Multi-temporal SAR Data Research Articles

Potential of TerraSAR-X and Sentinel 1 imagery to map deforested areas and derive degradation status in complex rain forests of Ecuador

This study explores the use of TerraSAR-X (X-Band) and Sentinel-1 (C-Band) SAR data for detecting degraded forests and deforestation in the Sumaco Biosphere Reserve, Ecuador. The study area shows a complex topography and permanent cloud cover, which complicates forest monitoring from optical remote sensing. Thresholds of canopy cover for degraded forest were identified by field data. Then a straightforward object-based approach was applied. Statistics on SAR backscatter were estimated from the objects and used as input to a random forest classification to map forest and non-forest areas. Finally, forest degradation was assessed using a moving window procedure applying the earlier identified thresholds on the forest / non-forest map. The SAR data were found adequate to map deforestation (overall accuracy of 81% for forest/non forest maps) but less accurate to fully assess forest degradation. However, multi-temporal SAR data can be useful to construct baseline data of the forest extent. Polarization and interferometric information could enhance the presented results for classifying forest / non forest areas and mapping forest degradation.

InSAR observations of the 2009 Racha earthquake, Georgia

Abstract. Central Georgia is an area strongly affected by earthquake and landslide hazards. On 29 April 1991 a major earthquake (Mw = 7.0) struck the Racha region in Georgia, followed by aftershocks and significant afterslip. The same region was hit by another major event (Mw = 6.0) on 7 September 2009. The aim of the study reported here was to utilize interferometric synthetic aperture radar (InSAR) data to improve knowledge about the spatial pattern of deformation due to the 2009 earthquake. There were no actual earthquake observations by InSAR in Georgia. We considered all available SAR data images from different space agencies. However, due to the long wavelength and the frequent acquisitions, only the multi-temporal ALOS L-band SAR data allowed us to produce interferograms spanning the 2009 earthquake. We detected a local uplift around 10 cm (along the line-of-sight propagation) in the interferogram near the earthquake's epicenter, whereas evidence of surface ruptures could not be found in the field along the active thrust fault. We simulated a deformation signal which could be created by the 2009 Racha earthquake on the basis of local seismic records and by using an elastic dislocation model. We compared our modeled fault surface of the September 2009 with the April 1991 Racha earthquake fault surfaces and identify the same fault or a sub-parallel fault of the same system as the origin. The patch that was active in 2009 is just adjacent to the 1991 patch, indicating a possible mainly westward propagation direction, with important implications for future earthquake hazards.

Forest Type Classification Based on Multi-temporal SAR and SPOT Remote Sensing Data in Pangu Forest Farm

Information extraction of forest type is difficult in remote sensing image classification. Daxing’an Mountains is an important forestry area in China mainly covered with natural forests,rich with a wide range of plants resources,which makes it difficult to accurately identify the forest types in this region. In order to compare and improve the accuracy of classification result,taking Pangu Forest Farm in Daxing’an Mountains as the study area,we proposed three methods to classify forest types by the maximum likelihood method combining with SPOT-5 and two different temporal RADARSAT-2 fully polarimetric SAR remote sensing data. We designed three schemes to classify the forest types and compared the accuracy. In the three schemes,SPOT image was only used to distinguish forest types,some descriptive parameters extracted from SAR polarimetry( POLSAR)images and the SPOT data were used for classification,and the integration of parameters extracted from multitemporal of full polarimetric SAR( Pol SAR) images with SPOT data was used for classification. The most effective method to identify white birch,larch,Pinus sylvestris and spruce among three proposed schemes was the third using multi-temporal SAR and SPOT remote sensing image. The classification accuracy and the Kappa coefficient were 84. 64% and 0. 79,respectively. However,the accuracy of forest type classification by using SPOT data individually was the lowest of 76. 66% with the Kappa coefficient of 0. 70.

