environmental applications have been developed namely land use and land cover change monitoring, urban sprawling, disaster management, vegetation mapping, crop production prediction and so on. Since surface water source is an indispensable resource for socio-economic development, its detailed mapping and monitoring is essential for sustainable use of water resources. Water scarcity and food security are the major consequences of climate change that we are going to deal with in the near future. Various parts of Bangladesh have already started facing such problems in different magnitude. In this regard, integrated water resource management is of critical importance for sustainable development as well as for achieving sustainable development goals. Surface watermapping may play a pivotal role as it will provide us with the amount and spatial distribution of surface water resources. Remote sensing satelliteimages are extensively used in this regard since such images offer synoptic vision of the region with various spatial, temporal and spectral resolutions. There are several image processing techniques that may be used for identifying water bodies from satelliteimages. For example, single-band technique extract water features through the use of a threshold value. But such method misclassifies features due to mixed pixel issues. Mixed-band techniques such as Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI) or Modified Normalized Difference Water Index (MNDWI) may be used for the extraction of water features from satelliteimages. In some cases, NDWI involves error in identifying waterbody as it generates false positive alarm in case of built-up area. On the other hand, MNDWI could overcome such error thereby achieving higher accuracy in identifying water bodies while suppressing errors from vegetation, soil as well as built-up land (Rokni, Ahmad, Selamat & Hazini, 2014). Sarp & Ozcelik,
Drought Monitoring by remotely sensed moisture vegetation indexes is being an active research subject as the vegetation spectral responses are showed to be highly correlated to water content. The MODIS (MODerate resolution Imaging Spectro-radiometer) sensor of the Terra satellite provides MOD09A1 product of BRDF (Bidirectional Reflectance Distribution Function) used in computing moisture vegetation indexes (MVI). The exploration of an MVI time-series in the Kroumirie forest in Northern Tunisia showed important noise due to both clouds contamination and sensor defaults that had to be removed. Amongst methods for removing these imperfections, TIMESAT tool was designed for correcting time-series of satellite data and also to retrieve seasonal parameters from smoothed vegetation indexes. The methodology of smoothing functions to fit the timeseries data is based on two stages. First, a least square fit to the upper envelope of the vegetation indices series is ap- plied. The second stage is achieved by local and adaptive fitting functions. The corrections have been made by spikes removal due to abrupt change of MVI variations and by fitting the MVI time-series to the upper envelop to correct the negative biases of remote sensing vegetation indexes. The adaptive Savitsky- Golay function filter compared to local filtering process produces variations that conserve local variations for all the tested MVI. Seasonal vegetation pa- rameters were extracted for each year of the time-series analysis and com- pared to the Standardized Precipitation Index (SPI) calculated at meteorolog- ical station level and for different time scales. Positive relations were found between SPI and the seasonal parameters expressed by the length and the am- plitude of the season, indicating MODIS derived MVI sensitivity to water def- icit or surplus conditions. The 6-month SPI showed the best performance when related to water sensitive indexes suggesting that MODIS derived in- dexes are more correlated to the precipitation variations over seasons.
There is a clear potential to upscale this information to national, European or even global scale. Similar upscaling to global scale was recently done for forest  and for surface water bodies . Main condition is that agricultural fields are bare after harvesting and/or before seeding, which may not be the case everywhere. Other auxiliary data than used in this study is necessary for masking forest, built-up, water and natural areas. For European scale, candidates are the Hansen Global Forest Change dataset  for forest and for water, wetlands, natural grassland and built-up it is possible to use data from the Copernicus Land Monitoring Service, respectively, the Permanent Water Bodies dataset, the Wetlands dataset, the Natural Grassland dataset, and the Imperviousness dataset. All datasets are available in 20 and 100 m resolution. For global scale, the same data may be used for masking forest. For water, the JRC Global Surface WaterMapping Layers  are available and for built-up the Global Urban Footprint . There are no readily available sources for masking natural (vegetated) areas at high spatial resolution and global extent but available land-cover products provide feasible alternatives. Soil properties could be predicted based on Harmonized Global Soil Profile Dataset (HGSPD ). The HGSPD is one of the global datasets used to create the Harmonized World Soil Database (HWSD ) and consists of 10,250 soil profiles, with some 47,800 horizons, from 149 countries. The dataset contains, however, several extended areas with lacking or a very low sampling density and the profiles are not uniformly sampled, described, and analysed, but vary according to methods and standards in use in the originating countries.
