Estimation kriging techniques were used to analyse data on electrical loads measured at the nodes of 220 and 400 kV transmission network and a 110 kV closed network over the whole area of Poland. The data were for a selected moment in time, i.e. 11 a.m. in the summer season (a Wednesday in July) in 2001. A sample population of N=103 and a much larger sample population of N=1029 were analysed for respectively the 220 and 400 kV transmission network and the 110 kV network. Original databases containing information on: the numbers of successive measurements, the values of coordinates X and Y (specifying the locations where the power measurements were performed) and the investigated parameter - electrical power). First semivariogram function ?(h) was used to analyse the degree and character of electric load variation. Then different kriging estimators, such as ordinary kriging and lognormal kriging, were used to represent the superficial variation in estimated electrical load averages Z over the whole territory of Poland. Isotropic semivariograms, computed for the given moment in time, are best approximated by spherical, exponential and cubic models. Empirical semivariograms based on the power data are characterized by a large share of random component UL (the C0 nugget effect) in the overall load variability, regardless of the power network variant. Then the whole territory of Poland was covered with a grid of 10 km x 10 km elementary blocks in order to estimate electrical power averages Z. For 5776 elementary block centres different geostatistical parameters, including coordinates X and Y (estimated averages Z and standard deviations estimation ? k in their number), were computed. Also the effectiveness and quality of the estimation of the poweraverages Z were examined. Besides the basic geostatistical parameters Z and ? k, also other major parameters, including the sum of positive weights wi assigned to power measurements in kriging, correlation r between the original Z values and estimated averages Z and covariance C of original Z values and estimated averages Z were computed. As a result, new, very comprehensive databases were obtained. They were used to compute raster maps, isoline maps and spatial block diagrams. The spatial visualization of the obtained pictures of power variation for the area of Poland reveals subareas of highly reliable estimation of averages Z and subareas for which the kriged estimates (averages Z) are rather unreliable.