Finally, the strengths of TerraSAR-X for wetland assessments lie in its high sensitivity to soil moisture and surface roughness on low-biomass surfaces. In particular, a good performance of TerraSAR-X for mapping salt marshes, including fish or mussel bed classification, as well as the depiction of degraded swamp forests and herbaceous vegetation, was highlighted in the reviewed studies. With respect to multi-temporal analyses, vegetation species showing fast growing vertical structures could be best distinguished by using X-band. However, the usage of TerraSAR-X data for areas with high-biomass conditions such as swamp forests with dense canopies showed limitations. Here, a better performance of L-band radar was discovered in several studies. On that basis, the combined usage of TerraSAR-X with other sensors yielded promising results for wetland mapping. However, while SAR can improve the accuracy of wetland classifications in inundated areas, the partially limited availability and often high cost of radar imagery still limits its broader use [ 58 ]. Despite being free of charge for scientific and non-commercial users, the access to TerraSAR-X products is not straightforward. As a first step, data need to be ordered via a detailed science proposal describing research objectives, work plan, data requirements, and detailed information about the “scientific use” criterion. Each proposal is evaluated by a scientific and technical committee, which judge the scientific priority. If accepted, the data will be delivered electronically. However, depending on the project, commercial users sometimes receive higher priority, conflicting with orders from scientific users, which could result in a reduction of the data or even in a complete rejection. This may change in the near future due to transitions in data policies, indicating a trend towards freely available remote sensing data. Moreover, the synergistic use of different SAR data (e.g., TerraSAR-X with L-band radar sensors such as ALOS-PALSAR-2) is a promising approach in which both the advantage of high-resolution X-band radar for observing herbaceous wetland vegetation and the suitability of L-band radar to map forested wetlands by penetrating canopies are combined. Further, data fusion of TerraSAR-X with high-resolution optical data (e.g., RapidEye) proved to be extremely valuable for a detailed wetland classification, as high-resolution SAR data might be influenced by speckle noise, which degrades the true spatial resolution. In the future, multi-sensor approaches may become increasingly important for wetland assessment due to the growing availability of different sensors with the advantage of complementing each other.
of Birds (RSPB), United Nations Development Programme (UNDP), and nationally with the Nigerian Conservation Foundation (NCF) (Akinsola et al., 2000; Kaugama and Ahmed, 2014). Other National and local administrators include the Ministry of the Federal Government, the River Basin Development Authorities (RBDAs), the State governments, Local government, Sasakawa 2000, JEWEL Project, National NGOs such as the Miyetti Allah organisation representing Fulɓe pastoralists, Al- Hayah representing the Shuwa and Koyam pastoralists, trade associations and other smaller groups within the wetlands (Blench, 2013). Although, the IUCN is the custodian of the HNWs, the Hadejia- Nguru Wetlands Conservation Project (HNWCP) was locally established by the IUCN with the main objective; “to promote sustainable use of the wetlands resources to the benefit of both local communities and waterbird populations”, this gradually led into the goal “to maintain the natural resources and functions of the wetlands” (Akinsola et al., 2000), this have the following objectives (i) to maintain both the economic and ecological functions of the wetlands, (ii) promote public awareness and education, (iii) monitoring, decision making, and cooperation, (iii) developing appropriate technologies with local farmers to enhance sustainable utilization of wetland resources, and (iv) improving management of protected areas and providing guidelines for the wise utilization of natural resources (Odada et al. 2005; Akinsola et al., 2000). However, climatic changes and altered hydrology in the wetland created multifaceted agreement between member states and other non- governmental organizations to have interests in the management and conflicts resolution in the management of water resources (Eaton and Sarch, 1997; Olalekan et al., 2014). For example, there are two River Basin Development Authorities within the basin, one for upstream and one for downstream (Goes, 2002). In 1996, World Bank aided Agricultural Development Projects (ADPs) conducted in the HNWs basin enabled flooded wetlands (fadama) farming through the creation of fadama Users Association (FUA) with 25 fadama farmers
Since its launch in 2007, TerraSAR-X has continuously revealed synthetic aperture radar (SAR) images of unprecedented high resolution from space. This has brought life to the once obscure and sometimes inscrutable SAR images that deterred many researchers. Figure 1 shows a comparison of the medium resolution ERS image and a high resolution TerraSAR-X spotlight image of the same area in Las Vegas. Individual buildings are for the first time interpretable by the naked eye from spaceborne SAR images, because the 1-m resolution in spotlight mode is well beyond the inherent scale of the 3-m floor height typical of urban infrastructure. This marks the start of an era of urban infrastructure monitoring using spaceborne SAR images. Currently, the staring spotlight mode provides images with a resolution up to 25 cm, from which the mapping of individual window edges is even possible. This breakthrough in spatial resolution, together with the precise orbit determination with sub-centimeter accuracy [ 1 , 2 ], positions TerraSAR-X images as a perfect dataset for long-term repeated monitoring of large areas with precision and high resolution.