3.2.1 Interference from clouds, shadows, and SLC-off The influence of clouds, shadows, and SLC failure on esti- mated flooded areas from Landsat imagery over 1999–2014 is illustrated in Figs. 10 and 11 based on field data from one small reservoir. Detected on 28 % of the 546 Landsatimages, clouds lead essentially to commission errors (false positives) and overestimation, as a result of visible wavelengths being reflected while much of the electromagnetic energy is ab- sorbed by the droplets in clouds. Their effect was not sys- tematic and proportional (Fig. 11), as shown by underesti- mations in 30 % of cases where clouds were detected across the whole lake cell. The diversity in the nature and proper- ties (temperature, thickness, water content, etc.) of clouds is indeed firstly responsible for the heterogeneous reflectance observed and the resulting classification difficulties. The in- fluence of shadows is more moderate and less frequent, re- sponsible for overestimations in only 5 % of images. This results partly from the greater difficulties in discriminating shadows and the overlap from clouds. Pixel loss from SLC failure varies across Landsat tiles, and for a given lake, 30 % of the 287 Landsat 7 SLC-off images suffered from minor pixel loss (<10 %). SLC-off pixel loss led to a systematic un- derestimation of flooded areas, on average by 35 % (Fig. 11). 3.2.2 Optimizing image availability vs. interferences Optimizing on R 2 tends towards removing all cloud and SLC interferences, and therefore reduces the number of available observations. Considering the objective of hydrological mon- itoring and maintaining sufficient temporal repetitivity, im- ages with up to 25 % SLC-off pixels and 40 % cloud and shadow pixels over the studied lake cell were retained here. These thresholds were found to minimize the mean squared error on surface area aggregated over 15 years, which gives importance to both the number and quality of the Landsat observations over time.
The use of persistent scatterers SAR interferometry in the detection of urban subsidence has gained a lot of popularity throughout the world after the first results produced by Ferretti et al. (1999). Ground subsidence is a well-known problem in Finland also, but extensive information about the phenomenon is usually missing. Local authorities and private companies carry out levelling surveys to monitor certain buildings that are, for example, historically remarkable. Therefore, any subsidence information provided by PSI could be valuable to local authorities or owners of the buildings. In publication V, the goal was to evaluate the usability of PSI in Finland. The cities of Helsinki and Turku were selected as test areas and a timeseries of ERS-1 and ERS-2 SAR were obtained through a Category-1 project of ESA. However, the comparison of subsidence rates of PSI and levelling was only possible in Turku, where the levelling surveys were more comprehensive than in Helsinki. Moreover, the Turku case is very favourable for testing the accuracy of PSI based methods, because the rate of the subsidence has been very stable throughout the years. The Coherent Target Monitoring software module developed by the Vexcel Canada (nowadays owned by the MDA Geospatial) was used in the PSI analysis. The Real Estate Department of Turku provided us with levelling surveys. In total, there were 10 buildings, for which both PSI and levelling subsidence rates were available. The number of coherent targets (PSIs) and levelling benchmarks varied from one building to another, but nevertheless, there were several points for comparison for each building. According to the results, very good agreement of the PSI subsidence values in comparison to levelling measurements was achieved. An example of two buildings is presented in Figure 9. When all 10 test buildings were considered, the RMSE of 0.82 mm/year between PSI and levelling subsidence rates was obtained.
