Search

Multi-frequency synthetic aperture radar (SAR) data was acquired over the Macquarie Marshes Nature Reserve in north central New South Wales (NSW), Australia, to demonstrate the potential of imaging radar for mapping and monitoring wetland extent and inundation patterns in an inland, semi-arid wetland environment. A sequence of ALOS PALSAR and TerraSAR-X images acquired between January 2007 and April 2008 captured the wetlands in three successive phases: dry, wet and transitional. Visual observation of multi-temporal and multi-frequency SAR backscatter data confirmed the capacity of SAR to respond to changes in surface water, soil moisture and biomass. Longer wavelength L-band SAR provides a tool for discriminating wetlands, mapping surface water and detecting below-canopy inundation. Shorter wavelength X-band SAR provides a tool for detecting flushes in growth of vegetation in response to higher soil moisture from flooding. PALSAR data demonstrated a high capacity for discrimination of different wetland classes, while TerraSAR-X data was less suited to discriminating between cover types. The integration of X- and L-band data revealed the extent of floodplain inundation and presence of aquatic vegetation in ponded areas. Given a stable, well-calibrated time-series of SAR data, change analysis provides a mechanism for understanding the hydrological and/or ecological change in an area. In the longer term, it is envisioned that radar will provide a valuable tool within an operational system for monitoring the impact of environmental flows in NSW inland wetlands.

This paper describes, and illustrates using documented applications, a general framework methodology for wide-area forest and land use mapping and change detection using Synthetic Aperture Radar (SAR) remote sensing. Consideration is given to implementation of the SAR-based methodology using both commercial and free/open-source software. Our experience shows that constructing a complete processing chain requires either a variety of expensive commercial software packages, or recourse to considerable bespoke software development, and in either case yields a fragmented solution not readily accessible to non-experts. It is argued that this poses a barrier to the uptake of satellite SAR by tropical forest countries, preventing them from exploiting a reliable and predictable source of remote sensing data suited to forest monitoring. Construction of a free software tool for use with SAR and other remotely sensed data is considered and it is argued that a single tool encompassing the complete framework methodology is both feasible and desirable.

This paper compares segmentation results of synthetic aperture radar (SAR) images using Gaussian-Markov random field (MRF) and Gamma-MRF models. A Gamma distribution function is more accurate and proper to trace the multilook SAR intensity data distribution. However, it is found that, at least from examples used in the paper, when the distribution function is incorporated with the MRF model to implement SAR image segmentation, the Gamma-MRF model is not necessarily shown to be superior to the Gaussian-MRF model. Occasionally the Gamma-MRF model wrongly merges a few small segments, suggesting that the Gaussian-MRF model might be more stable and reliable.

An effective algorithm for digital image noise smoothing using wavelet transform techniques is presented in this paper. This algorithm is more powerful when compared to other existing filtering algorithms in terms of speckle suppression for synthetic aperture radar images where the presence of speckle makes the ratio of standard deviation to mean (STM) very high. Examples show that the original STM of about 0.30 (equivalent to three-look images) can be reduced to 0.05-0.03 (equivalent to more than 100-look images), with a possible small sacrifice of losing some details and narrow edges. The quantitative analysis is carried out and compared with the results of some existing filtering algorithms including median, K nearest neighbour averaging, Lee's multiplicative and Crimmins' geometric filters, showing that imagery filtered by the wavelet transform is the smoothest.

An integrated remote sensing/field ecology project linked the use of synthetic aperture radar (SAR) and aerial photography to studies of landscape spatial heterogeneity and bird community ecology. P-, L-, and C-band SAR data, collected over a section of Kakadu National Park in Australia's Northern Territory during the Joint NASA/Australia DC-8 data acquisition campaign, were analyzed in light of field data integrating vegetation structure and floristics with bird abundances across a heterogeneous study site. Results indicate that SAR data are able to discern structural differences relevant to bird habitat quality within floristically homogeneous stands, while multispectral sensors successfully identified floristic differences among habitat types. Simplifying indices of bird diversity showed ambiguous changes across the site; however, the abundances of individual species were observed to change significantly across both floristic and structural gradients. These results suggest that efforts to map bird diversity should focus on species-specific habitat relationships and that some measure of vegetation structure is needed to understand bird habitat. The approach employed here advances the use of SAR data in the three-dimensional mapping of animal habitats from remotely sensed data, and extends current capabilities for mapping and modeling large-scale patterns in the distribution of biological diversity.

