Denitsa Borisova

Clustering of spectral images using Echo state networks

In the present work we applied a recently developed procedure for multidimensional data clustering to processing of spectral satellite images. The core of our approach lays in projection of multidimensional image to a two dimensional one. The main aim is to discover points with similar characteristics. This was done by clustering of the resulting image. The processing technique exploits equilibrium states of a kind of recurrent neural network - Echo state network (ESN) - that are obtained after intrinsic plasticity (IP) tuning of the ESN using multidimensional data as inputs. The proposed in our previous work automated procedure for multidimensional data clustering is further refined and tested on the satellite image data. The obtained number and position of clusters of a multi-spectral image of a mountain region in Bulgaria is compared with the classification of the region landscape given by the Ministry of Regional Development and Public Works.

Recurrent neural networks for automatic clustering of multispectral satellite images

In the present work we applied a recently developed procedure for multidimensional data clustering to multispectral satellite images. The core of our approach lays in projection of the multidimensional image to a two dimensional space. For this purpose we used extensively investigated family of recurrent artificial neural networks (RNN) called “Echo state network” (ESN). ESN incorporates a randomly generated recurrent reservoir with sigmoid nonlinearities of neurons outputs. The procedure called Intrinsic Plasticity (IP) that is aimed at reservoir output entropy maximization was applied for adapting of reservoir steady states to the multidimensional input data. Next we consider all possible combinations between steady states of each two neurons in the reservoir as two-dimensional projections of the original multidimensional data. These low dimensional projections were subjected to subtractive clustering in order to determine number and position of data clusters. Two approaches to choose a proper projection among the all possible combinations between neurons were investigated. The first one is based on the calculation of two-dimensional density distributions of each projection, determination of number of their local maxima and choice of the projections with biggest number of these maxima. The second one applies clustering to all projections and chooses those with maximum number of clusters. Multispectral data from Landsat 7 Enhanced Thematic Mapper Plus (ETM+) instrument are used in this work. The obtained number and position of clusters of a multi-spectral image of a mountain region in Bulgaria is compared with the regional landscape classification.

Assessment and mapping of the current state of the landscapes/ecosystems in Haskovo region (southeastern Bulgaria) in relation to ecosystem services using remote sensing and GIS

Assessment and mapping of the ecosystems state in the context of ecosystem services that they supply are important tasks to improve human well-being, especially in regions with considerable land degradation. Haskovo region is situated in the Southeastern part of Bulgaria and is considered as an extremely sensitive to land degradation in terms of climate change and human activities in result of unappropriated land management practices. In order to improve the conservation activities and ecosystem services of the region, rapid and available technics are needed in addition to the used analytical methods. The study presents the potential of remote sensing methods (satellite data from different sensors Sentinel and Landsat) and GIS for assessment of the current state of the landscapes to supply ecosystem services and allows a comprehensive evaluation of the main indicators for assessment of ecosystem services to be performed. The proposed methodology includes application of vegetation indices (NDVI, NDWI and MSAVI2) and SAR data. The results show that the referred technics can be used for a rapid and accurate assessment of the main indicators showing the state of the terrestrial ecosystems such as: soil degradation, land use and impact of human activities, responsible for the ecosystem services supply.

Application of remote sensing for ecosystems monitoring and risk assessment

In recent years on the territory of Bulgaria it has been observed the existence of events with extreme character – floods, forest fires, etc.- that have a negative effect on ecosystems and ecosystem services. The purpose of the present research is the application of remote sensing for ecological monitoring implementation for the ecosystems upon the appearance of natural hazards. In this paper a methodology for ecological monitoring in different temporal intervals has been proposed and additionally the results from the application of remote sensing for the purpose of ecosystem monitoring and risk assessment in case of events that induce negative effect on ecosystems have been presented. The methodology and criteria have been implemented in observing different types of ecosystems. For the purpose of the present investigation satellite data with different spatial, temporal and spectral resolution from Sentinel 2, Landsat and air photo images have been used. Terrestrial data have been used for results verification and validation. The introduced results have been obtained for different temporal intervals from ecological monitoring, on which base criteria for optimization of the temporal characteristics of the ecological monitoring have been suggested. The present research is with conformance of Directive 92/43/EEC on the conservation of natural habitats and of wild fauna and flora and Directive 2009/147/EC on the conservation of wild birds. The results from the completed research can be of benefit for defining concrete actions for the implementation of measurements appointed in the Action Plan for nature, people and the economy of 27.4.2017 COM(2017) 198.