Hyperspectral data were obtained from the 1-nm-wide narrowband FieldSpec Pro FR spectroradiometer manufactured by Analytical Spectral Devices (ASD) measuring spectra over a spectral range of 0.4–2.5 µm (ASD, 2007). Gathering spectra at any given location involved optimizing the integration time (typically set at 17 ms), providing foreoptic information, recording dark current, collecting white reference reflectance, and then obtaining target reflectance. A 18 0 field of view (FOV) was used. At each sampling location, target reflectance was measured from wetland vegetatin species. Using handhold GPS, 30 field plots (30 by 30 m) were located within the south- east section of the Terkos Lake. There was difficulties to sample the plots randomly because of the water depth problem some areas was not reachable by boot for representing a range of wetland vegetation characteristics. Plots constrained to be easily reachable and distributed according to region properties by existing high resolution satelliteimages. Field sampling took place in the middle of the day (i.e. 2 hours before and after true midday) during the summer season (i.e. from June to July 2006 and 2007), while plants were still fresh; there was no evidence of wilting or senescence.
II. S TUDY A REA AND D ATASETS
Throughout this study, we use Landsat TM/ETM+ images from three heterogeneous glacier regions in the world to cover a range of ground conditions and mapping challenges. In view of rapid ground changes related to snow fall and melt, and due to the frequent cloud cover over glacier areas, a considerable limitation of the Landsat missions is the relatively low tempo- ral resolution (16-day revisit time). However, one can simulate higher temporal resolution usingimages from adjacent years (e.g., ±2 years) as glacier changes over such a short-time span are typically small. In this study, we test the time-series map- ping potential in three glaciated regions: 1) north in the Pamirs, on the border between Tajikistan/Kyrgyzstan; 2) the Chugach Mountains in Alaska; and 3) northern Patagonia in Chile. The study sites have been selected because they cover not only cold and dry continental high mountain areas (the Pamirs), warm and wet maritime areas (Chugach Mountains), and maritime areas in the rain shadow of mountains (parts of the northern Patagonian ice field, Chile) but also pure maritime conditions (south of the northern Patagonian ice field). The focus in this study will particularly be on the Chilean and Pamir study areas as they are two opposite endmembers of typical glacier condi- tions in terms of climate conditions. The local Pamir study area covers a region of 344 km 2 , with glaciers at an altitude within approximately 2800–5500 m a.s.l. (centerpoint: Lat: 39.516, Lon: 70.648). The regional study area in northern Patagonia, Chile covers a region of 9450 km 2 and have glaciers in the alti- tude range between ca. 200 and 2400 m a.s.l. (centerpoint: Lat: −47.278, Lon: −72.724).
Abstract. With dense SAR satellite data timeseries it is pos- sible to map surface and subsurface glacier properties that vary in time. On Sentinel-1A and RADARSAT-2 backscat- ter timeseriesimages over mainland Norway and Svalbard, we outline how to map glaciers using descriptive methods. We present five application scenarios. The first shows poten- tial for tracking transient snow lines with SAR backscatter timeseries and correlates with both optical satelliteimages (Sentinel-2A and Landsat 8) and equilibrium line altitudes derived from in situ surface mass balance data. In the sec- ond application scenario, timeseries representation of glacier facies corresponding to SAR glacier zones shows potential for a more accurate delineation of the zones and how they change in time. The third application scenario investigates the firn evolution using dense SAR backscatter timeseries together with a coupled energy balance and multilayer firn model. We find strong correlation between backscatter sig- nals with both the modeled firn air content and modeled wet- ness in the firn. In the fourth application scenario, we high- light how winter rain events can be detected in SAR timeseries, revealing important information about the area extent of internal accumulation. In the last application scenario, av- eraged summer SAR images were found to have potential in assisting the process of mapping glaciers outlines, especially in the presence of seasonal snow. Altogether we present ex- amples of how to map glaciers and to further understand glaciological processes using the existing and future massive amount of multi-sensor timeseries data.