Continuity of earth observation data is paramount to operational forest information monitoring in support of national accountability of forest and carbon stocks. Following the loss of ALOS PALSAR in May 2011, the minimum 2 year gap before PALSAR-2 comes online, and the superiority of L-band data for forest monitoring, this paper addresses the issue of interoperability of SAR data through comparison of forest/non-forest, land cover and deforestation derived from independent and combined classifications of PALSAR and RADARSAT-2 data. Forest/non-forest mapping accuracies were highest using PALSAR data alone. The combined use of C- and L-band data resulted in higher forest/non-forest mapping accuracies compared to using C-band data alone. Attributed largely to an increased capacity for penetration of the vegetation canopy and interaction with woody structures, L-band SAR is most suited to forest cover mapping. Without a dense time-series, the separation of cleared or deforested land and regrowth areas is more difficult using C-band alone.

Developing a large-scale forest monitoring system able to take advantage of the complementary nature of optical and radar remote sensing data presents a number of technical and conceptual challenges. This paper investigates the issue of sensor interoperability in a time series of Landsat and ALOS-PALSAR data for purposes related to forest mapping and monitoring. The proposed approach relies on the processing methods developed in the frame of an existing and operational Landsat-based forest monitoring system. These methods are here applied to a PALSAR dataset within a bioregion of north-eastern Tasmania, Australia. Particular attention is given to the selection of training data in an attempt to generate results comparable to those obtained with the original Landsat-only time series, thereby allowing for a relevant assessment of interoperability. Results are presented in the form of forest maps and areal forest estimates. Despite similar gross amounts of forest extents, these results highlight differences in the forest (and change) classifications produced using different sensors. Combinations of sensors should therefore be carefully considered in light of what is required of the monitoring system.

The arid zone of Australia is an environment with ‘boom and bust’ dynamics; where years of drought can be followed by periodic flash flooding. Major floods occurred along 500km of the Paroo River in March 2010 at levels unequaled since 2000. Dual polarization (HH/HV) L-Band synthetic aperture radar (SAR) data was obtained from the Japanese Advanced Land Observing Satellite (ALOS) to evaluate its use for mapping flooding over an area of high conservation value floodplain wetlands. The data were shown to be highly informative for mapping flooded classes and provide an alternative data source to optical imagery in the presence of cloud cover. Due to radar speckle, segmentation was used to partition the scenes within eCognition Developer. Classification was performed on image objects using decision rules based on field verified radar backscatter thresholds of classes and change detection. The methods developed for single and multiple images provide a framework for operational mapping and monitoring water over time, as well as a rapid response strategy for flood emergencies. Future work on image transforms and research on the interoperability of SAR and optical data will aim to improve differentiation of wetland classes.

The Kokoda Track area of Papua New Guinea is geographically diverse, with terrain varying from coastal areas to steep mountains, and biomes that extend from mangroves, to grasslands and savannas, and primary forests. The area is the subject of conservation measures using remotely sensed data in a collaboration between the Australian and Papua New Guinea governments. We describe this Kokoda Track initiative and the land cover mapping to be used as a basis for land use planning. We discuss hydromorphological analysis and land cover classification extracted from Interferometric Synthetic Aperture Radar (SAR) imagery. Our land cover analysis has employed two classification techniques, both object-based, and has included texture measures calculated from SAR imagery. The output is a high resolution classification product, to be used by local government stakeholders.

