Self-organizing Maps (SOM) applied to prediction of free air anomalies obtained by aero and terrestrial gravity
20/09/2016 | 16:45 | Session 3: Local/regional geoid determination methods and models
Author(s): Ana Cristina Oliveira Cancoro de Matos, Denizar Blitzkow, Cleyton de Carvalho Carneiro, Stephen Fraser, Ilce de Oliveira Campos and Gabriel Do Nascimento Guimaraes
Ana Cristina Oliveira Cancoro de Matos, Denizar Blitzkow, Cleyton de Carvalho Carneiro, Stephen Fraser, Ilce de Oliveira Campos and Gabriel Do Nascimento Guimaraes
An initial analysis of a computational tool, Self-organizing Maps (SOM), applied to prediction of free air anomalies obtained by aero and terrestrial gravity is presented. The SOM enables to analyze and visualize data in an n-dimensional space, based on principles of vector quantization and vector similarity measures. The vectors, characterized not only by the values, but also by the orientation in space, are the basis for the analysis of the comparison among them, preserving the respective topological relationship. In this paper, the n-dimensions were built by a grid of samples with spatial resolution of 5’ composed by the following kind of data: (i) digital terrain model (SAM3s_v2); (ii) gravity disturbance of global geopotential model (EIGEN_6C4 to degree and order 200); (iii) free air gravity anomaly obtained from aero and terrestrial gravity. The data cover a region located between 24S to 20S in latitude and 53W to 48W in longitude. This area is of great importance due to recent and high quality terrestrial gravity data available with a spatial resolution of 5-8 km. This research was developed in two experiments where selected windows from available free air anomaly data were removed in the terrestrial information. The SOM method was applied to predict the values in the gaps and the results were compared with the known values. In the first experiment, the information was omitted within the coordinates 23S to 21S in latitude and 51W to 49W in longitude (21.7% of the total area). In the second the information was removed in three areas (21S-20S/51W-50W, 23S-22S/52W-51W and 24S-23S/49W-48W, 17% of the total area). The mean, RMS difference, maximum and minimum value of the difference between the aero and terrestrial free air anomaly in the first test are 2.04, 1.58, 12.28 and -5.79 mGal, respectively. After the application of SOM method, the statistical results from the difference between the terrestrial free air anomaly known and predicted were -0.17, 2.21, 9.65 and -10.44 mGal respectively. In the second experiment, the statistical results of the aero and terrestrial free air anomaly were 2.04, 1.58, 12.28 and -5.79 mGal respectively. The results showed that the difference between the terrestrial free air anomaly known and predicted were -0.03, 2.56, 10.17 and -7.94 mGal, respectively. In the first and second experiments, the determination coefficient (R2) applied to the original data (aero and terrestrial) shows a linear fit of 96%. After applying the SOM method (terrestrial and estimated), the results were 94% (first experiment) and 95% (second experiment) in terms of R2. It will be necessary to generate more experiments increasing the region without gravity information to check the highest percentage of acceptable gaps to predict values using the SOM method. These experiments aim to verify the possibility to expand the use of the method to other regions where the National Petroleum Agency (ANP) has undertaken aerogravity measurements and where no terrestrial gravity data are available due to difficult access, in order to help in the precise geoid model computation for Brazil.