Complex Analysis of Optic and Radar Data Observed for Forest Landscapes
Savorskiy, Victor1; Zakharov, Aleksandr1; Panova, Olga1; Ermakov, Dmitry1; Zakharova, Lyudmila1; Kaevitser, Vladilen1; Maklakov, Sergey1; Chumachenko, Sergey2

The forests growing on the territory of Russia contain more than on fifth (22%) of the total amount of the world wood. Their conservation and rational management is a necessary condition to ensure environmental security and sustainable development of the world community. At the same time, the provision of information on the status and dynamics of Russian forests still remains insufficient, as the inventory data acquisition for the essential part of them is characterized by relatively low accuracy and frequency. Although informational efficiency of the remote sensing data from satellites for a wide range of applications for forest monitoring is of no doubt among the majority of scientists and experts in the field, the estimation of growing stock still remains a challenge especially on local forest inventory scales for plot sizes lower 1 km. So, the work is devoted to the solution of one of the urgent problems of forest monitoring the applicability of polarization radar measurements in L-band to estimations of the average growing stock on local inventory scales and to combine radar data with optical ones to get better statistics.

The data used in the work are those obtained in observations of forest sites with the L-Band radar with synthetic aperture PALSAR mounted on the Japanese Erath explorer satellite ALOS (Advanced Land Observing Satellite) in 2008-2010. The sessions of measurements in detailed mode in two polarizations HH and HV were selected for analysis. SAR synthesis was performed with the surface spatial resolution of 25 m (compared with potentially achievable one of 10 m) to reduce speckle noise.

First testing site is the territory of an experimental forestry (Dankovskoye and Otradnenskoye district forestry), located at the south of Moscow region along the left bank of Oka River in the two districts Serpukhov and Stupino. The region belongs to the subzone of mixed coniferous and deciduous forests. Second testing site is forestry located in Kirov region with dominated coniferous forests and larger plot sizes than in first case.
There were selected the homogeneous plots of the forest site with the following forest taxation parameters:
1. Species: spruce, pine, birch.
2. Percentage of dominant species in the plot: at least 80% (pure and conditionally pure plots).
3. Completeness of the growing stock (a degree of density of standing trees in the stock, characterizing the percentages of the used space): at least 0.8
4. Age: category one 20 through 60 years; category two 60 through 100 years.
Homogeneous areas were selected according to forest inventory data. Theses areas were accomplished by detailed images of Dankovskoye forestry retrieved from GoogleEarth. It allowed visually to control reliability of forest taxation data by homogeneity and to exclude border pixels, which cover the neighboring areas with significantly different forest type, from analysis.
Obtained results received in summer dry period reveal that RCS values allow easy discriminating between deciduous and coniferous forest plots with a contrast of about 2 dB. The contrast between different coniferous (pine / spruce) is much lower. Nevertheless on HH polarization it is about 0.5 dB. These data were obtained on 31.08.2010 immediately after extremely dry summer of 2010. The presence of radar contrast on HH polarization between deciduous and coniferous plots should be noted for almost all sessions of observations, characterizing the seasonal and interannual variability. The maximum overall variation of radar contrast on HH polarization was registered for areas of spruce. Minimum contrast on HH polarization between the areas of the same species but different ages was shown by pines at winter season. A stable presence of the radar contrast on HV polarization (unlike HH polarization) should be noted between deciduous and coniferous plots. Maximum overall variation of the radar contrast on HV polarization was obtained for young birch. Maximum contrast on HV polarization between the plots of the same species of different age was given by birch plots in summer season.

Retrieved dependencies of averaged RCS values on growing stock reveal that RCS on HH polarization in L-band grows approximately on 0,75 dB in winter season and on 1 dB in summer season when growing stock increases from 0 to 400 cubic meters per Ha. In the same conditions RCS on HV polarization increase on 1,5 dB in summer season. It should be remarked drastic drop of RCS on HV polarization in mid-range growing stocks lower than 50 cubic meters per Ha.

Important testing element is regarded to be an estimation of average contrast between two cases: in presence of controlled object and in its absence. In our case it is observation of forested plots with respect of forestless ones. We chose forestless plots as those which have growing stock lower than 5 cubic meters per Ha. The analysis shows essential contrast about 5 dB for both polarizations both in summer and in winter season. It indicates on potential of L-band SAR for forest exploration in European part of Russia.

It is shown that L-band radar polarimetry imagery of the Earth allows classification of surface in small scales into types of: field, forest, and city. Also it is possible to apply L-band radar polarimetry for small scale discrimination of forests according to dominant tree species (coniferous and deciduous) and delimitation of areas overgrown with bushes. The developed algorithms for determining the parameters of forest ecosystems (type of vegetation, growing stock, age composition, soil moisture etc) could be, according to presented results on test areas, successfully applied for mapping of the indicative characteristics of forests.