Generation and Evaluation of Digital Elevation Model Derived Through Spaceborne Synthetic Aperture Radar Interferometry

Many success stories of spaceborne InSAR have been reported all over the world in various earth sciences applications. However, only a limited number of articles were reported for applications in tropical regions. Therefore the objectives of this research are: i) To generate DEM from repeat pass spaceborne InSAR by using SAR data from European Earth’s Resource Satellite (ERS-1 and ERS-2), Canadian Earth Resource Satellite (Radarsat) which are both in C-band systems and L-band Japanese Earth Resource Satellite (JERS-1) with emphasise on coherence optimisation; ii) To increase the coherence level through slope enhancement technique; and iii) To assess the potential, suitability and limitations of InSAR DEM generated from the above data and conducting accuracy assessment of the InSAR DEM in Malaysian condition. Two test sites (Site A-Selangor and Site B-pahang) were selected for their variability in vegetations cover and landforms. Three pairs of ERS tandem data (pair 1, pair 2 and pair 3), one pair of JERS-1 data (pair 5) and one pair of standard mode Radarsat S7 data (pair 7) were acquired for site A-Selangor. On the other hand, one pair ERS tandem data (pair 4) and one pair JERS-1 data (pair 6) were acquired for site B-Pahang. Despite of having many pairs of data, the generation of DEM for site A-Selangor was not very successful. The ERS tandem data has revealed a very low coherence level ( 0.5). The clearly visible interferogram pattern could be seen in higher coherence areas. However, the atmospheric artifacts can be observed clearly on the interferogram through ‘fuzzy’ fringe pattern especially along the coastline. The Radarsat and JERS-1 data have shown a poorer coherence and interferogram quality, eventhough slightly better results for JERS-1 comparatively, presumably due to the long temporal separation of 24 days and 44 days between the two successive data acquisitions. ERS tandem data over site B-Pahang has shown a promising result with 65% of the ERS scene especially in urban and areas with low vegetations was successfully generated. A long dry month and non-windy condition was assumed to be the contributing success factor. The forested areas especially in high relief areas remain ingenerated. The use of Radarsat and JERS-1 data did not improve the DEM generation. The temporal separation between the data pair was long enough to observe changes within the canopy and hence the data pairs were totally uncorrelated. DGPS field observations were conducted to improve the accuracy of InSAR derived DEM and for the accuracy assessment. The third order polynomial warping with thirteen GCP (X, Y) and fourteen GCP (Z) were used for the transformation of DEM from relative to absolute height with the RMSE of 2.4 m, 0.52 m and 1.78 m in X, Y and Z respectively. Quantitative accuracy assessments by using RMSE and R2 regression correlation analysis were performed on the InSAR DEM against DGPS points, spot height check points, contour generated DEM, SRTM Mean Sea level DEM and SRTM WGS84 DEM. A Root Mean Square Error (RMSE) of 12.619, 15.662, 15.421, 15.016 and 15.694 were reported for the analysis accordingly. The corresponding R2 results of 0.8598, 0.913, 0.914, 0.8948 and 0.8938, obtained from similar evaluation pairs proved they are highly correlated. The results were satisfactory and agreeable to the results from most researches. The qualitative assessment was performed and presented graphically to further emphasis on the quality of the results. ERS tandem data is capable to provide a medium scale DEM at 1: 100 000 or smaller with accuracy within 20-25 m and contour interval of 40 m. This is absolutely complementing the present techniques of producing DEM especially in areas that is constraint by the persistent cloud cover through out the year. This complementary approach is considered as the best approach to expedite the generation of DEM for the whole country. Spaceborne ERS tandem data eventhough with one day temporal separation is still insufficient to ensure a highly correlated image pairs for 100% success in generating DEM. The DEM quality will be tremendously improved with single pass InSAR where the data was acquired by the two sensors simultaneously which is currently available only in airborne and shuttle environment. Therefore, single pass data acquisition should ideally be the preference for this part of the world such that have been proven by airborne Airsar Topsar Pacrim I and II missions in 1996 and 2000 and SRTM