Remote sensing-based assessment of impact of Phailin cyclone on rice in Odisha, India

Assessing the nature and extent of damage due to natural calamities remains one of the thrust areas in monitoring resource inventory through remote sensing. The effect of the cyclone Phailin and the post-incessant rains during second fortnight of October 2013 on coastal Odisha was studied in terms of rice area flooded, submerged and damaged. Multi-temporal SAR data were analysed to obtain the rice mask, and from this rice mask, the flood affected rice area was determined. Taluka-wise and district-wise crop loss proportion was estimated, and the overall production loss has been estimated. SAR data aided in delineation of flooded regions, while AWiFS NDVI data of subsequent dates showed both continued inundation and crop vigour status post-flood time period. The ground truth indicated that a major portion of the inundated region was not rice but was typha grasses and harvested rice field which should not be accounted as damage to rice crop. The damage on crop yield was difficult to assess; however, the inundation of the crop at panicle initiation and flowering would have impact on grain filling (results in chaffiness) and was considered as completely damaged. Most of the current inundated rice regions fall in this category. It was estimated that a total of 0.167 million hectares and 0.37 million tons of rice crop was lost in the cyclone and floods. The district-level percentage area of rice flooded was communicated to State Remote Sensing Centre in four days timeframe. The overall accuracy obtained for the validation of the ground truth sites was 91.5 %.

Impact of refugee camps on their environment. A case study using multi-temporal SAR data

Monitoring the environment is a key task of remote sensing in particular in areas whose access is difficult or dangerous or where dense cloud cover obscures optical information. This study proposes an assessment of landscape changes related to large refugee camps, where information about environmental conditions is needed by both humanitarian organizations and regional administrations. Our intention is to provide a robust workflow which is applicable for an operational use. The study area is located in Western Kenya hosts a total number of 350.000 people. Images of ERS-2 and Sentinel-1 are used for the assessment of land degradation in a semi-arid savannah between 1997 and 2014. We expect a relationship between the existence of the refugee camps and the degradation of surrounding landscapes. For this purpose we present an approach which objectively reveals developments in natural resources based on six land-use / land cover classes integrating their relative importance for the ecosystem given by expert-based weights.

CLUSTERING OF MULTI-TEMPORAL FULLY POLARIMETRIC L-BAND SAR DATA FOR AGRICULTURAL LAND COVER MAPPING

Abstract. Recently, the unique capabilities of Polarimetric Synthetic Aperture Radar (PolSAR) sensors make them an important and efficient tool for natural resources and environmental applications, such as land cover and crop classification. The aim of this paper is to classify multi-temporal full polarimetric SAR data using kernel-based fuzzy C-means clustering method, over an agricultural region. This method starts with transforming input data into the higher dimensional space using kernel functions and then clustering them in the feature space. Feature space, due to its inherent properties, has the ability to take in account the nonlinear and complex nature of polarimetric data. Several SAR polarimetric features extracted using target decomposition algorithms. Features from Cloude-Pottier, Freeman-Durden and Yamaguchi algorithms used as inputs for the clustering. This method was applied to multi-temporal UAVSAR L-band images acquired over an agricultural area near Winnipeg, Canada, during June and July in 2012. The results demonstrate the efficiency of this approach with respect to the classical methods. In addition, using multi-temporal data in the clustering process helped to investigate the phenological cycle of plants and significantly improved the performance of agricultural land cover mapping.

PSINSAR IMPROVEMENT USING AMPLITUDE DISPERSION INDEX OPTIMIZATION OF DUAL POLARIMETRY DATA

Abstract. Persistent Scatterer Interferometry for SAR data (PSInSAR) improves the ability of conventional InSAR time-series methods by detecting and analysing pixels where the portion of spatiotemporal decorrelations on the phase is negligible. Using dual/quad polarized SAR data provide us with an additional source of information to improve further the capability of InSAR analysis. In this paper, we present a method to enhance PSInSAR using polarimetric optimization method on multi-temporal polarimetric SAR data. The optimization process has been implemented to minimize the Amplitude dispersion Index (ADI) of pixels in SAR images over the time based on the best scattering mechanism. We evaluated the method on a dataset including 17 dual polarization SAR data (HH/VV) acquired by TerraSAR-X data from July 2013 to January 2014 over Tehran plain, Iran. The area has been affected by high rate (> 20 cm/yr.) of surface subsidence due to groundwater overexploitation. The effectiveness of the method is compared for both agricultural and urban regions affected by land subsidence. Furthermore single pole and optimized polarization results are compared together and with external observations from GPS measurements. The results reveal that using optimum scattering mechanism decreases the ADI values in urban and non-urban regions and increase the PS Candidate pixels (PSC) about three times and subsequently improves the PS density about 50% more than using single channel datasets.