The SAR backscatter, strongly related to the sea surface roughness, can be used to derive wind fields at spatial resolution that no other instrument can provide, thereby potentially revealing horizontal wind variability unavail- able by other means. Terrain-induced low-level jet and wake patterns are particularly conductive to the SAR-instrument examination. A case in point is bora, a cold and dry downslope wind blowing from north-easterly directions on the eastern side of the Adriatic Sea. Bora subtle- ties, like its varying strength or interactions of different mechanisms, are not fully understood yet, but solid understanding of its basic nature for the strong to severe cases does exist. A recent review of advances in the understandings of severe bora flows is found in GRISOGONO & BELUŠIĆ (2009) . Bora onset and development
The four water classification models created using grey-level thresholding can be seen in Figure 6. Areas of black represent water and grey areas represent non-water. Table 2 outlines the percentages of the two classes through time, as well as the threshold value used. Water classification changes from 8% to 10% throughout the four scenes, disagreeing with seasonal trends, which should show an increase in temperature causing a decrease in water. However, several other processes are occurring to account for this change. Ice can be seen in only the first scene (blue box in Figure 6A), as it is classified as both other and water, and decreases the amount of total water classified. The marsh land shown in the red box in Figure 6 is flooded in the first scene (A), but dried out by the last scene (D). This change agrees with seasonal change, despite the overall trend of water classification showing an increase. Counteracting this seasonal drying is an increase in overall misclassification in the first scene (A), as a result of ice and snow, and last scene (D), as a result of vegetation growth (shown in the yellow box in Figure 6). Although the single-polarization methodology was able to see seasonal changes in some wetlands, flooded vegetation was not classified, and misclassification errors occurred as a result of ice and tall vegetation causing an incorrect interpretation of the total surface water change in the area, a clear limitation of using single-polarization data only.
The classification nomenclature chosen in this study is more precise than those currently used in SAR studies for wetland mapping. Thus, many studies using SAR data, in particular intensity parameters (Hess et al., 2003), textural indexes (Gosselin et al., 2012) or polarimetric parameters (Schmitt et al., 2012), only consider main land-cover types (e.g. water, woods, flooded forest, non-flooded forest). Regarding the grassland classes, the classification obtained in this study is as accurate as those obtained in studies comparing vegetation and micro-topography using LIDAR data to classify wetland areas (Moeslund et al., 2011). Several studies indicated that vegetation species in wetlands can be identified using hyperspectral data (Schmidtlein et al., 2007). From the perspective of this study, this method could be assessed to classify wetland vegetation at the dominant-species level using TerraSAR-X data. However, this suggests conducting more accurate field work, for instance in collaboration with botanists, as recommended by Pettorelli et al. (2014), which highlights the need for interdisciplinary studies between remote sensing and ecology communities.
