global forest cover identification and forest biomass estimation. More importantly, the easy availability of data on a regular basis from operational satellites such as ERS, JERS, RADARSAT, SPOT and LANDSAT has created the opportunity for multisensory monitoring of tropical forests. While optical sensors have been successfully exploited for such studies, their use in tropical areas is severely limited by prevalent cloud covers. This limitation has been somewhat alleviated by the use of the cloud penetrating synthetic aperture radar (SAR). Additional information can also be derived from the interferometric coherence of the radar backscatter signals. This property has proved useful in discriminating deforested areas and it can be applied to yield useful information for landuse classification and for monitoring of changes in landuse pattern, such as the conversion of forest to agricultural lands. ERS-1/2 Interferometry: Some Results on Tropical Forest - presented at FRINGE96. Landuse Study in Coastal Jambi, Sumatra, using SAR/InSAR - presented at 3rd. ERS Symposium, 1997.

Vegetation Mapping of Singapore Singapore has been transformed into an urbanised island during the past few decades. However, a considerable proportion of its land aea is still under dense vegetation cover. These areas include the Bukit Timah Nature Reserve, the Central and Western Catchment areas and some of the many nature parks. In this study, vegetation maps of Singapore are derived from SPOT and LANDSAT multispectral images. A vegetation basis decomposition technique is being developed for landuse classification. Vegetation Mapping in Singapore