ASSESSMENT OF TEMPORAL VARIABILITY FOR AGRICULTURAL LAND BY FRACTAL ANALYSIS OF SATELLITE IMAGERY PUBLISHEDFlorin Sala, Cosmin Alin Popescu, Mihai Valentin Herbei None email@example.com
The study used fractal analysis and remote sensing technique to analyze and describe the temporal variation of an agricultural area. The satellite images (10 images) were achieved in the Rapid Eye system (RGB - 321, False Color - 532), between 28.03 - 31.10 2017. The study was carried out over a total time interval (T) of 218 days, with several partial intervals between the moments of the satellite images acquisition (t), that varied between 10 and 49 days. The fractal analysis was performed on the binarized images, using the box-counting method. Fractal dimensions (D) were obtained in conditions of statistical accuracy (R2 for D=0.999). Fractal dimensions had values ranging between D=1.735 (trial 9, data 29.09) and D=1.810 (trial 1, data 28.03). ANOVA test, single factor, highlighted the existence of variance in the experimental data set and statistical accuracy of the data (F>Fcrit, p<<0.001), under conditions of Alpha=0.001. There were identified very high negative correlations between D and T (r=-0.942). The variation of the fractal dimension (D) with respect to T has been accurately described by a polynomial model of degree 2, under conditions of R=0.951, p<<0.001, and by a model of smoothing spline, under conditions of statistical certainty ( ). In the framework of PCA analysis, PC1 has explained 87.845% of variance, and PC2 has explained 10.688% of variance. Cluster analysis, based on fractal dimensions (D), led to the grouping of the studied cases, associated with the ten moments of time, according to the Euclidean distances, under statistical accuracy conditions (Coph.corr=0.895). Variants were found to be grouped into two distinct clusters. With a high degree of affinity, variants were associated as follows, 4 with 5 (sample data 28.06, 08.07), subcluster C1, variants 7 with 8 (sample data 19.08, 10.09) and variants 9 with 10 (sample data 29.09, 31.10), subcluster C2, respectively variants 2 with 3 (sample data 15.05, 03.06), subcluster C3. Estimating a moment in time T depending on the values of the fractal dimensions (D) was possible on the basis of a model expressed by a polynomial equation of 3rd degree, under conditions of statistical accuracy R2=0.947, p=0.00031.
agricultural land, fractal analysis, fractal dimension, PCA, temporal variability