HVE=102 cm; BIO=3.9 kg/m²
t b = 6 h; t e = −10 h; Wd=30 cm
The analysis showed a higher radar signal at locations with water bodies than at locations without water bodies. The brightest radar returns were caused by double-bounce scattering between the water surface and the vertical stems and leaves of the vegetation. The difference in the radar signal level ( ) between the flooded areas and the unflooded areas is generally two times greater in HH compared to HV ( HH~5.5 dB and HV~3.5 dB). This is due to the attenuation of the backscattered radar signal by the vegetation, which is more significant at HV polarization than at HH polarization. Baghdadi et al.  found also that the potential of HH polarization is higher than HV and VV polarizations in a study mapping wetlands from C-band SAR data. Our results also showed that the penetration depth of the radar wave in the X-band is high, even for dense and tall vegetation. For HVE between 20 and 55 cm and water bodies with depths between 4 and 10 cm, flooded areas are clearly visible on the images (Figure 8a,c). A strong penetration was also observed in other training grassland plots with HVE between 71 and 102 cm and water bodies with depths of approximately 30 cm (Figure 8b,d). These plots corresponded to wet biomass (BIO) values up to 3.9 kg/m².
Remote Sens. 2018, 10, x FOR PEER REVIEW 16 of 29
Figure 13. Histogram of TerraSAR-X related fields of research, expressed in percent based on the analysis of 2850 publications.
A noticeably high percentage of papers deal with methodological and technical issues (36 percent), followed by a large number of papers in the hydrosphere group (22 percent). The geosphere (13 percent) represents the second largest geoscientific group. Whereas the fractions of anthroposphere (11 percent) and biosphere (10 percent) are very similar, the cryosphere (8 percent) shows a relative small number of publications. It should be noted that there are overlaps between the different research fields, particularly between hydrosphere and cryosphere as well as anthroposphere and geosphere. Snow and glaciers are at the center of attention in cryosphere research but at the same time they also play an important role in the hydrological cycle, e.g., as water source and reservoir. The observations of volcanos and earthquakes improve the understanding of the tectonically active areas of the Earth in terms of location and areal extent, dynamics and intensity. Many of those areas are densely populated and, thus, relevant for anthropogenic research as well.
4.3 Environmental studies
Remote sensing methods are very often employed in environmental studies. The main advantages of TerraSAR-X in this discipline, similarly with other SAR systems like ERS-2, Envisat, RADARSAT etc. are big spatial coverage, independency on weather conditions, as well as night and day possible acquisitions. For the X-band used by TerraSAR- X the penetration depth in any type of material is lower than for lower frequencies (C/L-band). This feature can be an advantage while considering evaluating land coverage with TerraSAR-X images, flood mapping or identification of burned areas. There are many applications of InSAR in environmental researches, starting with small deformations caused by mining industry,as well as wetlands, ice and snow coverage monitoring. One of the limitations for using SAR data in environmental studies was the resolution of images. Up to now SAR images with the resolution of app. 20m couldn’t provide accurate results for monitoring small, local phenomenon. Additionally, long temporal baseline for some areas was making the interferometric processing impossible. Moreover, TerraSAR-X data are provided with precise orbit information reducing the effort for baseline estimation, and baseline error itself.
Deep learning has been widely used in recent years in computer vision, biology, medical imaging, and remote sensing. Although the theory of deep learning is not yet mature, its capabilities shown in numerous applications have attracted the attention of many researchers. Let us simply review the development of image translation with deep learning. In 2016, Gatys et al. demonstrated the power of Convolutional Neural Networks (CNNs) in creating fantastic artistic imagery. With a good understanding of pretrained VGG networks, they achieved style transfer and demonstrated that semantic exchange could be made using neural networks. Since then, Neural Style Transfer has become a trending topic both in academic literature and industrial applications . To accelerate the speed of Neural Style Transfer, a lot of follow-up studies were conducted. A typical one is Texture networks. With the appearance of GANs, several researchers turned to GANs to find more general methods without defining the texture. In this paper, we examine three typical methods, the method of Gatys et al. , Texture Networks , and Conditional GANs . By analyzing their advantages and disadvantages in SAR image translations, we propose a new GAN-based framework which is the combination of the definition of SAR image content and the GAN method.
Abstract: A Synthetic Aperture Radar (SAR) sensor with high geolocation accuracy greatly simplifies the task of combining multiple data takes within a common geodetic reference system or Geographic Information System (GIS), and is a critical enabler for many applications such as near-real-time disaster mapping. In this study, the geolocation accuracy was estimated using the same methodology for products from three SAR sensors: TerraSAR-X (two identical satellites), COSMO-SkyMed (four identical satellites) and RADARSAT-2. Known errors caused by atmospheric refraction, plate tectonics and the solid-Earth tide were modeled and compensated during the analysis. Of the products analyzed, TerraSAR-X provided the highest absolute and relative geolocation accuracy.