Finally, this proof-of-concept is supporting the interest for satellite mission proposals such as IRSUTE (Seguin et al., 1999) that would provide estimates of the surface temperature with a typical daily revisit period. Indeed, with current satellite revisit capabilities at the 100 m pixel resolution (a resolution compatible with most agronomical applications), successive acquisitions of TIR images are interspaced with large periods of time (15 days) which can include several rainfall events. Some satellites offer a higher revisit frequency, but for a much coarser resolution (of the order of 1 km) often incompatible with the scale of application. Therefore, using radiation fluxes, albedo, emissivity, and LAI values deduced from the existing frequent (every 2–3 days) high and low resolution Remote-Sensing data would help to derive maps of unstressed temperatures only if it could be combined with series of observed surface temperature images from a high-resolution IRT sensor such as IRSUTE. Then, estimates of pixel-to-
MODIS-based results in this study were largely insensitive to subtle canopy damages from selective logging. Any integrated approach that requires several years of satellite imagery will not replace the need for operational deforestation monitoring (INPE, 2006) or global burn scar mapping (e.g., Roy et al. 2008; Giglio et al. 2009). As presented here, the BDR algorithm can only map canopy damages from fire with high confidence on a two-year delay in order to track 570
Slow-moving landslides are widespread in many landscapes with significant impacts on the topographic relief, sediment transfer and human settlements. Their area-wide mapping and monitoring in mountainous terrain, however, is still challenging. The growing archives of optical remote sensing images offer great potential for the operational detection and monitoring of surface motion in such areas. This study proposes a multiple pair wise image correlation (MPIC) technique to obtain a series of redundant horizontal displacement fields, and different multi- temporal indicators for a more accurate detection and quantification of surface displacement. The technique is developed and tested on a series of monoscopic and stereoscopic Pléiades satelliteimages at a test site in the South French Alps. Empirical tests confirm that MPIC signifi- cantly increased detection accuracy (F−measure = 0.85) and that the measurement error can be reduced by averaging velocities from all pair combinations covering a given time-step (i.e. when stereo- pairs are available for at least one date). The derived inventory and displacement fields of 169 slow-moving landslides show a positive relationship between the landslide size and velocities, as well as a seasonal acceleration of the largest landslides in response to an increase in effective precipitation. The processing technique can be adapted to better exploit increasingly available time-series from a variety of optical satellites for the detection and monitoring of landslide displacement .
vegetation phenology (Verbesselt et al., 2012; Watts and Laffan, 2014). However, this approach requires high quality data corrected for radiometric disturbances (e.g., atmospheric effects and illumination) as well as geometric properties (pixel registration and elevation effects). Since tropical forests are characterized by season-specific spectral response, a method failing to account for that can generate mapping inaccuracies of more than 40% (Langner et al., 2014). BFAST accommodates for seasonal changes using harmonic modeling and therefore allows the use of all available timeseries observations. The method can be used across sensors and was successfully applied for mapping small-scale changes in the tropics usingLandsat data (DeVries et al., 2015b) and in combination with RADAR data (Reiche et al., 2015). Deforestation monitoring is influenced by many factors, among them the different types and drivers of forest change and their impact on the spectral signal influence mapping accuracies, and the amount of available observation data (Souza et al., 2009). Furthermore, timeseries contaminants introduced by atmospheric and/or topographic features need to be suppressed, as they are recognized sources of noise (Hansen and Loveland, 2012; Song et al., 2001; Wulder and Franklin, 2003) Removing clouds from images is compulsory, as it determines the success of any change detection mapping effort (Hilker et al., 2012). Regardless of the applied monitoring, method accounting for these effects will determine the success of the deforestation mapping. Landsat preprocessing suitability was identified by a performance-based comparison of various preprocessing schemes. Most commonly used correction methods for such applications are summarized in (Vanonckelen et al., 2013). Often used are Landsat Ecosystem Disturbance Adaptive Processing System (LEDAPS) (Vermote et al., 1997), Moderate spectral resolution atmosphere transmittance algorithm (MODTRAN) (Berk et al., 2008), or Dark Object
Monitoring plastic-mulched land cover through per-pixel techniques is attracting a great deal of attention nowadays, particularly in China. In this way, Hasituya et al. [ 16 ] monitored plastic-mulched farmland by a single Landsat 8 OLI image using spectral and textural features. Working with multi-temporal imagery, Lu et al. [ 17 ] achieved a very simple but consistent decision tree classifier for extracting transparent plastic-mulched land cover from a very short Landsat 5 TM timeseries composed by only two images during an agricultural season. They proposed a plastic-mulched land cover index (PMLI) by using the reflectance values of red band (b3) and SWIR1 (b5). Furthermore, large timeseries of MODIS surface reflectance daily L2G (250 m ground sample distance, GSD) covering the cotton crop period from the 85th day to the 150th day were also used by Lu et al. [ 18 ] for plastic-mulched land cover extraction. A simple threshold model based on the temporal-spectral features of plastic-mulched land cover in the early stage of a growing season after planting was designed with the number of days (d) when the normalized difference vegetation index (NDVI) value was larger than a threshold value (x) as the discriminator. This rule achieved very good results along three different years.