The potential of ALOS PALSAR backscatter and InSAR coherence for forest growing stock volume estimation in Central Siberia

Abstract The full potential of ALOS PALSAR L-band interferometric (InSAR) coherence data for the estimation of forest growing stock volume ( GSV ) in the boreal forest has rarely been investigated. Moreover, ALOS PALSAR backscatter and InSAR coherence have yet to be used together to delineate GSV . Due to the observation strategy and the high acquisition success rate over Eurasia, a large amount of high quality ALOS PALSAR L-band data is available over Siberia. Consequently, this paper investigates the capability of ALOS PALSAR backscatter and InSAR coherence for the estimation of GSV in Central Siberia, Russia. The potential of backscatter and coherence are directly compared using the same inventory data. Altogether, 87 PALSAR images are used and eleven forest inventory sites are investigated. Based on this large dataset it was observed that InSAR coherence acquired in frozen conditions offers the highest potential for GSV estimation. The saturation level for single coherence images was on average 230 m 3 /ha, with an average R 2 between coherence and GSV of 0.58. PALSAR backscatter acquired in unfrozen conditions could also estimate GSV ; however, the saturation levels (75–100 m 3 /ha) and the average R 2 (0.42–0.48) were lower. HV backscatter offered only a slightly greater potential than HH backscatter. A simple inversion approach aiming at the delineation of forest GSV maps based on the multitemporal SAR data was developed and applied to five forest inventory sites. This approach combines HV backscatter data acquired in unfrozen conditions and InSAR coherence data acquired in frozen conditions. In general, the produced maps feature a corrected relative RMSE corr of 2 between inventory data and SAR data based maps varied between 0.54 and 0.83.

Detecting the Source Location of Recent Summit Inflation via Three-Dimensional InSAR Observation of Kīlauea Volcano

Starting on 21 April 2015, unusual activity on the summit of Kīlauea was detected. Rapid summit inflation and a rising lava lake in Halema‘uma‘u crater were interpreted as early signs of imminent magma intrusion. We explored the three-dimensional (3D) surface motion accompanying this volcanic event using the Interferometric Synthetic Aperture Radar (InSAR) stacking method. Multi-temporal COSMO-SkyMed X-band SAR data collected from ascending and descending orbits were processed for the time period encompassing the unrest behavior. The 3D displacement maps retrieved by integrating the stacked InSAR with Multiple-Aperture Interferometric SAR (MAI) measurements revealed the deformation patterns and areal coverage of this volcanic activity. The observed maximum displacements were approximately 8.2, −13.8, and 11.6 cm in the east, north, and up directions, respectively. The best-fit model for the mechanism causing the surface deformation was determined via ten thousand simulations using the 3D surface deformation as the input. When compared to the results of a previous study, the 3D-based modeling produced more precise model parameter estimates with markedly lower uncertainties. The optimal spheroid magma source was located southwest of the caldera, lying at a depth of approximately 2.8 km below the surface. Precise model parameter estimates produced using the 3D-based modeling will be helpful in understanding the magma behavior in Kīlauea’s complex volcanic system.

