Conference Presentations by Shridhar Jawak
Existing digital elevation models (DEMs) of Antarctic region in the Antarctic digital database (A... more Existing digital elevation models (DEMs) of Antarctic region in the Antarctic digital database (ADD) indicate elevation variations of up to hundreds of meters, which necessitates the construction of local DEMs. The present study focuses on the achieved improvements in construction of accurate DEMs by fusing multi-sensor, multisource, multitemporal, and multi-technological elevation datasets. The study focuses on two spatially different study regions, (a) Larsemann hills and environ and (b) Schirmacher oasis and environ, where India's Antarctic research stations are established. The study focuses on synergetic fusion of three sensor technologies, (a) photogrammetry, (b) laser altimetry, and (c) radar altimetry. The constructed DEMs were evaluated using differential global positioning system (DGPS) validation points. We also analyzed a suite of interpolation methods for constructing DEMs from multitemporal and multi-technology derived point elevation datasets, in order to determine the level of confidence with which the interpolation techniques can generate a better interpolated continuous surface, and eventually improve the elevation accuracy of DEMs from synergistically fused datasets. This is one of the first international attempts of fusing multitemporal, multisensory and multisource elevation data to generate a DEM of any part of Antarctica, in order to address the ice elevation change and the ice mass balance. This research experiment demonstrates the stability (w.r.t. multitemporal datasets), performance (w.r.t best interpolation technique) and consistency (w.r.t all the experimented interpolation techniques) of synergistically fused DEMs over existing DEMs. On the basis of error analyses, the newly constructed DEMs may serve as a benchmark for future elevation models such as from the ICESAT-II mission to spatially monitor ice sheet elevation. Our approach focuses on the strengths of each elevation data source to produce an accurate elevation model.
The study discusses the design and implementation of novel methods to extract essential for semia... more The study discusses the design and implementation of novel methods to extract essential for semiautomatic extraction of geo-scientific information in cryospheric environments by using very high resolution (VHR) WorldView-2 (WV-2) satellite data the main rationale of this study is to explore the varied applications of VHR imagery in Antarctic earth sciences. We focus on three novel methods developed for such information extraction; (a) an ensemble classification approach for land cover mapping, (b) spectral index ratio-based information mining, and (c) lake feature extraction using novel normalized difference water index (NDWI). The study evaluates the range of applications of spectral remote sensing for rapid information extraction, which would be useful for various polar research applications such as glaciology, environmental and land-cover change monitoring, limnology studies, supraglacial processes, etc. Nevertheless, the semi-automatically extracted information would also be useful for planning Antarctic logistic activities and for providing advisories for safe field campaigns. The study also evaluates 8-band image data as an influential tool for satellite image classification and consequent land cover mapping in Antarctic environment. This research provides new facets to image processing technology. The results indicate that the use of the novel methods on 8-band WV-2 data can significantly improve the semiautomatic extraction of cryospheric features, which can ultimately contribute to an enhanced perceptive of the Antarctic geospatial information in the context of climate change.
Using in-situ density profiles recorded by deploying Expendable CTD (XCTD) probes and satellite-b... more Using in-situ density profiles recorded by deploying Expendable CTD (XCTD) probes and satellite-based fields mapped during the Indian Expedition to Antarctica during austral summers of 2012 and 2013, we explore basic hydrodynamics of the Indian Ocean sector along the Indian Ocean sector of coastal Antarctica. 50 stations in 2012 and 59 in 2013 where occupied along the Antarctic coast from Prydz Bay 69° 9' S, 75° 36' E) to India Bay (69° 16' S, 13° 46' E) up to a depth of 1000m. Since remote sensing provides large domain with a repeated coverage, we have used sea surface temperature (SST) data of AVHRR, surface chlorophylla concentration from NASA's GeoEye Orbview2 (Sea WiFS sensor) and sea surface salinity (SSS) from Aquarius. Different signature/features in the profiles during 2012 and 2013 were identified which confirms the presence of winter water (WW) layer close to the coast. Vertical section of temperature from both the expeditions did show a well developed mixed layer. During 2012 traces of winter mixed layer (ML) was identified at India Bay region. During both the years, the WW layer was observed near Scott mountain region from 50 m to 300 m. During 2013 this WW layer extended down to 1000 m indicating a downwelling zone near Scoble glacier at around 60°E. In Prydz Bay region a WW layer was observed extending up to 1000m which may be due to the melting of Amery Ice Shelf. In 2012 a upwelling zone was observed at 40°E which is indicated by the warm water of temperatures ranging from 0.5°C to 1°C rising to the surface. During both the years warm waters (> 0°C) was observed from ~200 m depth below the ML and WW layer from 20°E to 40°E which represents Circumpolar Deep Water. Salinity section for 2012 and 2013 showed patches of low salinity regions with salinity less than 33.75 psu at the surface. In the deeper section the salinity also showed an increasing trend with salinity exceeding >34.5 psu, except for regions in which there was a presence of cold water with temperature <0.5° C salinity range varied between 34.25-34.5 psu. Low salinity at the surface indicates that the melting of sea ice during Antarctic summers.