Many studies have been performed with TerraSAR-X data, but the majority have focused either on the accuracy of the elevation reconstruction, the geolocation or other technical properties, describing it more as an instrument comparative with, e.g., ALS. Some of the earliest studies deriving forest related variables from TerraSAR-X data are described in [20–22]. They found that the accuracy of the derived DSMs were very high over bare ground while regions of forest were underestimated in the range of 25%–35% compared to the real canopy height, with an average underestimation of 27%. They also concluded that additional experiments over different types of forest are needed to establish if this underestimation would be reduced on a large scale. These early forest studies were extended in . As mainly dense stands of deciduous trees were used, it was also stated that the canopy height underestimation was expected to be larger for coniferous trees and for clearer stands. No seasonal effects (frozen/unfrozen) or the influence of weather conditions could be studied as all images used in the study were acquired during April-June and no weather differences were present. This was also the case in  where the authors carried out Random Forest evaluation of forest using 109 field plots with 8 m radius, using only spring-time images. They found that the forest stem volume could be predicted with 34.0% relative root mean square error (RMSE) and the mean forest canopy height with 14.0%. Overall, the authors concluded that the influences of tree species, seasonal effects, and weather conditions have to be studied further. A further study  from primarily the same authors utilized the same field and SAR data, but investigated plot-level estimations of AGB and stem volume compared to ALS estimates. The authors concluded that radargrammetry is a promising technique for this and that AGB and stem volume could be estimated with 29.9% and 30.2% relative accuracy in comparison with 21.9% and 24.8% for ALS. They also concluded that further studies on radargrammetry for large-area mapping are needed.
The usefulness of earth observation in crisis situations such as large-scale floods greatly depends on the timeliness of the first post-disaster satellite acquisition and the quality of subsequent data processing and product generation.
In this work we presented two fully automatic processing chains aimed to improve the timeliness of data handling and product dissemination through a combined use of both optical and radar data in flood monitoring. The classification output from systematically and daily-acquired MODIS data (monitoring mode) is used for an on demand triggering of a TerraSAR-X based flood mapping service (emergency response mode) to derive high-resolution information on the inundation extent. The methodology includes a computation and adaption of global auxiliary data (digital elevation models, topographic slope information, and reference water masks), an unsupervised initialization of the classification, a post-classification refinement, and dissemination of the crisis information via a web-based user interface.
Abstract. The calving fronts of many tidewater glaciers in Greenland have been undergoing strong seasonal and in- terannual fluctuations. Conventionally, calving front posi- tions have been manually delineated from remote sensing images. But manual practices can be labor-intensive and time-consuming, particularly when processing a large num- ber of images taken over decades and covering large areas with many glaciers, such as Greenland. Applying U-Net, a deep learning architecture, to multitemporal synthetic aper- ture radar images taken by the TerraSAR-X satellite, we here automatically delineate the calving front positions of Jakobshavn Isbræ from 2009 to 2015. Our results are con- sistent with the manually delineated products generated by the Greenland Ice Sheet Climate Change Initiative project. We show that the calving fronts of Jakobshavn’s two main branches retreated at mean rates of − 117 ± 1 and −157 ± 1 m yr −1 , respectively, during the years 2009 to 2015. The interannual calving front variations can be roughly divided into three phases for both branches. The retreat rates of the two branches tripled and doubled, respectively, from phase 1 (April 2009–January 2011) to phase 2 (January 2011– January 2013) and then stabilized to nearly zero in phase 3 (January 2013–December 2015). We suggest that the re- treat of the calving front into an overdeepened basin whose bed is retrograde may have accelerated the retreat after 2011, while the inland–uphill bed slope behind the bottom of the overdeepened basin has prevented the glacier from retreat- ing further after 2012. Demonstrating through this successful case study on Jakobshavn Isbræ and due to the transferable nature of deep learning, our methodology can be applied to many other tidewater glaciers both in Greenland and else- where in the world, using multitemporal and multisensor re- mote sensing imagery.