The effects of field plot size on model-assisted estimation of aboveground biomass change using multitemporal interferometric SAR and airborne laser scanning data

Remotely sensed data from airborne laser scanning (ALS) and interferometric synthetic aperture radar (InSAR) can greatly improve the precision of estimates of forest resource parameters such as mean biomass and biomass change per unit area. Field plots are typically used to construct models that relate the variable of interest to explanatory variables derived from the remotely sensed data. The models may then be used in combination with the field plots to provide estimates for a geographical area of interest with corresponding estimates of precision using model-assisted estimators. Previous studies have shown that field plot sizes found suitable for pure field surveys may be sub-optimal for use in combination with remotely sensed data. Plot boundary effects, co-registration problems, and misalignment problems favor larger plots because the relative impact of these effects on the models of relationships may decline by increasing plot size. In a case study in a small boreal forest area in southeastern Norway (852.6ha) a probability sample of 145 field plots was measured twice over an 11year period (1998/1999 and 2010). For each plot, field measurements were recorded for two plot sizes (200m2 and 300/400m2). Corresponding multitemporal ALS (1999 and 2010) and InSAR data (2000 and 2011) were also available. Biomass for each of the two measurement dates as well as biomass change were modeled for all plot sizes separately using explanatory variables from the ALS and InSAR data, respectively. Biomass change was estimated using model-assisted estimators. Separate estimates were obtained for different methods for estimation of change, like the indirect method (difference between predictions of biomass for each of the two measurement dates) and the direct method (direct prediction of change). Relative efficiency (RE) was calculated by dividing the variance obtained for a pure field-based change estimate by the variance of a corresponding estimate using the model-assisted approach. For ALS, the RE values ranged between 7.5 and 15.0, indicating that approximately 7.5–15.0 as many field plots would be required for a pure field-based estimate to provide the same precision as an ALS-assisted estimate. For InSAR, RE ranged between 1.8 and 2.5. The direct estimation method showed greater REs than the indirect method for both remote sensing technologies. There was clearly a trend of improved RE of the model-assisted estimates by increasing plot size. For ALS and the direct estimation method RE increased from 9.8 for 200m2 plots to 15.0 for 400m2 plots. Similar trends of increasing RE with plot size were observed for InSAR. ALS showed on average 3.2–6.0 times greater RE values than InSAR. Because remote sensing can contribute to improved precision of estimates, sample plot size is a prominent design issue in future sample surveys which should be considered with due attention to the great benefits that can be achieved when using remote sensing if the plot size reflects the specific challenges arising from use of remote sensing in the estimation. That is especially the case in the tropics where field resources may be scarce and inaccessibility and poor infrastructure hamper field work.

Improving Pixel-Based Change Detection Accuracy Using an Object-Based Approach in Multitemporal SAR Flood Images

Most of existing change detection methods could be classified into three groups, the traditional pixel-based change detection (PBCD), the object-based change detection (OBCD), and the hybrid change detection (HCD). Nevertheless, both PBCD and OBCD have disadvantages, and classical HCD methods belong to intuitive decision-level fusion schemes of PBCD and OBCD. There is no optimum HCD method as of yet. Analyzing the complementarities of PBCD and OBCD method, we propose a new unsupervised algorithm-level fusion scheme (UAFS-HCD) in this paper to improve the accuracy of PBCD using spatial context information through: 1) getting the preliminary change mask with PBCD at first to estimate some parameters for OBCD; 2) deriving the unchanged area mask to eliminate the areas without changes, reducing error amplification phenomenon of OBCD; and 3) obtaining the final change mask by means of OBCD method. Taking flood detection with multitemporal SAR data as an example, we compared the new scheme with some classical methods, including PBCD, OBCD, and HCD method and supervised manual trial-and-error procedure (MTEP). The experimental results of flood detection showed that the new scheme was efficient and robust, and its accuracy sometimes can even exceed MTEP.

Backscatter Analysis Using Multi-Temporal and Multi-Frequency SAR Data in the Context of Flood Mapping at River Saale, Germany

In this study, an analysis of multi-temporal and multi-frequency Synthetic Aperture Radar data is performed to investigate the backscatter behavior of various semantic classes in the context of flood mapping in central Europe. The focus is mainly on partially submerged vegetation such as forests and agricultural fields. The test area is located at River Saale, Saxony-Anhalt, Germany, which is covered by a time series of 39 TerraSAR-X data acquired within the time interval December 2009 to June 2013. The data set is supplemented by ALOS PALSAR L-band and RADARSAT-2 C-band data. The time series covers two inundations in January 2011 and June 2013 which allows evaluating backscatter variations between flood periods and normal water level conditions using different radar wavelengths. According to the results, there is potential in detecting flooding beneath vegetation in all microwave wavelengths, even in X-band for sparse vegetation or leaf-off forests.