Multi-temporal L-band SAR data for 2007 and 2010 was obtained from the phased array L-band synthetic aperture radar sensor aboard the advanced land observing satellite (ALOS PALSAR) acquired in Fine Beam Dual (FBD) mode (Shimada et al. 2010). All images were acquired in ascending mode with an incidence angle of 34.3° and were provided in Level 1.1 format. The swath width of FBD data is 70 km. The data set covers Track/Frame 116/90 and 117/90 (Figure 2.2) and consists of two FBD images per frame for 2007 and 2010 (Table 2.1). The ALOS PALSAR FBD images were pre-processed independently using the Gamma software package (Werner & Strozzi 2000), combined with SRTM (Shuttle Radar Topography Mission) version 4 DEM information (http://srtm.csi.cgiar.org/). Pre-processing included multi- looking, radiometric calibration using standard calibration coefficients (Shimada et al. 2009), topographic normalization as described in (Hoekman et al. 2010) as well as geocoding to 25 m pixel resolution (WGS84, UTM 21N). Visual comparison with Landsat and high resolution RapidEye as reference showed very good geocoding accuracy (around 1 pixel error, 25 m). Among the different pre-processed images for one tile no significant geolocation differences are visible. In addition to the general SAR pre-processing steps, adaptive multi-temporal SAR filtering (Quegan & Yu 2001) has been applied separately to the FBD image pairs (HH - horizontal transmit and horizontal receive polarisation; HV - horizontal transmit and vertical receive polarisation) for 2007 and 2010. In contrast to conventional bi-dimensional SAR speckle filter that result in a trade-off between speckle-reduction and decreased spatial resolution, multi-temporal SAR filter reduce the SAR speckle without losing radiometric accuracy and spatial resolution of the single channels (Trouvé et al. 2003; Quegan et al. 2000; Quegan & Yu 2001). A measured increase of the equivalent number of looks (ENL) from 4.58±0.03 to 8.45±0.46 for HH and from 5.13±0.35 to 9.05±0.51 for HV indicates a clear reduction of SAR speckle in the data. This is crucial for analysing the data at pixel level or small segment level. In addition, no significant changes in the mean radiometric characteristics were observed.
Figure 4 shows the flooding pattern for the single-season rice-shrimp crops aggregated overall sample areas. The percentage of water was more than 90% from March to August for the period of 2016- 2017. The water is pumped out of the floodplain compartments in January 2016. During the growing stage between October and February water is nearly absent. After this period, the water began to increase from February (48%) in October (74%). The water remained high at around 90% for 8 months then declined to 12% in November. The water percentage rose from 26.78% in January to 94.42% in March 2017. The peak was 99.22% in July. Afterward, it reduced to 60.89% in September. The single-season rice-shrimp cropping system depends on the salinity and the beginning of the rainy season time. In addition, the single rice-shrimp crops are near the coastlines. After harvesting shrimp in the middle of September, the rice was sowed. Those are the reasons why the percentage of water was always high from March to September.
DOI: 10.4236/jgis.2019.113020 332 Journal of Geographic Information System (SSR), is among the most important fungal diseases affecting soybean yields and represents a recurring annual threat to soybean production in South Dakota. In- itially reported in Poland in 1982 as a disease of local importance , white mold was, more than a decade later, ranked in the top ten diseases that suppress soy- bean yields . The apothecia of white mold generally appear after the crop ca- nopy develops, around mid to late July and the environmental conditions cor- responding to the development of white mold are cool (air temperature around 12˚C - 24˚C), wet and moist (enough rain: 70 - 120 hours of continuous wet- ness) conditions . These conditions are favorable for optimal yield; therefore, incidence of white mold has been negatively correlated with yields  because the disease is more likely to develop where there is high yield potential. Thus, mapping and quantifying the disease is crucial to understand its impact on yields, and two options can be used: field scouting represents an accurate as- sessment, but remains time-consuming and does not provide a global view of the variations in the field, while remote sensing represents the best solution because it provides a synoptic view and allows observations to span large areas in a short period .