Consistent estimation of carbon stocks at national level requires the integration of wall-to-wall, time-series satellite and in situ data of forest area, type and change. In this paper we demonstrate a consistent approach to the generation of wall-to-wall time-series mosaics using ALOS PALSAR data acquired over 2007 to 2009 for Tasmania, Australia. The project is part of a series of National Demonstrators initiated by the Group on Earth Observations (GEO) Forest Carbon Tracking (FCT) task that emphasize the contribution and operational use of satellite measurements for forest monitoring and national carbon accounting. Interoperability between optical and Synthetic Aperture Radar (SAR) derived forest measurements will also be demonstrated. The project will deliver a series of forest monitoring products and technical documentation, which will be made available as a guide to GEO member countries with a desire to develop their own national carbon accounting systems.

In preparation for the ESA ERS-1 mission, a series of multitemporal, multifrequency, multipolarization aircraft SAR data sets were acquired near Fairbanks in March 1988. P-, L-, and C-band data were acquired with the NASA/JPL Airborne SAR on five different days over a period of two weeks. The airborne data were augmented with intensive ground calibration data as well as detailed simultaneous in situ measurements of the geometric, dielectric, and moisture properties of the snow and forest canopy. During the time period over which the SAR data were collected, the environmental conditions changed significantly; temperatures ranged from unseasonably warm (1 to 9 C) to well below freezing (-8 to -15 C), and the moisture content of the snow and trees changed from a liquid to a frozen state. The SAR data clearly indicate the radar return is sensitive to these changing environmental factors, and preliminary analysis of the L-band SAR data shows a 0.4 to 5.8 dB increase (depending on polarization and canopy type) in the radar cross section of the forest stands under the warm conditions relative to the cold. These SAR observations are consistent with predictions from a theoretical scattering model.

The ALOS Kyoto & Carbon Initiative is an international collaborative project led by the JAXA Earth Observation Research Center (EORC). It forms the continuation of JAXA's on-going JERS-1 SAR Global Rain Forest and Global Boreal Forest Mapping project(GRFM/GBFM) into the era of the Advanced Land Observation Satellite (ALOS). The ALOS K & C Initiative has been set up to support the data and information needs required by international environmental Conventions; Carbon Cycle scientists and Environmental Conservation programs (referred below to as the CCCs). Led and coordinated by EORC JAXA, the Initiative is being undertaken by an international Science Team and focuses primarily on defining and optimizing the provision of data products and validated thematic information derived from in-situ and satellite sensors focusing on data acquired from the Phased Array L-band Synthetic Aperture Radar (PALSAR) on-board the Advanced Land Observing Satellite (ALOS). The objective of the ALOS K & C Initiative is to define, develop and validate thematic products derived from ALOS PALSAR data that can be used to meet specific information requirements relating to the CCCs. A key component of this work has been the development of a systematic data acquisition strategy for ALOS which comprises fixed, systematic global observation plans for PALSAR. The strategy is implemented as a top-level foreground mission and with a priority level second only to that of emergency observations. With emphasis on acquiring repetitive and consistent data over continental scales, it ensures that adequate data will be collected to allow the required thematic output products to be developed on a timely basis. The K & C Initiative is based on the three coordinated themes relating to global biomes; Forests, Wetlands, Deserts and Semi-Arid Regions, and a fourth theme dealing with the generation of regional ALOS PALSAR Mosaics. A key word for the Initiative is regional- scale applications and product development, with data requirements in the order of hundreds or thousands of PALSAR scenes for each prototype area. Each theme has identified key products that can be generated from the PALSAR data including land cover, forest change mapping and forest biomass and structure analysis; global wetland inventory compilation and landscape change determination; and freshwater resources and desertification. Each of the products developed will be generated using a combination of PALSAR imagery, in situ measurements and ancillary datasets.An international Scientific Advisory Panel has been established to maintain the scientific relevance of project design and to ensure alignment with other relevant international efforts (e.g. GOFC/GOLD, IGOL, GTOS/TCO). The panel consists of scientists active in the fields of carbon modeling and biophysical parameter retrieval; SAR and image processing, as well as and representatives from GOFC, TCO, FAO, space agencies, universities and public research institutions.

