DIGITALNA ARHIVA ŠUMARSKOG LISTA
prilagođeno pretraživanje po punom tekstu




ŠUMARSKI LIST 7-8/2015 str. 71     <-- 71 -->        PDF

At this stage, necessary values have been derived from the metadata from the image. Three different image processing methods were used on the Satellite image.
1. Using the reflection values of each band separately (Band 1, Band 2 etc.)
2. Simple ratio (Band 4/3, Band 5/3, Band 5/3 and Band 5/1, etc.),
3. Normalized ratios are (NDVI) values.
Besides, RapidEye is one of the first commercial earth observation programme utilizing a specific band on board, targeting the changes in chlorophyll contend of plants, Red-Edge (Band4). Thus, other variables derived from Band 4 and its derivatives were produced and used in equations.
After the reflection values of all the pixels of the image were calculated with three different methods, the image was converted to vector data format. As a result of the conversion operation, a point layer, with 3 964 929 pixels and each pixel representing a field of 5m × 5m had been created. 344 sample fields, with their terrestrial measurements complete, carbon stock capacity calculated and numerically stored in the GIS environment had been added as a separate data layer. Ranging from 400 to 800 m2 in size, each sample area held at least 16 and up to 32 pixels. For each field, a single image was needed. So, the averages of reflection values of the pixels that fall within the boundaries of the sample area were used. Reflection value of each sample field was identified by overlaying these layers in the ArcGIS.
RESULTS AND DISCUSSION
Rezultati I Rasprava
Reflection values of the sample areas in the investigated area were calculated with the specified formula and many variables (Band4)2, (Band4)3, (1/ Band4)2, (Band1/Band2), (Band 3/Band4), NDVI, etc.) were obtained as a result of the image enhancement process.
The relationship between these independent variables and the amount of carbon stock of the 258 sample fields was determined using regression analysis (SPSS 13.5 software). For establishing the regression models, curve estimation approach including Linear, Logarithmic, Inverse, Quadratic, Compound, Power, S, Growth, Exponential were all tested. 86 control points were used for controlling the regression model result. Table 1 shows the most successful parameter estimates models.
The Carbon stock of sample pixels (tones/pixels) and scaled reflectance value of (Band 4)2 for the 258 plots is shown in Figure 4.
After the equation producing the highest R2 was selected, this particular equation was first tested with the number of sample points reserved for control, 86 control points. Calculated values were compared to the ground measurements by employing pairwise sample t-test. Under P>0.05 confidence level, no difference was noticeable between them so it was decided the equation was viable. Above ground forest biomass (AGB) from SPOT-5 data, using a non-parametric