Monitoring changes in the distribution and density of plant species often requires accurate and h... more Monitoring changes in the distribution and density of plant species often requires accurate and highresolution baseline maps of vegetation. Detecting such change at the landscape scale is often problematic, particularly in remote areas. Vegetation mapping of plant communities at fine spatial scales is increasingly supported by remote sensing technology. Less frequent imaging with high spatial resolution satellite sensors enable more detailed analyses of vegetation change. This study is the first to use high-resolution WorldView-2 (WV-2) imagery to classify vegetation communities on Antarctic oases and to provide automated means to map vegetation as an important indicator for environmental change. Multispectral imagery (MSI) and panchromatic imagery (PAN) from very high resolution WV-2 have been used for the analysis of vegetation in different forms in Antarctic environment. A range of classification methods have been executed using pansharpened WV-2 data. This study comparatively evaluates vegetation mapping results using supervised and unsupervised classification methods to extract vegetation in Larsemann Hills and Schirmacher oasis, east Antarctica. We also focused on the use of supervised pixel-based classifiers and textural measures, in addition to standard multispectral information, to improve the classification of Antarctic vegetation communities. Classification results were validated with independent reference datasets.
Blue-ice areas (BIAs) cover 1% of the East Antarctic ice sheet. High resolution PAN-sharpened cal... more Blue-ice areas (BIAs) cover 1% of the East Antarctic ice sheet. High resolution PAN-sharpened calibrated images from WorldView-2 (WV-2) were used for extracting blue ice areas in Schirmacher Oasis, east Antarctica. The Schirmacher oasis is located between the edge of the Antarctic ice Sheet and the Novolazarevskaya Nivl Ice Shelf which extends from 70°45′ S to 70° 75′ S and 11°38′ E to 11° 38′ E. Blue ice extent may vary because of weather, seasonal effects, and climate change. Blue ice areas are generally visual evidence of long-term ablation. The amplitude of blue ice is lower than that of snow, because the ice surface is smoother than the latter. But difference is not so obvious when applying automatic extraction techniques. Extraction of BIAs in Antarctica deal with the total area of blue ice excluding the other features appearing on or near it. To achieve desirable results and support comparative analysis, multiband image combinations were generated from atmospherically corrected WV-2 data. For feature extraction process, regions of interest (ROI) were considered in which blue ice was used as target and white snow/ice appearing on the blue ice was considered as non-target. Various semiautomatic feature extraction methods, such as, target detection, mapping methods, etc, and many trials were used for extracting blue ice areas. Surface patterns of alternating snow and blue ice bands are found in East Antarctica which becomes obstacle to clearly extract blue ice feature. From the high resolution WV-2 data, reference data (digitized data) were prepared for blue ice area and extracted blue ice area was obtained from feature extraction methods. By comparing reference data and extracted data, bias and root mean square (RMS) error values were calculated. Accuracy assessment was done considering the entire necessary prior results of the blue ice area. Normalized difference spectral index for blue ice is obtained from the spectral profile from maximum and minimum value of band which helped to successfully create and implement the NDBI (normalized difference blue ice index) model. After processing all the 12 test tiles whole image was processed and final blue ice is extracted from the data.
Land cover classification is one of the widely used applications in the field of remote sensing. ... more Land cover classification is one of the widely used applications in the field of remote sensing. Accurate land cover maps derived from remotely sensed data is the major requirement for many geoscientific applications in polar regions. The present study explores the capabilities of C-band dual polarimetric (HH & HV) level1 SAR image data from Indian Radar Imaging Satellite (RISAT-1) for land cover feature mapping around Larsemann Hills and Schirmacher oasis, Antarctica. We used RISAT-1 Fine Resolution STRIPMAP (FRS-1) mode data having 3m spatial resolution. In order to increase the amount of information in dual polarized RISAT SAR data, a band HH+HV was introduced to make use of the original two polarizations. Transformed divergence (TD) procedure for class separability analysis was performed to evaluate the quality of the statistics prior to image classification along with data calibration. For most of the class pairs the TD values were comparative, which indicates that the classes have good separability. Nonparametric classifier Support Vector Machine (SVM) was used to classify RISAT1 data with optimized polarization combination into three land cover classes consisting of sea ice/snow/ice, rocks/landmass, and lakes/waterbodies. This study demonstrates that C-band FRS1 image mode data from RISAT1 mission can be exploited to identify, map and monitor land cover features in the polar regions, even in the dark winter period. The final results based on the SAR backscattered information will be presented during the conference, which is crucial in discriminating and delineating the land cover features.
High resolution pansharpened images from WorldView-2 were used for bathymetric mapping around Lar... more High resolution pansharpened images from WorldView-2 were used for bathymetric mapping around Larsemann Hills and Schirmacher oasis in east Antarctica. We digitized the lake features in which all the lakes from both the study areas were manually extracted. In order to extract the bathymetric values from multispectral imagery we used different models: (a) Stumpf model (b) Lyzenga model and (c) Jupp's model. We tried multiband combinations to improve the results. To test the effect of size and dimension of lake on bathymetry retrieval, we distributed all the lakes on the basis of size and depth (reference data), as some of the lakes were open, some were semi frozen and others were fully frozen. Several tests were performed on open lakes on the basis of size and depth wise distribution. Very shallow lakes provided better correlation (≈0.8) compared to shallow (≈0.6) and deep lakes (≈0.6) for depth wise lake distribution. In case of size wise distribution, large lakes yielded better correlation in comparison to medium and small lakes. The entire comparisons were done on the basis of derived values against the reference values. Both bathymetric map and contour maps are created. We planned to develop the Lyzenga and Jupp's model in IDL programming and compare the results for same lakes with results of Stumpf model in order to check which model yields better results. The results are awaited and will be published during the conference as trial and errors are been carried out.
Uploads
Conference Presentations by Shridhar Jawak