Northern Australia exhibits an extremely seasonal climate and this is reflected in the change in vegetation communities present in a given area of wetland. The monsoonal nature of this climate means a huge variation in the amount of water available to plant life both on and surrounding the floodplains of the major river systems. Kakadu National Park, a heritage listed and Ramsar identified wetland of international significance, has a long history of scientific study an ideal study site on which to conduct studies such as this. Changes within the wetland areas in the park have been noted since the early 1980s, but have been poorly quantified. Many of the changes are related to the timing of the onset of the monsoon, the period of inundation and the location of the major flooding. However neither multispectral or hyperspectral sensors are able to capture information at the onset of the monsoon due to perpetual cloud coverage. Radar has the unique ability to image through cloud and hence 27 Standard and ScanSAR images were collected over the region from 1996 to 2001. The majority of the images were S4 ascending, with ScanSAR narrow collected simultaneously. The images were co-registered allowing comparison of the same geographic location through time. The simultaneous ScanSAR images allowed assessment of the technique over a wider swath, yielding results on a regional rather than local scale. Radar reflection is intimately related to the moisture content of the target due to the influence of moisture upon the dielectric constant. A modified Gaussian Markov Random Field Model Segmentation was used to cluster areas exhibiting similar radar response at each successive date in the time series. The ability to segment and extract reflection statistics for the same geographic location for each image in the series allows the construction of temporal curves, indicating change in the target material's dielectric constant over time.

Constructive and destructive interference have long been a significant problem for the classification of radar imagery. Several authors have noted the ability of segmentation to work around the problem by incorporation of spatially adaptive filters and fuzzy logic. We present the methodology and results of such a study using multitemporal datasets in the place of multispectral data. For the study 27 standard and ScanSAR images were collected over the region from 1996 to 2001. The majority of the images were S4 ascending, with ScanSAR narrow collected simultaneously. The ScanSAR images allowed us to assess the methodology over a wider swath, yielding results on a regional rather than local scale. Northern Australia exhibits an extremely seasonal climate. The monsoonal nature of this climate means a huge variation in the amount of available water to plant life both on and surrounding the flood plains of the major river systems. Within this area is Kakadu National Park, a heritage listed and Ramsar identified wetland of international significance. Kakadu has a long history of scientific study, and is an ideal site on which to conduct studies such as this. The seasonal nature of this area presents a unique opportunity, that of a worst case scenario for classification schemes, whereby each location on the flood plain undergoes significant change over the seasonal cycle, similar in many respects to the change between bands of a multi-spectral dataset. A modified Gaussian Markov random field model segmentation routine was used to cluster areas exhibiting similar radar response (both numerical and textural) at each successive date in the time series. For each date cluster statistics were then generated. This allowed the construction of temporal curves due to the fact that as the target material dries the material's dielectric constant decreases. As dielectric constant is significantly dependent on water content, these temporal curves act as a proxy measure for water availability, and hence aid in the discrimination of wetland from non-wetland areas and any change in size and location of these areas throughout the season. The output of the segmentation routine is a series of three bands. The first and second bands are the statistics (mean and standard deviation) of the original image and the third a vector file of the edge locations. By outputting statistical information instead of arbitrary segment number standard classification routines may be used successfully on radar imagery. Classification of the results of the segmentation may be either supervised or unsupervised, however analysis indicated that the Isodata algorithm adequately addressed classification needs. As a combined system the segmentation and classification system used in this paper has shown some excellent results when applied to multi-date radar imagery of Kakadu National Park in Northern Australia. Results for a segmentation of multitemporal radar dataset are presented alongside the results of the subsequent classification. By creating a system that requires minimal user input a wide variety of applications may be addressed. Indeed, the registration of the imagery becomes the most user intensive portion of the process. Automated routines such as this allow for analysis of large areas on a regular basis, an exceptional result for monitoring change and establishing baselines for later comparison.

