Лямина В.А.   Зольников И.Д.   Глушкова Н.В.   Смоленцева Е.Н.   Королюк А.Ю.  

Methods of mapping and monitoring of heterogeneous natural-territorial complexes

Reporter: Лямина В.А.

When conducting research on natural systems of any territory, we are confronted with spatial heterogeneity of landscapes. In recent decades in the study of spatial organization of ecosystems at all levels of the hierarchy are widely used remote sensing data (RSD). With the increasing volume and variety of satellite imagery actual development of techniques to process them. To conduct field research and data collection was developed unified multi-parameter characteristics of natural-territorial complexes (NTC) [1]. As a result of unification of the standard description of a set of individual quantitative and semiquantitative indicators of features that characterize the typical landscape situation was defined.
The geodatabase, created as a result of complex field descriptions are supplemented by spectral libraries obtained from the satellite imagery of the selected spatial resolution [2, 3]. Spectral libraries are compiled for the contours with the complex describthion. Typically, research on the key sites are maintained for several years, which makes it possible to verify the geodatabase, and the spectral library compiled by the contours.
At the next stage the degree of internal homogeneity of NTC on all channels is estimated. In the case of homogeneity, NTC should have unimodal distribution of brightness characteristics over the entire range of the spectrum. Also analyzed and discarded anomalous values ​​associated with a  neighboring sites occupied by other landscapes, as well as local phenomena. Particular attention is paid to the spatial distribution of pixels in the preliminary results of the classifications.
In the results of this stage "spectral portraits" of elementary sites of the study area are formed. This allows to make a classification tables of reference objects with showing targeted signs that differ on this or that NTK. A refined classification of NTC taking into account of similarities and differences in spectral characteristics is made in the form of a spectral library of the landfill,  conjugated to complex geodatabase.
On the next stage a supervised classification is made with the training of the reference samples. Image processing can be performed using a mask in a few iterations. On large-scale level separately classified forests and grassy landscapes. The proposed technology involves the allocation of heterogeneous NTC using density grids. The use of density grids allows you to: 1) to exclude excessively small objects and to highlight the most significant areas and structures, 2) to use images of different spatial resolution for the joint analysis (Landsat, SPOT); 3) to provide heterogeneous NTC with different percentages of the components.
At allocation of  heterogeneous NTC for each class (elementary NTC), obtained in a supervised classification the density scheme are constructed. The sell size in each case is determined by the spatial resolution of processed satellite image, and the radius of the sliding window is chosen empirically according to the functional scale of the landscape map. In the next step to each pixel is assigned the values ​​obtained from the the density schemes, normalized to the sum of the densities in each pixel.
Thus, each pixel receives the characteristic as a percentage of the densities of NTC. After that distinguished types of combinations NTC. The proposed technology allows us not only to map but also to assess the contribution of different vegetation types (depending on their percentage) in the spectral characteristics of the images. This allows the monitoring of heterogeneous natural-territorial complexes.

To reports list