Estimation of Biomass Carbon Stocks over Peat Swamp Forests using Multi-Temporal and Multi-Polratizations SAR Data

Abstract. The capability of L-band radar backscatter to penetrate through the forest canopy is useful for mapping the forest structure, including above ground biomass (AGB) estimation. Recent studies confirmed that the empirical AGB models generated from the L-band radar backscatter can provide favourable estimation results, especially if the data has dual-polarization configuration. Using dual polarimetry SAR data the backscatter signal is more sensitive to forest biomass and forest structure because of tree trunk scattering, thus showing better discriminations of different forest successional stages. These SAR approaches, however, need to be further studied for the application in tropical peatlands ecosystem We aims at estimating forest carbon stocks and stand biophysical properties using combination of multi-temporal and multi-polarizations (quad-polarimetric) L-band SAR data and focuses on tropical peat swamp forest over Kampar Peninsula at Riau Province, Sumatra, Indonesia which is one of the most peat abundant region in the country. Applying radar backscattering (Sigma nought) to model the biomass we found that co-polarizations (HH and VV) band are more sensitive than cross-polarization channels (HV and VH). Individual HH polarization channel from April 2010 explained > 86% of AGB. Whereas VV polarization showed strong correlation coefficients with LAI, tree height, tree diameter and basal area. Surprisingly, polarimetric anisotropy feature from April 2007 SAR data show relatively high correlations with almost all forest biophysical parameters. Polarimetric anisotropy, which explains the ratio between the second and the first dominant scattering mechanism from a target has reduced at some extent the randomness of scattering mechanism, thus improve the predictability of this particular feature in estimating the forest properties. These results may be influenced by local seasonal variations of the forest as well as moisture, but available quad-pol SAR data were unable to show these patterns, since all the SAR data were acquired during the rainy season. The results of multi-regression analysis in predicting above ground biomass shows that ALOS PALSAR data acquired in 2010 has outperformed other time series data. This is probably due to the fact that land cover change in the area from 2007 – 2009 was highly dynamic, converting natural forests into rubber and Acacia plantations, thus SAR data of 2010 which was acquired in between of two field campaigns has provided significant results (F = 40.7, P < 0.005). In general, we found that polarimetric features have improved the models performance in estimating AGB. Surprising results come from single HH polarization band from April 2010 that has a strong correlation with AGB (r = 0.863). Also, HH polarization band of 2009 SAR image resulted in a moderate correlation with AGB (r = 0.440).