The aim of this paper is to compare radar backscatter response from imaging SAR systems operating in C-, L- and P-band wavelengths over wetland environments found in the Kakadu World Heritage Region, Northern Territory, Australia. Determining the spatial pattern of inundation in monsoonally domination wetlands and the sequence of floodplain drying is an important first step to understanding the hydrologic, geomorphic and ecological processes operating in these landscapes. SIR-C data acquired in 1994 and AIRSAR multiband imagery obtained over the study site in October 1993 and again in November 1996 were available for analysis along with MOMS-2P optical data. Radarsat imagery was acquired on a near monthly basis during 1998 for the study site. The data were calibrated, geo-referenced and overlain in order to determined water boundaries and to map the pattern of receding water levels. Both Standard Beam 4 and ScanSAR data has been analysis to evaluate the usefulness of C-band horizontally polarised data for wetland mapping. The identification of saturated condition under open canopies commonly found in Australian monsoonal forests can be demonstrated at all three wavelengths. Significantly multirate radarsat imagery proved used in detecting the distribution and extent of water under trees in all seasonal conditions; in mapping the pattern of floodwater recession and in assessing the drying out phases observed in this wetland environment.

This paper presents segmentation of radar imagery by two steps: 1. Initial segmentation using wavelet transform techniques and the watershed method; 2. Segment merging using the Gaussian Markov random field models. The method can be applied to both single-channel and multi-channel images.

A unique opportunity was afforded in November 1993 to acquire state-of-the-art radar data over Australian test sites. CSIRO-COSSA and the University of New South Wales Centre for Remote Sensing and GIS collaborated with NASA's Mission to Planet Earth Airborne Sciences Program, and the Jet Propulsion Laboratory (JPL) to bring AIRSAR, the world's most advanced synthetic aperture radar system to Australia. AIRSAR operates in a full polarimetric mode at L-band (24.5 cm), C-band (5.6 cm) and P-band (68 cm). In addition to this three-frequency polarimetric capability called POLSAR, the incorporation of extra C- and L-band antennas permits interferometric data known as TOPSAR to be acquired. During November 1993 this system acquired data over 55 sites on the Australian continent for 35 principal investigators including both US and Australian scientists and collaborators. This data has been processed and analysed and the research and science findings presented in Proceedings of the International Workshop on Radar Image Processing and Applications, Sydney, 6-8 November 1995. AIRSAR Australia 1993 has made a significant contribution, both in terms of advanced polarimetric and interferometric radar data not previously available and in fostering a research community interested in using the unique characteristics of radar to analyse a range of environmental surfaces and conditions. The program has also resulted in a significant increase in technology transfer in bringing together an Australian community not previously experienced in using polarimetric radar and in providing a knowledge base from which radar can be integrated into research aimed at determining geophysical parameters of surface materials.

In response to Australia's requirement for spatial information on greenhouse gas emissions from land use change and forestry, the past and present use and future potential of synthetic aperture radar (SAR) data for quantifying woodland biomass is reviewed.

Multi-temporal spatial datasets quantifying above ground biomass (AGB) are required for calculating national, regional and global carbon budgets, due to the large extent and rapidly changing nature of woodlands in Australia. This paper provides an overview of procedures for quantifying and scaling-up estimates of total AGB and component biomass from the tree to the landscape using remote sensing data and focusing on woodlands in Queensland.

Focusing on woodlands in Queensland, Australia, this paper outlines a multi-stage approach to quantifying and scaling-up field-based measurements of vegetation structure, biomass and community composition to the landscape. The approach utilises remotely sensed data from a range of instruments (including SAR, hyperspectral and lidar) operating at different spatial and spectral resolutions. The research is anticipated to benefit the calculation of greenhouse gas emissions, conservation of biodiversity, and sustainable utilisation of woodlands in Australia.

The study assesses the consistency of relationships between the component (leaf, branch, and trunk) biomass of Australian woodlands and synthetic aperture radar (SAR) backscatter data acquired by airborne TOPSAR and spaceborne SIR-C SAR. The strongest relationships were observed between TOPSAR P-band data and all components of the biomass, with P-band I-IV considered most suited for quantifying total above ground biomass (TAGB) due to the larger dynamic range of the data and greater saturation level (80-100 Mg ha/sup -1/) in the relationship with TAGB. The main similarities observed between sensors were the strength of the relationships between C- and L- band data and component biomass, and the TAGB saturation levels of 20-30 Mg ha/sup -1/ and 50-60 Mg ha/sup -1/ for C- and L-band respectively.