NDVI _ _ from NOAA.
For this study, satellite data from Meteosat Second Generation (MSG) and National Oceanic and Atmospheric Administration (NOAA) AVHRR were used. MSG is the new European system of geostationary meteorological satellites with the associated infrastructure; it was developed to succeed the highly successful series of original Meteosat satellites that has served the meteorological community since the first launch in 1977 (EUMETSAT, 2005). The advanced Spinning Enhanced Visible and Infrared Imager (SEVIRI) radiometer onboard the MSG satellites enables the Earth to be scanned in 12 spectral channels from visible to thermal infrared at 15 minute intervals. Each of the 12 channels has one or more specific applications, either when used alone or in conjunction with data from other channels. From these 12 channels, this research used Channels 1 and 2 to detect vegetation condition. These two visible channels are well known from similar channels of the AVHRR instrument flown on NOAA satellites and can be used in combination to generate vegetation indices, such as NDVI (EUMETSAT, 2005).
Abstract. Advances in remote sensing technologies have allowed us to send an ever-increasing number of satellites in orbit around Earth. As a result, Earth Observation data archives have been constantly increasing in size in the last few years, and have become a valuable source of data for many scientific and application domains. When EO data is coupled with other data sources many pionner applications can be developed. In this submission to the semantic web challenge we show how EO data, on- tologies, and linked geospatial data can be combined for the development of a wildfire monitoring service that goes beyond applications currently deployed in various EO data centers. The service has been developed in the context of European project TELEIOS that faces head in the chal- lenges of extracting knowledge from EO data, capturing this knowledge by semantic annotation encoded using EO ontologies, and combining these annotations with linked geospatial data to allow the development of interesting applications.
Including the spatial information in the classification process can increase its performance (Daya Sagar & Serra, 2010). The performance of crop mapping can also be improved by inclusion of the spatial information about the croplands in the classification. Accordingly, the first step of the proposed algorithm is to extract the spatial information for crop classification. There have been several methods proposed for extracting spatial features that model the spatial information of data. However, the proposed feature extraction algorithm takes advantage of image segmentation methods to extract spatial patterns from the time-series data. The purpose of image segmentation algorithms is to partition the time-series data into several homogeneous segments. Each segment contains a group of pixels which are spatially close to each other and share similar temporal characteristics. Several image segmentation methods have been proposed, which can be used in this step. However, in this paper, the multiresolution segmentation, implemented in eCognition developer software due to its good performance, was adapted as image segmentation algorithm (Definiens, 2009).
7.2.3. User feedback
The service we developed has a web based interface tailored for decision makers and crisis managers interested in real time wildfire monitoring. While new hotspots are detected the application animates the evolution of a fire-front along with useful auxiliary information (e.g., affected municipalities). Refinement operations are hidden from the end-user for simplicity. Also, a user-friendly option is offered to retrieve and watch past wildfires. The application was used during the fire seasons of 2012 and 2013 by the Greek Civil Protection Agency, the Fire Brigade and the Army both during the fire events for strategy planning and after them to assess the strategies that were followed. The service has also been thoroughly tested during the 3rd user workshop of the project TELEIOS. The spectrum of users which participated this test includes both end users which use fire monitoring products on an operational basis (e.g., civil protection agencies) and stakeholders from the EO and IT communities. In general the collected feedback was very encouraging: most users found the applications very useful, specifically when it concerned stakeholders that need fire monitoring products as part of their daily work practice (e.g. Greek Ministry of Environment, Energy and Climate Change, Italian civil defense agency, or foresters in local administrative units). The value of applying semantic queries for the thematic refinement of the hotspot products has been appreciated by EO community. It was also shown, and it is envisaged, that rapid mapping applications can be easily deployed using semantic technologies with distributed data. EO service providers are enabled to use