Effective radium-226 activity concentration (ECRa), the radon-222 source term, was measured in the laboratory with 724 topsoil samples collected over a ~110 km2 area located ~20 km south of Paris, France. More than 2100 radon accumulation experiments were performed, with radon concentration measured using scintillation flasks, leading to relative uncertainties on ECRa varying from 10% for ECRa= 2Bq.kg-1 to less than 6% for ECRa>5Bq.kg-1. Small-scale dispersion, studied at one location with 12 samples, and systematically at 100 locations with three topsoils separated by 1 m, was of the order of 7%, demonstrating that a single soil sample is reasonably representative. Agricultural topsoils (n=540) had an average (arithmetic)ECRa of 8.09±0.11 Bq.kg-1, and a range from 2.80±0.22 to 19.5±1.1 Bq.kg-1, whileforest topsoils (n=184), with an average of 3.21±0.14 Bq.kg-1and a range from 0.45±0.12 to9.09±0.55 Bq.kg-1, showed a clear systematic reduction of ECRa when compared with the closest agricultural soil sample. Large-scale organization of ECRa was impressive for agricultural topsoils, with homogeneous domains of several kilometers size, characterized by smooth variations smaller than 10%. These patches emerged despite heavy human remodeling; they are controlled by the main geographical units, but do not necessarily coincide with them. Valleys were characterized by larger dispersion and less organization. This study illustrates how biosphere and anthroposphere modify the soil distribution inherited from geological processes, an important baseline needed for the study of contaminated sites. Furthermore, the observed depletion of forest topsoils suggests an atmospheric radon signature of<br/>deforestation.