Baghdadi et al. (2008) examined the sensitivity of TerraSAR-X radar signals to surface soil parameters reflected by the roughest and the smoothest areas over agricultural fields, using HH polarization and various incidence angles (26°, 28°, 50°, 52°). Two study area sites were used: the first is in Villamblain, South of Paris and the second site is located at the Orgeval watershed, located to the East of Paris. Seven TerraSAR-X images were acquired in the spotlight mode with a spatial resolution of 1m during January and February 2008. Baghdadi et al. (2008) also analysed the potential of L-band of ALOS PALSAR images acquired in February 2008 with HH polarization at a 38° incidence angle and 6.25 m spatial resolution. A single ASAR image was also acquired in February 2008 with both HH and VV polarizations in the 1S2 mode (image swath) with 23° incidence and 12.5 m resolution. Ground truth measurements of soil moisture content, bulk density, and surface roughness were performed by collecting gravimetric soil moisture samples at depths in the range 0–5cm. They calculated the volumetric soil moisture by multiplying the gravimetric soil moisture by the dry soil bulk density. Each training field estimated the backscattering coefficient σ° by averaging the linear value of σ° for all pixels related to a given field followed by converting results into decibels (dB). The study arrived at the following conclusions: 1) TerraSAR-X radar signal is to some extent more sensitive at high incidence angles to surface roughness, 2) sensitivity increases in the L-band (low frequency) with PALSAR/ALOS data, 3) the dynamics of radar signals for frozen or very wet soils is reduced based on the roughness parameter (rms), 4) under very wet soil conditions radar signals decrease at high or low incidence angles, and 5) soils with high water content appear darker in TerraSAR-X imagery.
A. TerraSAR-X images
A unique data set consisting of numerous types of remotely sensed images over one single event hydrograph were acquired over the selected study area . From this dataset a stripmap TerraSAR-X image acquired on July 25 2007 (at 06:34 GMT, Wednesday) was selected (see Fig. 2). The image is a multi-look ground range spatially enhanced scene with 1.5 m pixel spacing and has a mean incidence angle of 24°. Its H/H polarization mode arguably allows for the best discrimination between a SAR image’s flooded and non-flooded parts . At the time of the satellite overpass and image acquisition, there was relatively low wind speed and no rain . Moreover, no rainfall was recorded in the 30 hours preceding the TerraSAR-X acquisition, as well as during the satellite overpass itself.
desministerisum für Bildung und Forschung ; BMBF), the
German Aerospace Center (Deutsches Zentrum für Luft- und
Raumfahrt ; DLR) and Astrium GmbH. The launch took place
on 15 June 2007, and the sensor has a nominal lifetime of 5 years, although it is currently expected to last up to 7 years. The payload of the satellite is an X-band SAR system with a 9.65 GHz center frequency, an electronically steerable phased-array antenna and a side-looking imaging capability within an off-nadir pivoting range of approximately 20 ◦ –55 ◦ . The satellite is in a near-polar dawn/dusk orbit at an altitude of 514 km. Using its active radar antenna, it is able to produce image data with a spatial resolution on the order of ∼1 m, regardless of weather conditions, cloud cover or absence of daylight.
This paper describes a method for monitoring winter wheat growth using multi-temporal TerraSAR-X dual-polarimetric data. Six TerraSAR-X HH/VV images were collected in Hokkaido, and the temporal responses to the winter wheat fields were analyzed. The height, moisture content and dry matter of the crops were measured at nearly the same time as TerraSAR-X data was acquired, and the relation- ships between these parameters and SAR data, including sigma naught and coherence, were studied. Quadratic relationships between the crop height and sigma naught were observed for HH polarization. The determination coefficient was 0.73 and the model had an RMS error of 0.17 dB for the validation data. Coherence is expressed as a regression equation with two explanatory variables: crop height and elongation. Next, the determination coefficient of 0.69 was observed for HH, while the RMS error of coherence was 0.01 for the validation data. The possibility of using the co-polarization ratio of TerraSAR-X to estimate the vegetation’s water content was also analyzed and a determination coef- ficient of 0.70 was obtained. The results confirm that X-band SAR data possess great potential for the development of an operational system for monitoring wheat growth.