The integration of JERS-1 SAR and Landsat TM data for quantifying and mapping carbon losses associated with vegetation clearance near Injune, central Queensland, Australia, is outlined. Above ground biomass (including stem and leaf components) was estimated for 70 plots by applying allometric equations to measurements of tree size taken in July, 1997. Significant Pearson's product moment correlation coefficients were derived between foliage biomass and Landsat TM NDVI and stem biomass and the JERS-1 SAR backscatter coefficient. To estimate above ground biomass, a simple regression model integrating both JERS-1 SAR backscatter and NDVI data was derived. For a property in central Queensland, spatial estimates of biomass were generated by applying the model to segmented JERS-1 SAR data. These estimates were combined with the Queensland Statewide Landcover And Trees Study (SLATS) land cover change data sets to estimate carbon losses associated with vegetation clearance between 1995 and 1997.

Speckle filtering is probably one of the important steps in post SAR (synthetic aperture radar) image processing. Speckle appears in all SAR products including single-polarimetric, multipolarimetric and interferometric image data. Many speckle filtering algorithms have been proposed aimed at speckle reduction. Texture, revealing the spatial-dependent information in images, may be as important as other polarimetric measures. Such information is not expected being distorted and degraded during speckle suppression. This paper examines various speckle filters including the Lee, Frost, Kuan and other filters in terms of texture preservation. The first and second order texture measures are used in the assessment.

SAR image segmentation is investigated by integrating both the Gamma distribution and the Gaussian distribution, appropriate data representations for uniform areas in SAR images, with the Markov random field model. Both models are able to distinguish spatially adjacent objects by up to their second order statistics and spatial textures. Overall results from the Gaussian model are slightly better, although the Gamma distributions are more appropriate to trace real radar image data distributions.

The potential of Synthetic Aperture Radar (SAR) for estimating the above ground and component biomass of woodlands in Australia is demonstrated using two case studies. Case Study 1 (In,june; central Queensland) shows that JERS-1 SAR L HH data can be related more to the trunk than the leaf and branch biomass of woodlands. A strong relationship between L HH and above ground biomass is obtained when low biomass pasture sites are included. Case Study I1 (Talwood, southern Queensland) determines that L and P band data can be related both to trunk and branch biomass, due to the similarity in the orientation and size of these scattering elements, and also to total above ground biomass. Saturation of the C. L and P band data occurred at approximately 20-30 Mglha; 60-80 Mglha and 80-100 Mglha. These preliminary results indicate that data from SAR are useful for quantifying changes in carbon stocks resulting from land use change in Australia's woodlands and for applications in rangeland assessment and management. Key words: remote sensing, biomass, woodlands

Classification of radar images remains a challenge due largely to the effects of speckle. In this paper the classification of radar images is accomplished in two steps. First, an image is partioned into uniform areas or segments and second, these segments are then classified into information classes. Both segmentation and classification are achieved by using the Gaussian Markov random field model. The results of segmentation and classification routines applied to the woodland areas in Queensland using both spaceborne and airborne SAR image data are presented. In terms of segmentation, regions whose mean differences are as small as 0.5dB and with ratios of the standard deviation to the mean as high as 0.35 are separated with accuracies approaching 90%. In terms of classification, there are more ambiguities in single-band data. Multi-band polarised data on the other hand provided better results. Relationships established between component biomass and the backscatter coefficient at all wavelengths and polarisations indicated a strong correspondence (r2>0.80) between AIRSAR L- and P-band backscatter and above ground, branch and trunk biomass. The highest correlation between component biomass and backscatter using JERS1 related to the estimation of stem biomass leading to the conclusion that more reliable estimates of total and component biomass will require multiband-multipolarimetric data.