Remote sensing measurements in creating thematic spectral library

In Earth observations the reference spectra of well-described objects are required for better object-oriented interpretation of remotely sensed data from laboratory, field, airborne, and satellite sensors. For this purpose measurements of spectra using laboratory and field spectrometers are performed. The acquired spectra are used in creating a thematic spectral library. The used spectral instruments work in the wavelengths (0.4 to 2.5 microns) covering the spectral ranges from the visible /VIS/ to the shortwave infrared /SWIR/. Two different spectrometers are used to measure spectra included in the library: (1) Thematically oriented multichannel spectrometer covering the spectral range 0.4 to 0.9 microns and (2) high resolution NIRQuest spectrometer covering the range from 0.9 to 2.5 microns, both models of Ocean Optics Inc. Spectrometric measurements of representative samples of minerals, rocks, related soils, vegetation, and their natural mixtures are made in laboratory and field conditions. In some cases, samples were purified, so that the unique spectral characteristics of the studied objects could be related to their typical structure. The relations between the spectra and the structures are important for interpreting remotely sensed data acquired in the field or from an air- or space-borne platform. In some cases for making easy wide use of the spectra in the library the obtained spectra have to resample to selected broadband multispectral sensors for example those based on the satellites Landsat and Sentinel. The obtained spectral data with the metadata and additional information are planned for including in files for better interpretation of images with different spatial resolution.

Multisensor Earth observation systems: data fusion

Advancements in modern technologies, such as remote sensing systems and instruments have led to rapid developments in the field of Earth observation /EO/. As a result, enormous volumes of EO data with various spatial and spectral resolutions are obtained. However, the expected enhancements in the classification accuracy still have not been reached, due to the complexity of the remote sensing measurements and the big volume of data that need to be processed. The last leads to the necessity of development and improvement of methods and techniques for data obtaining and analysis. The methods include the validation multi-sensor systems, the processing technique of big data, and the object identification and classification methods for improving information quality through data fusion. To achieve correct information with highest accuracy in data analyzing and interpreting, researchers have to apply these methods and to create technologies for obtaining and integrating data from different Earth Observation Systems /EOS/. For gathering and using all of the information a local and regional EOS of Systems needs to be established. By creating such local EOS of Systems more extensive information could be collected, analyzed and retrieved. In this paper a local system is presented, focusing on the description of the ground component. The main sensors embedded in the system are spectrometers. The working range of the multi-sensor system is VIS-NIR-SWIR. Thus, by applying the data fusion methods, combining images and spectral information, a more accurate thematic interpretation is achieved. Example illustrating the benefits of a multisensor system data fusing is presented and discussed.

Application of SAR and optical data from Sentinel satellites for spatial-temporal analysis of the flood in the region of Bregovo-Bulgaria, 11/03/2018

This research presents the results of a survey on the extent of the affected areas in the municipality of Bregovo from the flood of the Timok River, which occurred on 11.03.2018. The application of SAR and optical data for assessment of the spatial and temporal characteristics of the occurred flood is the objective of this paper. This methodology applies orthogonal transformation of different multispectral images from Sentinel 2 mission combined with SAR data from Sentinel 1. The assessment was made on the basis of the orthogonal transformation’ components of the bands from different multispectral Sentinel 2 imagery: Tasseled Cap Transformation TCT-brightness, TCT-wetness, TCT-greenness. Indicators of quantitative changes in areas affected by the floods have been obtained. Satellite images from Sentinel 1- SAR and Sentinel 2-MSI satellite missions, orthophotography, terrestrial and in-situ terrain data from the Bregovo municipality affected by flooding were used. The processed satellite images are from different sensors and are selected by different dates before and after the day of the natural disaster. Pseudocomposite radar images with different polarization (vv,vh) were used to clarify and more precisely visualize the territorial coverage of the flood in the surveyed area. Various normalized quantitative differentiation indices (vegetation, humidity, vapors, etc., NDGI,) are generated after image processing. Results are presented for correlation between the values obtained for the different data types. On the basis of the obtained data and results, a comparative analysis of the dynamics of the changes occurred as a result of the disaster was carried out and a quantitative assessment of the changes occurred and respectively the registered negative environmental impacts in the territorial extent of the flood.