Extreme Learning Machine-based Crop Classification using ALOS/PALSAR Images

Classification maps are required for agricultural management and the estimation of agricultural disaster compensation. The extreme learning machine (ELM), a newly developed single hidden layer neural network is used as a supervised classifier for remote sensing classifications. In this study, the ELM was evaluated to examine its potential for multi-temporal ALOS/PALSAR images for the classification of crop type. In addition, the k-nearest neighbor algorithm (k-NN), one of the traditional classification methods, was also applied for comparison with the ELM. In the study area, beans, beets, grasses, maize, potato, and winter wheat were cultivated; and these crop types in each field were identified using a data set acquired in 2010. The result of ELM classification was superior to that of k-NN; and overall accuracy was 79.3%. This study highlights the advantages of ALOS/PALSAR images for agricultural field monitoring and indicates the usefulness of regular monitoring using the ALOS-2/PALSAR-2 system. Discipline: Agricultural engineering Additional key words: Hokkaido, machine learning, sigma naught *Corresponding author: reysnb@gmail.com Received 12 May 2014; accepted 4 February 2015. Introduction Land-cover classification is one of the most common applications of remote sensing. Crop-type classification maps are useful for yield estimation and agricultural disaster compensation, in addition to the management of agricultural fields. Optical remote sensing remains one of the most attractive options for the accumulation of biomass information and forest monitoring (Samreth et al. 2012, Sarker and Nichol 2011). In addition, while optical satellites such as ALOS/AVNIR-2 (Sonobe et al. 2014a), Landsat (Hartfield et al. 2013), MODIS (Sakamoto et al. 2009), and NOAA (Hirano and Batbileg 2013) have been employed in the identification of species and conditions of vegetation, cloud cover significantly limits the number of available optical images, radar is unaffected by cloud cover or low solar zenith angles (Bindlish and Barros 2001). And significant information about soil and vegetation parameters can also be obtained through microwave remote sensing (Sonobe et al. 2014b). These techniques are employed increasingly to manage land and water resources for agricultural applications (Sonobe et al. 2014c). The number of studies on rice and wheat monitoring and mapping using SAR data has increased, and some studies utilizing multi-temporal SAR data have reported high correlations between backscattering coefficients, and plant height and age (Chakraborty et al. 2005, Sonobe et al. 2014d, Waisurasingha et al. 2008). These examples highlight possible uses in the area of agricultural management, specifically for the identification of paddy fields. They indicate the potential use of SAR data for the discrimination of crop types. Furthermore, a number of examples have shown the practical usefulness of supervised learning for classification (Sonobe et al. 2014b). This paper reports a comparison of crop classification using PALSAR data performed by extreme learning machine (ELM) and k-nearest neighbor algorithm (k-NN).

Complementarity of Two Rice Mapping Approaches: Characterizing Strata Mapped by Hypertemporal MODIS and Rice Paddy Identification Using Multitemporal SAR

Different rice crop information can be derived from different remote sensing sources to provide information for decision making and policies related to agricultural production and food security. The objective of this study is to generate complementary and comprehensive rice crop information from hypertemporal optical and multitemporal high-resolution SAR imagery. We demonstrate the use of MODIS data for rice-based system characterization and X-band SAR data from TerraSAR-X and CosmoSkyMed for the identification and detailed mapping of rice areas and flooding/transplanting dates. MODIS was classified using ISODATA to generate cropping calendar, cropping intensity, cropping pattern and rice ecosystem information. Season and location specific thresholds from field observations were used to generate detailed maps of rice areas and flooding/transplanting dates from the SAR data. Error matrices were used for the accuracy assessment of the MODIS-derived rice characteristics map and the SAR-derived detailed rice area map, while Root Mean Square Error (RMSE) and linear correlation were used to assess the TSX-derived flooding/transplanting dates. Results showed that multitemporal high spatial resolution SAR data is effective for mapping rice areas and flooding/transplanting dates with an overall accuracy of 90% and a kappa of 0.72 and that hypertemporal moderate-resolution optical imagery is effective for the basic characterization of rice areas with an overall accuracy that ranged from 62% to 87% and a kappa of 0.52 to 0.72. This study has also provided the first assessment of the temporal variation in the backscatter of rice from CSK and TSX using large incidence angles covering all rice crop stages from pre-season until harvest. This complementarity in optical and SAR data can be further exploited in the near future with the increased availability of space-borne optical and SAR sensors. This new information can help improve the identification of rice areas.

Stochastic Approach in Wet Snow Detection Using Multitemporal SAR Data

This paper introduces an alternative strategy for wet snow detection using multitemporal SAR data. The proposed change detection method is primarily based on comparison between two X band SAR images acquired during accumulation (winter) and melting (spring) seasons, in French Alps. The new decision criterion relies on local intensity statistics of SAR images by considering backscattering ratio as a stochastic process: probability that the intensity ratio fits into predetermined range of is larger than a defined confidence level. Both conducted snow backscattering simulations and state of art measurements indicate more complex relation between backscattering properties of two snow types, with respect to conventional assumption of augmented electromagnetic absorption associated to wet snow. Therefore, rather than adopting standard hypothesis, we analyse wet/dry snow backscattering ratio as a function of local incidence angle (LIA). After employing multi-layer snow backscattering simulator, calibrated with scatterometer measurements in C band, we modify, to some extent, range of ratio values indicating presence of wet snow, by including positive ratio values for lower LIA. By simultaneously accounting for speckle noise, proposed stochastic approach derives refined wet snow probability map. The performance analyses are carried out both through comparison with ground air temperature map and by comparing two co-polarized channels processed separately.

