This study presents a methodology based on nonlinear regression for landform attributes representation. The equations to estimate the parameters of a two-dimensional polynomial are shown and, for testing the methodology, it was used the data from landform attributes of the state of Minas Gerais (Brazil) obtained by the Digital Elevation Model (DEM) in GTOPO30 project. They form a regular grid, with spacing of approximately 900m. The presented methodology can be used to minimize time of sending landform attributes information through network, to minimize space by storing the parameters of the estimated polynomial, and to make possible the process distribution of the polynomial coefficients calculations to different CPUs over internet network.