MODIS satellite data for estimating actual evapotranspiration in Bulgaria (2000-2014)

In the present paper, the actual evapotranspiration for the territory of Bulgaria on an annual basis for the period 2000- 2014 has been modeled, using satellite data MOD16A3. The data was received by means of remote sensing from a MODIS sensor. Raw evapotranspiration (ET) data were integrated and processed in GIS environment in order to obtain the final goal of finding 2D distribution of the qualitative values of the actual evapotranspiration (АET) for the territory of Bulgaria. In order to generate the relevant values for the annual АET in the area, a model was developed using the MOD16A3 evapotranspiration dataset. In the present paper, the actual evapotranspiration was estimated as a function of the land cover and a digital elevation model. The results obtained show the relationship between the actual evapotranspiration, the land cover and a DEM. In the process of research, some trends for the annual quantity of АET were estimated. The model for the quantitative area estimation of the evapotranspiration developed in the study has already been applied to the catchment area of Lefedzha River (located in northern Bulgaria). It was established that the satellite data give a more representative and reliable information on the spatial distribution of the AET on a regional scale. These methods have less human interference in obtaining information about the individual physical parameters on which the process of evaporation depends. In the MOD16 product the temperature of the spreading surface that influences the evaporation processes was recorded.

Using SAR and optical data from Sentinel satellites for precise modeling of seasonal floating reed islands dynamics in Srebarna Lake

The present study is a continuation of the previous monitoring studies on floating reed islands based on remote sensing methods, but this time the study is much more precise in order to create a sustainable operating model for subsequent monitoring studies on this specific type of habitats. The aim of this study is to create a precise model for the movement and dynamic of the floating reed islands in Srebarna Lake. This was done by creating a hybrid model (based on optical and SAR data), assessing the actually condition of floating reed islands, and applying it to quantify of the movement of floating reed islands to perform an actual and seasonal habitats monitoring. To create the hybrid model, the advantages of SAR data – Sentinel-1 for the hydrological dynamics monitoring of Srebarna Lake were used. The SAR data used were obtained for different time periods, within the observed seasons. Multispectral satellite data from Sentinel-2 was also used in order to apply an orthogonal transformation model called Tasselled Cap Transformation (TCT). The Tasselled Cap model is a very effective method for classifying and analyzing processes related to the dynamics of changes affecting the main components of the Earths surface: soil, water, and vegetation. This model proved to be very effective in recognizing specific types of vegetation and habitats, such as floating reed islands and their transformation over a period of time. The results for the reconciliation of TCT images and SAR data define very well the precise boundaries of both the central water body in Srebarna Lake, and the floating reed islands. The results obtained by means of comparative analysis confirm both methods as being equally effective to determine the floating reed islands dynamics in the hybrid model proposed in this study.

Characterization of imaging spectrometers

The increasingly widespread applications of the Earth observation system information and the rapid development of modern technologies over the past decades have led to the imposition of remote sensing tools and, in particular, of the imaging spectrometers. These trends place high demands on the development and improvement of both the systems themselves and the characterization methods. This paper highlights features that are expected to be considered critical for imaging spectrometer performance: The basic steps for characterization of imaging spectrometer critical characteristics are indicated in the flow diagrams. The characterization methods and correspondence correction procedures are determined and the obtained results are presented.