Long-Term Land Subsidence Monitoring of Beijing (China) Using the Small Baseline Subset (SBAS) Technique

Advanced techniques of multi-temporal InSAR (MT-InSAR) represent a valuable tool in ground subsidence studies allowing remote investigation of the behavior of mass movements in long time intervals by using large datasets of SAR images covering the same area and acquired at different epochs. Beijing is susceptible to subsidence, producing undesirable environmental impacts and affecting dense population. Excessive groundwater withdrawal is thought to be the primary cause of land subsidence, and rapid urbanization and economic development, mass construction of skyscrapers, highways and underground engineering facilities (e.g., subway) are also contributing factors. In this paper, a spatial–temporal analysis of the land subsidence in Beijing was performed using one of the MT-InSAR techniques, referred to as Small Baseline Subset (SBAS). This technique allows monitoring the temporal evolution of a deformation phenomenon, via the generation of mean deformation velocity maps and displacement time series from a data set of acquired SAR images. 52 C-band ENVISAT ASAR images acquired from June 2003 to August 2010 were used to produce a linear deformation rate map and to derive time series of ground deformation. The results show that there are three large subsidence funnels within this study area, which separately located in Balizhuang-Dajiaoting in Chaoyang district, Wangjing-Laiguangying Chaoyang district, Gaoliying Shunyi district. The maximum settlement center is Wangsiying-Tongzhou along the Beijing express; the subsidence velocity exceeds 110 mm/y in the LOS direction. In particular, we compared the achieved results with leveling measurements that are assumed as reference. The estimated long-term subsidence results obtained by SBAS approach agree well with the development of the over-exploitation of ground water, indicating that SBAS techniques is adequate for the retrieval of land subsidence in Beijing from multi-temporal SAR data.

3차원 웨이블렛 변환을 이용한 다중시기 SAR 영상의 특징 추출 및 분류

이 연구에서는 다중시기 SAR 영상으로부터 3D 웨이블렛 변환을 통해 추출된 특징 정보를 이용하여 토지피복 분류를 수행하였고 그 적용가능성을 평가하였다. 분류를 하기 전 단계로 3차원 웨이블렛 변환기반 특징을 추출하였고, 이후 토지 피복 분류에 사용하였다. 비교를 목적으로 특징추출 단계가 들어가지 않는 원본 영상과 주성분분석 기반 특징들의 분류를 함께 수행하였다. 성능 검증을 위해 당진에서 촬영된 다중시기 Radarsat-1호 영상을 사용하였고 토지피복은 논, 밭, 산림, 수계, 도심지가 포함된 5개의 클래스로 구분하였다. 토지피복 식별 능력 분석에 따르면 밭과 산림은 매우 유사한 특성을 보이기 때문에 두 클래스를 구분하는 것은 매우 어렵다. 3차원 웨이블렛 기반 특징을 사용하는 경우, 도심지를 제외하고 모든 클래스의 분류 정확도가 향상되었다. 특히 밭과 산림의 정확도가 향상된 것을 확인할 수 있었다. 이러한 향상은 다중시기자료를 시간과 공간적으로 동시에 분석하는 3차원 웨이블렛 변환 과정에 기인한 것으로 판단된다. 이 결과로부터 3차원 웨이블렛 변환이 영상으로부터 특징을 추출하는데 이용 가능하다는 것을 확인할 수 있었고, 추후에 다른 센서나 다른 연구지역으로 추가 실험을 수행할 예정이다. In this study, land-cover classification was implemented using features extracted from multi-temporal SAR data through 3D wavelet transform and the applicability of the 3D wavelet transform as a feature extraction approach was evaluated. The feature extraction stage based on 3D wavelet transform was first carried out before the classification and the extracted features were used as input for land-cover classification. For a comparison purpose, original image data without the feature extraction stage and Principal Component Analysis (PCA) based features were also classified. Multi-temporal Radarsat-1 data acquired at Dangjin, Korea was used for this experiment and five land-cover classes including paddy fields, dry fields, forest, water, and built up areas were considered for classification. According to the discrimination capability analysis, the characteristics of dry field and forest were similar, so it was very difficult to distinguish these two classes. When using wavelet-based features, classification accuracy was generally improved except built-up class. Especially the improvement of accuracy for dry field and forest classes was achieved. This improvement may be attributed to the wavelet transform procedure decomposing multi-temporal data not only temporally but also spatially. This experiment result shows that 3D wavelet transform would be an effective tool for feature extraction from multi-temporal data although this procedure should be tested to other sensors or other areas through extensive experiments.

