Population data are generally provided by state census organisations at the pre- defined census enumeration units. However, these datasets very are often required at user- defined spatial units that differ from the census output levels. A number of population estimation techniques have been developed to address these problems. This article is one of those attempts aimed at improving county level population estimates by using spatial disaggregation models with support of buildings characteristic, derived from national topographic database, and average area of a flat. The experimental gridded population surface was created for Opatów county, sparsely populated rural region located in Central Poland. The method relies on geolocation of population counts in buildings, taking into account the building volume and structural building type and then aggregation the people total in 1 km quadrilateral grid. The overall quality of population distribution surface expressed by the mean of RMSE equals 9 persons, and the MAE equals 0.01. We also discovered that nearly 20% of total county area is unpopulated and 80% of people lived on 33% of the county territory.