From Previous C-Band to New X-Band SAR Systems: Assessment of the DInSAR Mapping Improvement for Deformation Time-Series Retrieval in Urban Areas

We investigate the capability improvement of the advanced differential interferometric synthetic aperture radar (DInSAR) techniques to map deformation phenomena affecting urban areas by exploiting multitemporal SAR data acquired by the new X-band sensors with respect to those of the previous C-band systems. In particular, we perform a comparative analysis of the deformation time-series retrieved by applying the full-resolution Small BAseline Subset DInSAR technique to selected sequences of SAR data acquired by the ENVISAT and RADARSAT-1 sensors (both operating at C-band) and by the X-band radar systems onboard the SAR sensors of the COSMO-SkyMed (CSK) constellation. This study, focused on the city of Napoli (Italy), allows us to quantify the dramatic increase of the DInSAR coherent pixel density achieved by exploiting the high-resolution X-band CSK SAR images (a few meters), resulting in an improvement factor of about 320% and 550%, with respect to the RADARSAT-1 and ENVISAT products, respectively. This improvement permits us to analyze nearly all the structures located within the investigated urbanized area and, in many cases, also portions of the same building. The improved coherent pixel spatial densities, combined with the reduced revisit times of the new X-band SAR missions, allow us to significantly increase the effectiveness of the advanced DInSAR methodologies, further extending the role of those Earth Observation data in the development of monitoring scenarios.

Multitemporal polarimetric RADARSAT-2 SAR data for urban land cover mapping through a dictionary-based and a rule-based model selection in a contextual SEM algorithm

This paper presents a dictionary- and rule-based model selection approach in an adaptive contextual semi-supervised algorithm for improving urban land cover classification using high-resolution multitemporal RADARSAT-2 polarimetric SAR (PolSAR) data. Six-date PolSAR data were acquired from June to September, 2008, over the Greater Toronto Area. Contextual information and the capabilities of different PolSAR distribution models were explored by the spatially variant Finite Mixture Model (FMM) with an adaptive Markov Random Field (MRF) in a Stochastic Expectation–Maximization (SEM) algorithm. This algorithm can obtain homogenous results while preserving shape details in the complex urban environment with high accuracy. Commonly used PolSAR distribution models such as Wishart, G0p, Kp, and KummerU were compared through the proposed approaches for urban land cover mapping. According to a Goodness-of-Fit test based on Mellin transformation, an accurate PolSAR distribution model could be selected with the dictionary-based classification. However, the results showed that improvement from the dictionary-based approach was limited. Therefore, further improvements were expected by exploring expert knowledge. The initial results showed that G0p and KummerU performed better for distinguishing between low density built-up areas and forest. G0p, Kp, and KummerU are better for the low scattering classes. The Wishart model has superior capacity in separating high density built-up areas and the adjacent roads. Based on such knowledge, a set of rules was developed to integrate the advantages of alternative models. Significant improvement on the overall classification accuracy could be observed by this rule-based approach. The biggest improvement was achieved using the HD–Road rule on the G0p model with the best overall classification accuracy at 89.99% (kappa: 0.87). This represented 4.1% (kappa: 0.045) improvement over that of G0p without model selection.